Sustainable non-automotive vehicles: The simulation challenges

Renewable and Sustainable Energy Reviews ∎ (∎∎∎∎) ∎∎∎–∎∎∎

Contents lists available at ScienceDirect

Renewable and Sustainable Energy Reviews journal homepage: www.elsevier.com/locate/rser

Sustainable non-automotive vehicles: The simulation challenges Ian Briggs n, Martin Murtagh, Robert Kee, Geoffrey McCulloug, Roy Douglas School of Mechanical & Aerospace Engineering, Queen's University Belfast, BT9 5AH, United Kingdom

art ic l e i nf o

Keywords: Fuel consumption Emissions Simulation Off-highway Hybrid

a b s t r a c t Simulation is a well-established and effective approach to the development of fuel-efficient and lowemissions vehicles in both on-highway and off-highway applications. The simulation of on-highway automotive vehicles is widely reported in literature, whereas research relating to non-automotive and off-highway vehicles is relatively sparse. This review paper focuses on the challenges of simulating such vehicles and discusses the differences in the approach to drive cycle testing and experimental validation of vehicle simulations. In particular, an inner-city diesel-electric hybrid bus and an ICE (Internal Combustion Engine) powered forklift truck will be used as case studies. Computer prediction of fuel consumption and emissions of automotive vehicles on standardised drive cycles is well-established and commercial software packages such as AVL CRUISE have been specifically developed for this purpose. The vehicles considered in this review paper present new challenges from both the simulation and drive-cycle testing perspectives. For example, in the case of the forklift truck, the drive cycles involve reversing elements, variable mass, lifting operations, and do not specify a precise velocity-time profile. In particular, the difficulties associated with the prediction of productivity, i.e. the maximum rate of completing a series of defined operations, are discussed. In the case of the hybrid bus, the standardised drive cycles are unrepresentative of real-life use and alternative approaches are required in the development of efficient and low-emission vehicles. Two simulation approaches are reviewed: the adaptation of a standard automotive vehicle simulation package, and the development of bespoke models using packages such as MATLAB/Simulink. & 2016 Elsevier Ltd. All rights reserved.

Contents 1. 2.

3.

n

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Automotive simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 2.1. Simulation packages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 2.1.1. AVL CRUISE. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 2.1.2. AMESim . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 2.1.3. ADVISOR. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 2.2. Other simulation packages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 2.2.1. ECOGest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 2.2.2. CMEM. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 2.2.3. Adams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 2.2.4. PSAT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 2.2.5. MSC Easy5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 2.2.6. Dynacar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 2.2.7. MapleSim . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 2.2.8. VHDL-AMS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 2.3. Automotive simulation summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 Non-automotive simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 3.1. On-highway vehicles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 3.2. Off-highway vehicles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

Corresponding author. E-mail address: [email protected] (I. Briggs).

http://dx.doi.org/10.1016/j.rser.2016.02.018 1364-0321/& 2016 Elsevier Ltd. All rights reserved.

Please cite this article as: Briggs I, et al. Sustainable non-automotive vehicles: The simulation challenges. Renewable and Sustainable Energy Reviews (2016), http://dx.doi.org/10.1016/j.rser.2016.02.018i

I. Briggs et al. / Renewable and Sustainable Energy Reviews ∎ (∎∎∎∎) ∎∎∎–∎∎∎

2

3.3.

Simulation modes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 3.3.1. Backward-facing simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 3.3.2. Forward-facing simulation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 3.3.3. Choice of simulation mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 3.4. Non-automotive simulation summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 4. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

1. Introduction Accurate simulation of vehicle powertrains is an essential part of the study into reducing fuel consumption and emissions of onand off-highway vehicles. The ability to simulate vehicles on a component level allows for models of complex architectures to be built and validated. Once validated, these models can then be used to predict how changes to the vehicle's powertrain affect fuel consumption. An accurate vehicle model leads to several advantages from a research and development viewpoint; the most important of which is a reduction in the amount of physical prototyping, as changes to powertrain components can be simulated based on physical data which may be obtained from manufacturers or bench testing of individual components. This reduction in wholevehicle physical prototyping leads to an associated reduction in development times and costs, and allows several components to be investigated in a virtual environment; the most suitable may then be selected for installation and testing on the real vehicle. Validation of these models may be carried out by comparing the performance of the whole vehicle, or individual systems within it, to experimentally-obtained data. This usually takes the form of drive-cycle analysis, where an instrumented vehicle is driven on a pre-defined course and parameters such as vehicle speed, driver inputs, and individual component data such as pressure, speed, position, etc. are recorded. The simulated vehicle can then be compared to the experimental data, thus highlighting areas where the model must be improved. A properly validated model allows the vehicle designer to then investigate how changes to the vehicle or its powertrain affect parameters such as fuel consumption and emissions by including these changes in the simulation and re-running the validated drive cycle. Within the automotive sector, simulation has been a widely used tool for decades, mainly stemming from the simulation of engines and their application to automotive vehicles [1–4]. This has allowed advanced models to be built, while a comprehensive database of literature and vast array of experimental data has allowed these models to be easily validated and compared. The amount of literature available also goes some way to defining a standard approach to modelling automotive vehicles. However, within the non-automotive sector, which in this paper will include buses, trucks and off-highway vehicles, the development of advanced simulations has been much more recent; therefore the volume of published literature is severely limited. This makes it all the more difficult for the non-automotive vehicle designer to use a previously tested approach to develop validated simulations. This paper begins by outlining the approach taken in a number of automotive simulations: the choice of software environments available, the approach taken and the challenges encountered. This leads on to a discussion of the various approaches taken to adapt these automotive models to non-automotive applications, and how bespoke models have been created for specific scenarios. The

simulation of hybrid vehicles is discussed throughout both automotive and non-automotive sections, and in particular the suitability of various software packages in adapting to the challenges associated with hybrid vehicle architecture is discussed. The paper focuses on the simulation of two particular cases studies in the non-automotive sector: hybrid city buses, and forklift trucks, both of which provide numerous simulation challenges. The conclusions of the paper aim to offer the reader a summary of the simulation approaches taken in the various applications discussed and some of the obstacles that can be expected when simulating non-automotive vehicles. This allows the reader to take a more direct approach to developing a simulation strategy depending on the application under study.

2. Automotive simulation Simulation of automotive vehicles is a well-established field of study and there is a wealth of commercial and open-source software available [5]. This section discusses some of the main packages available and shows how they have been applied to automotive simulations. The suitability of each package to simulate hybrid vehicles will also be discussed in each subsection where literature on the subject was available. 2.1. Simulation packages 2.1.1. AVL CRUISE This commercial automotive simulation software allows the user to create a graphical model of the vehicle, hiding from view all but the main mechanical connections and modules. Blackmore et al. [6] used CRUISE to create a model of an automotive vehicle. The basic vehicle model, including tyres, brakes, differential, transmission, engine, and a driver model, was constructed as shown in Fig. 1. This model was then compared to one of a hybrid vehicle where CRUISE was able to produce detailed analysis of battery state-of-charge and showed instantaneous battery consumption throughout the ‘Japanese 10–15’ drive cycle. It also allowed a detailed study of battery technology to be carried out, with three different battery types used in the vehicle model (NiMH, Lead-Acid and Li-ion). A similar parametric study was carried out by Oh et al. [7], who used CRUISE to model SI (spark ignition) and CI (compression ignition) automotive vehicles over seven types of automotive drive cycle. The authors mentioned CRUISE's prediction of CO2 emissions is not based on maps, and instead it assumes fuel is completely combusted. Therefore, the emissions rate is related to the type of fuel used and the calculated instantaneous rate of fuel burned. The results of this paper show excellent correlation between the simulations and experimentally obtained data. Parameters such as engine speed were predicted very accurately on all the drive cycles investigated, as were fuel consumption and CO2 emissions. In addition, the authors were able to perform a comprehensive parametric study of each vehicle analysed and

Please cite this article as: Briggs I, et al. Sustainable non-automotive vehicles: The simulation challenges. Renewable and Sustainable Energy Reviews (2016), http://dx.doi.org/10.1016/j.rser.2016.02.018i

I. Briggs et al. / Renewable and Sustainable Energy Reviews ∎ (∎∎∎∎) ∎∎∎–∎∎∎

Nomenclature ABS AVL CI CMEM CO2 CVT ECT FMI FTP HCCI HiL ICE Li-ion

Anti-lock braking system Anstalt für Verbrennungskraftmaschinen List GmbH Compression Ignition Comprehensive Modal Emission Model Carbon Dioxide Continuously Variable Transmission European Transient Cycle Functional Markup Interface Federal Test Procedure Homogeneous Charge Compression Ignition Hardware-in-the-Loop Internal Combustion Engine Lithium-ion

3

LPG MBD MLTB NEDC NiMH NOx PID PSAT SI SORT TNO

Liquefied Petroleum Gas Model-based Design Millbrook London Transport Bus New European Drive Cycle Nickel-metal Hydride Oxides of Nitrogen Proportional-Integral-Differential Controller Powertrain System Analysis Toolkit Spark Ignition Standardised On-road Test Nederlandse Organisatie voor Toegepast Natuurwetenschappelijk Onderzoek VDI Verein Deutscher Ingenieure VHDL-AMS VHSIC Hardware Description Language (analogue and mixed-signal extensions)

Fig. 1. Typical vehicle model structure using AVL CRUISE (adapted from [6]).

concluded that vehicle weight had the largest effect on fuel efficiency, regardless of vehicle type or drive cycle. As expected, rolling resistance, frontal area and air resistance only showed an effect on fuel efficiency on drive cycles that included periods of high travel speed. Pacheco et al. [8] created a simulation of a Ford Focus over the NEDC (New European Drive Cycle) using AVL CRUISE. In this study, the authors investigated the performance of a HCCI (Homogeneous Charge Compression Ignition) engine, designed to reduce the NOx emissions compared to the equivalent SI engine. A composite fuel consumption map was used to enable HCCI operation over a small operating window – HCCI was not implemented as a modification to the engine within CRUISE – instead, a portion of the fuel consumption and emissions maps was modified in the region where HCCI was active. Extra attention was required on the design of the exhaust system as the lower exhaust temperatures caused by HCCI required modifications to the catalyst model. This approach allowed a detailed study of toxic emissions and fuel consumption over the NEDC drive cycle to be carried out. CRUISE can also be used to analyse the performance of individual drivetrain components; Banjac et al. [9] created a model of a

1.6-litre diesel automotive vehicle to assess its thermal management. The model used a 0D engine model to study engine parameters on a crank angle-resolved basis. The simulation included turbocharger and catalyst models, and used a Wiebe combustion model. The authors included detailed models of oil system, cooling system and the overall gas path in the engine - a highly detailed model, all contained within CRUISE. The model accurately predicted fuel consumption and vehicle acceleration, as well as oil and coolant temperature over multiple drive cycles. Even details such as cylinder liner and head temperatures were predicted accurately, as was exhaust gas temperature, proving the ability of CRUISE to provide detailed and accurate results about individual powertrain components and overall vehicle performance. Srinivasan and Kothalikar [10] used CRUISE to enable fuel consumption and emissions prediction of a 3-cylinder SI-engine automotive vehicle over typical drive cycles. The authors also modified various components of the drivetrain and examined the effect this had on fuel consumption. Again, the authors showed not only good agreement with experimental fuel consumption data, but also in acceleration prediction, driveability and the maximum speed achieved in each gear. After the validation results

Please cite this article as: Briggs I, et al. Sustainable non-automotive vehicles: The simulation challenges. Renewable and Sustainable Energy Reviews (2016), http://dx.doi.org/10.1016/j.rser.2016.02.018i

4

I. Briggs et al. / Renewable and Sustainable Energy Reviews ∎ (∎∎∎∎) ∎∎∎–∎∎∎

were presented, the authors showed the effects on fuel consumption and vehicle performance of changing parameters such as final drive ratio, reduced aerodynamic drag and using a different engine. The authors' conclusions included a statement affirming their belief in the suitability of CRUISE to this task. Srinivasan et al. [11] examined the use of CRUISE on a 4cylinder diesel-powered automotive vehicle, again for the purposes of studying fuel consumption and greenhouse gas emissions. The model showed a similarly good agreement to experimentally measured drive cycle data. A similar parametric study to that of [10] was carried out on various components of the powertrain and vehicle, and demonstrated the performance of CRUISE. Optimisation of the powertrain models also showed a reduction in CO2 emissions. 2.1.2. AMESim Another commercial simulation package used for vehicle modelling is LMS AMESim developed by Siemens. This is another graphical modelling environment that allows the user to define a vehicle powertrain by splitting its components into a block structure. Sun et al. [12] presented a model of a 1.6-litre automotive vehicle created in AMESim. The vehicle parameters were specified, including tyre radius, vehicle mass and frontal area and fuel type, among others. The vehicle powertrain was then assembled graphically, selecting the required components from an inbuilt database of parts as necessary. Each component within AMESim allows the user to specify detailed information such as number of cylinders, swept volume, and individual gear ratios. The simulation investigated different shift strategies to be developed, and the authors presented an optimised shift pattern which produces a 12% reduction in fuel consumption over a simple acceleration test. Xie et al. [13] presented an automotive vehicle with a CVT (continuously variable transmission), modelled using a cosimulation approach with MATLAB/Simulink coupled to AMESim. In this instance, AMESim was used to provide a simple interface for the vehicle model, including the driver, engine, transmission and tyres, while Simulink was used to control all on-board hydraulic systems associated with the transmission; functionality which was not available within AMESim. Each component of the vehicle model was supplied to AMESim in equation form, while the hydraulic circuit was simplified for simulation within Simulink. The transmission control system was validated during a brief driving cycle; this showed the transmission ratio changed as the authors desired. This paper showed that the off-the-shelf software was unable to model all aspects of the vehicle, and where detailed control of subsystems is required Simulink often provides a better solution. This theory was also discussed in a similar co-simulation approach presented by Vijayagopal and Rousseau [14], who described a model of an automotive vehicle using a combination of an engine model created using GT-Power, a transmission model which used AMESim, and a vehicle dynamics model constructed in CarSim. It is suggested that it is better to use the advantages provided by these expert software packages than to create the full model from scratch using Simulink. The paper describes the integration of the above tools through a specially-developed tool called Autonomie, which allows for an almost completely automatic process to integrate the output of each model with the input of the next, as seen in Fig. 2. Autonomie takes advantage of the ‘export-to-Simulink’ function of each of the individual pieces of software and builds a wrapper around the output of each part of the model to integrate it in a common interface. Because there is a high amount of detail captured by each part of the model, it was suggested that

Fig. 2. Autonomie vehicle model (adapted from [14]).

investigations into fuel consumption could focus on smaller components within the powertrain, engine and overall vehicle platform. Tavares et al. [15] studied a power-split hybrid vehicle, similar to the Toyota Prius, except this concept used a hydraulic accumulator rather than electrical storage. As with the study presented by Vijayagopal and Rousseau [14], the hydraulic components had to be modelled externally; an AMESim model of a variabledisplacement turbocharged engine was coupled to a model of the hydraulic system, powertrain and vehicle modelled in Simulink, showing the ability of AMESim to be coupled in a multidisciplinary co-simulation. This Simulink model of the hydraulic circuit included equations to simulate the hydraulic accumulator and the hydraulic pump and motor. The fuel consumption of the hybrid vehicle was measured over the urban FTP-75 drive cycle and was compared to a model of a vehicle with a conventional powertrain and a fixed-displacement SI engine. The authors showed great improvement in fuel consumption via the use of the variable displacement engine and hydraulic hybrid platform, and state that is was necessary to follow the co-simulation route with AMESim and Simulink; this allowed for the inclusion of a suitable engine and energy storage controller within Simulink, which required a novel operating strategy to be implemented. 2.1.3. ADVISOR The open-source powertrain modelling package ADVISOR is based on a MATLAB/Simulink back-end and aims to provide the user with a simpler, graphical interface. A selection of ICEs is provided and can be selected from a drop-down menu, as can components such as transmission systems, clutch, axles and exhaust after-treatment. In a study of the suitability of ADVISOR to model vehicles, Markel et al. [16] say there is limited scope for user input in these parameters, though the in-built components can be edited by editing the Simulink block associated with each. The authors conclude that while ADVISOR is well-suited to analysing vehicle performance over drive-cycles and standardised tests, there are numerous limitations which can affect the design of dynamic systems. These include an inability to predict vibrations, vehicle dynamics and the performance of electrical systems, for which external or add-on software is required.

Please cite this article as: Briggs I, et al. Sustainable non-automotive vehicles: The simulation challenges. Renewable and Sustainable Energy Reviews (2016), http://dx.doi.org/10.1016/j.rser.2016.02.018i

I. Briggs et al. / Renewable and Sustainable Energy Reviews ∎ (∎∎∎∎) ∎∎∎–∎∎∎

MacBain et al. [17] also discussed the need for external software added on to an ADVISOR simulation of a series hybrid automotive vehicle. In particular, the authors discussed a collaborative effort to model the mechanical system using the ADVISOR software package, while both the high- and low-voltage electrical systems were co-simulated using Saber, a physical modelling domain developed by Synopsys, which allowed detailed analysis of the power flow within the vehicle. The authors suggested that this co-simulation arrangement allowed for better analysis of component and subsystem interaction, using purpose-built software to model each part of the vehicle. The vehicle modelling software did not provide an adequate solution for the equations governing the electrical performance of the vehicle, and hence a dedicated software package was used and was closely-coupled to the mechanical vehicle model. The vehicle model calculated the power needed to meet the drive cycle requirements, and this power demand was then sent to the electrical model on each timestep of the simulation to calculate the electrical motor torque and speed, which allows the motion of the vehicle to be calculated. ADVISOR was also used by Silva et al. [18] to model automotive vehicles under urban driving conditions. The authors showed that short-term events during the drive cycle, which produced a peak in toxic gas emissions, were difficult to simulate effectively. The prediction of toxic gas emissions on a light-duty cycle was within 10% of measured values. Sulaiman et al. [19] reviewed several papers which used ADVISOR to model hybrid vehicles. Several weaknesses of the software were exposed, including difficulties predicting energy losses throughout the drivetrain and errors in dynamic modelling. Complications due to the interaction of subsystems were highlighted as the main source of error in the models. This shows the difficulty of simulating complex multi-domain systems with commercial software that does not allow the scope to include detailed characteristics of each system. 2.2. Other simulation packages As well as the evaluation of ADVISOR discussed above, Silva et al. [18] also compared their model to those created in two other simulation packages: EcoGest [20] and CMEM [21]. 2.2.1. ECOGest This is a simulation method written in Visual Basic and uses twenty input parameters which together detail the vehicle's engine, transmission, exhaust after-treatment and occupancy, as well as environmental factors and road topography. A database of various engine maps that detail fuel consumption and toxic emissions is built in to the software. The software is designed to allow analysis of fuel consumption and emissions during different phases of a drive cycle, i.e. idle, acceleration, cruise and deceleration. 2.2.2. CMEM This package allows only very specific types of light-duty vehicle to be simulated, and operates on a similar principle to that of ECOGest, whereby vehicle characteristics and components are specified from a supplied database. Only gasoline or diesel fuels may be modelled, and Silva et al. [20] reported that it would not be trivial to model a different type of engine, since the input replies on experimental data to calibrate the model. Despite the limitations of these two packages, the authors showed they performed similarly to their ADVISOR simulation in terms of prediction of emissions.

5

2.2.3. Adams This multi-body dynamic simulation tool was developed by MSC Software. Dyer et al. [22] and Bowles et al. [23] show how Adams was used in automotive simulations, providing a dynamic vehicle model for co-simulations with other powertrain/component software. In both these studies, significant modifications were required to allow the powertrain software to interface with the Adams vehicle models. Both sets of authors reported the successful implementation of an Adams-based simulation to analyse individual powertrain or vehicle components, in conjunction with suitable separate powertrain models such as PSAT (Powertrain System Analysis Toolkit). 2.2.4. PSAT This is primarily a transmission model with a graphical frontend, working with a MATLAB-based forward-facing environment. Rousseau et al. [24] reported a PSAT-based powertrain model for hybrid automotive vehicles. With PSAT, the user selects the drivetrain components, but rather than graphically connecting them, the back-end of the software automatically compiles a model at the start of each simulation. This poses some limitations in that the order of drivetrain components cannot easily be modified, but the authors reported it allows for faster model creation and reduced user input. Manzetti and Mariasiu [25] also discussed the use of PSAT in studying hybrid vehicle emissions, battery performance and running costs. The software was ideally suited to switching simple components within the vehicle's architecture and coped well with the interactions between electrical and mechanical subsystems without the need for co-simulation. 2.2.5. MSC Easy5 Ricardo [26] described the use of this whole-system simulation tool for automotive modelling. As part of this study, models of various automotive vehicle platforms were developed, varying from a small A-segment car to a full-size pickup truck. Easy5 allows features such as hybridised powertrains to be simulated, and the authors showed how an electric hybrid architecture could be added to the vehicle, along with a study into optimised control systems, engines and transmission systems. 2.2.6. Dynacar This is a simulation tool based on National Instruments Labview software. Iglesias [27] studied an advanced driver assistance system using Dynacar, and highlighted one of the main features of the software: its Labview background allows for driver-in-the-loop simulation. In this study, a real driver was used and the modelling software showed how the optimisation of driver assistance systems could reduce fuel consumption. 2.2.7. MapleSim In a study focused on MBD (model-based design), Komoto and Masui [28] presented a multi-domain model of an automotive vehicle using MapleSim. Although primarily concerned with how to classify each subsystem to allow for fast analysis and design, the vehicle model was used to successfully design an electric automotive vehicle. 2.2.8. VHDL-AMS Originally developed to model electronic systems, but subsequently extended to include mechanical, magnetic, thermal and fluidic environments, Aydin et al. [29] applied this code to simulate an automotive powertrain. A model of the driver was built to compare desired and actual vehicle speeds, based on a drive cycle input, and adjusted accelerator, brake and gear commands accordingly. The engine was modelled as a mapped relationship between engine speed, accelerator pedal position and engine

Please cite this article as: Briggs I, et al. Sustainable non-automotive vehicles: The simulation challenges. Renewable and Sustainable Energy Reviews (2016), http://dx.doi.org/10.1016/j.rser.2016.02.018i

I. Briggs et al. / Renewable and Sustainable Energy Reviews ∎ (∎∎∎∎) ∎∎∎–∎∎∎

6

output torque. Detailed models of the clutch and gearbox were also included, which calculated parameters such as the friction torque of the clutch and power losses in the gears. The brake system was also modelled in a similar manner to the clutch; a series of equations to calculate braking torque applied to the vehicle. A simple model of the chassis was also included based on Newton's second law of motion. The authors presented a completed powertrain model and showed that it could predict vehicle behaviour over the NEDC drive cycle, but no validation results were presented. Availability of such an array of software allows fast development of a model for a particular application. In many automotive studies, including some of those discussed above, modifications to engine components [4,8], novel powertrains [13], new fuels [16], are all easily developed using these tools. Hybrid vehicles are also easily modelled in such packages as the vehicles become more advanced. In particular, some of the software packages listed above can easily be applied to electric hybrid powertrains, usually in conjunction with a MATLAB/Simulink module to control the hybrid aspects of the powertrain [17,30–33]. However, many studies of both electric and nonelectric hybrid automotive vehicles tend to use MATLAB/Simulink standalone models [34–40], due to the ease of working wholly in one software package, thus avoiding the need to use an interface between the vehicle model and the hybrid model or controller. This is particularly important if the vehicle modelling package does not have the capability to model certain types of hybrid component. Similarly, a number of authors justified the use of MATLAB/Simulink by the flexibility it allowed in constructing an accurate and customised model of the vehicle under study. Damiani et al. [35] constructed a modular hybrid vehicle model which allowed characteristic parameters to be altered in order to simulate various vehicle hardware configurations. This allowed an energy analysis to be performed for various vehicles over a selection of representative automotive drive cycles. Berard et al. [39] used a similar modular construction in Simulink but developed a control strategy, which modified shift strategy to optimise fuel consumption and catalyst performance over the NEDC drive cycle. Alzuwayer et al. [40] modelled a hybrid vehicle but despite attempting to modify control systems, no dynamic effects were included in their Simulink model; rather, steady state maps were used to investigate component performance. MATLAB/Simulink is also well-suited to the development of control systems. For hybrid vehicles, several sources showed that much of the performance gain in terms of fuel consumption is dependent on the development of an advanced control system [24,31,41,42]. 2.3. Automotive simulation summary The preceding section of the paper has focused on the simulation of automotive vehicles. For standard powertrains, the choice of software is vast; ranging from long-established commercial software to community-developed open-source packages. The choice of which package to use will depend largely on user experience, financial commitment and specific modelling requirements such as the prediction of emissions or fuel consumption. For more complex modelling situations, such as hybrid vehicles, some of these established packages may be used, occasionally with an external add-on to model hybrid architectures or external controllers in a so-called co-simulation routine. However, this added modelling complexity tends to drive researchers towards developing simulations in the MATLAB/Simulink environment where they can have full control over all aspects of the model; particularly in a hybrid vehicle simulation, where many small

elements of the vehicle architecture are under scrutiny, or an advanced control system is required.

3. Non-automotive simulation While automotive simulation has been shown above to be comprehensively documented, there is much less of a standardised approach available when simulating non-automotive vehicles. As will be described below, simulation packages are often twisted or contorted to perform in ways they were not intended, and add-ons, and co-simulation methods employed to make the models meet the simulation requirements. 3.1. On-highway vehicles Some of the commercial software designed for automotive simulation has been successfully applied to non-automotive applications. For example, in a study of long-haul trucks, Lutsey et al. [43] used ADVISOR to investigate the fuel consumption savings possible by using fuel cells as an auxiliary power source when the engine was idling. The authors reported that the software needed to be substantially modified to model the truck's architecture including the creation of representative drive cycles, creation of suitable engine fuel consumption maps, inclusion of fuel cell performance data, all specific to the vehicle under study. While primarily concerned with cost analysis, this study produced useful simulation results comparing a baseline vehicle to one that used fuel cells, reporting an 80% reduction in fuel consumption during avoidable idling periods; periods where the truck is stationary but the engine is needed to provide power to on-board equipment, climate control, etc. A study by Xiu-qin et al. [44] into the ABS (anti-lock braking system) of a multi-axled truck used ADAMS to simulate the vehicle in conjunction with a MATLAB/Simulink ABS control system; the use of MATLAB was due to the inability of ADAMS to model the ABS system by itself. Sandberg [45] used DYMOLA to model a heavy-duty truck. The modular construction of this simulation allowed each component of the vehicle, including chassis, trailer and cargo to be modelled individually. This allowed a study of rolling resistance to be conducted and fully validated. Lajunen [46] created a simulation of a hybrid heavy duty truck using Autonomie using some of the inbuilt pre-defined vehicle component libraries, including a range of six-cylinder CI engines and transmission components. The authors had to create the hybrid architecture from scratch as no inbuilt library components existed; this included batteries, an electric motor, and a torque coupler to combine the conventional and electrical power sources. A comprehensive parametric study into the hybrid components was conducted, though no validation results were presented. Ribau et al. [47] used ADVISOR to model an urban bus architecture. The authors reported that the bus could be modelled entirely within ADVISOR and showed good validation results over two urban drive cycles. Similarly, Li et al. [48] studied a fuel cell urban bus using the Easy5 simulation package. In this instance, the authors created a detailed model of the fuel cell, which was validated against experimental measurements. Models of the battery, motors and tyres were also included, as was a simplified vehicle model consisting of equations for external forces due to aerodynamic resistance and grade, and the fore and aft weight transfer of the vehicle during accelerations. A vehicle control strategy was also modelled, using battery SOC (state-of-charge) and fuel cell power limits as the main control parameters. Optimisation of the power distribution between battery and fuel cell provided the bulk of the work in this study to improve operational efficiency.

Please cite this article as: Briggs I, et al. Sustainable non-automotive vehicles: The simulation challenges. Renewable and Sustainable Energy Reviews (2016), http://dx.doi.org/10.1016/j.rser.2016.02.018i

I. Briggs et al. / Renewable and Sustainable Energy Reviews ∎ (∎∎∎∎) ∎∎∎–∎∎∎

While some of the non-automotive literature uses these tools, perhaps supplemented with some other software (usually Simulink) for modelling more complex architectures, i.e. battery modelling, hybrid vehicles, energy recovery, control systems, etc., most studies uncovered for this paper take the approach of using a custom-built Simulink model. Various studies of medium- or heavy-duty trucks were found which used a Simulink vehicle model. Feng and Huang [49] modelled a hydraulic hybrid vehicle including a hydraulic pump, motor and accumulator, co-simulated with a Simulink-based engine model. The authors reported good validation results over the FTP-72 drive cycle, and showed that by using engine stop/start, and using the accumulator to provide drive, the hybrid vehicle could reduce fuel consumption by up to 51%. A similar hydraulic hybrid truck using an accumulator was also studied by Wu et al. [50], who again used MATLAB/Simulink to model the vehicle and investigate power management strategies over drive cycles. The results showed that a 68% reduction in fuel consumption could be expected by carefully defining the gear change strategy and controlling the accumulator state-of-charge. Midgley et al. [51] also studied a hydraulic hybrid truck, this time with hydraulic regenerative braking. A Simulink vehicle model was used to show between 11% and 17% reduction in fuel consumption over three urban drive cycles. The analysis and simulation of city buses is becoming more common as operators and manufacturers seek to improve the performance of their vehicles over the various drive cycles on which the buses operate. The field of energy recovery is also growing rapidly with respect to city buses and the simulation of hybrid buses is becoming ever more important. Simpson et al. [52] and Berta et al. [53] both created computational models of diesel-electric hybrid buses using MATLAB/ Simulink and FORTRAN respectively. In both instances, a conventional powertrain was modelled as a baseline vehicle, before electrical storage was included in the model to produce a hybrid powertrain. The ability to model each system separately as a series of equations allowed considerable complexity and detail to be included in the models; elements such as battery performance, electrical drivetrain components and various hybrid layouts. The model developed by Simpson et al. [52] is shown in Fig. 3. This shows the data flow path and the interactions between various subsystems. This model simulated the mechanical components of the drivetrain as well as the electrical and thermal subsystems including batteries, electrical generator and components associated with regenerative braking. This highlights the level of detail that can be built in to a mathematical model of a complex hybrid vehicle. De Lorenzo et al. [54] also studied a hybrid bus using Simulink and included a polymer electrolyte fuel cell as the hybrid power

Fig. 3. Hybrid bus model architecture (adapted from [52]).

7

source; battery modules, fuel cell and electric motor units were all included in the model. Sulaiman et al. [19] reviewed a series of simulation methodologies, including a MATLAB/Simulink study of the braking strategy of an urban hybrid bus, and an energy management strategy for hybrid electric vehicles. In both instances, the suitability of the software was shown in the control it offered the authors to create bespoke models of complex vehicle architectures. When considering energy recovery, it is also possible to use an external engine-modelling package connected to a MATLAB/ Simulink vehicle/powertrain model. This approach allows changes to the engine to be analysed in the context of the drive cycle performance of the bus. Trajkovic et al. [55] used MATLAB/Simulink vehicle model connected to a GT-Power engine model to simulate a pneumatic hybrid bus, while Briggs et al. [56,57] used a similar approach, using a MATLAB/Simulink vehicle model indirectly connected to a Ricardo WAVE engine model. In this case, exhaust energy recovery was modelled in the engine simulation package, and the resulting fuel consumption maps were supplied to the vehicle model to represent the engine. Bao and Stobart [58] studied a pneumatic hybrid bus which also recovered energy through regenerative braking. The simulations were conducted using MATLAB/Simulink and the authors showed the potential to reduce fuel consumption by up to 7% over the Braunschweig [59] and MLTB [60] drive cycles; both these cycles are typically used for bus analysis. Briggs et al. [61] discussed the performance of a hybrid bus with exhaust energy recovery using the SORT (standardised on-road test) cycles [62]. These cycles are split into three separate tests at various speeds; the first repetition of each cycle is shown in Fig. 4a. Typically on a vehicle test, the SORT cycles are repeated up to ten times to allow a valid comparison with other cycles. However, these cycles are not representative of real-life applications and consequently, operators and manufacturers have attempted to create better tests, such as MLTB [60,63] (based on London bus route 159) as shown in Fig. 4b. This cycle, the first 200 seconds of which are shown, is much more realistic of real world conditions and includes multiple stop-start events. Other similar cycles have been developed for bus use, including the ETC [64] and TNO [65] cycles. These standardised tests allow manufacturers and operators to analyse performance, as well as fuel consumption and emissions of different vehicles over comparable test cycles. This detailed analysis of emissions is particularly important with the adoption of ever more strict emissions limits, i.e. EURO VI standards [65,66]. 3.2. Off-highway vehicles For this category of vehicle, there is a relative scarcity of material to review. However, the need to produce accurate models of off-highway vehicles has been discussed for some time. Schutz et al. [67] examined the need to produce accurate drive cycle simulations when designing off-highway earthmoving vehicles, and the inherent difficulties and inaccuracies associated with attempting to standardise such an unpredictable drive cycle. Savaresi et al. [68] presented a paper on the control of a CVT for agricultural tractors, but no direct simulation of the powertrain was conducted. Bietrasto et al. [69] conducted an equation-based simulation of an agricultural tractor using Newton's second law to express equations for the vehicle's dynamics. The aim of the study was to calculate transmission efficiency for three different tractors over an acceleration test. However, in the authors' dynamic vehicle model they considered the tractor as a particle, rather than modelling each component of its powertrain; efficiencies presented were therefore mean values. Rahmfeld and Klocke [70] simulated two types of mobile machinery: a 20 t harvester and a 15 t crawler, with a focus on the

Please cite this article as: Briggs I, et al. Sustainable non-automotive vehicles: The simulation challenges. Renewable and Sustainable Energy Reviews (2016), http://dx.doi.org/10.1016/j.rser.2016.02.018i

8

I. Briggs et al. / Renewable and Sustainable Energy Reviews ∎ (∎∎∎∎) ∎∎∎–∎∎∎

Fig. 4. Typical bus drive cycles. (a) Standardised on-road test cycles (b) MLTB drive cycle.

efficiency of the hydraulic pump used in the transmission. For this, the authors created a model of each vehicle using a validated inhouse simulation package. The effect of different pumps on the overall system efficiency was demonstrated. Wang and Wang [71] modelled a hybrid hydraulic excavator using a modular equation-based simulation constructed in AMESim. This included a model of the vehicle's hydraulic system, controllers and a super-capacitor for energy storage. A similar study into an excavator using more conventional battery energy storage was conducted by Wang et al. [72] using Simulink to study three different powertrain configurations: a baseline conventional powertrain, a series hybrid and a parallel hybrid. The area of forklift trucks provides a little more literature on vehicle simulation. Murtagh et al. [73] studied a forklift truck in order to analyse fuel consumption and drive cycle performance. AVL CRUISE was used to simulate the vehicle's powertrain and MATLAB/ Simulink was used to model the hydraulic architecture of the vehicle's lifting equipment and the various control systems used. The CRUISE and Simulink components of the model were dynamically linked via co-simulation to allow the systems to interact in real time. The authors of this paper also discussed the difficulties in simulating an off-highway drive cycle to measure performance; unlike automotive cycles, which use a velocity versus time profile to specify the cycle, forklift truck cycles, such as VDI 2198 [74], specify the route to take in the form of distances between checkpoints, the load to carry on the forks, and how many repetitions to complete in a certain time. In addition, the drive cycle includes a number of hoisting, tilting and lifting operations that do not occur in automotive drive cycles. The VDI 2198 drive cycle is shown in Fig. 5. Forklift productivity is another important aspect of performance; this is the maximum possible number of cycles performed in a given time. However, no standardised cycles exist to measure productivity and they are instead generally created by manufacturers or operators to compare truck performance. This may take the form of a VDI 2198 cycle driven as fast as possible, or alternatively a cycle more representative of an operator's requirements may be created. Stein et al. [75] developed a validated model of diesel and LPG forklift trucks using Simulink as part of a procedure to allow developers to correctly select engines based on application and highlighted one of the challenges associated with the truck's operation; typically a forklift truck's engine operates across the full range of its torque map depending on the functions it is performing, as shown in Fig. 6.

Fig. 5. Aerial view of VDI 2198 drive cycle (adapted from [73]).

Fig. 6. Typical forklift truck operating profile (adapted from [75]).

Thus, when operating at low engine speed and high torque, for example when lifting or tilting with heavy loads on the forks, the engine must not be allowed to stall. Likewise, acceleration performance requirements often demand the engine operates at full speed. Often, in order to develop hybrid forklift trucks such as those described by Ogawa et al. [76], a bespoke Simulink model must be constructed. Bohler et al. [77] studied a diesel-electric forklift truck using a Simulink model and showed that the control strategy required a lot of development due to the interactions of the multiple truck systems – powertrain, lifting mechanisms, regenerative braking, electrical storage, etc. Minav et al. [78] modelled an electro-hydraulic forklift truck using MATLAB/Simulink. The authors described a comprehensive model for the creation of the hydraulic lifting and lowering mechanisms; this included the electrical servo, hydraulic cylinder and hydraulic pump and motor. This work was extended in [79] to include an analysis of energy

Please cite this article as: Briggs I, et al. Sustainable non-automotive vehicles: The simulation challenges. Renewable and Sustainable Energy Reviews (2016), http://dx.doi.org/10.1016/j.rser.2016.02.018i

I. Briggs et al. / Renewable and Sustainable Energy Reviews ∎ (∎∎∎∎) ∎∎∎–∎∎∎

storage on the forklift truck via electrical and hydraulic storage methods. Hosseinzadeh et al. [80] modelled a fuel cell/battery hybrid forklift truck using MATLAB/Simulink. The drive cycle used for the analysis was a customised time-based cycle consisting of periods of acceleration, deceleration and constant speeds, as well as hoisting operations and travelling loaded and unloaded. Development of the forklift truck's control strategy allowed for optimisation of fuel cell size depending on the vehicle speeds specified in the drive cycle. A similar fuel cell/battery hybrid truck was studied by Keranen et al. [81] using MATLAB/Simulink. The authors analysed a ‘triple-hybrid platform’ that included batteries, fuel cells and an ultra-capacitor. Again, a custom-designed drive cycle was employed which involved periods of driving with the forks loaded and unloaded, and hoisting operations at the end of each driving phase. The model allowed a comprehensive parametric study of the effects of battery temperature on the performance of the hybrid system to be undertaken and the sizing and operation of the ultra-capacitor to be analysed. As highlighted above, the lack of industry-accepted standardised off-highway drive cycles often leads to researchers developing their own cycle to suit their particular needs. While VDI 2198 provides such a cycle, it has not currently been adopted in enough studies to allow valid comparisons to be made between works. In other non-automotive applications, authors such as Pedersen [82] employed co-simulation techniques such as functional mockup interface (FMI) to combine physical and computational domains. This allowed a full analysis of the total system's dynamic performance to be undertaken along with an investigation into the communication between subsystems. FMI [83] is a standardised and structured method of co-simulation where multiple computational domains (mechanical, electrical, thermal, etc.) are linked. This differs from the co-simulation methods listed above due to its standardised computational language and framework. The process can be extended in a similar manner to other co-simulation methods to include Hardware-in-the-Loop simulations (HiL) where real life hardware is added to the simulated control systems. For non-automotive, off-highway vehicles, it is much more challenging, but not impossible, to use commercial software. In most cases, it is necessary to supplement the models with some external custom software, usually in the form of a Simulink add-on [71,73]. This is usually for some element of control or operation that is absent from the commercial software. 3.3. Simulation modes In most of the simulations discussed above, the authors chose whether to employ a forward- or backward-facing simulation. To address the differences between these simulation modes, a description of each is discussed in the subsections that follow. 3.3.1. Backward-facing simulation The backward-facing approach assumes the vehicle was able to meet the criteria laid out in the drive-cycle under analysis [16]. Assuming this criterion is met, the backward-facing model determines how each component performed during the cycle. There is no attempt to model a driver in this approach; rather, the acceleration force required is calculated from the drive-cycle speed trace. This force is then translated into a torque of the component directly upstream in the simulation, i.e. wheels, transmission and engine respectively. The flow of information in a backward-facing simulation is shown in Fig. 7. This allows for the calculation of the fuel or electrical energy used to meet the drive cycle speed trace, meaning the backward-facing model is ideal for fuel consumption prediction. The backward-facing approach takes advantage of each component's measured efficiency against its torque and speed

9

Fig. 7. Backward-facing simulation methodology.

inputs. This means backward-facing simulations are usually quick to run. However, the main drawback is the assumption that the vehicle can meet the required drive-cycle speed trace; it is implied that all components in the vehicle model can perform so as to achieve the measured drive cycle. There is no direct way of telling if the required accelerations exceed the capabilities of the powertrain [73]. Also, the measured efficiency tables do not include dynamic effects, since the values in each efficiency table were obtained during steady-state testing. Backward-facing systems also do not lend themselves to control system integration or development since control systems are typically discrete systems where information flows from inputs to outputs, and cannot always be reverse engineered. Instead, backward-facing models lend themselves well to more simplified investigations of fuel economy performance rather than overall vehicle performance or drivability. 3.3.2. Forward-facing simulation Forward-facing simulations include a model of the driver, usually through a PID controller. The commanded throttle position is translated into a torque from the engine, which is used as an input to the transmission system. The computed torque is transmitted forward through the model in the direction of physical power flow until it reaches the wheels and results in a tractive force on the road as shown in Fig. 8; the resultant vehicle acceleration can be calculated using Newton's second law of motion, and the vehicle's motion is fed back to the driver and the accelerator and brake pedal inputs are adjusted to meet the drive cycle profile. Rather than component performance being dictated by a series of experimentally-obtained performance maps, as is the case with backward-facing models, forward-facing models use a governing set of equations to simulate each component. Thus the architecture of a forward-facing model is significantly more complex and necessarily more detailed. The forward-facing approach allows for development of control systems since driver demands are used as simulation inputs and the data flows in the direction of physical power flow. This type of simulation also allows the maximum acceleration rates to be calculated since the maximum tractive force is easily obtained. This means that forward-facing models not only perform well in fuel consumption simulations, but also allow a more quantitative assessment of vehicle drivability and performance to be undertaken. However, forward-facing simulations tend to run slower than backward-facing models since the speed and torque of drivetrain components must be calculated on each timestep and fed back to the driver model.

Please cite this article as: Briggs I, et al. Sustainable non-automotive vehicles: The simulation challenges. Renewable and Sustainable Energy Reviews (2016), http://dx.doi.org/10.1016/j.rser.2016.02.018i

10

I. Briggs et al. / Renewable and Sustainable Energy Reviews ∎ (∎∎∎∎) ∎∎∎–∎∎∎

fuel economy without the need to employ a full transient mathematical model as that used by Ma et al. [88]. 3.4. Non-automotive simulation summary

Fig. 8. Forward-facing simulation methodology.

3.3.3. Choice of simulation mode The choice of whether to use a backward- or forward-facing model largely is defined by the application and user preference. Where a time-based drive cycle exists, and the vehicle under study has been proven to achieve this cycle, a backward-facing approach provides a fast and computationally inexpensive approach. However, where the drive cycle is not time-based, as is the case with most off-highway cycles, or if the vehicle may not achieve the velocity specified in the cycle, a forward-facing simulation provides a more realistic evaluation of performance. One of the drawbacks of using Simulink in a forward-facing model is that it requires the user to model the driver as one or more simple PID controllers, altering the accelerator pedal input to meet the demands of a particular drive cycle. While this is acceptable in many cases, if the drive cycle does not consist of a velocity-time trace, the use of such a PID controller is not possible. This is one of the advantages provided by commercial software such as AVL CRUISE [73]; it allows the use of a realistic driver model which includes driver characteristics such as shifting behaviour and driving behaviour (tuneable levels of aggression on the pedal inputs) [84]. Mohan et al. [85] states that forward-facing simulations allow the physical limits of the system to be captured, but also points out that if any of the powertrain components are altered, the driver model used in the simulation, which is typically based on PID controllers, may need to be re-tuned. Indeed, in their work, the authors showed that the ability to tune the driver model was key to accurate prediction of fuel consumption and allowed powertrain optimisation to be carried out. Markel et al. [16] described a model which used a combination of backward- and forward-facing modelling procedures. Largescale demands such as a change in vehicle speed were propagated backward through the system to the engine. This defines the ICE power output required to achieve the desired performance. This power output was then fed forwards through the various components and subsystems in the model to minimise the error between the driver's demand and the vehicle output. This approach has also been adopted by other authors [86,87]. If control systems are to be developed, a forward-facing model is generally the most suitable approach since it allows the limits of the vehicle to be simulated and the hardware characteristics and responses are included in the simulation. This also allows detailed predictions of drivability and vehicle performance to be made; this cannot be achieved with a backward-facing model, which is better-suited to studies focusing on fuel economy alone. This can also go some way to investigating the influence of driving style on

This section has highlighted the differences between the approaches to automotive and non-automotive simulation. In particular, the wealth of commercial simulation software is not always suited to non-automotive applications, and it is particularly difficult to apply to off-highway simulations. One of the main challenges of non-automotive simulation is the lack of standardised drive cycles, which makes comparing results from different studies difficult. Defining drive cycles for individual applications is also demanding as the nature of non-automotive applications rarely results in repeatable cycles. For on-highway applications such as city buses, standardised drive cycles have gone some way to allowing operators and manufacturers to compare vehicle emissions. For off-highway vehicles, if standardised tests exist at all, they are still unrepresentative of all realworld applications, since every application is inherently different. As with automotive vehicles, but particularly important in the non-automotive sector, the development of advanced control systems is extremely important when analysing hybrid vehicles [57,71,89,90], and for forklift truck applications, complex control strategies are required for various modes of operation such as inching, power reversal and hoisting [73]. Due to these complexities in operation, the design of a suitable control system also has a large impact on the performance of the vehicle, and a poorlyconstructed control model can be hugely detrimental to performance figures. For control system development MATLAB/Simulink is often preferred.

4. Conclusions This paper has outlined several approaches to automotive and non-automotive simulation. In the automotive sector, several commercial and open-source software packages are available to offer a wide choice of simulation methods, depending on the parameters under investigation. Many of the challenges associated with automotive simulation have been encountered by others and the modelling methodologies are well documented. This provides the researcher with a solid foundation when embarking on an automotive simulation project. In the non-automotive sector, there is considerable literature available documenting the simulation of on-highway vehicles such as trucks and buses. Standardised drive cycles allow different vehicles to be directly compared over the same test. Several papers were discussed which reported the simulation of hybrid vehicles, a rapidly growing area in the bus market. However, for off-highway vehicles, there is considerably less literature available for reference. This means the researcher will often be forced to make some blind choices when modelling an off-highway vehicle, including making several assumptions about the drive cycles used by others. The lack of standardised drive cycles also makes it difficult to compare and validate the performance of different vehicles. This paper has shown a number of simulation approaches, and that different subject areas, as well as different user requirements, may benefit from different simulation strategies. Below is a summary of the key challenges to consider when simulating an offhighway vehicle along with various suggestions to aid in the implementation a simulation strategy.

 Vehicle architecture: with non-automotive vehicles, due to often complex drivetrains, it is not possible to model various

Please cite this article as: Briggs I, et al. Sustainable non-automotive vehicles: The simulation challenges. Renewable and Sustainable Energy Reviews (2016), http://dx.doi.org/10.1016/j.rser.2016.02.018i

I. Briggs et al. / Renewable and Sustainable Energy Reviews ∎ (∎∎∎∎) ∎∎∎–∎∎∎







components of the architecture using existing vehicle modelling software. Therefore, a Simulink model of these components can be created separately and connected to the vehicle model. This allows the complex physical parameters of add-on components to be captured and co-simulated with a simpler drivetrain model. This may be extended further to model the full vehicle in Simulink to avoid software interfaces. In this case, the architecture of the vehicle must be modelled as a series of blocks containing equations Driver model: the use of specific vehicle modelling software is useful when an accurate driver model is required. This means the bulk of the vehicle model can be created in the vehicle modelling software, and other elements of the powertrain or vehicle hardware can be modelling in Simulink Drive cycles: drive cycles for non-automotive vehicles should be carefully chosen to allow comparisons to both other, similar vehicles, and to real-life operations. Where possible, a standardised drive cycle should be used for analysis, and where not possible, one should be defined based on available information. This often takes the form of a profile constructed from operations or duty cycles Prediction of vehicle performance: care should be taken when considering whether to use a forward- or backward-facing model. The choice depends on the modelling criteria (whether vehicle drivability or component performance is to be investigated, or purely fuel consumption performance), simulation time, and whether a velocity versus time drive cycle is available for analysis. Backward-facing simulations tend to be faster to build and run, and give good assessment of fuel consumption under known driving conditions, but do not reveal the achievable performance of drivetrain components. This is due to a lack of detailed component modelling; components are modelled as a series of steady-state maps. If control system development is the objective, then forward-facing is the required method of simulation due to the increased fidelity of component modelling, and the data flow through the model allows for the vehicle to feedback performance criteria, such as acceleration rates, to the driver model

As simulation methods become more advanced, the ability to accurately capture complex powertrains improves. Off-highway vehicles often require bespoke models to simulate non-standard drivetrain components and configurations. This paper has shown several approaches to overcome the challenges of simulating non-automotive vehicles using both commercial powertrain modelling software and MATLAB/Simulink. The latter was shown to be the more widely used approach in simulating non-automotive off-highway vehicles due the flexibility it affords in creating application-specific models, but several studies showed it could be used effectively in co-simulation with commercial engine modelling software or driver models.

Acknowledgements The authors would like to thank AVL LIST GmbH for their technical support and provision of software. This project is partfunded by Invest Northern Ireland and the EU European Regional Development Fund through the Investment for Growth and Jobs Programme 2014-2020.

References [1] Watson N, Marzouk M. A non-linear digital simulation of turbocharged diesel engines under transient conditions. Warrendale, PA: SAE International; 1977. [2] Watson N, Pilley A, Marzouk MA. A combustion correlation for diesel engine simulation. 1982.

11

[3] Watson NA, Janota MS. Turbocharging the internal combustion engine. United Kingdom: Macmillan Publishers Limited; 1982. [4] Rychter TJ, Teodorczyk A, Stone CR, Leonard HJ, Ladommatos N, Charlton SJ. A theoretical study of a variable compression ratio turbocharged diesel engine. Proc. Inst. Mech. Eng. A: J. Power Energy 1992;206:227–38. http://dx.doi.org/ 10.1243/PIME_PROC_1992_206_039_02. [5] Ehsani M, Ahmadi A, Fadai D. Modeling of vehicle fuel consumption and carbon dioxide emission in road transport. Renew. Sustain. Energy Rev. 2016;53:1638–48. [6] Blackmore K, Wong CH, Thring R, Rama P. Fuel cell vehicle powerplant optimisation. Proceedings of the avl advanced simulation technologies international user meeting. Graz, Austria: AVL LIST GmbH; 2007. [7] Oh Y, Park J, Lee J, Eom MD, Park S. Modeling effects of vehicle specifications on fuel economy based on engine fuel consumption map and vehicle dynamics. Transp. Res. Part D: Transp. Environ. 2014;32:287–302. [8] Pacheco AF, Martins MES, Zhao H. New European Drive Cycle (NEDC) simulation of a passenger car with a HCCI engine: emissions and fuel consumption results. Fuel 2013;111:733–9. http://dx.doi.org/10.1016/j.fuel.2013.03.060. [9] Banjac T, Wurzenberger JC, Katrašnik T. Assessment of engine thermal management through advanced system engineering modeling. Adv. Eng. Softw 2014;71:19–33. http://dx.doi.org/10.1016/j.advengsoft.2014.01.016. [10] Srinivasan P, Kothalikar UM. Performance fuel economy and CO2 prediction of a vehicle using AVL cruise simulation techniques. Proceedings of the powertrains, fuels and lubricants meeting. Birmingham, UK: SAE International; 2009. [11] Srinivasan P, Umashankar KS, Kothalikar UM. Optimization of performance and CO2 emission of a common rail DI diesel operated vehicle using AVL cruise. Proceedings of the international powertrains, fuels & lubricants meeting. Birmingham, UK: SAE International; 2010. [12] Sun G, Wei M, Shao J, Pei M. Automotive powertrain modeling and simulation based on AMESim. Proceedings of the asia pacific automotive engineering conference. Hollywood, USA: SAE International; 2007. [13] Xie F, Wang J, Wanga Y. Modelling and co-simulation based on AMESim and simulink for light passenger car with dual state CVT. Proc. Eng. 2011;16:363–8. http://dx.doi.org/10.1016/j.proeng.2011.08.1096. [14] Vijayagopal R, Rousseau AP. System analysis using multiple expert tools. SAE world congress & exhibition. Detroit, USA: SAE International; 2011. [15] Tavares F, Johri R, Filipi Z. Simulation study of advanced variable displacement engine coupled to power-split hydraulic hybrid powertrain. J. Eng. Gas Turbines Power 2011;133:122803. [16] Markel T, Brooker A, Hendricks T, Johnson V, Kelly K, Kramer B, et al. ADVISOR: a systems analysis tool for advanced vehicle modeling. J. Power Sources 2002;110:255–66. http://dx.doi.org/10.1016/S0378-7753(02)00189-1. [17] MacBain J, Conover J, Brooker A. Full vehicle simulation for series hybrid vehicles. SAE future transportation technology conference & exposition. Portland, USA: SAE International; 2003. [18] Silva CM, Farias TL, Frey HC, Rouphail NM. Evaluation of numerical models for simulation of real-world hot-stabilized fuel consumption and emissions of gasoline light-duty vehicles. Transp. Res. Part D: Transp. Environ. 2006;11:377–85. http://dx.doi.org/10.1016/j.trd.2006.07.004. [19] Sulaiman N, Hannan MA, Mohamed A, Majlan EH, Wan Daud WR. A review on energy management system for fuel cell hybrid electric vehicle: issues and challenges. Renew. Sustain. Energy Rev. 2015;52:802–14. http://dx.doi.org/ 10.1016/j.rser.2015.07.132. [20] Silva CM, Farias TL, Mendes-Lopes JMC. EcoGest – numerical modelling of the dynamic, fuel consumption and tailpipe emissions of vehicles equipped with spark ignition engines. WIT transactions on the built environment. . Southampton, UK: WIT Press; http://dx.doi.org/10.2495/UT020511. [21] Barth M, An F, Younglove T, Scora G, Levine C, Ross M, et al. Development of a comprehensive modal emissions model. United States: Transportation Research Board of the National Acadamies; 2000. [22] Dyer A, Pagerit S, Datar M, Mehr D, Negrut D. A co-simulation environment for virtual prototyping of ground vehicles. SAE commercial vehicle engineering congress & exhibition. USA: SAE Technical Paper Rosemont; 2007. [23] Bowles P, Tiller M, Elmqvist H, Brück D, Mattsson SE, Möller A, et al. Feasibility of detailed vehicle modeling. Proceedings of the SAE world congress & exhibition. Detroit, USA: SAE International; 2001. [24] Rousseau A, Sharer P, Besnier F. Feasibility of reusable vehicle modeling: application to hybrid vehicles. SAE world congress and exhibition. Detroit, USA: SAE International; 2004. [25] Manzetti S, Mariasiu F. Electric vehicle battery technologies: From present state to future systems. Renew Sustain. Energy Rev. 2015;51:1004–12. http://dx.doi. org/10.1016/j.rser.2015.07.010. [26] Ricardo Inc. Computer simulation of light-duty vehicle technologies for greenhouse gas emission reduction in the 2020–2025 timeframe. United States: United States Environmental Protection Agency; 2011. [27] Iglesias I, Isasi L, Larburu M, Martin A, Peña A. Networked Clean Vehicles, How the Environment Information will Improve Fuel Efficiency and CO2 Emissions. SAE Int. J. Fuels Lubr. 2009;2:167–71. http://dx.doi.org/10.4271/2009-01-0599. [28] Komoto H, Masui K. Classification of design parameters with system modeling and simulation techniques. CIRP Ann. – Manuf. Technol. 2014;63:193–6. http://dx. doi.org/10.1016/j.cirp.2014.03.098. [29] Aydin T, Egel T, Ash-Rafzadeh A. Drivetrain modeling, simulation, and analysis using VHDL-AMS. SAE world congress and exhibition. Detroit, USA: SAE International; 2003. [30] Winkler D, Gühmann C. Hardware-in-the-Loop simulation of a hybrid electric vehicle using Modelica/Dymola. In: Proceedings of the 22nd International

Please cite this article as: Briggs I, et al. Sustainable non-automotive vehicles: The simulation challenges. Renewable and Sustainable Energy Reviews (2016), http://dx.doi.org/10.1016/j.rser.2016.02.018i

12

[31]

[32]

[33]

[34] [35] [36] [37]

[38]

[39]

[40] [41]

[42]

[43]

[44]

[45] [46]

[47] [48]

[49] [50] [51]

[52]

[53] [54]

[55]

[56]

[57]

[58]

[59]

I. Briggs et al. / Renewable and Sustainable Energy Reviews ∎ (∎∎∎∎) ∎∎∎–∎∎∎ Battery, Hybrid and Fuel Cell Electric Vehicle Symposium. Yokohama, Japan; 2006. pp. 1054–63. Cheng C, McGordon A, Jones RP, Jennings PA. Development of a comprehensive and flexible forward dynamic powertrain simulation tool for various hybrid electric vehicle architectures. Proc. Inst. Mech. Eng. Pt. D: J. Automobile Eng. 2012;226:385–98. http://dx.doi.org/10.1177/0954407011417764. Schulze M, Mustafa R, Tilch B, Eilts P, Küçükay F. Energy Management in a Parallel Hybrid Electric Vehicle for Different Driving Conditions. SAE Int. J. Altern. Power 2014;3:193–212. http://dx.doi.org/10.4271/2014-01-1804. Yong JY, Ramachandaramurthy VK, Tan KM, Mithulananthan N. A review on the state-of-the-art technologies of electric vehicle, its impacts and prospects. Renew. Sustain. Energy Rev. 2015;49:365–85. http://dx.doi.org/10.1016/j. rser.2015.04.130. Finesso R, Spessa E, Venditti M. Layout design and energetic analysis of a complex diesel parallel hybrid electric vehicle. Appl. Energy 2014;134:576–88. Damiani L, Repetto M, Prato AP. Improvement of powertrain efficiency through energy breakdown analysis. Appl. Energy 2014;121:252–63. Gao Y, Ehsani M. A torque and speed coupling hybrid drivetrain – architecture, control and simulation. IEEE Trans. Power Electron. 2006;21:741–8. Jackey R, Smith P, Bloxham S. Physical system model of a hydraulic energy storage device for hybrid powertrain applications. SAE world congress & exhibition. Detroit, USA: SAE International; 2005. Stelson KA, Meyer JJ, Alleyne AG. Hencey B. Optimization of a passenger hydraulic hybrid vehicle to improve fuel economy. In: Proceedings of the 7th JFPS international symposium on fluid power. Toyama, Japan: JFPS; 2008. Berard F, Cotta A, Stokes J, Thring R, Wheals J. An Integrated Powertrain (IPT) model – stage one. SAE world congress & exhibition. Detroit, USA: SAE International; 2000. Alzuwayer B, Abdelhamid M, Pisu P, Giovenco P, Venhovens P. Modeling and Simulation of a Series Hybrid CNG Vehicle. SAE Int. J. Altern. Power 2014;3:20–9. Baer K, Ericson L, Krus P. Modeling of a series hybrid hydraulic drivetrain for a light-duty vehicle in hopsan. In: Proceedings of the 13th scandinavian international conference on fluid power-SICFP2013. Linköping, Sweden; 2013. pp. 107–12. Johri R, Filipi Z. Low-Cost Pathway to Ultra Efficient City Car: Series Hydraulic Hybrid System with Optimized Supervisory Control. SAE Int. J. Engines 2010;2:505–20. http://dx.doi.org/10.4271/2009-24-0065. Lutsey N, Brodrick C-J, Lipman T. Analysis of potential fuel consumption and emissions reductions from fuel cell auxiliary power units (APUs) in long-haul trucks. Energy 2007;32:2428–38. http://dx.doi.org/10.1016/j.energy.2007.05.017. Xiu-qin Z, Bo Y, Chao Y, Guan-neng X. Research on ABS of multi-axle truck based on ADAMS/Car and Matlab/Simulink. Proc. Eng. 2012;37:120–4. http://dx. doi.org/10.1016/j.proeng.2012.04.213. Sandberg T. Modeling and validation of traveling resistance for heavy trucks. SAE world congress & exhibition. Detroit, USA: SAE International; 2001. Lajunen A. Fuel economy analysis of conventional and hybrid heavy vehicle combinations over real-world operating routes. Transp. Res. Part D: Transp. Environ. 2014;31:70–84. http://dx.doi.org/10.1016/j.trd.2014.05.023. Ribau JP, Silva CM, Sousa JMC. Efficiency, cost and life cycle CO2 optimization of fuel cell hybrid and plug-in hybrid urban buses. Appl. Energy 2014;129:320–35. Li Z, Tian G, Zhou W, Chen Q, Zhu Y. An Easy5 based research on simulation of fuel cell city bus. In: Proceedings of the 21st international battery, hybrid and fuel cell electric vehicle symposium. Monaco; 2005. Feng D, Huang D. Modeling and control of a series hydraulic hybrid vehicle. J. Comput. Inf. Sci. 2012;8:1805–19. Wu B, Lin C-C, Filipi Z, Peng H, Assanis D. Optimal power management for a hydraulic hybrid delivery truck. Veh. Syst. Dyn. 2004;42:23–40. Midgley WJB, Cathcart H, Cebon D. Modelling of hydraulic regenerative braking systems for heavy vehicles. Proc. Inst. Mech. Eng. Pt. D: J. Automob. Eng. 2013;227:1072–84. http://dx.doi.org/10.1177/0954407012469168. Simpson A, Fleck R, Kee R, Douglas R, Steele D, Hanna A, et al. Development of a heavy duty hybrid vehicle model. Proceeding of the SAE commercial vehicle engineering congress & exhibition. Rosemont, USA: SAE International Congress and Exhibition; 2009. Berta GL, Durelli E, Nymann I. Simulation models for hybrid buses. IMechE, Part D: J. Automob. Eng. 1998;212:59–72. De Lorenzo G, Andaloro L, Sergi F, Napoli G, Ferraro M, Antonucci V. Numerical simulation model for the preliminary design of hybrid electric city bus power train with polymer electrolyte fuel cell. Int. J. Hydrog. Energy 2014;39:12934–47. http://dx.doi.org/10.1016/j.ijhydene.2014.05.135. Trajkovic S, Tunestal P, Johansson B. Vehicle driving cycle simulation of a pneumatic hybrid bus based on experimental engine measurements. SAE world congress & exhibition. Detroit, USA: SAE International; 2010. Briggs I, McCullough G, Spence S, Douglas R, O’Shaughnessy R, Hanna A, et al. Waste heat recovery on a diesel-electric hybrid bus using a turbogenerator. SAE commercial vehicle engineering congress. Rosemont, USA: SAE International; 2012. Briggs I, McCullough G, Spence S, Douglas R. Whole-vehicle modelling of exhaust energy recovery on a diesel-electric hybrid bus. Energy 2014;65:172–81. http://dx. doi.org/10.1016/j.energy.2013.11.075. Bao R, Stobart R. Study on optimization of regenerative braking control strategy in heavy-duty diesel engine city bus using pneumatic hybrid technology. SAE world congress & exhibition. Detroit, USA: SAE International; http://dx.doi.org/10.4271/2014-01-1807. Barlow TJ, Latham S, McCrae IS, Boulter PG. A reference book of driving cycles for use in measurement of road vehicle emissions. United Kingdom: UK Department of Transport; 2009.

[60] Eastlake A. The latest developments in heavy duty vehicle aftertreatment testing for real world emissions and fuel economy. SAE international congress & exhibition. Detroit, USA: SAE International; 1999. [61] Briggs I, McCullough G, Spence S, Douglas R, O’Shaughnessy R, Hanna A, et al. A Parametric study of an exhaust recovery turbogenerator on a diesel-electric hybrid bus. San Antonio, USA: ASME TurboEXPO; 2013. [62] “SORT” testing methodology: a new protocol for hybrid buses. UITP Advancing Public Transport n.d. 〈http://www.uitp.org/%E2%80%9Csort%E2% 80%9D-testing-methodology-new-protocol-hybrid-buses〉 [accessed 30.01.15]. [63] Doelling WS, Zhang Y, Wang S, Li J, Margraf J, Mrosek M, et al. NH3 Generator for Enhanced Low-Temperature SCR Performance; An Option also for Retrofit. Proc. SAE-China Congr. 2014: Sel. Pap. 2015;328:191–203. [64] European Parliament. COMMON POSITION (EC) No /991999. [65] Department for Transport. Low Carbon Vehicle Partnership. Vehicle Accreditation Requirements: Low Carbon Emission Buses 2011. 〈http://goo.gl/1sVjEx〉 [accessed 25.02.15]. [66] European Commission. Emissions from heavy duty vehicles. 2014. [67] Schutz PW, Klokkenga DA, Stattenfield DB. Vehicle mission simulation. SAE earthmoving industry conference. Peoria, USA: SAE International; 1970. [68] Savaresi SM, Taroni F, Prevedi F, Bittanti S. On the design and tuning of the controllers in a power-split continuously variable transmission for agricultural tractors. Proc. Inst. Mech. Eng. Pt. D: J. Automobile Eng. 2003;217:745–59. [69] Bietresato M, Friso D, Sartori L. Assessment of the efficiency of tractor transmissions using acceleration tests. Biosyst. Eng. 2012;112:171–80. http://dx.doi. org/10.1016/j.biosystemseng.2012.03.009. [70] Rahmfeld R, Klocke C. The importance of pump/motor efficiencies for mobile machines operating costs. Las Vegas, USA: IFPE; 2011. p. 264–9. [71] Wang T, Wang Q. Design and analysis of compound potential energy regeneration system for hybrid hydraulic excavator. Proc. Inst. Mech. Eng. Part I J. Syst. Control Eng. 2012;226:1323–34. http://dx.doi.org/10.1177/0959651812456642. [72] Wang D, Guan C, Pan S, Zhang M, Lin X. Performance analysis of hydraulic excavator powertrain hybridization. Autom. Constr. 2009;18:249–57. http://dx. doi.org/10.1016/j.autcon.2008.10.001. [73] Murtagh M, Kee R, McCullough G, Stuart C, Bradley C, Trimble S, et al. Development and validation of a forklift truck powertrain simulation. SAE Int. J. Commer. Veh. 2013;6:150–60. [74] Verein Deutscher Ingenieure. VDI 2198: Type Sheets for Industrial Trucks; 2002. [75] Stein G, Meitz K, Kriegler W, Wichmann M. Powertrain optimization using simulation – example of engine selection for forklift trucks. SAE commercial vehicle engineering congress & exhibition. Rosemont, USA: SAE International; 2004. [76] Ogawa K, Futahashi K, Teshima T, Akahane Fumihiro. Development of the World's First Engine/Battery Hybrid Forklift Truck. Mitsubishi Heavy Ind. Tech. Rev. 2010;47(1). [77] Bohler F, Thiebes P, Geimer M, Santoire J, Zahoransky R. Hybrid drive systems for industrial applications. 9th International conference on engines and vehicles. Naples, Italy: SAE International; 2009. [78] Minav TA, Filatov DM, Pyrhonen JJ, Pietola Matti. Modelling of a working cycle of an electro-hydraulic forklift in matlab simulink. Int. Rev. Model. Simul. 2007:6. [79] Minav TA, Schimmel T, Murashko K, Åman R, Pyrhönen JJ, Pietola M. Towards a better energy efficiency through a systems approach in an industrial forklift system. Proc. Inst. Mech. Eng. Pt. D: J. Automobile Eng. 2015;229:273–82. http://dx. doi.org/10.1177/0954407014539672. [80] Hosseinzadeh E, Rokni M, Advani SG, Prasad AK. Performance simulation and analysis of a fuel cell/battery hybrid forklift truck. Int. J. Hydrog. Energy 2013;38:4241–9. [81] Keranen TM, Karimaki H, Viitakangas J, Vallet J, Ihonen J, Hyotyla P, et al. Development of integrated fuel cell hybrid power source for electric forklift. J. Power Sources 2011;196:9058–68. [82] Pedersen N, Madsen J, Vejlgaard-Laursen M. Co-Simulation of Distributed Engine Control System and Network Model using FMI & SCNSL. IFACPapersOnLine 2015;48:261–6. http://dx.doi.org/10.1016/j.ifacol.2015.10.290. [83] Modelisar. Functional Mock-up Interface for Model Exchange. Modelisar Consortium; 2010. [84] A.V.L. LIST GmbH. AVL CRUISE v2013 User Guide. n.d. [85] Mohan G, Assadian F, Longo S. Comparative analysis of forward-facing models vs backward facing models in powertrain component sizing. IET Hybrid and Electric Vehicles Conference 2013 (HEVC 2013). IET; 2013. pp. 1–6. http://dx. doi.org/10.1049/cp.2013.1920. [86] Wang X, Wei J, Chen X, Zhong Z. Modeling of FCV powertrain using a combined forward/backward approach. 2011 International Conference on Electric Information and Control Engineering (ICEICE); 2011. pp. 5106–5109. http://dx. doi.org/10.1109/ICEICE.2011.5776856. [87] Dixon G, Steffen T, Stobart R. Unified backwards and forwards simulation of a hybrid electric vehicle in the A-segment using MATLAB Simscape. SAE 2015 World Congress & Exhibition, Detroit, USA: SAE International; n.d. [88] Ma H, Xie H, Huang D, Xiong S. Effects of driving style on the fuel consumption of city buses under different road conditions and vehicle masses. Transp. Res. Part D: Transp. Environ. 2015;41:205–16. http://dx.doi.org/10.1016/j.trd.2015.10.003. [89] Shen W, Jiang J, Su X, Reza Karimi H. Control strategy analysis of the hydraulic hybrid excavator. J. Frankl. Inst. 2015;352:541–61. http://dx.doi.org/10.1016/j. jfranklin.2014.04.007. [90] Wang L, Zhang Y, Yin C, Zhang H, Wang C. Hardware-in-the-loop simulation for the design and verification of the control system of a series–parallel hybrid electric city-bus. Simul. Model. Pract. Theory 2012;25:148–62. http://dx.doi. org/10.1016/j.simpat.2012.02.010.

Please cite this article as: Briggs I, et al. Sustainable non-automotive vehicles: The simulation challenges. Renewable and Sustainable Energy Reviews (2016), http://dx.doi.org/10.1016/j.rser.2016.02.018i