- Email: [email protected]
Man, Decision-Making, Control and Performance: A Complementary View to Process-Oriented approach
IFAC
[: 0
Copyright IFAC Intelligent Manufacturing Systems, Poznan, Poland, 200 I
[>
Publications www.elsevier.comllocate/ifac
MAN, DECISION-MAKING, CONTROL AND PERFORMANCE: A COMPLEMENTARY VIEW TO PROCESS-ORIENTED APPROACH
David Chen, Bruno Vallespir and Guy Doumeingts
LA P/GRA I, University Bordeaux I 351, Cours de la liberation, 33405 Talence cedex France Tel : +33556846530; Fax : +33556846644 [email protected]
Abstract : This paper tentatively presents some basic concepts and constructs for the modelling of enterprise control and decision-making. It is proposed that man-centred decision-making oriented enterprise control be a complementary view to computer executed process-oriented approach . Based on some theoretical arguments from automatic control, the paper discusses the set of concepts on the role of man in enterprise to make decisions and to control operational processes so that global objectives of the company can be reached. Copyright © 2001 IFAC Keywords : Enterprise modelling, Enterprise Integration, Control, Decision-making, performance
I. INTRODUCTION
the architecture implemented enables to get the level of performance required by the strategy of the company (Chen, 1996, 1999, 2000). Enterprise control plays a major role for matching performances and then to keep competitive. However, most of current enterprise modelling techniques are mainly information, function and process oriented and the control aspect is reduced to an information processing issue. To match current industrial challenges, control needs to be in an equivalent position than the others. To control a system that cannot be entirely formalised and modelled, it is necessary to provide a set of decision-centres with co-ordination links between them (Doumeingts, 1998, 1992). Focusing on control is focusing on decision-making and on the role of man in industrial systems as well. Man-based decision-making oriented enterprise control is proposed as a complementary viewpoint with respect to computer executed and process oriented approach. Main items of the decision-making are (I) the valued objective or set of objectives the decision has to match; (2) the decision variables enabling the decision-maker to know what it can act on and until where
The approach presented in this paper fits within the framework of IF AC Technical Committee on Enterprise Integration. It deals with the elaboration of enterprise reference architecture and methodologies including concepts, principles and constructs for the analysis and design of enterprise systems. The IF ACIFIP Task Force on Enterprise Integration has developed GERAM (Generalised Enterprise Reference Architecture and Methodologies) (lfac-Ifip task force, 1997). GERAM is based on the three main approaches: GRAI (GRAI Integrated Methodology) (Doumeingts, 1992, 1998), CIMOSA (CIM Open System Architecture) (Amice, 1993) and PERA (Purdue Enterprise Reference Architecture) (Williams, 1994). An International Standard is adopted based on the input of GERAM (IS, 1998). At the European level, CEN (European Committee for Standardisation) is working on the finalisation of two standards dealing with the framework and constructs for enterprise modelling (ENV 40003, 1990; ENV 12204, 1995). This paper addresses the necessity to model enterprise control viewpoint in order to ensure that
59
(constraints); (3) the target performance that an enterprise wants to achieve.
are defined and delimited by a co-ordinator module. Generally, the co-ordinator is in the same situation than that of the co-ordinated modules: it is coordinated too. Then, the complex, multi-objectives systems, such as industrial systems, are controlled by a hierarchical, multi-level structure which owns the important property of decentralised decision-making.
In the first part of the paper, some theoretical arguments from the point of view of automatic control theory are presented. Then enterprise wide decision-making and control are discussed. In the third part of the paper, basic concepts to develop constructs for enterprise modelling taking this aspect into account are tentatively stated. Future work and perspectives are given as part of conclusion.
2.
oflhe Control syslem
THEORETICAL ARGUMENTS
The traditional approach in system engineering coming from Automatic Control considers that a system has some significant states that can be formalised . Formalising these states enables to take them into account , i.e. to translate them into a model which is used to control the system. To consider that the significant states can also be formalised is remarkable. This assumption is the basis of automatic control theory . It leads the system to run "automatically" when it is in a formalised state or to be blocked or insensitive in other states. This approach is used when the risk to go out the set of formalised states is weak because the system is quasi -exhaustively modelled. However, in a complex enterprise system where human takes part in, the situations and events influencing the system cannot be entirely identified and modelled. That system is rarely in a nominal mode. Very often, the model of the system is an ideal process corresponding rarely to the reality. How to control the system when it is in a state which is significant but not formalised ? A principle coming from cybernetics gives a piece of solution : a system is really controlled only if the control system proposes at least as many states (control system variety) as the controlled system (controlled system variety) . Because this variety does not exist a priori, it is necessary to have inside the control system some variety generators, i.e. organs that are capable to react to a situation at the time the situation appears. The only organ known to be versatile and quick enough to play this role is the human being (see figure I) .
Sel of Ihe Slales oflhe Controlled Syslem
Sel of Ihe non· significant SI ales of Ihe syslem
Figure I . Increase of the control system variety by the presence of human beings The cognitive limitation of decision-makers is also an argument for hierarchical , decentralised control systems. A decision-maker has a cognitive limitation which can be expressed in term of quantity of information. Beyond this limit, the decision-maker is submerged by the information : he is not able to interpret the information he is supposed to use in order to make decision . The quantity of information is proportional to : • •
the size of the domain in which the decisionmaker is supposed to make decision , and the detail of the information handled by the decision-maker.
The information system also needs to be aggregated as it goes up in the structure, in order to provide the good granularity of information to decision-makers. To summary, to control a system that cannot be formalised enough, it is necessary to provide a set of decision-makers with co-ordination links between them.
The human being can play the role of variety generator if his freedom to make decision is high. This leads to a difficulty: a high freedom can lead the decision-maker to have a behaviour that is not compatible with the objectives of the system. Then, the decisional freedom needs to be restricted by a pre-defined decision frame . The problem grows when the number of decision-makers grows. There is a need to ensure that local decisional freedoms dealing with specific objectives be consistent with global objective(s) of the system. The solution is coordination in the meaning that decisional freedoms
Focusing on control is focusing on decision-making and on the role of man in industrial systems as well. Then, it is a necessity to avoid mechanistic approaches and to situate human decision into the system . To match current industrial challenges, control needs to be in an equivalent position than information and communication (figure 2).
60
aggregated to be used at any level of the hierarchy of control. Secondly, because decision is mainly made on a basis of objectives and decision variables, this information must be expressed in a way close to these two items. This is the notion of performance indicator (figure 4).
.ol::-
~
. e;'b-
~
~
01$'
V L -_ _ _ _ _~
information
DECISION CENTRE Objectives
Figure 2. Control point of view with communication and information points of view (ICIMS, 1998)
Performance
At the level of decision-making itself, it is considered that the main items influencing the decision-making are : • the valued objective or set of objectives the decision centre has to meet, • the decision variables enabling the decision centre to know what it can act on and until where (constraints ).
Decision variables
Decision
agreytion Operations and environment
At the minimum level, decision-making corresponds to the search for a position in a space defined by decision variables and restricted by constraints in order to process the received information and to match the objectives (figure 3).
Figure 4. Decision centre and {Objectives, Decision Variables, Performance Indicators} triplet Performance indicators must be consistent with objectives because it is necessary to compare performances targeted (objectives) and performances reached (indicator) (observability). Performance indicators must be consistent with decision variables because these ones must have an effect on the performance monitored (controllability). Then the main issue is to ensure the internal consistency inside a decision centre in terms of the triplet presented. This consistency is ensured if the performance indicators allow to verifY the achievement of the objective, and are influenced by actions on decision variables (figure 5).
DECISION FRAME Objectives Decision
Max (V D2 ) 1
VII (VD2)
~space offreedom 1
Min (VD2)
- - r -
'!!!! _ "!! _!"'!_ !!"'!:! _.c~~_ ~ _"'!! _:_
-+_....1 _ _ _--:-·-:-_ Min(VD I)
,. 1 _ - . Decision variable I
. . Max(VD I) VII (VDI)
Figure 3. Principle of decision-making in conformity with a decision frame (here: 2 decision variables)
Figure 5. Consistency of the {Objectives, Decision Variables, Performance Indicators} triplet
A structure can be really controlled only if some feed-back loops are implemented (and particularly when the modelling of the system is weak). The source of this feed-back is all the raw data coming from the operational processes and from the environment of the system. These raw data can be relevant for the lowest levels of control because these levels are close to the real system. However, because the upper levels have a broader and more conceptual view of the operational process, the raw data must be
3.
MAN, CONTROL, DECISION-MAKING
3.1. Impact on control and performance An industrial system is an artefact. In this way, an industrial system must run in accordance to the objectives defined by the strategy of the company.
61
one criteria. Two systems, with the same resources, products, etc. may match a very different level of performances in accordance with the quality of decisions made inside their control system. That is why it is very important for an industrial system to create the best environment for decision-making.
This aspect has to be taken into account during the design phase and during the operation phase. Our interest in this paper is the operation phase. An industrial system has most of time a high level of flexibility . This flexibility is absolutely necessary to match the requirements coming from the market. On the other hand, the risk is to use this flexibility in a bad way, leading to non-match of the objectives of the company. That is why control is today a key issue for reaching desired performances.
Today, the essential role of man inside an industrial system is not to operate any more but to decide. Then , to be able to focus on decision-making in system engineering is also to focus on the position of men in this system.
The main purpose of enterprise control is to ensure that operational activities run in a consistent way and that objectives coming from upper level (corporate strategy) are correctly implemented. For that purpose, control must on one hand co-ordinate tasks in relation to these objectives and to enable the deployment of this objectives into the structure and, on the other hand, follow up performances of operations in order to compare them to objectives. However, many industrial cases show that the follow-up at every level of the hierarchy is sometimes absent and very often not usable. Then, the main issue is to merge two main approaches as shown figure 6 : • •
The first effect of this consideration is to avoid any model where the man would be considered outside the system (man = customer of the system) as in techno-centered approach such automatic control, computer sciences, etc. but as a part of the system and involved in processes of the system (figure 7). The first approach remains relevant when important automatic processes exist. However, it is not often the case in industrial systems for the reasons shown before. SYSTEME STUDIED
a top-down approach corresponding to the deployment of objectives, a bottom-up approach dealing with the follow-up of operations and aggregation of information.
SYSTEME STUD IED
One
process
Is I
automatic activIty
STRATEGY
..
human [ ] activity
Objectives ....._ . _ - Decomposition
Figure 7. Presence of a man in processes: customer of the system and between two processes (A) or actor of the system and inside the process (8)
Aggregation .. ---tl.~1
The second effect is that is not expectable to model and formalise the internal running of decisional activities. In fact, in a modelling activity, it is attractive to do so because these activities belong to the studied process. This trend could be problematic because:
Raw data
PROCESS
Fig. 6. The issue of matching decomposition of objectives and aggregation of information
•
3.2. Impact on decision-making and "the man in the system" Very often, objectives on quality, costs and lead times are simultaneously enunciated. The main problem is that all these objectives are looked for at the same time. Control is always multi-objective. That is why the problem of decision-making is of a tremendous importance because, most of time, to make decision is not reducible to the optimisation of
•
62
first , it will be generally disappointing because, despite the efforts to model these activities, the man will present a numerous behaviours outside the model. Then, it is wiser in many cases to consider real decisional activities as black boxes whose the main inputs and outputs are known in order to situate them inside the process; second, it could be dangerous if the activity of modelling is understood as a formalising activity in order to constraint the man to a nominal behaviour. The capacity to generate variety of the decision-maker could be partially or completely
the necessity to control processes execution so that the global objectives and strategies of the enterprise can be met.
inhibited by this constraint. This aspect is perfectly illustrated in some structures where the standardisation of activities has led to the loss of responsibility and imagination In abnormal situations.
------------- - - --- ------------~ I I
,--- --- -1
I
CONTROL VIEW
To take our statement into account, it is necessary to provide a control view at two levels:
I
I I
: Enterprise: ____ I
I I I I
I I I
I I I
• •
a local view enabling to define the environment of decision-making (information required, performance indicators, etc.), a global view in order to implement co-ordination and to ensure the deployment of corporate objectives in all the industrial system.
4.
I I I I
I I
Man-based I
enterprise control
MODELLING CONCEPTS AND CONSTRUCTS executed by
provides
..
Basic concepts to de velop constructs (in italic ) for enterprise modelling taking the above discussion into account are tentati vely stated : From enterprise control and decision-making point of view, an enterprise is controlled by a set of Decisions taken by decision-making Activity performed by human Resource according to Objective within imposed Constraints. A Decision has a Type and is characterised by its Horizon/Periode. Activity to make decision consists to choose (according to criteria) among a set of known alternatives (Variable) the one which meets best the Objective within Constraints. Decisions of the same type having the same Horizon/period form a Decision Center. Decision Centers are related by Decision-Links to form a Decision Structure. A minimum requirement on that decision structure is to have some basic Types of decisions and some basic Levels of decision-making.
composed of
triggers
Process-based operational control
Figure 8. Some basic constructs for modelling enterprise control and decision-making It is considered that the two approaches (Man-based Decision-making / Enterprise Control and Computerbased operational process execution) are complementary rather than contradictory. To summary, the proposed approach is based on the following considerations : The scope of enterprise modelling must deal with the entire domain of enterprise, starting from strategy formulation and business plan down to the level of daily operational monitoring and control. While most of the process-oriented approaches only focus on lower levels of management and manufacturing. In an enterprise like any complex system where human involvement is important, only part of the system can be formalised in terms of processes which are computer executable. This process oriented approach is largely developed these years. To represent the non-formalised part of the enterprise where some higher level management activities are not algorithmic (not process-based), it is necessary to take human factors into account. The role of man in the enterprise system is to take decisions in order to control some significant states which are important for the company to meet its global objectives and strategies.
From operational control and information processing point of view, an enterprise is run by pre-defined Processes which are triggered by pre-identified Events. A Process is composed of Activities which are executed by Resources (Man or Machine including computer programme) providing needed Capabilities. Inputs and Outputs of an Activity are Object Views which are Information elements describing Enterprise Objects. Procedure Rules define the sequence of execution of Processes and Activities with predefined ending status so that the model can be interpreted by computer to control the start and termination of each process. Figure 8 shows some constructs to model enterprise control / decision-making and its relationship to process-oriented approach . In particular, it underlines
5.
63
CONCLUSIONS
evolution and engineering, May 24-26, 2000, Berlin. Doumeingts, G., Vallespir, B., Zanettin and M., Chen, D. (1992) . GIM : GRAI Integrated Methodology for designing CIM systems. GRAIILAP, University Bordeaux I, version 1.0, May. Doumeingts, G., Vallespir, B. and Chen, D. (1998). Decision modelling GRAI grid. Chapter in : Handbook on architecture for Information Systems (Peter Bernus, Kai Mertins, Gunter Schmidt. (Ed» , Springer. ENV 40003 ( 1990). Computer Integrated Manufacturing System Architecture Framework for Enterprise Modeling. European Pre-standard, CENICENELEC. ENV 12204 (1995 ). Advanced Manufacturing Technology - System Architecture - Constructs for Enterprise Modeling, European Pre-standard, CEN TC 3 IOIWG I . ICIMS ( 1998). ICIMS-NOE Scientific meeting, Brussels, Belgium. November 24, 1999, in ICIMS NEWS, March 1999. IFAC-IFIP Task Force (1997). GERAM : Generalized Enterprise Reference Architecture and Methodology. Version 1.4, ISOTCI 84 ISCSIWGI , N398. IS 15704 (1998). Requirements for Enterpri se Reference Architecture and Methodologies. ISO TC 1841SCSIWG 1. Vernadat, F. B. (1996). Enterprise Modelling and Integration Principles and Applications. Chapman & Hall, ISBN 0-412-60550-3 . Williams, TJ. (1994). The Purdue Enterprise Reference Architecture, In: Computer In Industry, vol. 24. 1994.
The rapid development and penetration of information and communication technology inside the enterprise enable individuals to work in a more autonomous and co-operative way. However, enterprise itself should remain controllable in order to reach its global objectives and strategies. Face on various complex situations that an industrial system may encounter during its operation, man-based decision-making control is proposed as a complementary viewpoint to computer executable process oriented approach . This paper has tentatively presented the basic concepts and constructs to model this aspect. The added value to conventional approaches would be the focus on human role (as a decision-maker) in the system. Future work would be concerned with the precise definition of constructs and a case study to validate the approach. REFERENCES AMICE (1993) . CIMOSA: Open System Architecture for CIM. Springer- Verlag, Berlin, 1993. Chen, D. and Doumeingts, G. (1996). The GRAIGIM reference model , architecture and methodology. In : Architecture f or Enterprise integration (P. Bernus, L. Nemes and T. Williams (Ed» , Chapman & Hall. Chen, D. and Vallespir, B. (1999) . General theories of design and enterprise modelling. ASf'99 annual Confe rence, Leuven, Belgium, September 24-26. Chen, D. , Vallespir, B. and Doumeingts, G. (2000) . Enterprise modelling and engineering: some complementary requirements on standardisation. In CEN TC310 WGI workshop on Enterprise
64
[: 0
Copyright IFAC Intelligent Manufacturing Systems, Poznan, Poland, 200 I
[>
Publications www.elsevier.comllocate/ifac
MAN, DECISION-MAKING, CONTROL AND PERFORMANCE: A COMPLEMENTARY VIEW TO PROCESS-ORIENTED APPROACH
David Chen, Bruno Vallespir and Guy Doumeingts
LA P/GRA I, University Bordeaux I 351, Cours de la liberation, 33405 Talence cedex France Tel : +33556846530; Fax : +33556846644 [email protected]
Abstract : This paper tentatively presents some basic concepts and constructs for the modelling of enterprise control and decision-making. It is proposed that man-centred decision-making oriented enterprise control be a complementary view to computer executed process-oriented approach . Based on some theoretical arguments from automatic control, the paper discusses the set of concepts on the role of man in enterprise to make decisions and to control operational processes so that global objectives of the company can be reached. Copyright © 2001 IFAC Keywords : Enterprise modelling, Enterprise Integration, Control, Decision-making, performance
I. INTRODUCTION
the architecture implemented enables to get the level of performance required by the strategy of the company (Chen, 1996, 1999, 2000). Enterprise control plays a major role for matching performances and then to keep competitive. However, most of current enterprise modelling techniques are mainly information, function and process oriented and the control aspect is reduced to an information processing issue. To match current industrial challenges, control needs to be in an equivalent position than the others. To control a system that cannot be entirely formalised and modelled, it is necessary to provide a set of decision-centres with co-ordination links between them (Doumeingts, 1998, 1992). Focusing on control is focusing on decision-making and on the role of man in industrial systems as well. Man-based decision-making oriented enterprise control is proposed as a complementary viewpoint with respect to computer executed and process oriented approach. Main items of the decision-making are (I) the valued objective or set of objectives the decision has to match; (2) the decision variables enabling the decision-maker to know what it can act on and until where
The approach presented in this paper fits within the framework of IF AC Technical Committee on Enterprise Integration. It deals with the elaboration of enterprise reference architecture and methodologies including concepts, principles and constructs for the analysis and design of enterprise systems. The IF ACIFIP Task Force on Enterprise Integration has developed GERAM (Generalised Enterprise Reference Architecture and Methodologies) (lfac-Ifip task force, 1997). GERAM is based on the three main approaches: GRAI (GRAI Integrated Methodology) (Doumeingts, 1992, 1998), CIMOSA (CIM Open System Architecture) (Amice, 1993) and PERA (Purdue Enterprise Reference Architecture) (Williams, 1994). An International Standard is adopted based on the input of GERAM (IS, 1998). At the European level, CEN (European Committee for Standardisation) is working on the finalisation of two standards dealing with the framework and constructs for enterprise modelling (ENV 40003, 1990; ENV 12204, 1995). This paper addresses the necessity to model enterprise control viewpoint in order to ensure that
59
(constraints); (3) the target performance that an enterprise wants to achieve.
are defined and delimited by a co-ordinator module. Generally, the co-ordinator is in the same situation than that of the co-ordinated modules: it is coordinated too. Then, the complex, multi-objectives systems, such as industrial systems, are controlled by a hierarchical, multi-level structure which owns the important property of decentralised decision-making.
In the first part of the paper, some theoretical arguments from the point of view of automatic control theory are presented. Then enterprise wide decision-making and control are discussed. In the third part of the paper, basic concepts to develop constructs for enterprise modelling taking this aspect into account are tentatively stated. Future work and perspectives are given as part of conclusion.
2.
oflhe Control syslem
THEORETICAL ARGUMENTS
The traditional approach in system engineering coming from Automatic Control considers that a system has some significant states that can be formalised . Formalising these states enables to take them into account , i.e. to translate them into a model which is used to control the system. To consider that the significant states can also be formalised is remarkable. This assumption is the basis of automatic control theory . It leads the system to run "automatically" when it is in a formalised state or to be blocked or insensitive in other states. This approach is used when the risk to go out the set of formalised states is weak because the system is quasi -exhaustively modelled. However, in a complex enterprise system where human takes part in, the situations and events influencing the system cannot be entirely identified and modelled. That system is rarely in a nominal mode. Very often, the model of the system is an ideal process corresponding rarely to the reality. How to control the system when it is in a state which is significant but not formalised ? A principle coming from cybernetics gives a piece of solution : a system is really controlled only if the control system proposes at least as many states (control system variety) as the controlled system (controlled system variety) . Because this variety does not exist a priori, it is necessary to have inside the control system some variety generators, i.e. organs that are capable to react to a situation at the time the situation appears. The only organ known to be versatile and quick enough to play this role is the human being (see figure I) .
Sel of Ihe Slales oflhe Controlled Syslem
Sel of Ihe non· significant SI ales of Ihe syslem
Figure I . Increase of the control system variety by the presence of human beings The cognitive limitation of decision-makers is also an argument for hierarchical , decentralised control systems. A decision-maker has a cognitive limitation which can be expressed in term of quantity of information. Beyond this limit, the decision-maker is submerged by the information : he is not able to interpret the information he is supposed to use in order to make decision . The quantity of information is proportional to : • •
the size of the domain in which the decisionmaker is supposed to make decision , and the detail of the information handled by the decision-maker.
The information system also needs to be aggregated as it goes up in the structure, in order to provide the good granularity of information to decision-makers. To summary, to control a system that cannot be formalised enough, it is necessary to provide a set of decision-makers with co-ordination links between them.
The human being can play the role of variety generator if his freedom to make decision is high. This leads to a difficulty: a high freedom can lead the decision-maker to have a behaviour that is not compatible with the objectives of the system. Then, the decisional freedom needs to be restricted by a pre-defined decision frame . The problem grows when the number of decision-makers grows. There is a need to ensure that local decisional freedoms dealing with specific objectives be consistent with global objective(s) of the system. The solution is coordination in the meaning that decisional freedoms
Focusing on control is focusing on decision-making and on the role of man in industrial systems as well. Then, it is a necessity to avoid mechanistic approaches and to situate human decision into the system . To match current industrial challenges, control needs to be in an equivalent position than information and communication (figure 2).
60
aggregated to be used at any level of the hierarchy of control. Secondly, because decision is mainly made on a basis of objectives and decision variables, this information must be expressed in a way close to these two items. This is the notion of performance indicator (figure 4).
.ol::-
~
. e;'b-
~
~
01$'
V L -_ _ _ _ _~
information
DECISION CENTRE Objectives
Figure 2. Control point of view with communication and information points of view (ICIMS, 1998)
Performance
At the level of decision-making itself, it is considered that the main items influencing the decision-making are : • the valued objective or set of objectives the decision centre has to meet, • the decision variables enabling the decision centre to know what it can act on and until where (constraints ).
Decision variables
Decision
agreytion Operations and environment
At the minimum level, decision-making corresponds to the search for a position in a space defined by decision variables and restricted by constraints in order to process the received information and to match the objectives (figure 3).
Figure 4. Decision centre and {Objectives, Decision Variables, Performance Indicators} triplet Performance indicators must be consistent with objectives because it is necessary to compare performances targeted (objectives) and performances reached (indicator) (observability). Performance indicators must be consistent with decision variables because these ones must have an effect on the performance monitored (controllability). Then the main issue is to ensure the internal consistency inside a decision centre in terms of the triplet presented. This consistency is ensured if the performance indicators allow to verifY the achievement of the objective, and are influenced by actions on decision variables (figure 5).
DECISION FRAME Objectives Decision
Max (V D2 ) 1
VII (VD2)
~space offreedom 1
Min (VD2)
- - r -
'!!!! _ "!! _!"'!_ !!"'!:! _.c~~_ ~ _"'!! _:_
-+_....1 _ _ _--:-·-:-_ Min(VD I)
,. 1 _ - . Decision variable I
. . Max(VD I) VII (VDI)
Figure 3. Principle of decision-making in conformity with a decision frame (here: 2 decision variables)
Figure 5. Consistency of the {Objectives, Decision Variables, Performance Indicators} triplet
A structure can be really controlled only if some feed-back loops are implemented (and particularly when the modelling of the system is weak). The source of this feed-back is all the raw data coming from the operational processes and from the environment of the system. These raw data can be relevant for the lowest levels of control because these levels are close to the real system. However, because the upper levels have a broader and more conceptual view of the operational process, the raw data must be
3.
MAN, CONTROL, DECISION-MAKING
3.1. Impact on control and performance An industrial system is an artefact. In this way, an industrial system must run in accordance to the objectives defined by the strategy of the company.
61
one criteria. Two systems, with the same resources, products, etc. may match a very different level of performances in accordance with the quality of decisions made inside their control system. That is why it is very important for an industrial system to create the best environment for decision-making.
This aspect has to be taken into account during the design phase and during the operation phase. Our interest in this paper is the operation phase. An industrial system has most of time a high level of flexibility . This flexibility is absolutely necessary to match the requirements coming from the market. On the other hand, the risk is to use this flexibility in a bad way, leading to non-match of the objectives of the company. That is why control is today a key issue for reaching desired performances.
Today, the essential role of man inside an industrial system is not to operate any more but to decide. Then , to be able to focus on decision-making in system engineering is also to focus on the position of men in this system.
The main purpose of enterprise control is to ensure that operational activities run in a consistent way and that objectives coming from upper level (corporate strategy) are correctly implemented. For that purpose, control must on one hand co-ordinate tasks in relation to these objectives and to enable the deployment of this objectives into the structure and, on the other hand, follow up performances of operations in order to compare them to objectives. However, many industrial cases show that the follow-up at every level of the hierarchy is sometimes absent and very often not usable. Then, the main issue is to merge two main approaches as shown figure 6 : • •
The first effect of this consideration is to avoid any model where the man would be considered outside the system (man = customer of the system) as in techno-centered approach such automatic control, computer sciences, etc. but as a part of the system and involved in processes of the system (figure 7). The first approach remains relevant when important automatic processes exist. However, it is not often the case in industrial systems for the reasons shown before. SYSTEME STUDIED
a top-down approach corresponding to the deployment of objectives, a bottom-up approach dealing with the follow-up of operations and aggregation of information.
SYSTEME STUD IED
One
process
Is I
automatic activIty
STRATEGY
..
human [ ] activity
Objectives ....._ . _ - Decomposition
Figure 7. Presence of a man in processes: customer of the system and between two processes (A) or actor of the system and inside the process (8)
Aggregation .. ---tl.~1
The second effect is that is not expectable to model and formalise the internal running of decisional activities. In fact, in a modelling activity, it is attractive to do so because these activities belong to the studied process. This trend could be problematic because:
Raw data
PROCESS
Fig. 6. The issue of matching decomposition of objectives and aggregation of information
•
3.2. Impact on decision-making and "the man in the system" Very often, objectives on quality, costs and lead times are simultaneously enunciated. The main problem is that all these objectives are looked for at the same time. Control is always multi-objective. That is why the problem of decision-making is of a tremendous importance because, most of time, to make decision is not reducible to the optimisation of
•
62
first , it will be generally disappointing because, despite the efforts to model these activities, the man will present a numerous behaviours outside the model. Then, it is wiser in many cases to consider real decisional activities as black boxes whose the main inputs and outputs are known in order to situate them inside the process; second, it could be dangerous if the activity of modelling is understood as a formalising activity in order to constraint the man to a nominal behaviour. The capacity to generate variety of the decision-maker could be partially or completely
the necessity to control processes execution so that the global objectives and strategies of the enterprise can be met.
inhibited by this constraint. This aspect is perfectly illustrated in some structures where the standardisation of activities has led to the loss of responsibility and imagination In abnormal situations.
------------- - - --- ------------~ I I
,--- --- -1
I
CONTROL VIEW
To take our statement into account, it is necessary to provide a control view at two levels:
I
I I
: Enterprise: ____ I
I I I I
I I I
I I I
• •
a local view enabling to define the environment of decision-making (information required, performance indicators, etc.), a global view in order to implement co-ordination and to ensure the deployment of corporate objectives in all the industrial system.
4.
I I I I
I I
Man-based I
enterprise control
MODELLING CONCEPTS AND CONSTRUCTS executed by
provides
..
Basic concepts to de velop constructs (in italic ) for enterprise modelling taking the above discussion into account are tentati vely stated : From enterprise control and decision-making point of view, an enterprise is controlled by a set of Decisions taken by decision-making Activity performed by human Resource according to Objective within imposed Constraints. A Decision has a Type and is characterised by its Horizon/Periode. Activity to make decision consists to choose (according to criteria) among a set of known alternatives (Variable) the one which meets best the Objective within Constraints. Decisions of the same type having the same Horizon/period form a Decision Center. Decision Centers are related by Decision-Links to form a Decision Structure. A minimum requirement on that decision structure is to have some basic Types of decisions and some basic Levels of decision-making.
composed of
triggers
Process-based operational control
Figure 8. Some basic constructs for modelling enterprise control and decision-making It is considered that the two approaches (Man-based Decision-making / Enterprise Control and Computerbased operational process execution) are complementary rather than contradictory. To summary, the proposed approach is based on the following considerations : The scope of enterprise modelling must deal with the entire domain of enterprise, starting from strategy formulation and business plan down to the level of daily operational monitoring and control. While most of the process-oriented approaches only focus on lower levels of management and manufacturing. In an enterprise like any complex system where human involvement is important, only part of the system can be formalised in terms of processes which are computer executable. This process oriented approach is largely developed these years. To represent the non-formalised part of the enterprise where some higher level management activities are not algorithmic (not process-based), it is necessary to take human factors into account. The role of man in the enterprise system is to take decisions in order to control some significant states which are important for the company to meet its global objectives and strategies.
From operational control and information processing point of view, an enterprise is run by pre-defined Processes which are triggered by pre-identified Events. A Process is composed of Activities which are executed by Resources (Man or Machine including computer programme) providing needed Capabilities. Inputs and Outputs of an Activity are Object Views which are Information elements describing Enterprise Objects. Procedure Rules define the sequence of execution of Processes and Activities with predefined ending status so that the model can be interpreted by computer to control the start and termination of each process. Figure 8 shows some constructs to model enterprise control / decision-making and its relationship to process-oriented approach . In particular, it underlines
5.
63
CONCLUSIONS
evolution and engineering, May 24-26, 2000, Berlin. Doumeingts, G., Vallespir, B., Zanettin and M., Chen, D. (1992) . GIM : GRAI Integrated Methodology for designing CIM systems. GRAIILAP, University Bordeaux I, version 1.0, May. Doumeingts, G., Vallespir, B. and Chen, D. (1998). Decision modelling GRAI grid. Chapter in : Handbook on architecture for Information Systems (Peter Bernus, Kai Mertins, Gunter Schmidt. (Ed» , Springer. ENV 40003 ( 1990). Computer Integrated Manufacturing System Architecture Framework for Enterprise Modeling. European Pre-standard, CENICENELEC. ENV 12204 (1995 ). Advanced Manufacturing Technology - System Architecture - Constructs for Enterprise Modeling, European Pre-standard, CEN TC 3 IOIWG I . ICIMS ( 1998). ICIMS-NOE Scientific meeting, Brussels, Belgium. November 24, 1999, in ICIMS NEWS, March 1999. IFAC-IFIP Task Force (1997). GERAM : Generalized Enterprise Reference Architecture and Methodology. Version 1.4, ISOTCI 84 ISCSIWGI , N398. IS 15704 (1998). Requirements for Enterpri se Reference Architecture and Methodologies. ISO TC 1841SCSIWG 1. Vernadat, F. B. (1996). Enterprise Modelling and Integration Principles and Applications. Chapman & Hall, ISBN 0-412-60550-3 . Williams, TJ. (1994). The Purdue Enterprise Reference Architecture, In: Computer In Industry, vol. 24. 1994.
The rapid development and penetration of information and communication technology inside the enterprise enable individuals to work in a more autonomous and co-operative way. However, enterprise itself should remain controllable in order to reach its global objectives and strategies. Face on various complex situations that an industrial system may encounter during its operation, man-based decision-making control is proposed as a complementary viewpoint to computer executable process oriented approach . This paper has tentatively presented the basic concepts and constructs to model this aspect. The added value to conventional approaches would be the focus on human role (as a decision-maker) in the system. Future work would be concerned with the precise definition of constructs and a case study to validate the approach. REFERENCES AMICE (1993) . CIMOSA: Open System Architecture for CIM. Springer- Verlag, Berlin, 1993. Chen, D. and Doumeingts, G. (1996). The GRAIGIM reference model , architecture and methodology. In : Architecture f or Enterprise integration (P. Bernus, L. Nemes and T. Williams (Ed» , Chapman & Hall. Chen, D. and Vallespir, B. (1999) . General theories of design and enterprise modelling. ASf'99 annual Confe rence, Leuven, Belgium, September 24-26. Chen, D. , Vallespir, B. and Doumeingts, G. (2000) . Enterprise modelling and engineering: some complementary requirements on standardisation. In CEN TC310 WGI workshop on Enterprise
64