- Email: [email protected]
A ubiquitous manufacturing network system
Robotics and Computer-Integrated Manufacturing ∎ (∎∎∎∎) ∎∎∎–∎∎∎
Contents lists available at ScienceDirect
Robotics and Computer-Integrated Manufacturing journal homepage: www.elsevier.com/locate/rcim
A ubiquitous manufacturing network system Yu-Cheng Lin a,n, Toly Chen b a b
Department of Computer-aided Industrial Design Overseas Chinese University, Taiwan Department of Industrial Engineering and Systems Management Feng Chia University, Taiwan
art ic l e i nf o
a b s t r a c t
Article history: Received 3 July 2015 Received in revised form 24 September 2015 Accepted 11 October 2015
In this study, a ubiquitous manufacturing network system was constructed. In this system, a customer places an order for an action figure by using a client-side app or a Web-based interface and pays online. The system server then assigns the order to the convenience store nearest the customer's location to print the required action figure. For determining the most suitable convenience store, a fuzzy integernonlinear programming model was proposed and solved using two modified fuzzy Dijkstra algorithms. Subsequently, the customer is informed of the location of and route to the recommended convenience store. Two illustrative cases were used to verify the applicability of using the proposed methodology. In addition, compared with an existing mobile guide, the proposed methodology effectively recommended the shortest path for obtaining the required action figure and reduced the waiting time at the convenience store. & 2015 Elsevier Ltd. All rights reserved.
Keywords: Ubiquitous manufacturing Location-aware service 3D printing Action figure
1. Introduction Ubiquitous computing is a concept in software engineering and computer science in which computing is performed at any location. As an application of ubiquitous computing in the manufacturing sector, ubiquitous manufacturing (UbiM) features an environment in which manufacturing services are provided at any location. A concept similar to UbiM is cloud manufacturing (CM; [26]). Both UbiM and CM enable ubiquitous, convenient, on-demand network access to a shared pool of configurable manufacturing resources (e.g., software tools, equipment, and manufacturing capabilities). However, in contrast with CM, UbiMemphasizes the mobility and dispersion of manufacturing resources and users. In general, manufacturing a product ubiquitously is impossible. Previously, UbiM implied that products could be supplied ubiquitously. Some advanced manufacturing technologies, such as lean manufacturing [4], CM, manufacturing grids[15], global manufacturing [15], virtual manufacturing [21,7], agile manufacturing [16], Internet manufacturing [17], and particularly additive manufacturing (AdM; [19]) have increased UbiM opportunities. An AdM network can be used to transfer the three-dimensional (3D) model of a product (in .STL or .OBJ file format) to a 3D printer at any locationand any time over the Internet, at which point the product isbuilt layerbylayer. At the end of the manufacturing process, excess soft resin is cleaned using a chemical n
Corresponding author.
bath [5]. Thus, manufacturing a product ubiquitously has become possible. Developments in AdM have enabled printing extremely complex 3D models. Various methodologies, such asthat developed by Ding et al. [13], have been proposed for generating deposition paths and arcs for such products. Keating and Oxman [18] reconfigured a 6-axis robotic arm into an integrated platform for 3D printing, milling, and sculpting and proposed the concept of “compound fabrication.” This study established a UbiM network system, in which a customer requests that a product be manufactured at a 3D output locationand arrives atthe 3D output location for the just in time (JIT; [8]) manufacturing of the requested product. UbiM network systems are used to manufacture action figures of popular cartoon or movie characters. The business feasibility of using a 3D printer to manufacture action figures was mentioned by Berman [5] and has been successfully demonstrated [10]. Information companies, 3D photo studios, supermarket chains, and other shops provide (Table 1) limited services for the printing of 3D action figures. In addition, by mass-producing action figures, 3D printing service providers such as the Shenzhen Xin Ju Xin Toy Design Company [24] have contradicted the conventional belief that 3D printers can be used only for small-scale trial production. However, most of the existing service providers have a single 3D output location;a systematic UbiM network remains undeveloped. As shown in Table 1, some companies have substantial waiting periods for receiving desired action figures. Such waiting is worthwhile if 3D printing enables customers to receive an action figure that they want as soon as possible, because even the largest
http://dx.doi.org/10.1016/j.rcim.2015.10.009 0736-5845/& 2015 Elsevier Ltd. All rights reserved.
Please cite this article as: Y.-C. Lin, T. Chen, A ubiquitous manufacturing network system, Robotics and Computer Integrated Manufacturing (2016), http://dx.doi.org/10.1016/j.rcim.2015.10.009i
Y.-C. Lin, T. Chen / Robotics and Computer-Integrated Manufacturing ∎ (∎∎∎∎) ∎∎∎–∎∎∎
2
Table 1 Examples of 3D printing service providers. Service provider
Region
Service items
Size of action figure
Price (USD)
Required time
Reference
Asda [1]
UK
Scan & print
Unknown
45
Chih-Mao information
Taiwan Print; Scan & print Taiwan Scan & print China Print
Height: 10 cm
32–39 /h
A week afterward 4.83 h
http://www.3D-printing.net/content/asda-launching-3D-printingservice-tomorrow http://www.cmic.com.tw/Production_process.aspx
13–15 cm Unknown
415 2–20
1–2 weeks 8–15 days
http://www.ego3d.com.tw/price.html http://szjxsy.en.alibaba.com/product/60239779940-212431068/ 3d_printing_anime_action_figure.html#!
EGO3D Shenzhen Xin Ju Xin Toy Design Co.
toy makers do not stock all types of action figure. In addition, transporting an action figure from an overseas toy maker to a customer's location may involvea longer waiting time. After the ordered action figure is printed, the customer visits the 3D output location and collects the action figure. Presently, the 3D output location may be located a considerable distance from the customer's location. Therefore, for overcoming this problem, this study proposed a UbiM network system. The customer can be guided to the nearest 3D output location to reduce travel time, a procedure called static UbiM. In another situation [6], dynamic UbiM was used for further elaborating the effectiveness of UbiM network systems. In this study's procedure, a customer receives directions for obtaining a pre ordered action figure from a preferred 3D output location while traveling according to the following objectives: (1) The customer should promptly leave for the 3D output location. (2) The 3D output location should be located near the customer's route. (3) When the customer arrives at the 3D output location, the requested action figure should be ready. In other words, the customer's waiting time at the 3D output location should be minimal. Fig. 1. Distribution of convenience stores in a 2-km2 region.
For achieving these objectives, the proposed methodology designated a chain of convenience stores to be the 3D output locations; the stores are suitable for the following reasons:
2. 3D printing 2.1. Platforms
(1) Some regions contain a high concentration of convenience stores that provide nearly identical services. For instance, a 2-km2 region in the Zhongzheng District of Taipei City in Taiwan (Fig. 1) contains 36 convenience stores of the same chain. (2) Most convenience stores provide printing services, which could serve as a basis for 3D printing services. (3) Convenience store chains in Taiwan regularly introduce promotional campaigns that allow customers to exchange points, accumulated after a fixed amount of purchases, for an action figure. The remainder of the paper is organized as follows: Section 2 presents a literature review of 3D printing. Section 3 describes the architecture and process flow of UbiM network systems. Section 4 presents a fuzzy integer-nonlinear programming (FINLP) model for solving the JIT output location and path problems. For solving the FINLP problems for static UbiM and dynamic UbiM cases, two modified fuzzy Dijkstra algorithms were proposed. In addition, two cases were used to illustrate the applicability of the proposed methodology. Section 5 presents a comparison of the performance of the proposed methodology with that of an existing method. Finally, Section 6 concludes the paper.
3D printing can be seamlessly integrated with CAD/CAM operations [5]. Therefore, the existing CAD/CAM software vendors are typically the pioneers in this field. For example, Autodesk recently developed an open 3D printing platform named Spark [3] on which application program interfaces (APIs) are provided for each stage of 3D printing. Consequently, 3D printing functionality can be quickly added to an application. By using Spark, a customer can design and even optimize a 3D model that can be printed by a specific 3D printer. There are also websites acting as hubs for gathering 3D models from volunteers worldwide. For example, My Mini Factory test prints an uploaded 3D model before it is publicized [22]. Through its online catalog, shapeways shows the 3D models it has collected for printing and purchasing, thus providing the company with a 3.5% commission from the designer of each 3D model printed or sold. Equipped with more than 100 3D printers of various sizes, i. materialise can be used to print 3D models uploaded by customers or passed from websites that provide 3D printing services but do not have printers [14]. APIs are provided for connecting a service website to i.materialise, in addition to third-party logistics that directly ship the printed 3D model to the service website's customer.
Please cite this article as: Y.-C. Lin, T. Chen, A ubiquitous manufacturing network system, Robotics and Computer Integrated Manufacturing (2016), http://dx.doi.org/10.1016/j.rcim.2015.10.009i
Y.-C. Lin, T. Chen / Robotics and Computer-Integrated Manufacturing ∎ (∎∎∎∎) ∎∎∎–∎∎∎
2.2. Applications
3
Customer browses the online catalog
3D printing has been used to produce prototypes and mockups, replacement parts, dental crowns, artificial limbs, and even bridges [5]. 3D printing has considerable potential for producing complex internal and external porous structures [2]. For example, Lam et al. [20] used a blend of starch-based polymer powders (cornstarch, dextran, and gelatin) for making 3D porous scaffolds with a 3D printer. Asadi-Eydivand et al. [2] optimized the parameters of 3D printing by using design-of-experiment techniques when making calcium-sulfate-based scaffolding prototypes. Building 3D models from the substantial number of images in existing databases remains the primary endeavor of this field. For example, Silva et al. [25] manipulated the tomographic images of a dry skull, thus enabling the craniofacial skeleton to be accurately reproduced using 3D printing. Rengier et al. [23] produced 3D objects for surgical planning according to volumetric medical images produced by computed tomography (CT) and magnetic resonance imaging (MRI). Websites like shapeway and i.materialise manage substantial online databases of 3D models uploaded by users worldwide. However, whether users are the copyright holders of the 3D models that they upload and whether downloading infringes on the rights of copyright holders are two problems that warrant further examination.
Customer chooses an action figure System server estimates the customer’s location and speed System server determines the convenient store
System server responds to the customer
Customer makes the payment online
System server assigns the task to the convenient store
Customer goes to the convenient store
The convenient store prints the action figure
3. System architecture and process flow A four-tier architecture was configured for the UbiM network system (Fig. 2). The UbiM network system comprises four major components:
Customer takes the action figure Fig. 3. Process flow of the UbiM network system.
(1) The customer or client browses the online product catalog by using a client-side app or system website, selects a desired action figure, makes an online payment, and receives instructions regarding the pick-up location of the action figure. (2) The system server prepares the online product catalog, receives and processes the customer's request, confirms receiving the online payment, and assigns the task to a convenience store. (3) A bank or an online payment service provides interfaces with the customer by displaying information relayed from the system server regarding online payment.
(4) The convenience store prints the required action figure according to the instructions from the system server and serves the customer upon their arrival. The UbiM network system exhibits the following process flow (Fig. 3): (1) A customer browses the online product catalog and chooses the desired action figure. (2) By using data from the global positioning system (GPS) system
Supply Locations
Tier 4 action figure
customer arrival data time
Tier 3
chosen action figure
System Server possible action customer location chosen figures data speed action figure
Tier 2
output details
Bank
Communication Service Provider possible action customer location chosen figures data speed action figure
payment
payment output details
Customers
Tier 1 use browser
use app
Fig. 2. Four-tier architecture of the proposed UbiM network system.
Please cite this article as: Y.-C. Lin, T. Chen, A ubiquitous manufacturing network system, Robotics and Computer Integrated Manufacturing (2016), http://dx.doi.org/10.1016/j.rcim.2015.10.009i
Y.-C. Lin, T. Chen / Robotics and Computer-Integrated Manufacturing ∎ (∎∎∎∎) ∎∎∎–∎∎∎
4
Fig. 4. Using a 3D printer to print an action figure.
(3) (4) (5) (6)
(7)
(8)
in the customer's smartphone, the system server estimates the location and speed of the customer. The system server determines the convenience store nearest the customer. The customer makes an online payment. The system server assigns the task to a convenience store. The system server provides the customer with details on the recommended convenience store and the time and route to reach the store. (Concurrent operations) The customer leaves for the recommended convenience store as the convenience store prints the desired action figure. The customer reaches the convenience store and collects the action figure.
Using a 3D printer to print an action figure involves five steps (Fig. 4): proofreading the 3D model file, printing, unearthing the model, removing the gypsum powder, and postproduction hardening. In the proposed UbiM network system, the 3D model files of all of the action figures in an online catalog were prepared; therefore, the first step may be unnecessary. In addition, certain steps require human intervention. Therefore, the processing times of such steps are variable. However, developments in AdM technology are expected to resolve this problem. The manufacturing cost for the 10-cm-high action figure depicted in Fig. 4 is approximately US$65–81, and the processing time is approximately 4.83 h.
location). The node farthest from the customer is node n. The length of the path connecting nodes i and j is lij , where i and j¼1 n; i≠j; and lij = ∞ if no connection exists between the two nodes. The shortest distance from the customer's location to node i is denoted as di . Obviously, d1 = 0, dn = max di , and di can be dei
rived according to the shortest distances of the nodes connected to di :
di = min (dj + l ji ) j
(1)
Assume a convenience store is located at node k. The problem of determining the shortest path to the convenience store can be formulated as an integer-nonlinear programming (INLP) problem as follows:
(2)
Min dk s.t.
di =
min (dj + l ji ), i = 1~k
j < i, l ji ≠∞
(3)
Eq. (3) can be replaced by
di ≤ dj + l ji , i = 1~k; j < i; l ji ≠ ∞
∑
di =
xji (dj + l ji ), i = 1~k
j < i, l ji ≠∞
∑
(4)
(5)
xji = 1, i = 1~k
j < i, l ji ≠∞
(6)
xji ∈ { 0, 1}, i = 1~k; j < i; lij ≠ ∞
(7)
4. Determining the most suitable convenience store 4.1. Static UbiM case The “most suitable” convenience store for printing an action figure is the easiest to reach, thus minimizing the waiting time. Two steps are required for determining the most suitable convenience store. First, the shortest path to each convenience store is determined. A shortest path problem can be solved in polynomial time only in graphs without negative cycles (Cormen et al., 2001). Subsequently, the waiting time for each convenience store is estimated. The nearest convenience store that minimizes the waiting time is recommended. The service region is abstracted into a traffic network, in which only the major roads and corners are considered to accelerate the reasoning process. Assume that there are n nodes in the abstracted traffic network. A customer is located at node 1 (i.e., the starting
However, determining the global optimal solution to this INLP problem is challenging. For managing this difficulty, many algorithms such as the Dijkstra, Bellman–Ford, A* search, Floyd–Warshall, and Johnson algorithms have been proposed [12,9]. Google Maps, which is relatively easy to use with online applications but which can provide inaccurate results, was used for identifying the shortest path. However, considering individual differences in the results obtained using Google Maps is inconvenient. Regarding the uncertainty of a customer's speed, the shortest distance to node i can be modeled using a triangular fuzzy number d˜ i = (di1, di2, di3 ). For example, if the speed of a customer is 30, 45, or 50 km/h, then the times required for the customer to traverse a 1-km path are 0.02, 0.022, and 0.033 h, respectively. After such uncertainty is considered, the following FINLP
Please cite this article as: Y.-C. Lin, T. Chen, A ubiquitous manufacturing network system, Robotics and Computer Integrated Manufacturing (2016), http://dx.doi.org/10.1016/j.rcim.2015.10.009i
Y.-C. Lin, T. Chen / Robotics and Computer-Integrated Manufacturing ∎ (∎∎∎∎) ∎∎∎–∎∎∎
5
Fig. 5. Experimental region.
problem is formulated:
Min d˜k
(8)
s.t. ~ ~ di ≤ dj ( + ) l ji , i = 1~k; j < i; lij ≠ ∞
~ di =
∑
˜ q = max {0, (wq1, wq2, wq3 )} w = max {0, (tc ( + ) p˜ )( − )(ts ( + ) d˜ q )}
(9)
~ xji (dj ( + ) l ji ), i = 1~k
j < i, l ji ≠∞
the waiting time at the convenience store located at node q is estimated as follows:
(10)
= max {0, (tc + p1 − ts − dq3, tc + p2 − ts − dq2, tc + p3 − ts − dq1)}
where tc is the current time, ts is the time when the customer departs for the recommended convenience store, and ( − ) in˜ q , Eq. (15) is used: dicates a fuzzy subtraction. For minimizing w
(tc + p1 − ts − dq3, tc + p2 − ts − dq2, tc + p3 − ts − dq1) = 0
∑
xji ∈ {0, 1} , i = 1~k; j < i; lij ≠ ∞
(11)
(12)
where (þ ) indicates a fuzzy addition. Instead of directly solving the FINLP problem, the problem is solved using several approximation algorithms including the fuzzy Dijkstra algorithm proposed by Deng et al. [11] and the fuzzy Dijkstra algorithm developed by Chen [6]. Let
d˜ q = min d˜k k∈Q
(15)
Applying the center-of-gravity method [27] to Eq. (15) yields
xji = 1, i = 1~k
j < i, l ji ≠∞
(14)
(13)
where Q is the index set of convenience stores. Accordingly, the convenience store located at node q is recommended.
tc + p1 − ts − dq3 + tc + p2 − ts − dq2 + tc + p3 − ts − dq1 =0 3
(16)
Thus, we obtain
ts =
3tc + p1 − dq3 + p2 − dq2 + p3 − dq1 3
(17)
In other words, the customer departs at ts, and the requested action figure is ready when the customer reaches the convenience store at node q. In other words, the customer reaches the store JIT. Solving the aforementioned steps is difficult. To facilitate determining the solution for an online application, Chen’s modified fuzzy Dijkstra algorithm [6] is described in Section 4.3.
4.3. Fuzzy Dijkstra algorithm for solving the static UbiM case 4.2. JIT production The processing time is represented using p˜ = (p1, p2 , p3 ), and
Chen's fuzzy Dijkstra algorithm was modified for determining the most suitable convenience store.
Please cite this article as: Y.-C. Lin, T. Chen, A ubiquitous manufacturing network system, Robotics and Computer Integrated Manufacturing (2016), http://dx.doi.org/10.1016/j.rcim.2015.10.009i
Y.-C. Lin, T. Chen / Robotics and Computer-Integrated Manufacturing ∎ (∎∎∎∎) ∎∎∎–∎∎∎
6
14 B
1 2
18 N
D
3
4
M
8
G 9
E
C
10
5
H
K
13
15
19
16
20
David Wang nnnnnnn 565 2015/05/23 02:35:22 PM (24.139535°, 120.660353°) ANI-SES-OP-001 Monkey D. Luffy 10cm 83 USD Paid XXXX-XXXX-XXXX-5780
˜ k = max {0, (wk1, wk2, wk3 )} w = max {0, (tc ( + ) p˜ )( − )(ts ( + ) d˜1k ( + ) d˜kn )} = max {0, (tc + p1 − ts − d1k3 − dkn3, tc + p2 − ts − d1k2
6 F
− dkn2, tc + p3 − ts − d1k1 − dkn1)} 11 J
L
21 O
7
Algorithm 1. (i) Set d˜1 = 0, and d˜ i = ∞ ∀ i ≠ 1. Mark node 1 as “visited” and the successors of node 1 as “unvisited.” Set i¼2. (ii) Evaluate the suitability of node i as
si = 1 − (di1 + di2 + di3 )/(p1 + p2 + p3 )
(18)
According to the fuzzy Dijkstra algorithm proposed by Deng et al.,
si = 1 − (di1 + 4di2 + di3 )/(p1 + 4p2 + p3 )
(21)
The modified fuzzy Dijkstra algorithm for the dynamic UbiM case is as follows: Algorithm 2.
12 Fig. 6. Abstracted road map.
(19)
(iii) If there are no “unvisited” convenience stores in the remaining path of the current node, proceed to Step (iv); otherwise, consider all of the current node’s successors. For each successor, calculate the distance and evaluate the suitability. Update the distance and suitability if the suitability increases. (iv) Mark the current node as “visited.” (v) If all of the convenience stores have been visited or if the suitability levels of all of “unvisited” nodes are lower than those of some convenience stores, proceed to Step (vi); otherwise, move to the “unvisited” node with the highest suitability and return to Step (ii). (vi) The convenience store with the highest suitability is the most suitable for printing the action figure. Use Eq. (17) to determine the start time, facilitating JIT production for the customer. (vii) Stop. 4.4. Dynamic UbiM case Assume a customer begins from nodes 1 to n at time ts . This route can be divided into two sections: the shortest path from node 1 to each convenience store and the shortest path from a convenience store to node n.
Min d˜1k ( + ) d˜kn
Customer name Cell phone number Date & time Location (latitude, longitude) Chosen action figure no. Chosen action figure name Price Payment status Credit card no.
estimated as follows: 17
I
A
Table 2 Request content.
(20)
where d˜ ij represents the shortest distance from nodes i to j. The waiting time at the convenience store located at node k is
i. Apply Algorithm 1 for determining d˜1k for each k ∈ Q. ii. For each k ∈ Q, set the start location to node k. Apply Algorithm 1 to determine d˜kn . iii. For each k ∈ Q, evaluate the suitability of the convenience store at node k as
sk = 1 − (d1k1 + dkn1 + d1k2 + dkn2 + d1k3 + dkn3 )/(p1 + p2 + p3 )
(22)
According to the fuzzy Dijkstra algorithm proposed by Deng et al.,
sk = 1 − (d1k1 + dkn1 + 4d1k2 + 4dkn2 + d1k3 + dkn3 )/(p1 + 4p2 + p3 )
(23)
iv. The convenience store with the highest suitability is the most suitable for printing the action figure. Use Eq. (17) to determine the start time, facilitating JIT production for the customer. v. Stop.
5. Illustrative cases In this section, two cases are presented for illustrating the applicability of the proposed methodology. The experimental region was a 4-km2 region in the Nantun District of Taichung City in Taiwan (Fig. 5). An abstracted road map (Fig. 6) shows 15 stores of the largest convenience store chain in Taiwan for printing 3D action figures in the UbiM network system. A customer sent a request to the system server. The request content is shown in Table 2. The latitude and longitude of the customer's location was detected by the GPS system in the customer’s smartphone; the customer was near the corner of Wuquan Road and Wuquan West 5th Street. According to multiple GPS detections, the customer's speed was estimated as 24.5, 25, and 25.5 km/h. Accordingly, the time required for printing the requested action figure was estimated to be 2.5, 2.75, and 3 h. Algorithm 1 was then applied to identify the convenience store most suitable for the customer. The process is summarized in Table 3. The convenience store most suitable for the customer was L or O, each of which achieved the highest satisfaction level (0.990) and the shortest distance (0.7 km).
Please cite this article as: Y.-C. Lin, T. Chen, A ubiquitous manufacturing network system, Robotics and Computer Integrated Manufacturing (2016), http://dx.doi.org/10.1016/j.rcim.2015.10.009i
Y.-C. Lin, T. Chen / Robotics and Computer-Integrated Manufacturing ∎ (∎∎∎∎) ∎∎∎–∎∎∎
7
Table 3 Application process of Algorithm 1. (Round 1) Node
Start location
20
21
L
17
d˜i (km) si (h) Visited
0.00
0.21
0.58
0.70
0.37
1.000 Visited
0.997 Unvisited
0.992 Unvisited
0.990 Visited
0.995 Unvisited
(Round 2) Node
Start location
d˜i (km) si (h) Visited
0.00
0.21
0.58
0.70
0.37
0.95
1.000 Visited
0.997 Visited
0.992 Unvisited
0.990 Visited
0.995 Unvisited
0.986 Unvisited
20
21
L
17
19
(Round 3) Node
Start location
20
21
d˜i (km) si (h) Visited
0.00
0.21
0.58
0.70
0.37
0.95
1.11
0.80
1.000 Visited
0.997 Visited
0.992 Unvisited
0.990 Visited
0.995 Visited
0.986 Unvisited
0.984 Unvisited
0.988 Visited
(Round 4) Node Start location
d˜i (km) si (h) Visited
20
21
L
L
17
17
19
19
15
15
H
H
O
0.00
0.21
0.58
0.70
0.37
0.95
1.11
0.80
0.70
1.000 Visited
0.997 Visited
0.992 Visited
0.990 Visited
0.995 Visited
0.986 Unvisited
0.984 Unvisited
0.988 Visited
0.990 Visited
Fig. 7. Route planned using Algorithm 1.
Convenience store L was considered as an example; the route from the starting point is shown in Fig. 7. To achieve JIT production, Eq. (17) was calculated, and the customer was recommended to depart at 05:18:41 PM.
In this case, Google Maps was used for comparison purposes. The recommendation results of Google Maps are shown in Fig. 8: (1) According to the recommendation results, approximately
Please cite this article as: Y.-C. Lin, T. Chen, A ubiquitous manufacturing network system, Robotics and Computer Integrated Manufacturing (2016), http://dx.doi.org/10.1016/j.rcim.2015.10.009i
8
Y.-C. Lin, T. Chen / Robotics and Computer-Integrated Manufacturing ∎ (∎∎∎∎) ∎∎∎–∎∎∎
Experimental Region Fig. 8. Recommendation results obtained using Google Maps.
9 min were required by the customer to reach the nearest convenience store. The distance was 2.8 km, and the satisfaction level was 0.959, which was considerably lower than that achieved using the proposed methodology. (2) Unexpectedly, Google Maps led the customer to a convenience store that was outside of the experimental region. (3) Google Maps specified five transportation modes: on foot and by car, bus, bicycle, or airplane. However, the customer's vehicle speed was not considered. (4) Similar to the proposed methodology, Google Maps proposed several routes, which were modifiable, for reaching the same convenience store. The customer planned to arrive at the convenience store at
05:00 PM and then travel to a destination (24.148755° latitude and 120.657154° longitude) located on Zhongming South Road. For each convenience store, Algorithm 2 was used to determine the shortest paths from the starting location (denoted as “St”) to the convenience store (i.e., the 1st shortest path) and from the convenience store to the destination (denoted as “De”) (i.e., the 2 nd shortest path). The results are summarized in Table 4. The total duration was minimized with convenience store K as the midpoint (Fig. 9). The satisfaction level was 0.980. In other words, the customer initially travelled to convenience store K to collect the action figure and then continued toward his destination. The time at which the customer arrived at convenience store K was estimated as follows: 05:00 PM þ1.24/(24.5, 25, 25.5) h¼(05:02:55, 05:02:59,
Please cite this article as: Y.-C. Lin, T. Chen, A ubiquitous manufacturing network system, Robotics and Computer Integrated Manufacturing (2016), http://dx.doi.org/10.1016/j.rcim.2015.10.009i
Y.-C. Lin, T. Chen / Robotics and Computer-Integrated Manufacturing ∎ (∎∎∎∎) ∎∎∎–∎∎∎
9
Table 4 Shortest paths to and from each convenience store. Convenient store
1st shortest path
Path Duration
2nd shortest path
Path duration
Total duration
A B C D E F G H I J K L M N O
St-L-11-J-12-7-A St-17-H-10-C-4-3-2-1-B St-17-H-10-C St-17-16-13-9-G-3-2-D St-17-16-13-9-E St-L-11-F St-17-16-13-9-G St-17-H St-17-H-I St-L-11-J St-17-16-15-K St-L St-20-19-N-18-M St-20-19-N St-21-O
2.34 2.58 1.71 2.30 1.84 1.42 1.48 0.76 0.92 1.19 1.24 0.70 1.48 1.13 0.70
A-7-12-J-11-L-17-16-15-De B-1-8-G-K-15-De C-10-9-13-16-15-De D-2-3-G-K-15-De E-9-13-16-15-De F-11-L-17-16-15-De G-K-15-De H-17-16-15-De I-H-17-16-15-De J-11-L-17-16-15-De K-15-De L-17-16-15-De M-14-15-De N-19-15-De O-21- L-17-16-15-De
2.93 1.44 1.44 1.24 1.13 2 0.41 1.18 1.34 1.77 0.16 1.28 0.60 0.60 1.88
5.37 4.02 3.15 3.54 2.97 3.42 1.89 1.94 2.26 2.96 1.40 1.98 2.08 1.73 2.58
Fig. 9. Route planned using the proposed methodology.
05:03:02) PM The completion time of the action figure was estimated as follows: 02:35:22 PM þ(2.5, 2.75, 3) h¼ (05:05:22, 05:20:22, 05:35:22) PM Consequently, the customer had to wait (05:05:22, 05:20:22, 05:35:22) (–) (05:02:55, 05:02:59, 05:03:02) ¼ (2.33, 17.38, 32.45) min. To achieve JIT production and avoid a waiting time, the customer was recommended to depart at 02:52:45 PM, that is, 17.38 min after placing the order. The route planned using Google Maps is shown in Fig. 10; the
route length was 5.3 km, considerably longer than that obtained using the proposed methodology. The satisfaction level was 0.923. The customer was again led to a convenience store outside the experimental region; it was estimated that the customer would reach the convenience store at 05:06 PM and the destination at 05:09 PM. The waiting time of the customer was estimated as ( 0.63, 14.37, 29.37) min.
6. Conclusions 3D
printing
is
introducing
revolutionary
changes
to
Please cite this article as: Y.-C. Lin, T. Chen, A ubiquitous manufacturing network system, Robotics and Computer Integrated Manufacturing (2016), http://dx.doi.org/10.1016/j.rcim.2015.10.009i
10
Y.-C. Lin, T. Chen / Robotics and Computer-Integrated Manufacturing ∎ (∎∎∎∎) ∎∎∎–∎∎∎
Fig. 10. Route planned using Google Maps.
manufacturing. The prevalence of 3D printers connected by the Internet also form a UbiM environment in which customers can easily access a nearby 3D printing facility, a phenomenon creating considerable business opportunities. Although some websites have collaborated online to provide 3D printing services jointly, few 3D printing facilities have worked together to serve customers from different regions, which clearly requires the formation of a UbiM network. To this end, a UbiM network system was constructed in this study. In UbiM network systems, a customer places an order for an action figure by using a client-side app or Webbased interface and pays online. The system server then assigns the order to the convenience store nearest to the customer’s location to print the required action figure. This study developed an FINLP model to determine the most suitable convenience store. Two modified fuzzy Dijkstra algorithms were employed to solve the FINLP problems for static and dynamic UbiM cases. Subsequently, the customer is informed of the location of and route to the recommended convenience store. Compared with conventional manufacturing and logistics services, which are difficult to access directly or time consuming, the intensive distribution of convenience stores and the quick manufacturing capability of 3D printers enable a customer to order and receive an action figure from any location almost instantaneously. In addition, the proposed methodology is an innovative attempt that transforms a service network into a manufacturing network. The
established UbiM network system involves applying ubiquitous computing, mobile commerce, 3D printing, and location-based services. In two cases for illustrating static and dynamic UbiM, the proposed methodology showed high applicability and considerable potential for providing satisfactory recommendation results to customers who had not yet departed for or were traveling to the 3D output location. An existing mobile guide, Google Maps, was used for comparison. Compared with Google Maps, the proposed methodology is advantageous because it recommends the shortest path for obtaining the required action figure and thereby increases the customer satisfaction level. Introducing such a UbiM service in convenience stores would bring profitable business opportunities to convenience store chains. In the future, other UbiM services and networks could be designed and established. In addition, the UbiM network system constructed in this study is homogeneous; in other words, it is composed of 3D printing facilities that possess equal capability. In addition, a heterogeneous UbiM network system could be constructed using a similar method. References [1] 3d-printing.net, Asda is launching a 3D printing service – tomorrow! 〈http://
Please cite this article as: Y.-C. Lin, T. Chen, A ubiquitous manufacturing network system, Robotics and Computer Integrated Manufacturing (2016), http://dx.doi.org/10.1016/j.rcim.2015.10.009i
Y.-C. Lin, T. Chen / Robotics and Computer-Integrated Manufacturing ∎ (∎∎∎∎) ∎∎∎–∎∎∎ www.ettoday.net/news/20131021/284587.htm〉, 2015. [2] M. Asadi-Eydivand, M. Solati-Hashjin, A. Farzad, N.A.A. Osman, Effect of technical parameters on porous structure and strength of 3D printed calcium sulfate prototypes, Robot. Comput. Manuf. 37 (2016) 57–67. [3] Autodesk, Spark. 〈https://spark.autodesk.com/〉, 2015. [4] Baidu. 〈http://wenku.baidu.com/view/ac8334707fd5360cba1adb21.html〉, 2013. [5] B. Berman, 3-D printing: the new industrial revolution, Bus. Horiz. 55 (2) (2012) 155–162. [6] T. Chen, Creating a just-in-time location-aware service using fuzzy logic, Appl. Spat. Anal. policy (2015) 1-21. [7] T. Chen, Y.-C. Wang, Z. Lin, Predictive distant operation and virtual control of computer numerical control machines, J. Intell. Manuf. (2015) 1-17. [8] T. Chen, H.-C. Wu, Finding the just-in-time service location and path in a ubiquitous service network, Int. J. Internet Manuf. Serv. 3 (2) (2013) 137–147. [9] B.V. Cherkassky, A.V. Goldberg, T. Radzik, Shortest paths algorithms: theory and experimental evaluation, in: Proceedings of the ACM–SIAM Symposium on Discrete Algorithms, 1994, pp. 516–525. [10] Chih-Mao Information. 〈http://www.cmic.com.tw/Production_process.aspx〉, 2015. [11] Y. Deng, Y. Chen, Y. Zhang, S. Mahadevan, Fuzzy Dijkstra algorithm for shortest path problem under uncertain environment, Appl. Soft Comput. 12 (2012) 1231–1237. [12] E.W. Dijkstra, A note on two problems in connexion with graphs, numer. Math. 1 (1959) 269–271. [13] D. Ding, Z. Pan, D. Cuiuri, H. Li, A practical path planning methodology for wire and arc additive manufacturing of thin-walled structures, Robot. Comput. Manuf. 34 (2015) 8–19. [14] i.materialise, Factory 2.0 for designers. 〈https://i.materialise.com/factory-2-0for-designers〉, 2015. [15] Y. Fan, D. Zhao, L. Zhang, S. Huang, B. Liu, Manufacturing grid: needs, concept and architecture, Lect. Notes Comput. Sci. 3032 (2004) 653–656.
11
[16] L. Goldman, R.L. Nagel, K. Preiss, Agile Competitors and Virtual Organizations – Strategies for Enriching the Customer, Van Nostrand Reinhold, New York, NY, 1995. [17] D.F. Kehoe, N.J. Boughton, New paradigms in planning and control across manufacturing supply chains – the utilisation of Internet technologies, Int. J. Oper. Prod. Manag. 21 (5/6) (2001) 582–593. [18] S. Keating, N. Oxman, Compound fabrication: a multi-functional robotic platform for digital design and fabrication, Robot. Comput. Manuf. 29 (6) (2013) 439–448. [19] J.P. Kruth, M.C. Leu, T. Nakagawa, Progress in additive manufacturing and rapid prototyping, CIRP Ann. Technol. 47 (2) (1998) 525–540. [20] C.X.F. Lam, X.M. Mo, S.H. Teoh, D.W. Hutmacher, Scaffold development using 3D printing with a starch-based polymer, Mater. Sci. Eng. C. 20 (1) (2002) 49–56. [21] E. Lin, I. Minis, D.S. Nau, W.C. Regli, Virtual manufacturing, Lawrence Associates, Inc. 〈http://www.isr.umd.edu/Labs/CIM/virtual.html〉, 1997. [22] My Mini Factory. 〈https://www.myminifactory.com/〉, 2015. [23] F. Rengier, A. Mehndiratta, H. von Tengg-Kobligk, C.M. Zechmann, R. Unterhinninghofen, H.U. Kauczor, F.L. Giesel, 3D printing based on imaging data: review of medical applications, Int. J. Comput. Assist. Radiol. Surg. 5 (4) (2010) 335–341. [24] ShenzhenXin Ju Xin Toy Design Co., 3d Printing Anime Action Figure. 〈http:// szjxsy.en.alibaba.com/product/60239779940-212431068/3d_printing_anime_ action_figure.html#!〉, 2015. [25] D.N. Silva, M.G. De Oliveira, E. Meurer, M.I. Meurer, J.V.L. da Silva, A. SantaBárbara, Dimensional error in selective laser sintering and 3D-printing of models for craniomaxillary anatomy reconstruction, J. Cranio-Maxillofac. Surg. 36 (8) (2008) 443–449. [26] D. Wu, M.J. Greer, D.W. Rosen, D. Schaefer, Cloud manufacturing: strategic vision and state-of-the-art, J. Manuf. Syst. 32 (4) (2013) 564–579. [27] H.J. Zimmermann, Fuzzy Set Theory and its Applications, Kluwer Academic Publishers, Dordrecht, 1991.
Please cite this article as: Y.-C. Lin, T. Chen, A ubiquitous manufacturing network system, Robotics and Computer Integrated Manufacturing (2016), http://dx.doi.org/10.1016/j.rcim.2015.10.009i
Contents lists available at ScienceDirect
Robotics and Computer-Integrated Manufacturing journal homepage: www.elsevier.com/locate/rcim
A ubiquitous manufacturing network system Yu-Cheng Lin a,n, Toly Chen b a b
Department of Computer-aided Industrial Design Overseas Chinese University, Taiwan Department of Industrial Engineering and Systems Management Feng Chia University, Taiwan
art ic l e i nf o
a b s t r a c t
Article history: Received 3 July 2015 Received in revised form 24 September 2015 Accepted 11 October 2015
In this study, a ubiquitous manufacturing network system was constructed. In this system, a customer places an order for an action figure by using a client-side app or a Web-based interface and pays online. The system server then assigns the order to the convenience store nearest the customer's location to print the required action figure. For determining the most suitable convenience store, a fuzzy integernonlinear programming model was proposed and solved using two modified fuzzy Dijkstra algorithms. Subsequently, the customer is informed of the location of and route to the recommended convenience store. Two illustrative cases were used to verify the applicability of using the proposed methodology. In addition, compared with an existing mobile guide, the proposed methodology effectively recommended the shortest path for obtaining the required action figure and reduced the waiting time at the convenience store. & 2015 Elsevier Ltd. All rights reserved.
Keywords: Ubiquitous manufacturing Location-aware service 3D printing Action figure
1. Introduction Ubiquitous computing is a concept in software engineering and computer science in which computing is performed at any location. As an application of ubiquitous computing in the manufacturing sector, ubiquitous manufacturing (UbiM) features an environment in which manufacturing services are provided at any location. A concept similar to UbiM is cloud manufacturing (CM; [26]). Both UbiM and CM enable ubiquitous, convenient, on-demand network access to a shared pool of configurable manufacturing resources (e.g., software tools, equipment, and manufacturing capabilities). However, in contrast with CM, UbiMemphasizes the mobility and dispersion of manufacturing resources and users. In general, manufacturing a product ubiquitously is impossible. Previously, UbiM implied that products could be supplied ubiquitously. Some advanced manufacturing technologies, such as lean manufacturing [4], CM, manufacturing grids[15], global manufacturing [15], virtual manufacturing [21,7], agile manufacturing [16], Internet manufacturing [17], and particularly additive manufacturing (AdM; [19]) have increased UbiM opportunities. An AdM network can be used to transfer the three-dimensional (3D) model of a product (in .STL or .OBJ file format) to a 3D printer at any locationand any time over the Internet, at which point the product isbuilt layerbylayer. At the end of the manufacturing process, excess soft resin is cleaned using a chemical n
Corresponding author.
bath [5]. Thus, manufacturing a product ubiquitously has become possible. Developments in AdM have enabled printing extremely complex 3D models. Various methodologies, such asthat developed by Ding et al. [13], have been proposed for generating deposition paths and arcs for such products. Keating and Oxman [18] reconfigured a 6-axis robotic arm into an integrated platform for 3D printing, milling, and sculpting and proposed the concept of “compound fabrication.” This study established a UbiM network system, in which a customer requests that a product be manufactured at a 3D output locationand arrives atthe 3D output location for the just in time (JIT; [8]) manufacturing of the requested product. UbiM network systems are used to manufacture action figures of popular cartoon or movie characters. The business feasibility of using a 3D printer to manufacture action figures was mentioned by Berman [5] and has been successfully demonstrated [10]. Information companies, 3D photo studios, supermarket chains, and other shops provide (Table 1) limited services for the printing of 3D action figures. In addition, by mass-producing action figures, 3D printing service providers such as the Shenzhen Xin Ju Xin Toy Design Company [24] have contradicted the conventional belief that 3D printers can be used only for small-scale trial production. However, most of the existing service providers have a single 3D output location;a systematic UbiM network remains undeveloped. As shown in Table 1, some companies have substantial waiting periods for receiving desired action figures. Such waiting is worthwhile if 3D printing enables customers to receive an action figure that they want as soon as possible, because even the largest
http://dx.doi.org/10.1016/j.rcim.2015.10.009 0736-5845/& 2015 Elsevier Ltd. All rights reserved.
Please cite this article as: Y.-C. Lin, T. Chen, A ubiquitous manufacturing network system, Robotics and Computer Integrated Manufacturing (2016), http://dx.doi.org/10.1016/j.rcim.2015.10.009i
Y.-C. Lin, T. Chen / Robotics and Computer-Integrated Manufacturing ∎ (∎∎∎∎) ∎∎∎–∎∎∎
2
Table 1 Examples of 3D printing service providers. Service provider
Region
Service items
Size of action figure
Price (USD)
Required time
Reference
Asda [1]
UK
Scan & print
Unknown
45
Chih-Mao information
Taiwan Print; Scan & print Taiwan Scan & print China Print
Height: 10 cm
32–39 /h
A week afterward 4.83 h
http://www.3D-printing.net/content/asda-launching-3D-printingservice-tomorrow http://www.cmic.com.tw/Production_process.aspx
13–15 cm Unknown
415 2–20
1–2 weeks 8–15 days
http://www.ego3d.com.tw/price.html http://szjxsy.en.alibaba.com/product/60239779940-212431068/ 3d_printing_anime_action_figure.html#!
EGO3D Shenzhen Xin Ju Xin Toy Design Co.
toy makers do not stock all types of action figure. In addition, transporting an action figure from an overseas toy maker to a customer's location may involvea longer waiting time. After the ordered action figure is printed, the customer visits the 3D output location and collects the action figure. Presently, the 3D output location may be located a considerable distance from the customer's location. Therefore, for overcoming this problem, this study proposed a UbiM network system. The customer can be guided to the nearest 3D output location to reduce travel time, a procedure called static UbiM. In another situation [6], dynamic UbiM was used for further elaborating the effectiveness of UbiM network systems. In this study's procedure, a customer receives directions for obtaining a pre ordered action figure from a preferred 3D output location while traveling according to the following objectives: (1) The customer should promptly leave for the 3D output location. (2) The 3D output location should be located near the customer's route. (3) When the customer arrives at the 3D output location, the requested action figure should be ready. In other words, the customer's waiting time at the 3D output location should be minimal. Fig. 1. Distribution of convenience stores in a 2-km2 region.
For achieving these objectives, the proposed methodology designated a chain of convenience stores to be the 3D output locations; the stores are suitable for the following reasons:
2. 3D printing 2.1. Platforms
(1) Some regions contain a high concentration of convenience stores that provide nearly identical services. For instance, a 2-km2 region in the Zhongzheng District of Taipei City in Taiwan (Fig. 1) contains 36 convenience stores of the same chain. (2) Most convenience stores provide printing services, which could serve as a basis for 3D printing services. (3) Convenience store chains in Taiwan regularly introduce promotional campaigns that allow customers to exchange points, accumulated after a fixed amount of purchases, for an action figure. The remainder of the paper is organized as follows: Section 2 presents a literature review of 3D printing. Section 3 describes the architecture and process flow of UbiM network systems. Section 4 presents a fuzzy integer-nonlinear programming (FINLP) model for solving the JIT output location and path problems. For solving the FINLP problems for static UbiM and dynamic UbiM cases, two modified fuzzy Dijkstra algorithms were proposed. In addition, two cases were used to illustrate the applicability of the proposed methodology. Section 5 presents a comparison of the performance of the proposed methodology with that of an existing method. Finally, Section 6 concludes the paper.
3D printing can be seamlessly integrated with CAD/CAM operations [5]. Therefore, the existing CAD/CAM software vendors are typically the pioneers in this field. For example, Autodesk recently developed an open 3D printing platform named Spark [3] on which application program interfaces (APIs) are provided for each stage of 3D printing. Consequently, 3D printing functionality can be quickly added to an application. By using Spark, a customer can design and even optimize a 3D model that can be printed by a specific 3D printer. There are also websites acting as hubs for gathering 3D models from volunteers worldwide. For example, My Mini Factory test prints an uploaded 3D model before it is publicized [22]. Through its online catalog, shapeways shows the 3D models it has collected for printing and purchasing, thus providing the company with a 3.5% commission from the designer of each 3D model printed or sold. Equipped with more than 100 3D printers of various sizes, i. materialise can be used to print 3D models uploaded by customers or passed from websites that provide 3D printing services but do not have printers [14]. APIs are provided for connecting a service website to i.materialise, in addition to third-party logistics that directly ship the printed 3D model to the service website's customer.
Please cite this article as: Y.-C. Lin, T. Chen, A ubiquitous manufacturing network system, Robotics and Computer Integrated Manufacturing (2016), http://dx.doi.org/10.1016/j.rcim.2015.10.009i
Y.-C. Lin, T. Chen / Robotics and Computer-Integrated Manufacturing ∎ (∎∎∎∎) ∎∎∎–∎∎∎
2.2. Applications
3
Customer browses the online catalog
3D printing has been used to produce prototypes and mockups, replacement parts, dental crowns, artificial limbs, and even bridges [5]. 3D printing has considerable potential for producing complex internal and external porous structures [2]. For example, Lam et al. [20] used a blend of starch-based polymer powders (cornstarch, dextran, and gelatin) for making 3D porous scaffolds with a 3D printer. Asadi-Eydivand et al. [2] optimized the parameters of 3D printing by using design-of-experiment techniques when making calcium-sulfate-based scaffolding prototypes. Building 3D models from the substantial number of images in existing databases remains the primary endeavor of this field. For example, Silva et al. [25] manipulated the tomographic images of a dry skull, thus enabling the craniofacial skeleton to be accurately reproduced using 3D printing. Rengier et al. [23] produced 3D objects for surgical planning according to volumetric medical images produced by computed tomography (CT) and magnetic resonance imaging (MRI). Websites like shapeway and i.materialise manage substantial online databases of 3D models uploaded by users worldwide. However, whether users are the copyright holders of the 3D models that they upload and whether downloading infringes on the rights of copyright holders are two problems that warrant further examination.
Customer chooses an action figure System server estimates the customer’s location and speed System server determines the convenient store
System server responds to the customer
Customer makes the payment online
System server assigns the task to the convenient store
Customer goes to the convenient store
The convenient store prints the action figure
3. System architecture and process flow A four-tier architecture was configured for the UbiM network system (Fig. 2). The UbiM network system comprises four major components:
Customer takes the action figure Fig. 3. Process flow of the UbiM network system.
(1) The customer or client browses the online product catalog by using a client-side app or system website, selects a desired action figure, makes an online payment, and receives instructions regarding the pick-up location of the action figure. (2) The system server prepares the online product catalog, receives and processes the customer's request, confirms receiving the online payment, and assigns the task to a convenience store. (3) A bank or an online payment service provides interfaces with the customer by displaying information relayed from the system server regarding online payment.
(4) The convenience store prints the required action figure according to the instructions from the system server and serves the customer upon their arrival. The UbiM network system exhibits the following process flow (Fig. 3): (1) A customer browses the online product catalog and chooses the desired action figure. (2) By using data from the global positioning system (GPS) system
Supply Locations
Tier 4 action figure
customer arrival data time
Tier 3
chosen action figure
System Server possible action customer location chosen figures data speed action figure
Tier 2
output details
Bank
Communication Service Provider possible action customer location chosen figures data speed action figure
payment
payment output details
Customers
Tier 1 use browser
use app
Fig. 2. Four-tier architecture of the proposed UbiM network system.
Please cite this article as: Y.-C. Lin, T. Chen, A ubiquitous manufacturing network system, Robotics and Computer Integrated Manufacturing (2016), http://dx.doi.org/10.1016/j.rcim.2015.10.009i
Y.-C. Lin, T. Chen / Robotics and Computer-Integrated Manufacturing ∎ (∎∎∎∎) ∎∎∎–∎∎∎
4
Fig. 4. Using a 3D printer to print an action figure.
(3) (4) (5) (6)
(7)
(8)
in the customer's smartphone, the system server estimates the location and speed of the customer. The system server determines the convenience store nearest the customer. The customer makes an online payment. The system server assigns the task to a convenience store. The system server provides the customer with details on the recommended convenience store and the time and route to reach the store. (Concurrent operations) The customer leaves for the recommended convenience store as the convenience store prints the desired action figure. The customer reaches the convenience store and collects the action figure.
Using a 3D printer to print an action figure involves five steps (Fig. 4): proofreading the 3D model file, printing, unearthing the model, removing the gypsum powder, and postproduction hardening. In the proposed UbiM network system, the 3D model files of all of the action figures in an online catalog were prepared; therefore, the first step may be unnecessary. In addition, certain steps require human intervention. Therefore, the processing times of such steps are variable. However, developments in AdM technology are expected to resolve this problem. The manufacturing cost for the 10-cm-high action figure depicted in Fig. 4 is approximately US$65–81, and the processing time is approximately 4.83 h.
location). The node farthest from the customer is node n. The length of the path connecting nodes i and j is lij , where i and j¼1 n; i≠j; and lij = ∞ if no connection exists between the two nodes. The shortest distance from the customer's location to node i is denoted as di . Obviously, d1 = 0, dn = max di , and di can be dei
rived according to the shortest distances of the nodes connected to di :
di = min (dj + l ji ) j
(1)
Assume a convenience store is located at node k. The problem of determining the shortest path to the convenience store can be formulated as an integer-nonlinear programming (INLP) problem as follows:
(2)
Min dk s.t.
di =
min (dj + l ji ), i = 1~k
j < i, l ji ≠∞
(3)
Eq. (3) can be replaced by
di ≤ dj + l ji , i = 1~k; j < i; l ji ≠ ∞
∑
di =
xji (dj + l ji ), i = 1~k
j < i, l ji ≠∞
∑
(4)
(5)
xji = 1, i = 1~k
j < i, l ji ≠∞
(6)
xji ∈ { 0, 1}, i = 1~k; j < i; lij ≠ ∞
(7)
4. Determining the most suitable convenience store 4.1. Static UbiM case The “most suitable” convenience store for printing an action figure is the easiest to reach, thus minimizing the waiting time. Two steps are required for determining the most suitable convenience store. First, the shortest path to each convenience store is determined. A shortest path problem can be solved in polynomial time only in graphs without negative cycles (Cormen et al., 2001). Subsequently, the waiting time for each convenience store is estimated. The nearest convenience store that minimizes the waiting time is recommended. The service region is abstracted into a traffic network, in which only the major roads and corners are considered to accelerate the reasoning process. Assume that there are n nodes in the abstracted traffic network. A customer is located at node 1 (i.e., the starting
However, determining the global optimal solution to this INLP problem is challenging. For managing this difficulty, many algorithms such as the Dijkstra, Bellman–Ford, A* search, Floyd–Warshall, and Johnson algorithms have been proposed [12,9]. Google Maps, which is relatively easy to use with online applications but which can provide inaccurate results, was used for identifying the shortest path. However, considering individual differences in the results obtained using Google Maps is inconvenient. Regarding the uncertainty of a customer's speed, the shortest distance to node i can be modeled using a triangular fuzzy number d˜ i = (di1, di2, di3 ). For example, if the speed of a customer is 30, 45, or 50 km/h, then the times required for the customer to traverse a 1-km path are 0.02, 0.022, and 0.033 h, respectively. After such uncertainty is considered, the following FINLP
Please cite this article as: Y.-C. Lin, T. Chen, A ubiquitous manufacturing network system, Robotics and Computer Integrated Manufacturing (2016), http://dx.doi.org/10.1016/j.rcim.2015.10.009i
Y.-C. Lin, T. Chen / Robotics and Computer-Integrated Manufacturing ∎ (∎∎∎∎) ∎∎∎–∎∎∎
5
Fig. 5. Experimental region.
problem is formulated:
Min d˜k
(8)
s.t. ~ ~ di ≤ dj ( + ) l ji , i = 1~k; j < i; lij ≠ ∞
~ di =
∑
˜ q = max {0, (wq1, wq2, wq3 )} w = max {0, (tc ( + ) p˜ )( − )(ts ( + ) d˜ q )}
(9)
~ xji (dj ( + ) l ji ), i = 1~k
j < i, l ji ≠∞
the waiting time at the convenience store located at node q is estimated as follows:
(10)
= max {0, (tc + p1 − ts − dq3, tc + p2 − ts − dq2, tc + p3 − ts − dq1)}
where tc is the current time, ts is the time when the customer departs for the recommended convenience store, and ( − ) in˜ q , Eq. (15) is used: dicates a fuzzy subtraction. For minimizing w
(tc + p1 − ts − dq3, tc + p2 − ts − dq2, tc + p3 − ts − dq1) = 0
∑
xji ∈ {0, 1} , i = 1~k; j < i; lij ≠ ∞
(11)
(12)
where (þ ) indicates a fuzzy addition. Instead of directly solving the FINLP problem, the problem is solved using several approximation algorithms including the fuzzy Dijkstra algorithm proposed by Deng et al. [11] and the fuzzy Dijkstra algorithm developed by Chen [6]. Let
d˜ q = min d˜k k∈Q
(15)
Applying the center-of-gravity method [27] to Eq. (15) yields
xji = 1, i = 1~k
j < i, l ji ≠∞
(14)
(13)
where Q is the index set of convenience stores. Accordingly, the convenience store located at node q is recommended.
tc + p1 − ts − dq3 + tc + p2 − ts − dq2 + tc + p3 − ts − dq1 =0 3
(16)
Thus, we obtain
ts =
3tc + p1 − dq3 + p2 − dq2 + p3 − dq1 3
(17)
In other words, the customer departs at ts, and the requested action figure is ready when the customer reaches the convenience store at node q. In other words, the customer reaches the store JIT. Solving the aforementioned steps is difficult. To facilitate determining the solution for an online application, Chen’s modified fuzzy Dijkstra algorithm [6] is described in Section 4.3.
4.3. Fuzzy Dijkstra algorithm for solving the static UbiM case 4.2. JIT production The processing time is represented using p˜ = (p1, p2 , p3 ), and
Chen's fuzzy Dijkstra algorithm was modified for determining the most suitable convenience store.
Please cite this article as: Y.-C. Lin, T. Chen, A ubiquitous manufacturing network system, Robotics and Computer Integrated Manufacturing (2016), http://dx.doi.org/10.1016/j.rcim.2015.10.009i
Y.-C. Lin, T. Chen / Robotics and Computer-Integrated Manufacturing ∎ (∎∎∎∎) ∎∎∎–∎∎∎
6
14 B
1 2
18 N
D
3
4
M
8
G 9
E
C
10
5
H
K
13
15
19
16
20
David Wang nnnnnnn 565 2015/05/23 02:35:22 PM (24.139535°, 120.660353°) ANI-SES-OP-001 Monkey D. Luffy 10cm 83 USD Paid XXXX-XXXX-XXXX-5780
˜ k = max {0, (wk1, wk2, wk3 )} w = max {0, (tc ( + ) p˜ )( − )(ts ( + ) d˜1k ( + ) d˜kn )} = max {0, (tc + p1 − ts − d1k3 − dkn3, tc + p2 − ts − d1k2
6 F
− dkn2, tc + p3 − ts − d1k1 − dkn1)} 11 J
L
21 O
7
Algorithm 1. (i) Set d˜1 = 0, and d˜ i = ∞ ∀ i ≠ 1. Mark node 1 as “visited” and the successors of node 1 as “unvisited.” Set i¼2. (ii) Evaluate the suitability of node i as
si = 1 − (di1 + di2 + di3 )/(p1 + p2 + p3 )
(18)
According to the fuzzy Dijkstra algorithm proposed by Deng et al.,
si = 1 − (di1 + 4di2 + di3 )/(p1 + 4p2 + p3 )
(21)
The modified fuzzy Dijkstra algorithm for the dynamic UbiM case is as follows: Algorithm 2.
12 Fig. 6. Abstracted road map.
(19)
(iii) If there are no “unvisited” convenience stores in the remaining path of the current node, proceed to Step (iv); otherwise, consider all of the current node’s successors. For each successor, calculate the distance and evaluate the suitability. Update the distance and suitability if the suitability increases. (iv) Mark the current node as “visited.” (v) If all of the convenience stores have been visited or if the suitability levels of all of “unvisited” nodes are lower than those of some convenience stores, proceed to Step (vi); otherwise, move to the “unvisited” node with the highest suitability and return to Step (ii). (vi) The convenience store with the highest suitability is the most suitable for printing the action figure. Use Eq. (17) to determine the start time, facilitating JIT production for the customer. (vii) Stop. 4.4. Dynamic UbiM case Assume a customer begins from nodes 1 to n at time ts . This route can be divided into two sections: the shortest path from node 1 to each convenience store and the shortest path from a convenience store to node n.
Min d˜1k ( + ) d˜kn
Customer name Cell phone number Date & time Location (latitude, longitude) Chosen action figure no. Chosen action figure name Price Payment status Credit card no.
estimated as follows: 17
I
A
Table 2 Request content.
(20)
where d˜ ij represents the shortest distance from nodes i to j. The waiting time at the convenience store located at node k is
i. Apply Algorithm 1 for determining d˜1k for each k ∈ Q. ii. For each k ∈ Q, set the start location to node k. Apply Algorithm 1 to determine d˜kn . iii. For each k ∈ Q, evaluate the suitability of the convenience store at node k as
sk = 1 − (d1k1 + dkn1 + d1k2 + dkn2 + d1k3 + dkn3 )/(p1 + p2 + p3 )
(22)
According to the fuzzy Dijkstra algorithm proposed by Deng et al.,
sk = 1 − (d1k1 + dkn1 + 4d1k2 + 4dkn2 + d1k3 + dkn3 )/(p1 + 4p2 + p3 )
(23)
iv. The convenience store with the highest suitability is the most suitable for printing the action figure. Use Eq. (17) to determine the start time, facilitating JIT production for the customer. v. Stop.
5. Illustrative cases In this section, two cases are presented for illustrating the applicability of the proposed methodology. The experimental region was a 4-km2 region in the Nantun District of Taichung City in Taiwan (Fig. 5). An abstracted road map (Fig. 6) shows 15 stores of the largest convenience store chain in Taiwan for printing 3D action figures in the UbiM network system. A customer sent a request to the system server. The request content is shown in Table 2. The latitude and longitude of the customer's location was detected by the GPS system in the customer’s smartphone; the customer was near the corner of Wuquan Road and Wuquan West 5th Street. According to multiple GPS detections, the customer's speed was estimated as 24.5, 25, and 25.5 km/h. Accordingly, the time required for printing the requested action figure was estimated to be 2.5, 2.75, and 3 h. Algorithm 1 was then applied to identify the convenience store most suitable for the customer. The process is summarized in Table 3. The convenience store most suitable for the customer was L or O, each of which achieved the highest satisfaction level (0.990) and the shortest distance (0.7 km).
Please cite this article as: Y.-C. Lin, T. Chen, A ubiquitous manufacturing network system, Robotics and Computer Integrated Manufacturing (2016), http://dx.doi.org/10.1016/j.rcim.2015.10.009i
Y.-C. Lin, T. Chen / Robotics and Computer-Integrated Manufacturing ∎ (∎∎∎∎) ∎∎∎–∎∎∎
7
Table 3 Application process of Algorithm 1. (Round 1) Node
Start location
20
21
L
17
d˜i (km) si (h) Visited
0.00
0.21
0.58
0.70
0.37
1.000 Visited
0.997 Unvisited
0.992 Unvisited
0.990 Visited
0.995 Unvisited
(Round 2) Node
Start location
d˜i (km) si (h) Visited
0.00
0.21
0.58
0.70
0.37
0.95
1.000 Visited
0.997 Visited
0.992 Unvisited
0.990 Visited
0.995 Unvisited
0.986 Unvisited
20
21
L
17
19
(Round 3) Node
Start location
20
21
d˜i (km) si (h) Visited
0.00
0.21
0.58
0.70
0.37
0.95
1.11
0.80
1.000 Visited
0.997 Visited
0.992 Unvisited
0.990 Visited
0.995 Visited
0.986 Unvisited
0.984 Unvisited
0.988 Visited
(Round 4) Node Start location
d˜i (km) si (h) Visited
20
21
L
L
17
17
19
19
15
15
H
H
O
0.00
0.21
0.58
0.70
0.37
0.95
1.11
0.80
0.70
1.000 Visited
0.997 Visited
0.992 Visited
0.990 Visited
0.995 Visited
0.986 Unvisited
0.984 Unvisited
0.988 Visited
0.990 Visited
Fig. 7. Route planned using Algorithm 1.
Convenience store L was considered as an example; the route from the starting point is shown in Fig. 7. To achieve JIT production, Eq. (17) was calculated, and the customer was recommended to depart at 05:18:41 PM.
In this case, Google Maps was used for comparison purposes. The recommendation results of Google Maps are shown in Fig. 8: (1) According to the recommendation results, approximately
Please cite this article as: Y.-C. Lin, T. Chen, A ubiquitous manufacturing network system, Robotics and Computer Integrated Manufacturing (2016), http://dx.doi.org/10.1016/j.rcim.2015.10.009i
8
Y.-C. Lin, T. Chen / Robotics and Computer-Integrated Manufacturing ∎ (∎∎∎∎) ∎∎∎–∎∎∎
Experimental Region Fig. 8. Recommendation results obtained using Google Maps.
9 min were required by the customer to reach the nearest convenience store. The distance was 2.8 km, and the satisfaction level was 0.959, which was considerably lower than that achieved using the proposed methodology. (2) Unexpectedly, Google Maps led the customer to a convenience store that was outside of the experimental region. (3) Google Maps specified five transportation modes: on foot and by car, bus, bicycle, or airplane. However, the customer's vehicle speed was not considered. (4) Similar to the proposed methodology, Google Maps proposed several routes, which were modifiable, for reaching the same convenience store. The customer planned to arrive at the convenience store at
05:00 PM and then travel to a destination (24.148755° latitude and 120.657154° longitude) located on Zhongming South Road. For each convenience store, Algorithm 2 was used to determine the shortest paths from the starting location (denoted as “St”) to the convenience store (i.e., the 1st shortest path) and from the convenience store to the destination (denoted as “De”) (i.e., the 2 nd shortest path). The results are summarized in Table 4. The total duration was minimized with convenience store K as the midpoint (Fig. 9). The satisfaction level was 0.980. In other words, the customer initially travelled to convenience store K to collect the action figure and then continued toward his destination. The time at which the customer arrived at convenience store K was estimated as follows: 05:00 PM þ1.24/(24.5, 25, 25.5) h¼(05:02:55, 05:02:59,
Please cite this article as: Y.-C. Lin, T. Chen, A ubiquitous manufacturing network system, Robotics and Computer Integrated Manufacturing (2016), http://dx.doi.org/10.1016/j.rcim.2015.10.009i
Y.-C. Lin, T. Chen / Robotics and Computer-Integrated Manufacturing ∎ (∎∎∎∎) ∎∎∎–∎∎∎
9
Table 4 Shortest paths to and from each convenience store. Convenient store
1st shortest path
Path Duration
2nd shortest path
Path duration
Total duration
A B C D E F G H I J K L M N O
St-L-11-J-12-7-A St-17-H-10-C-4-3-2-1-B St-17-H-10-C St-17-16-13-9-G-3-2-D St-17-16-13-9-E St-L-11-F St-17-16-13-9-G St-17-H St-17-H-I St-L-11-J St-17-16-15-K St-L St-20-19-N-18-M St-20-19-N St-21-O
2.34 2.58 1.71 2.30 1.84 1.42 1.48 0.76 0.92 1.19 1.24 0.70 1.48 1.13 0.70
A-7-12-J-11-L-17-16-15-De B-1-8-G-K-15-De C-10-9-13-16-15-De D-2-3-G-K-15-De E-9-13-16-15-De F-11-L-17-16-15-De G-K-15-De H-17-16-15-De I-H-17-16-15-De J-11-L-17-16-15-De K-15-De L-17-16-15-De M-14-15-De N-19-15-De O-21- L-17-16-15-De
2.93 1.44 1.44 1.24 1.13 2 0.41 1.18 1.34 1.77 0.16 1.28 0.60 0.60 1.88
5.37 4.02 3.15 3.54 2.97 3.42 1.89 1.94 2.26 2.96 1.40 1.98 2.08 1.73 2.58
Fig. 9. Route planned using the proposed methodology.
05:03:02) PM The completion time of the action figure was estimated as follows: 02:35:22 PM þ(2.5, 2.75, 3) h¼ (05:05:22, 05:20:22, 05:35:22) PM Consequently, the customer had to wait (05:05:22, 05:20:22, 05:35:22) (–) (05:02:55, 05:02:59, 05:03:02) ¼ (2.33, 17.38, 32.45) min. To achieve JIT production and avoid a waiting time, the customer was recommended to depart at 02:52:45 PM, that is, 17.38 min after placing the order. The route planned using Google Maps is shown in Fig. 10; the
route length was 5.3 km, considerably longer than that obtained using the proposed methodology. The satisfaction level was 0.923. The customer was again led to a convenience store outside the experimental region; it was estimated that the customer would reach the convenience store at 05:06 PM and the destination at 05:09 PM. The waiting time of the customer was estimated as ( 0.63, 14.37, 29.37) min.
6. Conclusions 3D
printing
is
introducing
revolutionary
changes
to
Please cite this article as: Y.-C. Lin, T. Chen, A ubiquitous manufacturing network system, Robotics and Computer Integrated Manufacturing (2016), http://dx.doi.org/10.1016/j.rcim.2015.10.009i
10
Y.-C. Lin, T. Chen / Robotics and Computer-Integrated Manufacturing ∎ (∎∎∎∎) ∎∎∎–∎∎∎
Fig. 10. Route planned using Google Maps.
manufacturing. The prevalence of 3D printers connected by the Internet also form a UbiM environment in which customers can easily access a nearby 3D printing facility, a phenomenon creating considerable business opportunities. Although some websites have collaborated online to provide 3D printing services jointly, few 3D printing facilities have worked together to serve customers from different regions, which clearly requires the formation of a UbiM network. To this end, a UbiM network system was constructed in this study. In UbiM network systems, a customer places an order for an action figure by using a client-side app or Webbased interface and pays online. The system server then assigns the order to the convenience store nearest to the customer’s location to print the required action figure. This study developed an FINLP model to determine the most suitable convenience store. Two modified fuzzy Dijkstra algorithms were employed to solve the FINLP problems for static and dynamic UbiM cases. Subsequently, the customer is informed of the location of and route to the recommended convenience store. Compared with conventional manufacturing and logistics services, which are difficult to access directly or time consuming, the intensive distribution of convenience stores and the quick manufacturing capability of 3D printers enable a customer to order and receive an action figure from any location almost instantaneously. In addition, the proposed methodology is an innovative attempt that transforms a service network into a manufacturing network. The
established UbiM network system involves applying ubiquitous computing, mobile commerce, 3D printing, and location-based services. In two cases for illustrating static and dynamic UbiM, the proposed methodology showed high applicability and considerable potential for providing satisfactory recommendation results to customers who had not yet departed for or were traveling to the 3D output location. An existing mobile guide, Google Maps, was used for comparison. Compared with Google Maps, the proposed methodology is advantageous because it recommends the shortest path for obtaining the required action figure and thereby increases the customer satisfaction level. Introducing such a UbiM service in convenience stores would bring profitable business opportunities to convenience store chains. In the future, other UbiM services and networks could be designed and established. In addition, the UbiM network system constructed in this study is homogeneous; in other words, it is composed of 3D printing facilities that possess equal capability. In addition, a heterogeneous UbiM network system could be constructed using a similar method. References [1] 3d-printing.net, Asda is launching a 3D printing service – tomorrow! 〈http://
Please cite this article as: Y.-C. Lin, T. Chen, A ubiquitous manufacturing network system, Robotics and Computer Integrated Manufacturing (2016), http://dx.doi.org/10.1016/j.rcim.2015.10.009i
Y.-C. Lin, T. Chen / Robotics and Computer-Integrated Manufacturing ∎ (∎∎∎∎) ∎∎∎–∎∎∎ www.ettoday.net/news/20131021/284587.htm〉, 2015. [2] M. Asadi-Eydivand, M. Solati-Hashjin, A. Farzad, N.A.A. Osman, Effect of technical parameters on porous structure and strength of 3D printed calcium sulfate prototypes, Robot. Comput. Manuf. 37 (2016) 57–67. [3] Autodesk, Spark. 〈https://spark.autodesk.com/〉, 2015. [4] Baidu. 〈http://wenku.baidu.com/view/ac8334707fd5360cba1adb21.html〉, 2013. [5] B. Berman, 3-D printing: the new industrial revolution, Bus. Horiz. 55 (2) (2012) 155–162. [6] T. Chen, Creating a just-in-time location-aware service using fuzzy logic, Appl. Spat. Anal. policy (2015) 1-21. [7] T. Chen, Y.-C. Wang, Z. Lin, Predictive distant operation and virtual control of computer numerical control machines, J. Intell. Manuf. (2015) 1-17. [8] T. Chen, H.-C. Wu, Finding the just-in-time service location and path in a ubiquitous service network, Int. J. Internet Manuf. Serv. 3 (2) (2013) 137–147. [9] B.V. Cherkassky, A.V. Goldberg, T. Radzik, Shortest paths algorithms: theory and experimental evaluation, in: Proceedings of the ACM–SIAM Symposium on Discrete Algorithms, 1994, pp. 516–525. [10] Chih-Mao Information. 〈http://www.cmic.com.tw/Production_process.aspx〉, 2015. [11] Y. Deng, Y. Chen, Y. Zhang, S. Mahadevan, Fuzzy Dijkstra algorithm for shortest path problem under uncertain environment, Appl. Soft Comput. 12 (2012) 1231–1237. [12] E.W. Dijkstra, A note on two problems in connexion with graphs, numer. Math. 1 (1959) 269–271. [13] D. Ding, Z. Pan, D. Cuiuri, H. Li, A practical path planning methodology for wire and arc additive manufacturing of thin-walled structures, Robot. Comput. Manuf. 34 (2015) 8–19. [14] i.materialise, Factory 2.0 for designers. 〈https://i.materialise.com/factory-2-0for-designers〉, 2015. [15] Y. Fan, D. Zhao, L. Zhang, S. Huang, B. Liu, Manufacturing grid: needs, concept and architecture, Lect. Notes Comput. Sci. 3032 (2004) 653–656.
11
[16] L. Goldman, R.L. Nagel, K. Preiss, Agile Competitors and Virtual Organizations – Strategies for Enriching the Customer, Van Nostrand Reinhold, New York, NY, 1995. [17] D.F. Kehoe, N.J. Boughton, New paradigms in planning and control across manufacturing supply chains – the utilisation of Internet technologies, Int. J. Oper. Prod. Manag. 21 (5/6) (2001) 582–593. [18] S. Keating, N. Oxman, Compound fabrication: a multi-functional robotic platform for digital design and fabrication, Robot. Comput. Manuf. 29 (6) (2013) 439–448. [19] J.P. Kruth, M.C. Leu, T. Nakagawa, Progress in additive manufacturing and rapid prototyping, CIRP Ann. Technol. 47 (2) (1998) 525–540. [20] C.X.F. Lam, X.M. Mo, S.H. Teoh, D.W. Hutmacher, Scaffold development using 3D printing with a starch-based polymer, Mater. Sci. Eng. C. 20 (1) (2002) 49–56. [21] E. Lin, I. Minis, D.S. Nau, W.C. Regli, Virtual manufacturing, Lawrence Associates, Inc. 〈http://www.isr.umd.edu/Labs/CIM/virtual.html〉, 1997. [22] My Mini Factory. 〈https://www.myminifactory.com/〉, 2015. [23] F. Rengier, A. Mehndiratta, H. von Tengg-Kobligk, C.M. Zechmann, R. Unterhinninghofen, H.U. Kauczor, F.L. Giesel, 3D printing based on imaging data: review of medical applications, Int. J. Comput. Assist. Radiol. Surg. 5 (4) (2010) 335–341. [24] ShenzhenXin Ju Xin Toy Design Co., 3d Printing Anime Action Figure. 〈http:// szjxsy.en.alibaba.com/product/60239779940-212431068/3d_printing_anime_ action_figure.html#!〉, 2015. [25] D.N. Silva, M.G. De Oliveira, E. Meurer, M.I. Meurer, J.V.L. da Silva, A. SantaBárbara, Dimensional error in selective laser sintering and 3D-printing of models for craniomaxillary anatomy reconstruction, J. Cranio-Maxillofac. Surg. 36 (8) (2008) 443–449. [26] D. Wu, M.J. Greer, D.W. Rosen, D. Schaefer, Cloud manufacturing: strategic vision and state-of-the-art, J. Manuf. Syst. 32 (4) (2013) 564–579. [27] H.J. Zimmermann, Fuzzy Set Theory and its Applications, Kluwer Academic Publishers, Dordrecht, 1991.
Please cite this article as: Y.-C. Lin, T. Chen, A ubiquitous manufacturing network system, Robotics and Computer Integrated Manufacturing (2016), http://dx.doi.org/10.1016/j.rcim.2015.10.009i