PRODUCTION PLANNING SYSTEM BASED ON SIMULATION (PPSS) : APPLICATION METHODOLOGY MARCEL B. DESCO, JONAS L. MAIA, ORIDES MORANDIN JR., EDILSON R. R. KATO Artificial Intelligence and Automation Laboratory, Computing Department Federal University of São Carlos Rod. Washington Luiz, Km 253, ZIP 13565-905, São Carlos, SP, BRAZIL E-mails: mdesco@comp.ufscar.br, jlmaia@comp.ufscar.br, orides@dc.ufscar.br, kato@dc.ufscar.br Abstract In the latest years, changes in the production systems and in the consumer market demanded the development of new techniques, including in the Production Planning and Control area. Meanwhile, discrete event simulation has gained strength in a broad range of applications, including Production Planning. However, regarding to such issue, it s not being used on a short-term basis. In order to solve this limitation, the research group proposed an approach called PPSS. This work aims at defining and application methodology for this approach, as well as developing an interactive interface to simplify dynamical data acquisition. In order to exemplify and validate the PPSS usage, the paper presents a case study where a combination of products processing sequences and production-line input sequences are evaluated and the impact of both factors on the total production is analyzed. Following, some conclusions are drawn about the methodology itself and about the case-study results. Keywords Production Control, Manufacturing Systems, Simulation, Methodology, Scheduling 1 Introduction In the latest years, various changes in the behavior of the consumer market, like products lifecycle reduction, more strict quality control issues, need to reduce costs, etc., demanded the development of technologies in order to support these new necessities. In such context, the efficient deployment of Production Planning and Control techniques is crucial so that the aimed targets can be reached [1][2][3]. Concomitantly, discrete event simulation techniques have presented an enormous development and have been more and more applied to solve problems in manufacturing systems, for example, lead-time reduction, identification of production bottlenecks, definition of stocking policy, plant layout validation, besides Production Planning and Control[4][8][9]. However, such techniques are rarely used to solve short-term questions in Production Planning and Control. Aiming at filling in this gap, our research group proposed in a previous work a technique called PPSS (Production Planning System based on Simulation)[5][10][11]. The work here presented aims at formalizing the application method of this technique, so that it can be correctly deployed in any corporation, achieving the expected results. The main steps of the methodologies proposed by [4][6] were explored. In order to aid the deployment and usage of the technique, a graphical interface was developed, which will actuate directly in some of the phases of the mentioned methodology. To illustrate the application of the presented methodology, a case-study was performed in which the impact of distinct production sequences was evaluated. The paper can be outlined as follows: Section 2 summarizes the main characteristics of the PPSS approach. Section 3 presents the methodology itself, describing all the steps that must be followed. Then comes the case study, exemplifying the deployment of the PPSS technique, aided by the methodology here proposed. The fifth section draws some conclusions about the methodology and about the results obtained in the case-study. 2 PPSS in a nutshell The main idea of the PPSS is to use a simulation software to, based on plant data (previously inserted into a database), verify the performance of distinct actions or factors combination. Whenever an event caused a rescheduling need in the plant, certain simulation scenarios can be chosen, in order to portrait all the possible relevant situations (either considering the plant current status or historical data). According to this selection, the performance of each one can be analyzed. The database must contain up-to-date information about the plant conditions (resource conditions, raw material and finished product stock levels, etc.), besides information concerning policies and strategies of the company. Some scenarios are picked up, using all the information mentioned above, and the simulation process will be carried out. Afterwards, the scenarios that achieved the specified goals will be classified according to their performance, aiding the decision-making process. Figure 1 exemplifies the information flow between the different PPSS modules.
Figure 1. Information flow between the PPSS modules The PPSS can be applied to evaluate the best combination of production sequences, products input sequences and batch sizes, in a user-friendly way, once that provides an easy-to-configure interactive interface. This way, the PPSS provides results that help the decision-making process, aiming to accomplish the production rescheduling in the most efficient way. 3 Application Methodology In order to apply this technique, some steps must be followed, which are described in detail below. Step 1: Problem Definition The customer gives a general characterization of the system, identifying the main questions to be solved. Also in this phase are established, the goals to be reached and the limits to be obeyed. Step 2: Conceptual Model Formulation Based on information acquired with the company staff, formal techniques are used to accomplish a previous modeling of the system (techniques such as block diagrams, Petri networks and so on). This modeling aims at assuring that the model actually represents the factory s plant and, using the tools provided by each technique (e.g. reachability tree in Petri Nets), a preliminary behavior analysis can be done. Step 3: Data Preparation At first the data that are relevant to the simulation are identified, based on the previous problem definition. It s carried out the gathering of such data and a statistical analysis of them, trying to verify their representativity and the most realistic manner of inserting them into the simulation. A fraction of the data must be obtained before the construction simulation model. However, some data must be dynamically gathered in order to improve the flexibility of the generation of distinct simulation scenarios. In the case study, it will be presented the interface to the PPSS applied to a plant that s been currently built in the Artificial Intelligence and Automation Laboratory, at the Federal University of São Carlos, plant which was already presented in a previous paper [7]. Step 4: Computational Model Construction A simulation engineer creates the simulation model that represents the factory to be analyzed. The model can be computationally developed using a simulation language (e.g.: GPSL, SIMAN and Sim- Script among others) or using a simulation tool (e.g.: Automod, Arena, ProModel, Taylor II, etc). Step 5: Verification Once that the computational model is concluded, it is proceeded a verification in its logic in order to avoid inconsistencies. Common testing techniques, like mutant analysis, can be applied to test the coding of the simulation model. Step 6: Validation Simulation rehearsals are carried out, so that their results can be compared to historical data in order to determine if the model actually reproduces the real system. Those rehearsals are based on scenarios that portray crucial conditions, like resources and storage areas concurrence, etc. Through statistical analysis, it s verified the degree of sensitivity of the output data. This analysis also allows verifying whether the results have the expected behavior for different input conditions. Step 7: Experiment Project Given the plant conditions, different simulation scenarios will be chosen in order to ensure the representativity of the simulation data. To accomplish a previous filtering of these scenarios, a Fuzzy filter is being developed and will be attached to the PPSS approach. Step 8: Simulation Runs Based on the data gathered in the previous steps the simulation of different scenarios is carried out. In the PPSS interface there s a module responsible for triggering the simulation process, besides allowing the user to quickly edit the model. Step 9: Interpretation and Analysis The results generated by the simulation runs are interpreted and analyzed in order to support the production rescheduling decision.
Once more, the graphical interface comes to support the decision-making, providing the user with comparative graphics, besides presenting the full report generated by the simulation software. Step 10: Reports and Documentation If necessary, documents and reports can be generated based on the results previously obtained in the simulation phase. 4 Case Study In order to validate the proposed methodology, a case study was carried out aiming to, using the proposed application methodology, evaluate the impact of different tooling sequences and production-line input sequences on products manufactured by the plant being built in the Artificial Intelligence and Automation Lab. Still regarding such occurrences, different input sequences and distinct batch sizes may impact on the plant total production capacity, within a predefined period of time. Step 2: Conceptual Model Formulation The plant was already modeled using the Petrinets formal technique, and the results were represented in [6]. As a compliment, the tooling sequences for products A, B and C were modeled by Petri-nets shown in figure 3. Step 1: Problem Definition As previously described, the main question to be solved is finding the best combination of different tooling sequences and products input sequences, in order to reduce the total production time spent by the manufacturing process. As a limitation, only one batch size is going to be rehearsed. This assumption can be considered true according to previous studies executed on the plant. Figure 2 shows a sketch of the mentioned plant. Figure 2. The Plant Layout Such plant is intended to produce three different products, each one with its own alternatives of tooling sequences. Needless to say, according to different occurrences in the plant, e.g machines breakdown, programmed maintenance and so on, one of the alternatives may accomplish a better performance than the others for each product. Figure 3. Petri-nets for Products A, B and C Step 3: Data preparation Based on the existing system and on the goals to be reached, the identified relevant data were: production time, machines setup time, machines breakdown, production time according to product and production sequences, batch size, production mix and total production Previously executed data analysis shown that the products production times can be approached by
a constant distribution, with times displayed (in minutes) in table 1. Mch1 Mch2 Mch3 Mch4 Mch5 Mch6 Product 1 (Seq1) 1.5 1.3 1.7 Product 1 (Seq2) 1.7 1.9 2 Product 2 (Seq 1) 2.1 0.9 1.6 1.85 Product 2 (Seq 2) 2.1 2.3 1.9 Product 3 1.1 1.4 1.3 2.1 1.9 2.3 Table 1. Production times (minutes) The machines suffered a programmed maintenance as follows: Machine 3 : Maintenance started in the second day, and lasted for 12 hours; Machine 4 : Maintenance started in the fourth day, and lasted for 12 hours; Figure 4. Input of Plant-related data Also, the values for the other parameters are shown below: Setup time: 30min (constant) Production Mix: A (45%), B (30%) and C (25%) Total Production: 4000 units Batch Size: 500 products As already mentioned, some data must be acquired at simulation time, fact that demanded an interactive solution, in order to make easier to the user the task of entering dynamical data in the simulation model. In order to provide such interactivity, an interface was developed and its screen is exemplified by figures 4,5 and 6. The module shown in figure 4 acquires data related to the products, like total production, production mix, batch sizes and input sequencing. Meanwhile, figure 5 illustrates the acquiring of data related to the plant, i.e. machines breakdown and production processes. Concluding, in order to make the system capable of using various production processes, the module shown in figure 6 allows the input of them, via the input of the time spent at every machine, considering a blank value as an identifier to the machines not included in the specified production process. Also, every combination must be assigned a probability of usage in the ratio field. Figure 5. Input of product-related data Figure 6. Entering production-processes Step 4: Computational Model Construction The plant, including its layout, operational logic and interfacing procedures was modeled with the Automod simulation software. Figure 4 shows displays a snapshot of the built software model
second, and product 3 its unique 1 alternative). The results were group according to the production-line input sequence used, 1-2-3 for example, means that product 1 was the first to enter the line, then product 2 and finally product 3, always obeying the batch sizes imposed. Total Production Time 18 Figure 7. The Plant Layout Step 5 & 6: Verification and Validation Previous work have already demonstrated that the model logic is consistent and it s actually reproduces the real system. Time (days) 16 14 12 10 8 6 4 2 0 P1:1 P2:1 P1:1 P2:2 P1:2 P2:1 P1:2 P2:2 1-2-3 2-1-3 3-2-1 Step 7: Experiment Project Production Processes Based on the data previously gathered, some simulation scenarios were built. The most representative scenarios were chosen and were used in the simulation run. Step 8: Simulation Runs Using the scenarios that were chosen, the simulation is carried out. It can be triggered easily by the interface, as shown in figure below: Figure 9. Total Production Time As can be seen, both variables have a major impact on the analyzed time. While the combination of the input-sequencing 2-1-3 and of the production processes 1-1-1 takes about 8 days and 7 hours to product the 4000 units, the 1-2-3 input-sequence and 2-1-1 tooling sequence combination takes about 16 days and 5 hours, almost two times the production time of the fastest one. Furthermore, it s not possible to assert that one input-sequencing is always better than the others, or that one production process will always show better performance. As shown in figure 10, the 1-2-3 input-sequence has a terrible performance in the 1-1-1 and 2-1-1 tooling sequences, but an acceptable one in the other production sequences. Total Production Time, Input 1-2-3 Figure 8. The Run Simulation module Step 9: Interpretation and Analysis When the simulation is finished, some graphical results are presented to the user so that they can be used as a support to the decision making process. A general comparison, between the products tooling sequences and the production-line input sequences is shown in figure 9. The X-axis represents the production sequence used for each product in the rehearsal (e.g. P1: 1, P2:2, means that product 1 used its first production alternative, product 2 its Time(days) 18 16 14 12 10 8 6 4 2 0 P1:1 P2:1 P1:1 P2:2 P1:2 P2:1 Production Processes P1:2 P2:2 Figure 10. Total Production Time, using the 1-2-3 input sequence Concluding this section, the analysis accomplished above could be done in order to meet a deliv-
ery date for a specific product. For example, suppose that product 3 must be delivered in a certain date that the plant occurrences (machines breakdown) wouldn t allow the delivery. This way, the rescheduling could use a combination that produced product 3 in the fastest manner, even though this combination may not be the best one to the whole production. Figure 11 exemplifies an analysis applied to product 3. The production process 1-2-1 with the 2-1-3 input sequencing produces all the units in 12 days, whereas the 1-2-3 takes 10 days and 12 hours to produce them all. As it could be expected, the 2-1-3 input sequencing accomplishes it in the shortest time (about 8 days) once that the product 3 is not the last one to enter the producing line. Time(days) 14 12 10 8 6 4 2 0 Product 3 Production Time, Production processes "1-2-1" 1-2-3 2-1-3 3-2-1 Production-line Input Sequences Figure 11. Production Time for product 3, using the 2-1-1 production processes Step 10: Reports and Documentation Although not necessary in this case, reports could be automatically generated containing graphics and the data that was defined at the beginning of the simulation. Once again, the interface provides all the information needed to compile the reports and other documents. 5 Conclusions Concluding, this paper revisited the PPSS approach, a solution to solve short-term rescheduling questions in manufacturing systems. The work here presented a methodology to deploy the PPSS in any production plant, as well as an interface to facilitate the usage of the approach, making any user capable of entering simulation data (with no knowledge on the simulation environment) and analyzing the results aided by automatically created graphics. A case study applied the PPSS approach and its application methodology to evaluate the best combination of plant characteristics like production-line input sequence and tooling sequences. The results were presented and the best combination was pointed out. Further work is being carried out in order to create a Fuzzy-filter able to analyze the simulation scenarios and determine which of them will produce representative results. Bibliographical References [1] Slack, N.; Chambers, S.; Harland, C.; et al. (1995) Operations Management. Pitman Publishing London. [2] Artiba, A.; Elmaghaby, S.E; et al. (1997) The Planning and Scheduling of Production Systems: Methodologies and Applications. Chapman & Hall. [3] Dorf, R. C., Kusiak, A. (1994) Handbook of Design, Manufacturing and Automation. John Willey Professional. [4] Banks, J. (1998) Handbook of Simulation- Principles, Methodology, Advances, Applications and Practice. John Willey& Sons. [5] Kato, E. R. R.; Morandin Jr, O.; Politano, P. R., et al. (2000) Production Planning System based on Simulation : Finding the Best Plant Stocking Policy. Proceedings of the IV IEEE Induscon. [6] Carson, J. S.; Nelson, B. L. et al. (2000) Discrete Event System Simulation. Prentice Hall. 3 rd Edition. [7] Morandin Jr, O.; Kato, E. R. R.; Politano, P. R.; Camargo, H. A., et al. (2000) A modular modeling approach for Automated Manufacturing Systems Based on Shared Resources and Process Planning Using Petri Nets. Proceedings of the IEEE International Conference on Systems, Man and Cybernetics. [8] Balogun, O. O.; Popplewell, K. (1999) Towards the integration of flexible manufacturing system scheduling. International Journal of Production Research, vol. 37, nº 15, 3399-3428. [9] Hill, J. A.; Berry, W. L.; Leong, G. K.; Schilling D. A., (2000) Master production scheduling in capacitated sequence-dependent process industries. International Journal of Production Research, vol. 38, nº 18, 4743-4761. [10]Reynolds, A. P.; Mckeown, G. P. (2000). Scheduling a Manufacturing Plant using Simulated Annealing and Simulation. Computers & Industrial Engineering, vol. 37, 63-67. [11]Byrne, M. D.; Bakir, M. A. (1999) Production Planning using a hybrid simulation-analytical approach. International Journal of Production Economics, vol. 59, 305-311.