Management Review (2006) 11(3),(2006) 133-139 Seong-HoonAsia ChoiPacific et al./Asia Pacific Management Review 11(3), 133-139
A Conceptual Framework of an Advanced Planning and Scheduling System for Printed Circuit Board Manufacturing Lines Seong-Hoon Choi a*, Moon-Won Park b, Dong-Ho Lee c, Keun-Chae Jeong d, Seung-Kil Lim e, Jae-Gon Kim f, and Geun-Cheol Lee g a
Department of Industrial and Information Systems Engineering, Sangmyung University, Chonan, KOREA b Research Planning Team, AIM System, Seoul, KOREA c
d
f
Department of Industrial Engineering, Hanyang University, Seoul, KOREA Department of Structural Systems and CAE, Chungbuk National University, Cheongju, KOREA
e Division of e-businessIT, Sungkyul University, Anyang, KOREA Department of Industrial and Management Engineering, University of Incheon, Incheon, KOREA g College of Business Administration, Konkuk University, Seoul, KOREA Accepted in April 2006
Available online
Abstract Advanced planning and scheduling (APS), which is one of the most advanced technologies in manufacturing management applications, is emerged to resolve the shortcomings of enterprise resource planning (ERP) systems and manufacturing execution systems (MES) as a planning and scheduling tool. In this paper, we suggest a conceptual design of an APS system, specific for printed circuit board (PCB) manufacturing lines. In general, the PCB line can be regarded as a hybrid flow shop in which there are serial workstations, each with unrelated parallel machines, and has special characteristics such as complex and re-entrant flows, high and sequence-dependent setup times, subcontracting in the operation level, etc. Although there are a number of commercially available APS packages, they are not directly applicable to the PCB lines due to these characteristics. Therefore, we suggest a new conceptual design of APS, called PCB-APS in this paper, which consists of five modules: work module, simulation run module, dispatching rule generation module, simulation model generation module, and master data input module. The PCB-APS suggested in this paper considers the characteristics of the PCB lines in more details, and hence can improve the system performances significantly. Keywords: Advanced planning and scheduling; Conceptual; Printed circuit board
than that in any other manufacturing systems (Geber, 2000; Lee et al., 2003; Liu et al., 2002).
1. Introduction The printed circuit board (PCB) is one of the products that play a key role in the electronics industry as a major component of most electronic products (Lee et al., 2003). According to the rapid growth of electronics industry, the diversity of new PCB products has been appeared and production volume has been steadily increased. Since PCBs are sub-parts of most electronic products, the PCB manufacturing system uses the make-to-order policy. Therefore, committing due dates quickly for customers’ orders and satisfying due dates are very important to maintain competitiveness (Cha et al., 2002)
Recently many PCB manufacturing companies have constructed enterprise resource planning (ERP) systems for making the company more information-oriented and productive. Although ERP systems increase productivity of companies, the production scheduling function of the ERP systems is not good enough since the ERP systems focus on elementary transaction processing. A survey result of AMR Research showed that more than 50% companies implementing ERP recognize ERP cannot resolve their planning and scheduling problems. Therefore we need to develop an APS system to generate an effective production planning and scheduling, which works together with the ERP system (Cha and Hwang, 2001; Kim et al., 2002; Turbide, 1998).
However, since the PCB manufacturing system can be regarded as a complicated reentrant hybrid flowshop consisting of a series of production stages with parallel machines, the production-scheduling problem is very difficult to solve (Choi et al., 2005). To make matters worse, the type of PCBs becomes more diversified with development of technology and the diversity of PCB product specifications complicates PCB production management as a result. Therefore, satisfying due dates of orders is more difficult *
Consequently, advanced planning and scheduling (APS) packages are emerged to resolve the shortcomings of ERP systems such as Rhythm of i2 Technology and APO of SAP. However many PCB companies cannot use these packages since they are very expensive and need to
Corresponding author’s e-mail:
[email protected] 133
Seong-Hoon Choi et al./Asia Pacific Management Review (2006) 11(3), 133-139
long construction time. Note that almost PCB manufacturers are small and medium (S&M) sized companies. Therefore we need to develop an efficient and effective APS package and hence construct the APS systems of the PCB manufacturing companies in more economical and speedy manner. We call the APS package for the PCB manufacturing companies as PCB-APS working together with the ERP system and manufacturing execution system (MES).
Start
Inner layer processing
Pressing
Drilling
Plating
Outer layer processing
End
In this study, we propose a development direction for the PCB-APS focused on S&M PCB manufacturing companies. PCB-APS should be developed by considering the characteristics of the PCB manufacturing lines in more detail and should generate simulation models automatically by reflecting changes occurred in the PCB manufacturing lines. In addition, by containing only the necessary functions for the PCB manufacturing companies in the package, PCB-APS should make it possible that the APS system can be constructed with more economical cost and in more short time and the APS system can generate better production plans and forecast due dates more accurately. Particularly, compared to previously used APS systems, which are mostly just implementations of commercial software packages, the proposed PCB-APS can generate more precise results and have more flexibility in managing the package by using a simulation model, which will be developed considering very details of PCB manufacturing lines.
Final inspection
Electrical testing
Routing
Coating
Figure 1. Manufacturing Process of PCB Products Based on the PCB product structure and its manufacturing processes explained earlier, the characteristics of PCB manufacturing systems could be summarized as following: 1) Basically, the PCB manufacturing system has a type of flow line with a large number of serial workstations. This results in long manufacturing lead times. In addition, there is much variation in the manufacturing lead times of different products. 2) There are various constraints on the identical or non-identical parallel machines at each stage. This requires a systematic method to control the amount of material flows among the parallel machines. 3) Some products may have the reentrant flows in that they visit the same workstation two or more times. This is a common characteristic in the products with multiple layers, such as wafer fabrication, and makes the material flows very complicated. 4) Setup operations are required between processing of different products. In general, the setup times are very long and also sequence-dependent, i.e., the setup time of an operation may differ according to the one just finished. Therefore, the PCB scheduling system should consider the setup times effectively since they affect the overall system performances significantly. 5) When the production capacity is not sufficient, most PCB manufacturing firms consider subcontracting. There are two forms of subcontracting in the PCB manufacturing systems. They are product subcontracting and operation subcontracting. Here, the operation subcontracting means that particular operations are subcontracted. 6) The plating operations can be done only after the necessary panels are prepared. This is because a PCB product is produced by assembling the panels required for the product. This is an important constraint to be considered in the scheduling system. 7) The drilling machines may have multiple axes, and perform the drilling operations using one of the axes after combining several panels required for a product. There are a large number of machines in the drilling workstation, and their processing times are generally longer than others. (For more details about PCB manufacturing process, please see LaDou (2006)).
The following section presents typical characteristics of PCB manufacturing system that make the production-scheduling problem complicated. In section three, we suggest a direction of the development of the PCB-APS. Finally, section four concludes the paper with a summary and discussion of expected detail influences and applications of the PCB-APS. 2. Characteristics of the PCB Manufacturing System The PCB manufacturing system can be regarded as a reentrant hybrid flowshop. In addition to such a typical feature of the PCB manufacturing system, the diversity of PCB product specifications complicates the production management by causing complicate material flows and frequent setup operations throughout the production line. Therefore, satisfying the due dates of orders in the PCB manufacturing system is more difficult than in any other manufacturing systems (Lee et al., 2003). An important feature of PCB products that affects material flows is that there are multiple layers on the boards of certain products. Such PCB products, called MLBs, are highly dense and thin PCBs made of more than four layers. In order to build more than four layers on a board, multiple panels are required for one MLB product. Figure 1 shows typical manufacturing processes for PCB products. Note that the workstation for each process in Figure 1 is composed of several serial and/or parallel machines for several sub-processes.
In general, the characteristics described above deteriorate the scheduling performances of PCB manufacturing systems, such as system throughput, work-in-process,
134
Seong-Hoon Choi et al./Asia Pacific Management Review (2006) 11(3), 133-139
facturing lines. In particular, PCB-APS should respond to changes of the manufacturing systems by improving usability and by generating the simulation model automatically, which can be possible by preparing the necessary basic simulation modules in advance. The previous packages tend to use algorithms which are not verified in terms of the performance and hence do not give good performance, therefore PCB-APS should use the more recent and verified algorithms such as the algorithm of Lee et. al. (2003), which can maximize the performance of the PCB lines. That is, PCB-APS finds the best combination of the dispatching rules by experimenting the candidate dispatching rules to the shop floor using the simulation model and applies the selected rules to the MES. By using this approach, the APS system can determine the schedules of the PCB manufacturing lines in real-time.
and due-date related measures, and make the production scheduling problem very complicated. Therefore, one of the factors that increase the competitiveness of PCB manufacturing firms is to develop a scheduling system that considers the above characteristics effectively. This is the main focus of this paper. 3. Advanced Planning and Scheduling for PCB Manufacturing Systems In this study, we propose a development direction for the PCB-APS focused on S&M enterprises. PCB-APS is different from the previous APS packages with respect to the two points: ‘this is a package for S&M enterprises’ and ‘this is a package for PCB manufacturing companies.’ PCB-APS should be developed by considering the characteristics of the PCB manufacturing lines in more detail and should generate simulation models automatically by reflecting changes occurred in the PCB manufacturing lines. In addition, by containing only the necessary functions for the PCB manufacturing companies in the package, PCB-APS should make it possible that the APS system can be constructed with cheaper cost and in shorter time and the APS system can generate better production plans and forecast due dates more accurately.
The objectives of the PCB-APS are minimization of the costs by maximizing the productivity and minimizing the work-in-process inventories and maximization of the order commit ratio by using the accurate due-date forecasting method. The previous APS package cannot achieve these objectives since they use general-purpose and relatively simple simulation models. These objectives can be achievable by using the approach that uses the APS designed for the PCB manufacturing company only. In addition to these objectives, PCB-APS should have other characteristics such as estimation function of outsourcing quantity, short construction time, short running time, quick response to the changes in the manufacturing lines.
The proposed PCB-APS can be represented as ‘an APS package for S&M PCB enterprises on the basis of simulation-based scheduling methods.’ The basic approach is that the APS system uses heuristics and dispatching rules for generating schedules and simulation models for reflecting the characteristics of the PCB manu-
PCB-APS SYSTEM USER PART
Work Module ERP System
Scheduling
Simulation Run Module
Order Releasing Order DB
Inventory DB
Production Simulation Run Subcontracting What-If Simulation Run Order Promising
MES
Management Information Analysis
Shop status DB Simulation Results DB
PDM
Product Design DB
Rule DB
Simulation Model DB
Process Plan DB
Dispatching Rule Generation Module
Simulation Model Generation Module
Master Data Input Module
Rule Input
Base Model Information Input
Master Data Input
Rule Parameter Tuning Optimal Rule Generation
Model Generation
DEVELOPER PART
Figure 2. A System View of the PCB-APS 135
Seong-Hoon Choi et al./Asia Pacific Management Review (2006) 11(3), 133-139
daily input order in the simulation model, we use factory-wide dispatching rules which considering the overall status of the production line for the better due date commitment and load balancing within each workstation.
We propose the PCB-APS that consists of five modules: work module, simulation run module, simulation model generation module, dispatching rule generation module, master data input module. Work module and simulation run module are user modules, whereas simulation model generation module, dispatching rule generation module and master data input module are included in developer (or administrator) modules. The proposed system configuration for PCB-APS is illustrated in figure 2.
The PCB products have various process plans according to their product types and each product type has a different workstation on which the product type has an effect in terms of the workload. Therefore, over-releasing of an order of one product type could cause a significant unbalance of workload on a specific workstation. Consequently, the factory-wide dispatching rules are needed when the order releasing is performed.
3.1 User Modules of PCB-APS Now we first explain the two user modules included in the proposed PCB-APS system in more detail. System users can build schedules, due-date quotation, and subcontract plan with Work Module, while Simulation Run Module can be used to decide the best plans and schedules using production and what-if simulation runs.
The subcontracting plan in the Work Module could be automatically produced by the subcontract dispatching rules and the predetermined subcontract policy for each product shift. In the simulation model, one subcontract partner is considered as a virtual machine and the subcontract amount is determined by the dispatching rule for the machine.
Work Module In the Work Module, the jobs related with the production and sales departments are performed, such as scheduling for each workstation, order releasing, outsourcing, and order promising. The output of the module is produced by using the results obtained from the simulation run. To generate the schedules of each workstation, simulation-based production scheduling method is used. The advantage of this method is that it can see the real-time status of the production line, such as materials and inventory status, whereas heuristic methods, which generate production planning with master production schedule (MPS), are static.
Determining orders and quantities for subcontracts is done step by step. First, the quantity for the subcontract is computed and then lots of orders are selected by the dispatching rule as much as the determined subcontract quantity. The quantity for the subcontract at a workstation can be computed by estimating the amount of the inventory beyond the capacity of the corresponding workstation and the lots of orders among the current inventory and the upcoming amount that will arrive within one shift. The fourth component of the Work Module is the order promising which automatically generates the estimated due date of an order when the order is quoted through the sales department by a customer. In this study, there are two methods of estimating due date, one is the simulation-based due date quoting method and the other one is based on the Estimated Machine Load (EML). The former is the method which determines the estimated due date of an order by running the simulation when the order arrives and taking the completion time of the order in the simulation results. The latter one uses the workload table for each machine and manipulates it statistically for determining the estimated due date of an order. Note that the workload table is updated periodically by using the latest simulation results. The simulation-based method can give more accurate results than the EML-based method, although it will take much more time. When the quick response for the order quoting is needed, the EML-based method is preferred.
In the simulation model, dispatching rules selected for each workstation are used when production schedules of each workstation are determined. Note that dispatching rules are implemented in the simulation model through the user-code writing and are stored in the Work Rule database (Rule DB in figure 2). There could be two types of the dispatching rules. In the first type, an available machine selects a lot to be processed on the machine (lot selection rule), while a waiting lot selects a machine on which the lot is loaded for processing in the second type (machine selection rule). Generally, all the workstations except for the drilling operation in PCB manufacturing use the lot selection rule. In this study, look-ahead dispatching rules are developed where not only the status of the current workstation but also those of the downstream and bottleneck workstations are considered and the rules are composed of several important criteria which are related with the system performance, such as setup times, machine load balancing index, due dates, waiting times, and etc.
The last component of the Work Module is the management information analysis. The component can provide tables and graphs of the daily trace and correlation of the performance evaluation indexes, such as customer satisfaction, output vs. plan, production rate, cycle time, utilization, inventory levels, amount of the accepted order, sales amount, profit rate and etc., and generate analysis
To generate the order releasing plan in the Work Module, like the production scheduling, we use the simulation results as well. That is, the amount of an order started to be produced for a day in the simulation results is used as the order releasing plan. When determining the 136
Seong-Hoon Choi et al./Asia Pacific Management Review (2006) 11(3), 133-139
The rule parameter tuning, the second component of the Dispatching Rule Generation Module, consists of two parts. The first part is the individual rule tuning in which the parameter values of the user-selected rule for a workstation are updated with the best combination result obtained from a series of the simulation experiments. The user can select a set of parameter values among the alternate combinations of the parameter values for a workstation through multiple simulation runs or perform a simulation for a specific set of parameter values. The overall rule tuning module, the second part of the optimal rule generation, let users automatically determine parameter values for all dispatching rules of all workstations. Simulations are performed in order of the bottleneck workstation and the parameter values of the corresponding workstation are determined according to its order. Also, the experimental design by the orthogonal array method can be used for determining the parameter values. Pre-given combinations of the parameter values can be used as the alternate combinations for the simulation experiments or parameter values can be inputted directly by user.
report automatically. CEOs can monitor management information in real time, evaluate the effect of the PCB-APS construction, and plan effective operation strategies going with the changes in the business environments. Simulation Run Module This module performs the simulation experiments using the dispatching rules selected by the user for an upcoming planning period. In the simulation experiments, the required data are obtained from MES and ERP in real time. This module consists of two submodules: production simulation run and what-if simulation run. The production simulation run sub-module performs simulation and generates the production schedule, the order release schedule, and the forecasted due dates for an upcoming planning period. The user can select the dispatching rules and the results are shown in a report form. Also, the what-if simulation run is done to analyze the effects of due-date changes, facility changes (machines, material handling systems, etc.), and order changes (due date and quantities), and changes in subcontracting.
The third component of the Dispatching Rule Generation Module is the optimal rule generation module. In the module, the optimal combination of the rules and the parameter values are obtained by a series of the simulation experiments for all alternate rule combinations at all workstation. Overall procedure of using the Dispatching Rule Generation Module can be presented as follows. First, the user determines the dispatching rule for each workstation using the optimal rule generation module. Next, more precise and accurate parameter values of each workstation are determined by using the overall parameter tuning module, and then apply the individual parameter tuning to a few of the primal workstations. Finally, in the rule input module, detail modifications and corrections can be done on the results obtained so far.
3.2 Developer Modules of PCB-APS Developer modules include dispatching rule generation module, simulation model generation module, and master data input module. System developer can customize PCB-APS with these modules according to PCB manufacturing company’s own requirements and environments. Details of each module are given below. Dispatching Rule Generation Module This module is composed of two functions, one for manual input of the dispatching rule for each workstation and the other for finding the optimal rule combination among the dispatching rules in the Rule Database for overall workstations, which can be automatically done by the simulation experiments.
Because the dispatching rule selection process needs a number of simulation runs and it takes very long time, thus it would be automatically performed after work hours or weekends. Although, early period of the system may require a significant amount of time to adjust the rules and parameter values, the user will use this module as a strong tool for controlling the shop floor effectively, once the rules are settled down.
The performance of the suggested system will be mainly affected by the dispatching rules embedded in the simulation model. Therefore, the development and maintenance of the dispatching rules are a core point in the system and this module is for helping the users to apply the dispatching rules to the system easily and effectively. The first component of the Dispatching Rule Generation Module is the rule input. Multiple dispatching rules are developed for each workstation and inputted through this component. A dispatching rule is a combined form of several scheduling-related terms, such as due date, waiting time, setup time, processing time and etc., and they are weighted by parameter values according to their relative effects. The user can find and input the optimal rule combination, i.e. selected terms and parameter values, through the simulation experiments. The selected terms and parameter values can be inputted manually or automatically.
Simulation Model Generation Module Developing a simulation model requires experienced experts and a very time- and cost-consuming task. Also, the developed simulation model should be updated whenever there are any changes in the systems such as process changes, facility changes, etc. To overcome these difficulties, therefore, we suggest a module to generate and update simulation models automatically. The module is the form of the user-computer interactive input method, i.e., wizard function, and generates a model using the minimum user 137
Seong-Hoon Choi et al./Asia Pacific Management Review (2006) 11(3), 133-139
data input. To do this, a database of standard models is developed and the user can develop an overall simulation model by combining the standard models so that the reusability of these standard models can be increased.
The master data, which includes the information on products, processes, facilities, process plans, subcontracting, raw materials, and shop floor state, is defined as the one indispensable for generating a simulation model.
This module consists of two submodules. The first one is the base model information input sub-module that receives the required data from the standard simulation model database, manufacturing information database, and standard dispatching rules database (Table 1 summarizes the required data.), and the second one is the model generation sub-module that generates a simulation model after loading the required data obtained from the base model information sub-module.
Among them, the information on raw materials and shop floor state change frequently in real time, and hence the PCB-APS should be designed to load the data automatically from the ERP and MES. On the other hand, the information on products, processes, facilities, process plans, subcontracting change infrequently and are saved in manufacturing information database and/or standard simulation model database after obtained manually from the master data input module. The master data are summarized in Table 2.
Master Data Input Module
Table 1. Data Required for Generating a Simulation Model Databases
Input data
Remarks
Standard simulation models -Standard process plans, etc.
Simulation Model DB
Facility information (photo, plating, press, drilling, coating, etc.) Manufacturing information DB Manual input in the case without existing Process information (inner layer processing, outer layer processing, (Product Design DB and Process information systems such as ERP or manudrilling, coating, etc.) Plan DB) facturing information system Product information (name, characteristic values, etc.) Standard dispatching rules DB (Rule DB)
Set of standard dispatching rules (for selecting the rules automatically)
Table 2. The Master Data Classification
Detailed information
Remarks
Product information
Product ID, Total processing time, etc.
Process information
Machines assigned to each operation, Various constraints on operation processing, etc.
Facility information
Operations performed on each machine , Machine capacity, Maintenance schedule, Standard processing time, Setup time, etc.
Process plan information
Operations required for each product, Conditions to select operations, etc.
Raw material information
Inventory levels, Warehousing information, etc.
Loaded from ERP DB
Production state information
Work-in-process , Production resource state , Progress of subcontracting, etc.
Loaded from MES DB
Subcontracting information
Subcontracting companies and their capacities, Subcontracting policy, etc.
138
Seong-Hoon Choi et al./Asia Pacific Management Review (2006) 11(3), 133-139
since those have several similar manufacturing characteristics to PCB manufacturing. On the other hand, it is necessary to study some research topics related to fast simulation, real-time dispatching and scheduling, and due-date quotation mechanisms for more advanced planning and scheduling.
4. Conclusions In this study, we investigated several distinct characteristics of the PCB production lines and suggested a direction of the development of the APS for PCB manufacturing companies. The suggested APS system for PCB production lines would overcome technical limits of the previous APS products, which results in a significant cost reduction of the APS implementation. Also, advanced scheduling algorithms embedded in the system would strengthen the competitiveness of the PCB manufacturing companies.
References Cha, C.N., Lim, S. and Jeong, Y.K. (2002). Single-machine job scheduling about a common due date with arbitrary earliness/tardiness penalties using a genetic algorithm. Asia Pacific Management Review, 7(2), 239-254.
We can summarize expected benefits resulted from the specialized APS system for PCB manufacturing as follows. First, time and cost needed for the development of APS system can be reduced since the suggested APS system has many built-in submodules specialized for PCB manufacturing. System developers and users can easily customize those modules according to their needs. Second, the suggested APS system can help improve critical performance measures in PCB business like on-time delivery, productivity and inventory level through more effective and efficient order promising, production planning and scheduling. APS system suggested in this study has functionality for supporting due-date quotation and production planning and scheduling with minimum human intervention. Finally, the APS system suggested here can be one alternative option for small-to-medium sized PCB manufacturing companies that want to build up integrated information systems since the APS system can play a role as a middle level decision support system to make a connection between enterprise level transaction systems like ERP and operational level execution systems like MES. The integrated information system can help improve productivity of those companies by reducing lead-time required from receiving customer orders until to fulfilling those orders.
______. and Hwang, H. (2001). Inventory ratio based production switching heuristic (RPSH) for the aggregate production planning problem. Asia Pacific Management Review, 6(1), 1-19. Choi, S. W., Kim, Y. D. and Lee, G. C. (2005). Minimizing total tardiness of orders with reentrant lots in a hybrid flowshop. International Journal of Production Research, 43(11), 2049-2067. Gerber, C. (2000). APS/ERP integration. ID Systems, 20(2), 41-49. Kim, H., Kang, J. and Park, S. (2002). Scheduling of Shipyard Block Assembly Process Using Constraint Satisfaction Problem. Asia Pacific Management Review, 7(1), 119-137 LaDou, J. (2006). Printed circuit board industry. to appear in International Journal of Hygiene and Environmental Health, 209(3), 211-219. Liu, W., Chua, T.J., Lam, J., Wang, F.Y., Cai, T.X. and Yin, X.F. (2002). APS, ERP and MES systems integration for semiconductor backend assembly. Seventh International Conference on Control, Automation, Robotics and Vision (ICARCV ’01), Singapore. Lee, G. C., Kim, Y. D., Kim, J. G. and Choi, S. H. (2003). A dispatching rule-based approach to production scheduling in a printed circuit board manufacturing system. Journal of the Operational Research Society, 54, 1038–1049. Park, M. W., Choi, S. H., Lee, G. C. and Kim, Y. D. (2002). An order promising procedure for simulation-based scheduling systems. Proceedings of 2002 KIIE Spring Conference, KAIST, Daejon, Korea. Turbide, D. (1998). New Concept of Scheduling, APS. Production Solu-
Research results of this study can be used to develop other specialized APS systems for specific manufacturing areas such as semiconductor and TFT/LCD manufacturing
tions.
139