University of Twente Student Theses


Lead time reduction by production planning and control

Noordhuis, RE.. (2018) Lead time reduction by production planning and control.

[img] PDF
Abstract:At Company X, printed circuit board assemblies (PCBAs) are mostly made in two sequential processes. The first process, SMD, is automated and places components on a bare board by a machine. For this process, a lead time of two weeks is used in planning. The second process, HMT, is a merely manual process and also includes the quality tests. For this process, a lead time of three weeks is used in planning. The total lead time is thus five weeks, which is quite high compared to competitors who offer two to three-week lead times. Another disadvantage in the current situation is a high Work-In-Process (WIP), leading to increased costs. Mainly for these reasons, the wish exists to lower the production lead time 50% of the current lead time, meaning lowering the total lead time used in planning to 2.5 weeks. We started our research by analysing the various components of the lead time, with special interest to the waiting times and actual lead times. We conclude that the waiting times are the biggest part of the lead time, generally more than 80% of the total lead time of five weeks. Especially the waiting until the first operation is performed is too high. After release, it takes on average more than five working days before the first operation is performed. Two remarks are needed here. Picking in the general warehouse is not counted in this measure and that normally takes two days. Second, in the SMD process, operations until the final processing station are not registered. Even with these remarks, the waiting time till the first operation is too high. With the use of a visual stream map and a problem bundle, we researched the causes of these waiting times. We notice that currently planning and the control of load is only done on a weekly basis, at the tactical planning level. At the operational level, load control is not used. Also, the current system is a push production system, meaning that orders are put in the system and expected to come out before the due date, without further control. For these reasons, we solve the problem of too high waiting times by proposing a new method for Production Planning and Control (PPC) on the operational level. We first research the literature for suggested method and did a qualitative analysis of the suggested method, before making a simulation model to do a quantitative analysis of the best PPC methods. During the literature research, we mainly looked into methods to reduce the lead time by limiting the WIP. As proven by Little’s law, the WIP is directly related to the waiting times. Limiting the WIP can be done by using a Production Planning and Control method on the operational level. During our research, we found three methods that according to the literature could work well at Company X. The first method is Workload Control (WLC), in which each station has a limit for the amount of work that is on the shop floor for that specific station. If the limit is reached and another order is available for processing, it may not be released to the shop floor. The effect of this method is that WIP is limited, reducing waiting times in the system. New orders are released on a daily basis and when a station starving, i.e. it has no orders left to process. The second method is Kanban. In this method, each station has a fixed number of cards and an order needs the card of the station it is moving to. In the case a card is not available, it may not progress to the next station and also does not release the card it is currently holding. Thus, when a station is full, it creates a chain reaction backwards, eventually to the point where new orders cannot be released due to a shortage of cards. The last method is Paired Overlapping Cells with Authorization (POLCA), which creates loops of two stations, which together get a single card. In order to allow processing, generally two cards are needed, one for the loop of the previous and current station, and one for the loop of the current and next station. Besides the PPC methods, we also researched some priority rules, for determining the order in which we release the orders and in which order the repair department should handle the orders brought in. Although these methods are recommended by the literature, it is unknown which method works best for the situation at Company X. We made a simulation model so that we can test the various methods and test their performance under the conditions of Company X. In this model we used order data of the second half year of 2018 and simulated the SMD and HMT departments based on their real-life properties. The best working method for Company X according to our simulation study is WLC. This method provides the best results with lowest lead times and is relatively insensitive to minor changes and variations. For the order interval, we suggest a 4-week period. A shorter interval leads to increased total lead time and a higher interval leads to higher individual processing times due to increased order size, which the system is less capable of handling efficiently. For the transition between SMD and HMT, we suggest to release the HMT order to picking, as soon as picking at the SMD department starts. This ensures the best alignment between the SMD part for the HMT process and the other HMT components coming from the warehouse. Regarding the dispatch rules, no significant difference was found between multiple options. Therefore, we advise for the rules that are easy to implement and understand. This means First Come, First Served with priority for continued SMD to HMT orders for order release and Earliest Due Date for the repair department. With these methods, it possible to reduce the lead for SMD to 4 working days, HMT to 14 working days and combined orders of these two to working 15 days. We also conducted a sensitivity analysis on this combination of methods. We found that the system is quite sensitive to incorrect estimations of processing times, if the mean is different between the estimated and real processing times. Variation between estimated and real processing times does not matter a lot, as long as the mean of both is equal. To implement these methods, a 4-5 month period is needed. In the beginning, acceptance needs to established under employees, such that they see the benefits of the methods. Afterwards, a new IT system needs to be developed in which the actual load of each station can be seen. The data for this system is available, although it is spread over multiple platforms, namely Baan and PFS. After implementation, the most important phase of evaluation and control starts. The methods need to be monitored continuously and adjusted where needed. Also the implementation process needs to evaluated, especially if the possibility for implementation at other departments exists. Further, we recommend to start tracking of processing times consistently, in order to increase the reliability of the data. At the moment this barely done, while WLC is quite sensitive for incorrect processing times. Finally, we recommend looking into the SMD Touch Up station. It no longer functions according to its original intent, as it is executed after the machine has finished instead of the same time, and it is meanwhile the biggest bottleneck in the SMD process. It could be a possibility to get rid of the separate station and merge it with the main HMT production station, which almost all products already pass anyway.
Item Type:Essay (Master)
Faculty:BMS: Behavioural, Management and Social Sciences
Subject:50 technical science in general, 85 business administration, organizational science
Programme:Industrial Engineering and Management MSc (60029)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page