5.1 Background of company X.
Company X is a global leader in the production of centrifuges for use in laboratories worldwide. It has been in existence for close to 100 years and over this period it has concentrated on building a variety of products to meet customers’ laboratory needs. It has also pioneered several technologies during this period. X is an ISO 9000 certified company and also meets several standards for companies dealing in centrifuges (IEC 1010-2-020, CE Mark, CSA). X’s supplier base has been approved on the basis of its past performance and the current operating systems. This has enabled the establishment of a vendor certification program that ensures delivery of quality parts to X. Company X is a wholly owned subsidiary of a larger scientific equipment manufacturer with annual revenues of about $ 400 million. The parent company handles some aspects of distribution and development for X. In addition to doing direct sales and using its parent company’s sales network, company X also distributes its products through Fisher Scientific International, the world’s leader in distributing scientific products. Fisher serves about 250 000 customers and boasted sales of $2.47 billion in 1999. Fisher uses a direct sales force and customer-service organization consisting over 2,600 technicians as well as a 2500 page catalog and a website that hosts a digital version of the catalog with real time information. These three distribution networks form the basis of company X’s forecasting function.
5.1.1 What do they make?
Company X’s primary product line is centrifuges. As of January 2000, it offered 36 different models broadly categorized into a table-top version and a floor based version. A typical centrifuge has about 200 different components (this value was reduced from 300 mainly due to design improvements) with some of these components being interchangeable between models. In addition to complete centrifuges, company X also sells centrifuge accessories such as rotors, buckets, adapters, heating jackets, tubes and carriers. Collectively, these accessories constitute 316 different parts with over one third of them being rotors. For each major piece of equipment, company X uses a serial number to monitor its motion through all the stages of the manufacturing cycle by using markings, labeling and accompanying documentation.
5.1.2 Who are their customers?
Company X has two categories of customers. The first is the US government which can be categorized as a single large customer that deals directly with company X and offers predictable forecasts. The second customer constitutes individuals and scientific institutions who place their orders with company X. Through its Fisher representatives, company X ‘s sales force also deals indirectly with customers. The prices for the items sold by company X range from less than $10 for some components to as much as $10 000 for a complete centrifuge. All orders are handled through the sales department which then creates the demand for the manufacturing department.
5.1.3 How long have they been using MRP and what did they have before?
Company X has had its manufacturing system set up for close to one century but it’s not until 1991 that they started running an MRP based system. Immediately before this, they used a card system to control their manufacturing process. These cards had different types of information such as lot purchasing details, the vendors involved in securing a part and the part number. These cards were stored in alphabetical order at a central location on the plant floor and they never traveled with the parts as they went through the various processes. There were three copies of each card with the first being maintained at the central location and the second being kept by a receiver. In addition to handling the cards, the receiver was also responsible for travelling with the raw material. The information generated in the system was usually run on computers and stored in the form of punched cards that were processed near the company premises.
5.1.4 Are they making any attempts to integrate with lean?
Company X has recently embarked on efforts to make its system lean. They already have a version of cellular manufacturing at the assembly stage of production although the cells do not operate strictly on the principles of lean manufacturing. Another area they have worked on is the reduction of inventory. This inventory has been halved from about 6 million to 3 million parts over the last 3 years. The company has the potential to convert its system from MRP to lean and there are several part families that would easily be used for a pilot study.
5.2 Description of Company X’s MRP II system based on one product line
The MRP system at company X was analyzed by studying the manufacturing processes for the base plate (part A) that goes into one of the table-top centrifuge models (model B). The model B centrifuge comes in twenty-two different specifications based on the operating voltage and the type of rotor used. As mentioned before, the centrifuge would has close to 200 different components including part A. When fully assembled it weighs about 11kg and has dimensions of about (30-60cm height, 35cm width and 40 cm depth).
5.2.1 Product types and sizes
Part A is made of aluminum and arrives at the shop floor in its cast form. It subsequently undergoes 5 different manufacturing processes before it is used in the final process of assembling model B. Part A moves in lot sizes of 125 pieces. A stack of these pieces can be seen in the attached photos in appendix 2. A planner work order accompanies each lot and every time an operation is completed, the machine operator signs off. The machines at the various workstations are arranged in a departmental format. Part A consequently has to travel a lot as it moves between workstations.
5.2.2 Value stream of part A and model B from forecast to shipping
The value stream for part A shows the various stages it goes through from the moment it is received from the vendor to the stage when it is assembled into the model B centrifuge. This information is presented in figure 6 which shows the value stream for part A as it goes through the different workstations on the shop floor. The information flow is also indicated.
Figure 6. Value Stream of part A of the Model B centrifuge
1. Purchase based on forecasts
The first step involves the purchase of the cast aluminum parts. These parts are obtained from the vendors by the buyers based on information received from the planners. The plans are developed based on the projected forecast of demand. At company X, the planners spend about 4hrs every 3 weeks doing long term plans and 4 hours a day re-planning and doing shorter term planning. Currently one planner is shared by several buyers but company X would like a system where each buyer is assigned their own planner. This relationship is summarized in the flow map shown in figure 7.
Figure 7. Relationship between Forecasts, the MPS and the MRP modules at company X.
The biggest problem facing company X right now is that the Forecast and MPS modules do not run the MRP module and hence the system is never in control. This problem is characteristic of push systems and particularly MRP as pointed out by Cochran in chapter 4. Company X has a ‘Dock-to-Stock’ time (the time interval between the arrival of the ordered part at the dock and its arrival at the first workstation for processing) of 1 day. All the inventory representing received stock is housed in one central section near the section of the shop floor that contains the various machines. There is usually a turnover of one month but some of the pieces have as much as a 6 month turnover. Order preparation by the buyer takes about 5 days. The sales team works with an average lead time of 14 days and therefore promise customers a delivery time of two weeks from the time they receive an order for a model B centrifuge. However, manufacturing works with a lead time of about 60 days which in reality reduces to 30-45 days. For the rest of this analysis I shall use the sales lead time of 14 days.
The first process involves turning part A on a lathe until it achieves the desired dimensions. The process uses an automated puma 10 Daewoo machine which is manned by a worker at all times. The entire lot of 125 parts takes up a total of 18.625 hours at the work station. This time includes a set up time of 3 hours and a run time of 0.125 hours per part.
The second process is a milling operation during which holes are drilled in part A and certain sections are milled to their finished quality. The process uses an automated Hitachi Seiko HC 500 machine which is also manned by a worker. The lot takes 55.5 hours to be processed which includes a 3 hour set up time and a run time of 0.42 hours per part.
Before part A proceeds to the next process, a small metallic bracket is fixed onto the milled part using a small portable rivet gun that is located on one of the worker’s tables. The time taken is negligible and is therefore ignored in this analysis.
This process uses a large industrial size Typhoon Proceco washer. Since it operates as a batch process, it is possible to handle more than 125 pieces at a go. This process has no set up time. Therefore the entire lot takes 2.4 hours, which roughly translates into a runtime of 0.0192 hours per part.
This process includes all the steps that have to be taken to prepare for painting. The run time is 0.06365 hours which translates into a processing time of 7.95625 hours.
Part A is spray painted by hand in special booths. Several booths lie adjacent to each other and the process just requires the lot to be wheeled to one of these booths. A worker manning the booth uses a spray gun to paint each part individually. There is no set up time for this process but the run time for each part is 0.0926 hours translating into a total processing time of 11.575 hours for the entire lot. Usually the parts are heated so as to reduce the drying time. When there is not enough capacity at company X, outside vendors are sometimes used to do the painting.
Once part A has been painted, it is delivered to the assembly area. There are 4 different assembly cells at company X. Each cell handles a different family of centrifuges and they are color coded as follows: yellow (large floor models), green (refrigerated and non- refrigerated models), red (large table top models) and blue (small non-refrigerated models). Model B is assembled in the blue cell. The assembly process is done in four steps. During all the stages, each centrifuge is accompanied by an inspection sheet that indicates any problems that are encountered during assembly. In the first step the base is assembled. This is the stage where part A is incorporated into the centrifuge. In refrigerated models, the refrigeration unit would be added next but this step is not necessary for model B. Next, the cabinet that forms the centrifuge housing is added followed by the transformer for voltage conversion. The final step involves testing the centrifuge for performance. Centrifuges that pass the test are packaged and sent to the shipping dock ready for delivery to the customer.
In case problems are encountered for a given part during assembly, the workers will try to resolve it themselves. If that fails, the part will be put aside and the manufacturing or design engineers will be called in to rectify the problem. This ensures uninterrupted production. The manufacturing engineer is also responsible for obtaining feedback from the workers in the cell and using that information to effect the necessary changes. The workers in the assembly cell are cross-functional and can consequently cover for a missing worker. However they remain tied to one assembly process in the cell at all times. It takes about one week to train a new worker. To facilitate this, a folder containing information about the assembly process is kept at the cell. Also within the cell is a computer terminal that is used to enter information into the MRP system. Typically, each worker will update the records every time they finish working on a part.
8. Shipping and Delivery
Parts are shipped and delivered on a daily basis. UPS makes deliveries in the morning and collects shipments in the afternoon. They pick up all the items that they find at the shipping dock and deliver them to the various locations based on the information provided by the sales office.
5.2.3 The MRP Process at company X
The MRP system at company X is designed to meet the demand for the finished product while minimizing inventory. It therefore has the role of determining the levels of intermediate good inventories needed to ensure that the finished product demand will be met. This premise is what has led to the design shown in figure 7. As mentioned before the sales forecasts and the MPS module at company X do not run the MRP module. Typically when new orders are received, they tend to use up all the forecast parts.
The MRP system at company X is based on the Glovia ERP system produced by Miracle Information Systems of the UK. The Glovia system incorporates both client/server technologies and object oriented standards. Company X has a modular system with financial, sales and operation modules. It runs on Windows and uses an Alpha Server system with an Oracle database. Company X made the upgrade to Windows in mid 1999. This is an example of how companies normally solve MRP problems through software improvements as discussed in Chapter 2. The operations module is further subdivided into three parts: Inventory Management, Master Production scheduling and Material Requirements Planning. This module integrates closely with the sales module and it provides data to the financial side. The following is a description of the key parts of the MRP system designed by Glovia showing how control is effected on the shop floor. This information is based on appendix 3 which shows the various interfaces of company X’s MRP system, a purchase order form, an exception report, the schedule of one of the cells and a work order.
Options and Location Table
There are two versions of this table, one for the MPS module (version i) and the other for the MRP module (version ii). Both versions indicate if there is inventory on hand.
The MPS module has a planning horizon of 365 days and a review time of (time bucket) of 7 days. An increase in stock is indicated as a replenishment while a decrease is noted as a requirement. Replenishment of the MPS is carried out using five documents; Purchase Order (PO), Computer Planned Order (CO), Work Order (WO), Forecast Planned Order (FPO) and Supplier (Sup). Requirements of the MPS are determined by four documents: Forecast (Fcst), Sales Order (SO), Forecast Planned Order (FPO) and Demand (Dem). Version i also indicates whether or not the forecast of demand has been used. It indicates when the MPS regeneration was done and indicates the person responsible.
The MRP module also has a planning horizon of 365 days but it has a time bucket of 14 days. The documents that determine replenishment of the MRP are identical to those of the MPS with the exception of the Computer Planned Order (CO). The documents for determining requirements are identical to those of the MPS.
MRP/MPS Planning Detail Inquiry
This table is used in netting so as to determine if there is sufficient material to meet the demand. It offsets the requirements from the replenishments to give the net number of items available. In addition it gives a reference number for the item(s) in question, the delivery date of the replenishments, the due date of the requirements and the quantity of each desired. The attached tables show the netting calculations for the model B centrifuge (version iii) and a component (part C) of the centrifuge, which is not the same as the aluminum base studied earlier (version iv).
Version iii uses the sales order and the forecasts to determine the requirements. Note that the net available stock for this case is always a negative value meaning that the shop floor is short of material. This is a characteristic of MRP that results form its effort minimize inventory such that material from the vendors is held back until the last possible moment.
Version iv uses the work order, the MPS and forecasts to determine requirements and purchase orders to do replenishments. Note that there is initially a net positive amount of stock as of the 10th of February but this quantity gradually declines to a negative value on the 7th of March as requirements are met. The first replenishment is not made for another 14 transactions until the 17th of March when an order for 135 pieces of part C are made. However the stock remains negative until the 6th of April when the second replenishment is made. The remaining appendices show copies of the purchase order form, the MRP exception report, the schedule for the red cell and a work order form. Note that the exception report is typically eighty pages long and the sheet shown is from page 44.
5.3 General observations
Company X does not pay attention to the capacity of the plant. Although a detailed capacity planning module that tells what the actual situation on the shop floor exists, it is not used. Instead they use an excel spreadsheet that is updated frequently to assign jobs to the various machines. This was pointed out in chapter 2 as one of the weaknesses of MRP system.
In spite of company X doing their planning a month ahead, it was noted that the mix of orders placed wasn’t usually met. However, their sales targets were often successfully met. This inaccuracy shows that the MRP system does a bad job of leveling the orders and this can be attributed to the large lot sizes and the long throughput time.
In summary, company X’s MRP system did not work very well because it had to deal with the randomness of the shop floor. To do this, it needed certain conditions to exist on the shop floor, which wasn’t necessarily the case at company X i.e.: A large backlog, excellent forecasts, vendors who didn’t miss quality targets, vendors who kept to schedule, existence of large inventories and short material lead times. These can easily be correlated to the various constraints highlighted using the decomposition in chapter 4.
Company X should try to establish a lean system on its shop floor. Based on the solutions suggested by Uday Karmakar in Chapter 2, company X could try implementing a ‘Tandem Push-Pull system’ or a ‘Requirements driven Kanban’ system. A proposed design that meets the requirements of a lean system is discussed later in the chapter. The infrastructure to enable this transition currently exists. A pilot cell can be based on a family of centrifuges for example those assembled at the blue cell.
The throughput time of part A is very high mostly as a result of lot delays and the long setup times. Based on the processing and setup times provided, the throughput time is 96.05625 hours (about 4 days). This does not take into account transport delay or machine downtime. The actual value adding time is less than an hour (0.72045hr). This could have been predicted by the analysis provided in chapter 4 using the Manufacturing System Design Decomposition. This information is shown in figure 5. This delay is clearly illustrated in figure 8 which compares the delay processing time for a run size of one and a run size of 125.
The performance of the MRP system at company X can be illustrated using the following real life data for a part A order. On February 10th, an order was opened for 136 bases. By the time the order was closed on April 6th, 132 bases had been completed and 4 remained. Starting April 14th, the parts started being used in the assembly process.
5.4 New and improved value stream map with linked cell system
A new and improved design for company X’s shop floor is presented in figure 9. In the lean design with linked cells, the plant floor is divided into three main processing areas. There is a set of cells that carry out the primary processing of part A i.e. milling and turning on a lathe. In this area, there are four almost identical sets of machines that contain either two mills and a lathe or two lathes and a mill. Before and after each of these machines, there exists a decoupler where the unprocessed / processed part is placed
Figure 8. Excel Spreadsheet showing the delay introduced into the system due to the large lot size (125 parts) compared to a lot size of 1 part.
in between cycles. When a part is placed on the decoupler located after one of the milling machines, the metal bracket that was described earlier in the chapter can be fixed as the part waits on the decoupler before it’s next process. The cell cycle time for this area will be determined by the slowest machine (or the longest process to be done on part A which in this case would be the milling operation). Since all the machines would be running in parallel, a complete part would be produced from the cell after each cycle time.
Figure 9. New and improved value stream map with linked cell system
The second area constitutes the washing process and the painting process. Sincere there is only one washer available and it has a large capacity, the parts can be brought to it in given batch sizes that ensure downstream processes are not kept waiting. Given the small processing time for the washer, this process can easily be coordinated to match the cycle time in the previous and subsequent cells. The part from the previous cell is carried from the last decoupler and taken to the decoupler next to the washer where other parts from the other cells on the shop floor are also assembled so as to build up volume. An ideal washer design would use a conveyer belt such that the part never has to wait on other parts in order to get processed.
The third area is the assembly area. This area would have all the four cells (blue, red, green and yellow). Parts coming from the washer would be placed on the appropriate decoupler in each of these cells. As soon as a part is pulled out of the cell to shipping, the part coming in from the washer would go into the cell and be assembled into the centrifuge. Company x currently has these cells set up and the only challenge would be to design their location so as to minimize transport time to the shipping dock.
The information system would work as follows. A customer would make an order to the sales office which would then coordinate with the manufacturing department to provide the right mix and volume of parts to meet the demand of the various customers for a given day. The customer can give this information electronically through the web based system discussed earlier in the chapter or using regular purchase order forms sent directly to company X. The ‘heijunka’ is a system that carries out the role of distributing demand in terms of volume and mix. Information from the ‘heijunka’ is then communicated to the four assembly cells using a signal kanban which tells the cells what to produce. The four assembly cells pass this information backwards to the washer and painting area and subsequently to the four primary processing cells. In this way production is driven by the customer demand. When the completed part is pulled out of the assembly cell and is sent to shipping, it is picked up daily whenever the delivery services comes by to collect parts to be delivered to the customer. The parts delivered from the vendor are also delivered daily to company X based on parts that have been pulled out of the four primary processing cells. This arrangement ensures that sales targets are met both in terms of mix and volume.
The customer demand can be used to determine the takt time which is simply the total customer demand divided by the available manufacturing time. The goal is to ensure that each of the cells on the shop floor has a cell cycle time that is equal to or less than the takt time. In this way, company X’s shop floor will be designed to keep pace with the customer demand. The arrangement in cells has the added advantage that various parts for any of the four centrifuges can be processed at any of the four primary processing cells provided the number of milling or turning operations is known. For processes requiring more than three operations, two cells can be combined to provide more machines for the additional operations. The speed of the cells can be varied by adding or reducing the number of workers operating the cell since they would each be cross-trained to handle any of the machines in their cell. Note that the workers are not tied to a given machine.