Chapter 4 - MRP II as a tool for shop floor control

4.1 How does MRP II attempt to achieve shop floor control?

MRP II is a push system with a type of feedback loop incorporated into its structure. Cochran consequently models MRP II as an open loop control system with a set of inputs and outputs connected by a transfer function (the MRP procedure) as shown in figure 2.

Note that the inclusion of the feedback loop in the model makes it appear to be a closed loop system. The feedback loop represents the machine counts that are taken at certain predefined times. Cochran points out that the feedback is independent of the manufacturing system’s operation since the sampling rate is too infrequent or too late. Hence, unlike a true closed loop system, MRP II doesn’t perform according to the plan immediately after the plan has been released. This happens because the state of the system cannot be controlled due to the lack of feedback.

INPUTS

MRP PROCEDURE

OUTPUTS

Figure 2. Modified open loop feedback control loop for MRP II shop floor control

Spearman and Hoff identify two dimensions that characterize shop floor control in MRP systems. First, MRP systems have to determine the appropriate production quantities for finished products requested through purchase orders and their component parts requested as jobs. Secondly, they need to establish production timing that will enable orders to be met by their due dates. This results into time being broken into intervals called buckets which range from a day to one week. The forecast demand is subsequently broken into discrete chunks based on these time buckets. Based on Cochran’s open loop control system model, Spearman and Hoff identify three groups of elements in MRP: inputs, the MRP procedure and outputs. The interaction of these three elements is what facilitates the control of the shop floor in MRP II. The following discussion defines each of these three elements and discusses how they interact to bring about control. This model also incorporates the hierarchy introduced in chapter 2 and spans the categories of intermediate-range planning and short term control.

4.1.1 MRP Input

Three items constitute input into the MRP control system: the forecast of demand for end items, the associated Bill of Materials (BOM) and the current inventory status. This information is obtained from 3 sets of documents generated by the MRP system:

- Item Master File – In its basic form, this document contains a description of the part being manufactured, its BOM information, its lot sizing information and the planning lead time for the part. The item master file is organized by part number.

- Master Production Schedule – The MPS was discussed earlier in chapter 2. It contains the part number, the quantity needed and the due date for each purchase order. The MPS uses the part number to link the Item Master File with records where other processing information is located.

- Inventory Status File – This document provides information about the inventory status. This information helps to determine the quantity of demand that is met by on hand inventory and scheduled receipts. On hand inventory contains information describing a part, the location of the part and the number of parts that are at hand. It is stored by part number. Scheduled receipts contain the part number, the current quantity, the desired quantity and the due date. It stored by job number.

4.1.2 MRP Procedure

Using the input discussed above, MRP goes through five steps for each level of the bill of material (hence covering both dependent and independent demand) starting with end items. The procedure is iterative and is repeated until the entire BOM for a given part has been analyzed. The procedure is conducted as follows:

1. Netting (Coverage Analysis) – determines the net demand that cannot be met by scheduled receipts and on hand inventory. The two quantities are subtracted from the gross requirements identified by the MPS or by previous MRP operations.

2. Lot sizing - determines how jobs are sized so as to balance the conflicting need of minimizing inventory by using smaller lots and that of increasing capacity by using larger lots to avoid frequent setups. The lot size provides ideal production quantities (jobs) to satisfy the net requirements. Different lot sizing rules exist and these include lot-for-lot, fixed Order Period and Period Order Quantity. ( Toomey 1996 )

3. Time Phasing – determines the lead time as an attribute of the part and the job. The job’s start time can then be calculated by subtracting the lead time from the due date. Note that the status of the shop floor is not taken into account during this step of the procedure.

4. BOM Explosion – determine the gross requirements for the next level of the BOM using the start times and the lot sizes. This information is used to carry out netting during the next iterative step

5. Iteration – The entire procedure is repeated for a new level of the BOM.

4.1.3 MRP Output

Three items are produced as outputs of the MRP control system as discussed below:

- Planned Order Release – This document contains the part number, the number of units needed and the due date for the job. The planned order release eventually becomes the jobs that are processed on the shop floor.

- Change notices – These documents exist in two forms and are used to indicate modifications of existing jobs, their due dates and priorities. The first form is used in expediting orders (making their due date earlier) while the other is used in deferring orders (make their due date later).

- Exception reports – These documents are used to notify users of MRP that there are discrepancies between what is expected and what actually transpires. Such differences would include job count differences, inventory discrepancies and defective parts.

4.2 What limitations and constraints are faced in using MRP II for shop floor control?

The limitations and constraints facing MRP II can be analyzed by contrasting its attributes to those of the ideal system identified in chapter 3. The Production System Design Laboratory at MIT uses a similar approach in studying different types of manufacturing systems. They have developed a diagnostic tool called the Manufacturing System Design Decomposition ( appendix 1) which identifies the functional requirements (FR) and design parameters (DP) of a manufacturing system that is designed to maximize the long term return on investments. The decomposition provides a breakdown of the functional requirements and the corresponding design parameters for different levels of a manufacturing system. The paths of the decomposition that relate to MRP II are highlighted in grey in the attached decomposition. The following discussion is broken down into seven subtopics and it highlights the limitations and constraints identified for MRP II using the decomposition. The first four are aimed at maximizing the sales revenue while the last three are aimed at decreasing the manufacturing costs.

1. Quality

An ideal control system will ensure that products are manufactured to within target design specifications. Before this can be achieved there must be an ability to assign causes of variation. However, owing to the delays associated with lot sizes and lead time, it takes a while before defects arising form such variations are identified in MRP. By this time, it is fairly hard to determine at which stage of the manufacturing process the defects were introduced and make the necessary corrections. This situation is true both for defects that arise as a result of the machine and those that may have been caused accidentally by the operator. In addition, the lack of control over upstream processes means that the downstream operator has to make do with defective parts until upper level management intervenes. Due to push nature, MRP systems do not lay a great emphasis on supplier quality programs and instead use a reactionary approach when they receive defective parts from their vendors.

Another weakness of the MRP II system is its failure to reduce the variation in the process output. As discussed in chapter 2, it is difficult to determine the source of problems in an MRP II system when the system gets loaded beyond its capacity. This occurs because of the systems inability to convert common causes into assignable causes. In addition, MRP II does not deal well with variation when this occurs. Professor Cochran points out that an MRP II system will usually oscillate out of control when a disturbance is introduced as opposed to pull systems which have self correcting capabilities (Cochran 1994, 226). This behavior is exhibited in the form system nervousness as identified in chapter 2 whereby small changes in the planned order releases is caused by small changes in the Master Production Schedule.

2. Identifying and resolving Problems

Here the goal is to ensure that products are delivered on time to the end customer. One way of achieving this requirement is by ensuring a quick response to production disruptions. MRP fails to respond quickly in three particular respects. There is a time lag between the occurrence of a disruption and its identification by the operator. This is a result of the infrequent counts done on the machines during production runs. Secondly, MRP II has a fairly complex material flow. Typically, parts move to different locations of the shop floor as they are transported from machine to machine and this makes it fairly hard to identify disruptions where they occur.

The third constraint is a consequence of the first two, namely, the feedback provided by MRP II is not context sensitive and is therefore not of much use.

3. Predictable output

A second way of ensuring that products are delivered on time is by minimizing the disruptions that occur to the system. This calls for an information system that is reliable and provides the relevant production information when needed. Unfortunately, the demand forecasts made by MRP II’s long term planning module are rarely accurate. Often, production of rush orders may have to be made at short notice and causing disruption when orders have to be expedited or deferred.

In cases where workers are tied to machines in MRP II, disruptions are likely to occur anytime an allowance has to be made for the worker. This problem could easily be solved by having cross trained workers and a system design that enables workers from one station to co-ordinate two machines at one go. The severity of the disruption is a usually a function of how quickly a replacement worker can be trained or the amount of overtime hours that can be used to make up for lost time.

Other disruptions will occur if there are problems with the delivery of parts by material handlers. Since down stream workstations have no control over how parts are delivered, there may be timing problems from the time a part is finished until the time a new one starts being processed. Unlike a true pull system, no standard WIP is maintained between workstations. Usually this situation may call for large inventories to be maintained at the workstation to ensure that the machine never stays idle.

4. Delay Reduction

The ideal control system should also be able to ensure that the throughput time is less than or equal to the customer expected lead time. MRP II however, does not make an attempt to accomplish this objective. Instead it uses the lead time as a buffer against the various delays imposed on the system. These delays arise in four different ways and leads to the accumulation of inventory on the plant floor.

There is lot delay arising from the relatively large lot sizes typical in MRP II. All the parts in one lot must be processed at one workstation before they can move on to a new process and this occasionally leads to periods when downstream machines are not being used as they wait for all the parts in a previous process to be completed.

MRP faces process delay due to parts piling up behind bottleneck processes. The effect of this is that the speed of all downstream processes is limited to the pace set by the bottleneck process. MRP II tries to overcome this by using various job dispatching rules and ensuring that the bottleneck machine is always kept busy. By incorporating a pull mechanism, this problem can be overcome by defining a takt time (the time characterizing the customer demand and calculated by dividing the total customer demand by the total available machine time). All the machine times would consequently be designed to be less than or equal to takt time.

There is run size delay due to the large number of parts of the same type that have to be processed before changeovers can be done. The changeovers are necessary in order to meet the desired quantity and mix during a demand interval. These large run sizes are typically aimed at minimizing the number of setups and material changeovers that must be done at one demand interval as this tend to take a lot of time. This problem can be overcome by designing all the machines to have quick (less than one minute) setup and material changeover times.

Finally, transportation delay occurs in MRP II due to the departmental arrangement of different machines on the shop floor. As a result, parts have to travel great distances across the shop floor as they move from one process to another. To overcome this shortcoming, the shop floor should have a material flow oriented layout design.

5. Direct Labor

One way of minimizing manufacturing costs is by reducing the waste caused by unutilized labor. With MRP II this type of waste is observed in three different instances involving the operator. Since the operator is tied to a particular machine, he / she has to wait on the machine until it gets its job done. This time could be utilized more effectively especially if the machine was automated and designed to have minimal failure. The operator could then leave the machine running and attend to other tasks elsewhere. Other operators may also tie up operators further downstream especially if they are inefficient or careless in doing their work. In MRP this coupling is built into the system since the operators are not given the responsibility of managing the lead time. They also lack ownership over the parts or family of parts they make owing to the departmental nature of the shop floor. The third instance of wasted labor time is caused by the wasted motion of the operators. A shop floor controlled by MRP as well as the machines used with it are usually not designed with the operator’s activities in mind. Consequently operators may have to walk long distances or repeat cumbersome routines as they work resulting in inefficiency. For example, in the case study presented in chapter 5, figure 7 (see appendix 2) shows the CNC lathe used at company X. In the foreground is a crate containing unprocessed parts. To load the parts onto the lathe, the worker has to walk back and forth and bend over to pick the heavy piece from the crate. An alternative design would have all the machines that process the part close to each other and between each machine would be a decoupler (conveniently designed so that the worker can load and unload it easily) The decoupler would hold parts that were not being processed. The lathe could also be designed for quick set up using SMED (Single Minute Exchange Dies) techniques. All these would simplify the worker’s tasks significantly.

6. Indirect Labor

In Chapter 3, one of the considerations that was identified for an ideal shop floor design was the span of control. Managers who are usually not directly involved in the actual production work nevertheless need to ensure work on the shop floor is executed smoothly. MRP II fails in this respect because information is designed to flow top down. Feedback from the operators is rarely utilized in making improvements to the plant’s performance. Consequently, a lot of the manager’s time is spent handling crises that arise whenever the system goes out of control. MRP II also wastes indirect labor because of the large human resources it requires to schedule the system. Often, the elaborately arranged plans end up not being used when the production system fails to keep up with the plans made for it.

7. Facilities cost

The computing infrastructure necessary to keep MRP systems running makes them to be fairly expensive. In addition to this, the departmental layout of machines causes them to use up a lot of space on the shop floor. If the machines are designed with the manufacturing process in mind and are also arranged in cells based on part families or individual parts, much greater efficiency can be achieved in using the space.

4.3 How do MRP hybrid systems overcome these limitations and constraints?

Various solutions were proposed in Chapter 2 for solving the problems that plague MRP. One of this solutions was the use of hybrid systems (Karmakar 1989 8). In all the three hybrid systems that were proposed, MRP II assumed the role of making general guidelines that were subsequently used to achieve smooth running in the long run. A Kanban based system was then used to handle the details of daily production. Kanban is the operation control system of JIT production. Benton points out that Kanban control when used with a JIT based system is designed to minimize the work-in-process inventories by eliminating or reducing discrete batches. He also highlights conditions proposed by Monden and based on the Toyota Production System that are necessary for the Kanban controlled system to succeed. They include: smoothed production; job standardization; reduction of set-up time; improvement of activities; design of machine layout; automation of processes taking into account the human touch (Monden 1983). Benton also highlights 4 reasons why Kanban provides a superior control mechanism: it has less complexity, the feedback is faster and it has a reduced production lead time. Production lead time refers to the duration of time allotted for the production of a part on a given line or routing (Spearmann and Hopp 1996, 224)

The MRP hybrid systems use an approach analogous to that of JIT by leaving MRP to handle the planning aspects of production while Kanban concentrates on control. Revisiting the three MRP hybrid versions introduced in Chapter 2, the complementary strengths of MRP II and Kanban can be identified as shown below:

JIT-MRP - The work is released by a pull mechanism thereby eliminating inventory. The system is designed to meet an overall daily or weekly demand instead of individual orders. To determine the inventory levels, a ‘back-flush’ is done. A backflush is an MRP technique that involves subtraction to allow for production that has taken place. Since they system does not keep track of individual orders, work is designed to flow along predictable paths and leave at predictable intervals. This arrangement is ideal for flow systems since it now incorporates flexibility that enables a different mix of products to be made with very quick turnarounds (Karmkar 1989, 9). JIT-MRP is shown in figure 3.

Figure 3. JIT-MRP

Tandem Push Pull - In this hybrid system, the purchase planning lead times are long and are therefore handled by MRP. However, the build routines are based on Kanban. Consequently, the assembly is run on pull and is characterized by great flexibility and short cycles. Whenever the floor’s schedule changes, the MRP databases are updated to reflect this (Karmkar 1989, 9). Tandem Push Pull is shown in figure 4.

Fig. 4 How MRP and Kanban relate to Tandem Push Pull

Requirement Driven Kanban - In this hybrid system, the entire shop floor is run on a cellular arrangement. It can therefore meet the highly variable and fast schedules demanded by parts with an unstable volume and mix. MRP is suitable for predicting the demand and therefore determining the work to be processed in the various cells. Due to the cellular arrangement that initiates production by pull, the MRP has no order releases and therefore doesn’t have to monitor the inventory level in the cell or match the demand with the available inventories. (Karmakar 1989, 9) Requirement Driven Kanban is shown in figure 5.

Fig. 5 How MRP and Kanban relate to requirement driven Kanban