This research is supported by Westinghouse Electronic Systems Group and the Maryland Industrial Partnerships (MIPS) Program.
Contents:
Production Management in the Westinghouse Feeders Shop
The feeder shop---the term "feeder"
denotes the logistical position of the shop;
it assembles radar chassis and other items
for the complete radar systems manufactured by Westinghouse ESG;
thus, it feeds the final assembly facility---of
Westinghouse Electronic Systems Group (ESG)
in Baltimore, Maryland, is a typical
example of a large make-to-order job shop.
over 10,000 different make items. The largest end items include up to 30
levels in the bill-of-materials, contain up to 1000 make items, and have
lead times that exceed 18 months. The major manufacturing processes
include machining, sheet metal stamping, plating, painting, and assembly.
These are performed by over 130 machines and workstations in an
86,000-square-foot facility.
While necessarily sophisticated, the production management procedures of Westinghouse ESG (Figure 1) are typical for a discrete-parts manufacturing firm (Vollmann, Berry, and Whybark, 1992). Given forecasts of arriving orders, production planning determines the appropriate levels of shop capacity and aggregate production to meet demand and minimize inventory and manufacturing costs. The master production schedule (MPS) is the disaggregated version of the production plan; it specifies the due dates of scheduled orders of individual end items. Program managers acting on behalf of the firm's customers create final assembly build schedules that are combined to form the master production schedule. The MPS is the input to the Material Requirement Planning (MRP) function. Using planned lead times for each component, MRP explodes the bills-of-materials of the end items to determine when to release purchase orders for raw materials and work orders for make items. MRP does not use capacity information while performing these calculations. This fact, coupled with the often optimistic delivery schedules set by the program managers, leads to uneven loading of the shop, overloading of the bottleneck resources and, as a consequence, high work-in-process inventory (WIP), long cycle times, and poor on-time delivery performance.
Due to the inability of MRP to handle this situation the shop uses a complementary procedure to manage the order release to the shop. The work orders that MRP has released collect in a pool of "approved" orders and wait in the pool until production control heuristically releases them based on work order due dates, program priority, and available capacity. The existing order release procedure attempts to balance the desire to keep excessive inventory off the shop floor against the need to meet due dates and the need to buffer work-in-process inventory to keep the machines busy and maintain throughput. The subjective and heuristic nature of the order release causes two problems: orders are released late and fail to meet their due dates, and the WIP may be unnecessary or insufficient. Inconsistent cycle times also contribute to the first problem: the shop has no accurate measure of when an order is behind schedule until it is past due. The second problem states that there may be too much WIP, or there may be too little. Due to the conservative nature of production scheduling, it is more likely that excessive WIP (safety stock) is present, but there is still no guarantee that this safety stock will keep all bottlenecks busy and thus maintain throughput.
Figure 1: Manufacturing Planning and Control System
Thus, the overall objective of this project is to develop an effective order release policy in order to minimize the average tardiness of work orders, to limit the amount of work-in-process inventory, and to maintain the throughput at high levels. In addition, proper order release will reduce the variability of the actual cycle times and lead to better production planning.
The impact of this study on the operation of the feeder area is expected to be substantial. The study will also yield both immediate quantifiable returns and long-term improvements in the overall production management practices of Westinghouse ESG. Because the order release policy complements the MRP system that is used by all manufacturing shops in the Baltimore facility and by other Maryland locations, the policy may be used by all facilities with similar problems.
Due to the significance of this project, Westinghouse ESG has provided seed funding to the University of Maryland CIM Lab. An exploratory phase was started in February 1995. In this phase we are concentrating on the machining area, identifying bottlenecks, and developing manual order release procedures for this area. Note that this work is part of a general reengineering effort within Westinghouse to improve production planning systems in Westinghouse ESG and financial and manufacturing management throughout the corporation.
Preliminary Results
The feeder shop of Westinghouse ESG
has been the focus of a previous MIPS project with the company
that improved the shop design by identifying group
technology cells, determining the layout of the machines in the cell,
finding a hybrid layout for the shop,
and determining an implementation plan for the cells that respects
budgetary constraints and maximizes the material handling benefits.
This work has been very successful, resulting in estimated annual
savings in material handling effort on the order of $300,000.
In the preliminary work on order release for the feeder shop, we have determined the bottlenecks in the shop, examined the current performance of the shop, and identified the current order release policies. Given the demand for products over a three-year period, the routings for each item, and the list of shop resources, we calculated the utilization of each resource in the shop. A number of machines in the MA (machining) area are among the most utilized ones in the shop.
In addition, we are building simulation models for use in evaluating order release policies. We are constructing these models in Simfactory Version 6.0, which allows us to simulate sophisticated sequencing and order release rules and to evaluate their effect upon the measures of throughput, cycle time, and on-time delivery. We are using previously developed software to convert Westinghouse routings to Simfactory models to facilitate this effort.
Implementation
We propose to implement
the order release policy as a series of automated steps that will support
the decisions of the Westinghouse production planning personnel.
The procedure is described in Figure 2.
The system inputs are provided by
the existing MRP and shop floor control systems.
The static MRP inputs include
the production routings and lead times for each make item and
the capacity of each bottleneck resource.
The dynamic MRP input is the list of approved orders that can be released.
The dynamic input from the shop floor control system is
the current workload of each machine in the area.
The system will use the approaches described above to identify, from
the list of approved orders, which orders should be released
at a given point in time.
We will begin with a manual implementation of the most promising order
release policies in order to test our results and to gain further
insight into the operational issues of the particular environment.
We will use a policy based upon a preliminary investigation and
modify it to suit the current needs of the feeder shop.
After refining the order release policy in order to address the feedback from the manual implementation and further analysis, we will begin developing and implementing software to automate the data collection and processing needed to perform the order release procedure. The PC-based software will include data gathering modules, preprocessing routines to form the desired inputs, a parameter-setting module for the scheduler to adjust the policy to current conditions, if necessary, and the order release module that suggests which work orders should be released and which others could be considered. The system may also include some capacity analysis functions that identify long-term bottlenecks.
The research plan is shown in the master schedule. The deliverables of this project are as follows: bottleneck-based simulation program of the feeder area, analysis of alternative order release procedure, and integrated software to include data collection, processing, and analysis functions.
Figure 2: Order Release System