A Pilot Production Planner and Scheduler System
The last chapter of the thesis focuses on the practical applicability of our theoretical results described in the foregoing. It presents how the aggregate planner and the constraint-based scheduler can be adjusted to the needs of real-life industries and fit together into an integrated production planner and scheduler system.
Although our system in general addresses make-to-order project-oriented manu-facturing systems, the problems were specified and the applied models were worked out with the cooperation of one specific industrial partner. Real-life test problem instances also originate from this company, a Hungarian plant of a multinational enterprise that manufactures complex mechanical products for the energy industry.
Beyond basic research, the scope of this project also covered the demonstration of the industrial applicability of the theoretical results through a software proto-type. The developed integrated production planner and scheduler system was named Proterv-II (Projekt tervez˝o – project planner, in Hungarian). Since it is a pilot system for experimental and demonstration purposes, it lacks many functionalities substantial for industrial software, e.g., advanced reporting, hand-tailoring and trou-bleshooting options. Nevertheless, it operated on the data directly received from the company’s enterprise information system, and its algorithms were successfully tested on these real-life problem instances. At the time of writing this thesis, our industrial partner shows interest towards the system, but it is still an open issue whether an extended version of Proterv-II will find application at that factory.
This chapter is organized as follows. In Sect. 4.1, we outline the current produc-tion environment at the targeted field of our applicaproduc-tion. Afterwards, in Sect. 4.2 we define the goals of the production planner and scheduler system. Then, in Sect. 4.3
77
78 4.1 Production Environment at the Target Enterprise
we briefly overview the tools we used for the development of Proterv-II, as well as the components of the factory’s information systems that we could rely on as data sources. Afterwards, we discuss in detail the models and algorithms applied in the production planner (Sect. 4.4) and the production scheduler (Sect. 4.5) sub-systems.
Finally, in Sect. 4.6, we investigate how the results of our mathematical models can bear up under the uncertainties and data inaccuracy or unavailability of a realistic production environment.
4.1 Production Environment at the Target Enterprise
The target enterprise manufactures mechanical products of high value for the energy industry. Its entire production addresses the fulfillment of customer orders, i.e., the factory is a typical make-to-order production environment. An order always refers to a single product, which belongs to one of the four product families. Each family contains product variants with the same structure but different parameters. Although products from different families significantly differ in their production technology, the problem cannot be decomposed, because the sets of resources used for producing them are not disjoint. At present, the overall number of different products is ca. 40, but this number may grow in the future.
Figure 4.1: A product of the enterprise.
Themaster production schedule specifies atime window, i.e., a release date (earli-est start time) and a deadline (lat(earli-est finishing time) for fulfilling each accepted order on the medium-term horizon. The length of this horizon ranges from three to six months, and the time unit is one week. Since raw material arrival is sometimes er-ratic, the enterprise endeavors to acquire all the raw material of a product before the release date of the project. On the other hand, deadline observance is an absolute must, even for unpredicted orders.
Each product is made in bundles called sets. The size of the set is fixed for each
product since the complete set will be built into the same larger unit. Members of a set are indistinguishable, and the production of a set is not divisible by forming smaller groups.
As usual, the assembly structure of a product is given by its bill of materials (BOM). The BOM is tree-structured. It is rooted at the end product, while its leaves stand for the raw materials that are purchased from outside of the factory.
Raw materials and intermediate products are all dedicated to specific projects, even if some of them may be technically equivalent. Consequently, there is no need to perform material planning.
To each intermediate product in the BOM tree, there is a sometimes lengthy sequence ofoperations assigned. This operation sequence is specified in theroutings. A typical project consists of 20 to 500 discrete manufacturing operations altogether, with operation processing times in the range of 0.5 to a maximum of 120 hours.
Subsequent operations cannot overlap in time, i.e., work must be finished on every member of the set before it can proceed to the machine where the next operation takes place. Most operations arenon-preemptive but breakable, i.e., the workpieces cannot be unmounted from and re-mounted to the machines but after completing an operation. On the other hand, the on-going operations can be interrupted, e.g., during the weekends, and continued later on without any extra cost. However, some operations, such as heat treatment, must not be broken due to technological reasons.
Some of the operations are preceded by a fixed-durationsetup and transportation.
Each operation employs a single machine for a given length of time. During this time, no other operation can be executed on the same machine tool. Amongst these machines, there are machining and welding centers, assembly and inspection stations, and other highly specialized machines as well. Machines with identical capabilities are organized into homogeneous machine groups. Currently, there are ca. 80 unique machines and 10 more homogeneous groups of machines with a capacity of between 2 and 8. However, the number of machines is growing as the facility is extended continuously.
Operations also need the attendance of a worker from a given group. Currently, each of the ca. 200 individual workers belongs to one of the 10 different worker groups, like welders, NC machinists, or inspectors. We consider workers within the same group to be of identical skills. When a sensitive (e.g., welding) operation requires proficient workforce, the problem of selecting of the appropriate individual from the worker group is left to the shop foreman.