Knowledge Based Production Planning and Scheduling

One of the topics of the EUREKA project PROTOS (Prolog Tools for Building Expert Systems) [Appelrath 87] is the development of a Prolog-based system that exploits heuristic knowledge for a concrete scheduling application from the chemical industry. The term scheduling designates the problem of chronological allocation of a given set of production orders to the apparatuses available taking into account certain secondary conditions and goals. By the approach pursued in PROTOS Prolog in general and special Prolog tools for the solution of specific production planning and control (PPC) problems are evaluated. Approaches to the scheduling problem suggested so far are mainly based on the conventional operations research methods [Bedworth 87]; first of all they have two shortcomings: High runtime complexity of the algorithms. This problem is due to the combinatorial complexity of the problem domain when practical quantities are involved and when at the same time an optimal solution is demanded. Low flexibility. There are great difficulties to integrate actual, task-specific secondary conditions into already existing, inflexibly implemented procedures. Such conditions are, for example, the knowledge of the production manager about the reasonable sequence of production operations or problem parameters changing at short notice (e.g. the drop-out of resources). Some other approaches try to use knowledge based techniques for scheduling (Fox 84, Kusiak 86, Kusiak 88, Tate 85), but most of the systems lack a user-interface. The approach pursued in PROTOS consists of two parts: 1. Support of the planning activities by a user interface providing both, the relevant information for the planning algorithm, and a 'work space' for manual planning. 2. Knowledge based scheduling: the heuristic knowledge of the production planner is used in order to obtain as good a plan as possible and to enable appropriate reactions to unexpected events. In chapter 1 of this paper the application problem in PROTOS is described. Chapter 2 contains some remarks on the modelling of the problem. The problem solving approach is introduced in chapter 3. In chapter 4 some comments on the first results are given. 1. Scheduling: The PROTOS Application Problem The main tasks of scheduling as a subsection of PPC are the generation and adjustment of a production schedule [Fox 87, Kusiak 88]. A schedule is the chronological allocation of apparatuses to production steps which must be executed in order to produce a product requiring certain conditions to be satisfied. The concrete application considered is scheduling in the chemical industry. The available initial data are a master schedule: It contains a list of orders to be executed for the manufacturing of certain products along with information about amount, start and due dates of the production. Figure 1 shows an example for an extract of a master schedule. Order Product Amount Start date Due date Priority 9S1 6 43200 03 07 89 07 08 89 1 9S1 8 36540 20 02 89 29 03 89 1 9S1 9 50400 02 10 89 05 12 89 1 9S1 10 53600 01 06 89 11 07 89 1 Fig. 1: Master schedule