Production to order. Models and rules for production planning (Q1202108)

This monograph deals with different control frameworks that can be set up to deal with Production to Order situations. Such situations exist in manufacturing firms that produce a wide variety of products and have a production process with one capacity bottle-neck, with large setup times. With uncertain demand for products, and not being able to carry substantial stock on hand, a due date is set for every order, in agreement with the customer. It then becomes important to properly schedule production so that the due dates are met as close as possible without incurring too many setups. The author gives examples of such (production to order) situations existing in the production of steel pipes, as well as in the chemical industry. A variation of the problem known as the assembly to order exists in truck industries and in telecommunications. The objective is to find good control rules for the production planning, and for establishing lead times, that satisfy both the firm and the client. The performance of these control rules are expressed in financial terms and consist of holding costs, production costs, setup costs, and costs due to changes in the capacity . These costs may be real costs, and also can serve as management variables which are used to indicate a level of satisfaction. To develop control rules for production planning and lead times and order acceptance, the monograph first analyzes a fixed production cycle, wherein, the sequence in which the types of products will be produced, as well as the available capacity for the production of a type is fixed. The production cycle is repeated over and over gain. Thus, every type of product ocurs at least once in the sequence, but the sequence may also contain some types more than once, interrupted by other types. Orders of a type will be produced according to the FCSF rule. Two different types of gating service (in which, we stop working on orders from a certain type at the end of a production interval) are considered. Under the extended gating service, work on all orders of a type can be done during the production interval of this particular type. Since, orders arriving during the production interval may be served immediately, this type of service yields shorter delivery times. However, this type of service is very expensive and some times impossible, since the sequence of orders should be known at the decision moment. Under the normal gating service, only orders that arrived before starting the setup of a particular type are worked on. Under both situations, expressions for expected delivery times \((E)\) are derived. Also, approximations to \(E\) to make calculations simpler are provided. Under the normal gating service, analysis is done with strict and flexible capacities. However, to obtain good results with a fixed production cycle, it is necessary for the demand to be stationary and that customers are able to provide good estimates of their future demands. The monograph then focuses on non-cyclic production methods. The concentration is on periodic review models, with usually a time- independent stochastic demand. First, for situations with no capacity restrictions, a Markov Decision Problem (MDP) based model is developed where the state vector is used to describe the required deliveries for the various periods and the action space represents the production- inventory decisions. After deriving an expression for an optimal policy for situations with smaller number of state spaces, heuristic approaches such as Silver-Meal and Wagner-Whitin are considered for solving larger problems. In these situations, the decision to produce depends on the number of orders for a type in the first period and not on the required deliveries for future periods. Secondly, for a given demand distribution the number of orders produced are always the same. Based on these facts, the author also introduces a new rule: the \((x,T)\)-rule. According to this rule, no production takes place if the number of late orders is less than \(x\); otherwise all orders are produced and delivered according to their due date, within \(T\) periods. The \(x\) and \(T\) values are decision variables. Comparing the performance of the heuristics, \((x,T)\)-rule with the optimal solution, one sees that the difference in average costs is very small, and the \((x,T)\)-rule is easier to use. Also, a dynamic programming decision rule for firm-initiated lead times is derived along with average profit. Next, the monograph extends the \((x,T)\)-rule to situations where several products are produced under strict capacity constraints. The order state is reduced through a weighting procedure based on the number of penalty points for a type, which captures the residual lead times for an order. Thus, the decisions on production are based on a largely reduced state space. This rule is compared to a two-step rule, wherein the \((x,T)\)-rule is split in two parts: a part for the individual types and a part on the aggregate level; and a semi-fixed cycle rule. The extended \((x,T)\)-rule performs well compared to the complex rules. In the following chapter, the analysis is directed to determining expected delivery times for orders of a particular type and priority. Thus, by creating a preliminary plan based on the expected number of penalty points for the various types in future period, the extended \((x,T)\)-rule is used for proposing lead times for newly arrived orders. Both, periodic as well as continuous review situations are explored. Finally, the \((x,T)\)-rule is applied to complex situations, that are combinations of simple situations such as: several product types with different demand rates, setups between types, orders with different priorities, backlogging, capacity constraints and overtime possibilities. Thus, the monograph does a comprehensive job of developing control rules for production planning for production to order situations. The performance of these rules are studied with the help of the demand process, lead time acceptance, lead time accuracy and the use of capacity.