A hierarchical production plan for a make-to-order steel fabrication plant

Journal: Production Planning & Control: The Management of Operations Authors: Brian D. Neureuthe, George G. Polak & Nada R. Sanders

1. Introduction
1.1 Forecasting, aggregate planning, disaggregate planning and sequencing are the principal components of a hierarchical production plan. The hierarchical production plan (HPP) approach and its benefits are well-established in the literature by Bitran and Hax (1977), Leong et al. (1989), Bowers and Jarvis (1992), Qui et al. (2001), and others.. Recent research has been conducted on production planning in hybrid make-to-stock and make-to-order environments. 1.2 In the current paper, a three-tiered hierarchical production plan (HPP) for a strictly make-to-order steel
fabrication plant with the objective of developing a production plan and master schedule for a set of product archetypes is implemented. Data are collected from an actual steel fabrication plant located in the Midwestern section of the US. An aggregate linear programming model, a non-linear disaggregate model and a master production schedule comprise the respective tiers.

1.3 The first tier consists of a linear programming model for an aggregate production plan that minimizes total costs. The output from the first-tier linear program, along with a monthly forecast for each product, becomes the input for the second tier, a nonlinear model, in which each product is assigned to a specific process plan so that set-up costs are minimized. The third tier is a master production schedule designed to sequence the processing of the products, grouped by archetype, on the machines throughout the factory. The goal for this third tier is to minimize makespan, and therefore maximize throughput. 1.4 General background of production facility The steel fabrication facility used to model the HPP is a century-old plant located in the industrial Midwest.Fourteen machines are used to fabricate thousands of different end products, from which 12 product archetypes have been identified (see Rogers et al. 1991).Each product is classified by class, as industrial or commercial; by type, as angle, beam or plate; and by weight, as light or heavy. Production sequences, set-up times and processing times vary among these archetypes. The plant operates on a five-day workweek with one eight-hour shift per day. Overtime can be scheduled on Saturday exclusively. Management follows a strictly make-to-order environment, scheduling production only when an order is received. This policy obviates the need to prioritize between make-to-stock and make-to-order products.

2. Application of Disaggregate plan
2.1 Aggregate Plan The aggregate linear programming model seeks to minimize total costs, consisting of inventory holding, regular time and overtime, and hiring and firing (Bowers and Jarvis 1992). The decision variables are the amounts of each product class to be produced, the inventory at the end of each month, the total number of workers employed in a given month, regular and overtime labour in hours and the number of hires and fires in a given month. The constraints enforce material balance of demand (2), total productive capacity (3), demand for labour (4), the physical storage limit (5), regular (6) and overtime (7) labour time availability, workforce smoothing (8), hours and workforce level (9), and workforce limits (10). LINDO software was used to solve the aggregate

which assigns production to the appropriate lines for each month. while a flange-cutting burner can take up to an hour or more. and yielded an objective value of $545. i. see Appendix B) with weight (light or heavy) now becoming relevant. Set-up costs therefore reflect labour and opportunity costs. Heavy materials require dedicated cutting and punching machines. The months of June–October required overtime due to the production capacity available during regular time 2.e.231. First. no matter how small the order Decision Variables .linear program (ALP) to optimality. For example. (1989) state that the key to any disaggregate model is to ensure that the production quantities determined in this model agree with those dictated by the aggregate model. the flange-cutting burner and cold saw for cutting and the 100-ton and greater punch presses for punching. there is no lower bound on production. This is the formal link that is required between the aggregate and disaggregate models. Two special features of the make-to-order environment under consideration should be noted. Nine distinct process lines are needed for the 12 product archetypes (certain products share common process plans. i.e. These specialized machines have considerable longer setup times and require more manual labour than those used for the light products.. The optimal solution maintained the workforce at the lower limit of 25 throughout the 12-month horizon.2 The disaggregate pan The optimal solution to the ALP becomes input into the disaggregate model. items can be produced.25. an abrasive cut saw or a band saw requires only 15 minutes for set-up. Leong et al.

4 The scheduling algorithm Scheduling problems are among the most difficult in combinatorial optimization: even the one-machine scheduling problem with release and due dates is ‘hard’ (i. 2.e. or sequencing. The authors constructed a scheduling program in C++ that was implemented on a Sun workstation running UNIX at Wright State University . see Adams et al. Results for a specific month resulted in an objective function value of $18. 1988.3 The master production schedule A master production schedule to minimize the makespan. Constraint system (12) is the link between the aggregate and disaggregate models while (13) expresses the limit imposed by process capacity. i.The objective function represents the minimization of total set-up costs. This schedule consists of a permutation. of the operations on each machine ( 2. Garey and Johnson 1979).846 yielding the optimal weekly product schedule. the time needed to finish all jobs in a shop. NP-complete in the strong sense.e.

although the cost is not readily quantifiable . (b) A slight modification in the solution by reducing the entire labour force from a lower bound of 25 to a lower bound of 17 (the labour required in the highest-volume month) would result in yearly savings of approximately $174.720 (c) The model can estimate and eliminate the cost of the backlog.95 in labor costs for the year due to abandoning its current labour policy and following the chase strategy outlined in the HPP model. Benefits (a) The company described in the paper should immediately realize approximate savings of $234.3.603.

improved shopfloor control and project planning analysis. All of these can drastically improve due date performance. there is no discussion of a more efficient plant layout. For example. 4. This benefit can be generalized to any make to-order facility that has highly customized parts in which product archetypes can be readily determined. (e) A principal benefit of the developed MPS is that an optimal scheduling plan was developed for all 12 product archetypes. due to the severely constrained system (c)The paper ignores other issues that can lead to further increase in net income. customer retention and customer loyalty. increased cash flow due to lower work-in-process inventory. . Limitations (a) Following a chase strategy is challenging due to the number of layoffs involved (b) A suboptimal scheduling plan may have developed a plan that could not be met.(d) Further cost savings are realized by the long-run effects of increased service (in terms of lead-time reduction).