Decision SupportA clearing function based bid-price approach to integrated order acceptance and release decisions
Introduction
In this paper, we develop a bid-price-based approach for the order acceptance and release decisions of a manufacturer facing stochastic demand in an order-driven, stochastic production system.
As an example of this setting, consider semiconductor back-end production. Fig. 1 depicts the production process of a typical semiconductor manufacturer, which comprises two stages: front-end and back-end production. In the front-end stage, the integrated circuits are produced on a raw wafer, which is a thin disc made of silicon. The finished wafers are stored in a die bank, which represents the system’s customer order decoupling point. In the back-end production stage, the functional circuits are cut from the wafers and assembled into finished chips, based on customer-specific requirements. The finished chips then undergo final testing.
In this setting, the manufacturer must decide which incoming orders to accept, considering e.g., achievable margins, requested delivery times, available wafers, and available assembly and testing capacity. In addition, processing of accepted orders has to be scheduled. These are the sets of decisions that we consider in this paper. We describe a concrete example of such a production process in Section 6.5.
Several factors complicate this planning task (Mönch, Fowler, & Mason, 2013). Unforeseen machine down times result in stochastic cycle times; expensive machines are used at multiple stages of the production process and for different products, leading to significant, sequence-dependent set-up times; on some equipment, sets of jobs are processed simultaneously in batches. In addition, demand is characterized by industry-wide peaks (see Mönch, Uzsoy, & Fowler, 2018). Capacity expansions are not a feasible mitigation option because they are time consuming and require high capital investments. Outsourcing is also infeasible during peak periods because all factories are working at capacity.
Generalizing from this motivating example, the present paper addresses integrated order acceptance and release decisions of a manufacturer in a flexible job shop with stochastic effective processing times. Production is order specific, and demand is stochastic. Orders differ with respect to their contribution margin (for example due to different prices), lead time requirements, and capacity requirements. Available capacity is exogenous. For such heterogeneous orders, the manufacturer has to decide which orders (with order-specific contribution margin) to accept. The objective is to maximize the difference between the contribution margin from accepted orders and the sum of work-in-process (WIP) holding costs, finished good inventory (FGI) holding costs, and backlog costs.
We consider a planning problem with stochastic demand and processing times. For discrete probabilities, a stochastic dynamic program can be developed to find expectation-optimal solutions. However, the state space must contain all possible demand states (backlogged orders) and production states (status of each machine and of each accepted order). Solving a stochastic dynamic program is computationally intractable already for much simpler problems, where only demand is stochastic and production is deterministic (Petrick, Steinhardt, Gönsch, & Klein, 2012).
Instead of modeling the production system via a stochastic dynamic program, we develop a heuristic solution approach by combining bid-price-based revenue management (RM) with order release planning based on clearing functions (CFs). Bid-price approaches approximate the value function with a linear function and are commonly used in RM-based order acceptance – in assemble-to-order (ATO) production systems (e.g. Gao, Xu, & Ball, 2012 and Guhlich, Fleischmann, & Stolletz, 2015), and make-to-order (MTO) production systems (see e.g. Hintsches, Spengler, Volling, Wittek, Priegnitz, 2010, Chevalier, Lamas, Lu, Mlinar, 2015, Guhlich, Fleischmann, Mönch, Stolletz, 2015, and the references therein).
In the existing literature, linear programming (LP) models with fixed lead times are used to model the production system. This is justified in deterministic production systems. However, in systems with stochastic influences, such as in the semiconductor setting discussed above, lead times depend non-linearly on the workload. CFs provide a means to capture this relationship. Specifically, CFs model the maximum output of a machine group as a non-linear function of the workload over a given period of time. CF-based LPs have been successfully applied to order release decisions in production systems with stochastic influences, e.g., in the semiconductor industry (cf. Asmundsson, Rardin, Uzsoy, 2006, Kacar, Mönch, Uzsoy, 2013). In the present paper, we integrate this approach into order acceptance planning, which is novel. Specifically, we use CFs to derive bid prices, reflecting opportunity costs of capacity usage.
Summarizing, our main contributions in this paper are as follows:
- 1.
We consider integrated order acceptance and release decisions in a stochastic production environment with stochastic, order-specific demand. This problem has not yet been studied in the literature.
- 2.
We propose a novel bid price approach using CFs to measure the opportunity costs of production capacity.
- 3.
In a numerical study, we show that the proposed approach significantly outperforms approaches that address variability in production systems by modeling fixed, workload-independent waiting times, which is a common practice in order release planning (Missbauer & Uzsoy, 2011).
From a managerial point of view, order acceptance decisions have to trade off incremental contribution margins against opportunity costs of resource consumption. In real-life production systems, assessing opportunity costs is challenging because orders require multiple resources and throughput times depend non-linearly on the workload, due to stochastic influences. Furthermore, opportunity costs vary over time and therefore depend on future order release decisions. In this paper, we present a method that can handle these complicating factors and thereby supports managers in making appropriate order acceptance decisions. The numerical study demonstrate, that especially for a high uncertainty in processing times and many production stages, this revenue management approach consistently outperforms approaches based on exogenous lead times as applied in real-world industrial settings. Therefore, it is worth spending additional effort for determining the shape of respective clearing functions for each machine group.
This paper is organized as follows. In Section 2, we review the related literature. In Section 3, we formally define the addressed problem. In Section 4, the derivation of CFs is discussed. The bid-price-based RM approach is developed in Section 5. In Section 6, a numerical study evaluates the performance of the presented approach. In Section 7, we outline our conclusion and provide possible avenues for further research.
Section snippets
Literature review
In this paper, we develop a novel demand fulfillment approach based on ideas of bid-price-based RM in order-driven manufacturing and of CF-based production planning. Therefore, we review literature from these streams. In addition, related literature on admission control and stochastic scheduling is discussed.
Problem description
In this section, we define the planning problem considered in this paper. In particular, we describe the production system, the production process, the demand, the decisions to be made, and the objective.
The planning horizon is divided in equidistant planning periods of length pl. I different order types denoted by can be produced. An order type is a class of orders that require a specific set of capacities and lead times and entails specific profit margins and backlog costs.
Assumption 3.1 Production system
CF-based capacity constraints
In this paper, we combine bid-price-based RM with CF-based order release planning. Therefore, in this section, we describe how CFs are used in production planning LPs in the literature. Specifically, we summarize the derivation of CF-based capacity constraints as described in Asmundsson et al. (2009). The production planning LP that we use in our approach is presented in Section 5.1. In addition, we describe how we estimate the parameters for the (linearized) CFs. Table 1 summarizes the
CF-based bid-price approach
In this section, we present a heuristic solution approach to the problem described in Assumption 3.5. Optimal solution approaches have been shown to be computationally intractable, even for simpler deterministic production systems (Petrick et al., 2012). We develop a bid-price-based RM approach that uses a production planning LP with CF-based capacity constraints to capture the non-linear relationship between workload and lead times in stochastic production systems.
We use this LP to compute bid
Numerical study
In the literature on order release planning as well as in the literature on RM, fixed lead time approaches are commonly used to model production systems (Guhlich, Fleischmann, Mönch, Stolletz, 2015, Missbauer, Uzsoy, 2011). In order release planning, it has been shown that CF-based models outperform approaches that assume fixed integer lead times in production systems with variability (see e.g. Kacar et al., 2013). In this section, to show that this result can also be reached in an RM context,
Conclusion and further research
In this paper, we consider order acceptance and release decisions of a manufacturer using a production system with stochastic influences and facing stochastic demand. We develop a novel bid-price-based RM approach using CFs to capture the non-linear relationship of workload and lead times. This algorithm combines the ideas of bid-price-based RM with CF-based order release planning. Bid prices are computed based on an LP that models the production system with CFs. The order acceptance and order
References (45)
- et al.
Estimating negative binomial demand for retail inventory management with unobservable lost sales
Naval Research Logistics
(1996) - et al.
Demand modeling with forecast evolution: an application to production planning
IEEE Transactions on Semiconductor Manufacturing
(2015) - et al.
Zero-order production planning models with stochastic demand and workload-dependent lead times
International Journal of Production Research
(2015) - et al.
Production planning with resources subject to congestion
Naval Research Logistics
(2009) - et al.
Tractable nonlinear capacity models for production planning part I : Modelling and formulations
Technical Report
(2004) - et al.
Tractable nonlinear production planning models for semiconductor wafer fabrication facilities
IEEE Transactions on Semiconductor Manufacturing
(2006) Minimizing earliness and tardiness costs in stochastic scheduling
European Journal of Operational Research
(2014)- et al.
Heuristic solution methods for the stochastic flow shop problem
European Journal of Operational Research
(2012) - et al.
Integrating order acceptance decisions with flexible due dates in a production planning model with load-dependent lead times
International Journal of Production Research
(2015) - et al.
Scheduling stochastic jobs with asymmetric earliness and tardiness penalties
Naval Research Logistics
(1997)
Revenue management for operations with urgent orders
European Journal of Operational Research
Towards a supply chain simulation reference model for the semiconductor industry
Proceedings of the 2011 winter simulation conference
The pattern of consumer purchases
Applied Statistics
On heuristic solutions for the stochastic flowshop scheduling problem
European Journal of Operational Research
Managing an available-to-promise assembly system with dynamic short-term pseudo-order forecast
Management Science
Revenue management in a multi-stage make-to-order production system
Working Paper
Revenue management approach to due date quoting and scheduling in an assemble-to-order production system
OR Spectrum
Revenue management in make-to-order manufacturing: case study of capacity control at Thyssenkrupp VDM
Business Research
Planning wafer starts using nonlinear clearing functions: A large-scale experiment
IEEE Transactions on Semiconductor Manufacturing
A comparison of multiple linear regression approaches for fitting clearing functions to empirical data
International Journal of Production Research
Estimating clearing functions for production resources using simulation optimization
IEEE Transactions on Automation Science and Engineering
Manufacturing lead times, order release and capacity loading
Handbooks in operations research and management science
Cited by (6)
Integrated and hierarchical systems for coordinating order acceptance and release planning
2022, European Journal of Operational ResearchCitation Excerpt :Their results highlight the need for an effective combination of order acceptance and order release to maintain acceptable cycle times and WIP levels. Guhlich, Fleischmann, Mönch, & Stolletz (2018) and Netzer, Haeussler, & Missbauer (2020) suggest hierarchical approaches that solve separate order acceptance and order release subproblems sequentially, using the ACF model for the order acceptance subproblem. Guhlich et al. (2018) use a clearing function model to derive bid prices to decide which orders to reject and includes an order release model using a clearing function.
A matheuristic for making order acceptance decisions in multi-product, multi-stage manufacturing systems
2021, Applied Soft ComputingCitation Excerpt :However, mathematical models are not analyzed in this paper. Guhlich et al. [35] propose a bid price-based revenue management approach to solve an integrated order acceptance and release problem. Piecewise linear CFs are used to model workload-dependent lead times.
A loosely-coupled deep reinforcement learning approach for order acceptance decision of mass-individualized printed circuit board manufacturing in industry 4.0
2021, Journal of Cleaner ProductionCitation Excerpt :In this circumstance, reinforcement learning on the known prediction model (i.e., model-based reinforcement learning) could be designed for order acceptance decisions. Another circumstance is the discrete stochastic dynamic order arriving with a relatively-large time intervals among PCB orders (i.e., both demand and production are stochastic (GUHLICH et al., 2018)). To realize the timely feedback of customers’ orders, real-time decision making should be realized for each order before the subsequent order arrived on, while the future reserve capacity for other orders should also be taken into consideration.
Order release in ETO systems - A basic optimization-based approach
2019, IFAC-PapersOnLineThe order selection strategy of polluting OEMs under environmental regulations
2021, Sustainability (Switzerland)