Katsuhisa Ohno

The optimal control of just-in-time-based production and distribution systems and performance comparisons with optimized pull systems

In just-in-time (JIT) production systems, there is both input stock in the form of parts and output stock in the form of product at each stage. These activities are controlled by production-ordering and withdrawal kanbans. This paper discusses a discrete-time optimal control problem in a multistage JIT-based production and distribution system with stochastic demand and capacity, developed to minimize the expected total cost per unit of time. The problem can be formulated as an undiscounted Markov decision process (UMDP); however, the curse of dimensionality makes it very difficult to find an exact solution. The author proposes a new neuro-dynamic programming (NDP) algorithm, the simulation-based modified policy iteration method (SBMPIM), to solve the optimal control problem. The existing NDP algorithms and SBMPIM are numerically compared with a traditional UMDP algorithm for a single-stage JIT production system. It is shown that all NDP algorithms except the SBMPIM fail to converge to an optimal control. Additionally, a new algorithm for finding the optimal parameters of pull systems is proposed. Numerical comparisons between near-optimal controls computed using the SBMPIM and optimized pull systems are conducted for three-stage JIT-based production and distribution systems. UMDPs with 42 million states are solved using the SBMPIM. The pull systems discussed are the kanban, base stock, CONWIP, hybrid and extended kanban.

The performance evaluation of a multi-stage JIT production system with stochastic demand and production capacities

This paper discusses a single-item, multi-stage, serial Just-in-Time (JIT) production system with stochastic demand and production capacities. The JIT production system is modeled as a discrete-time, M/G/1-type Markov chain. A necessary and sufficient condition, or a stability condition, under which the system has a steady-state distribution is derived. A performance evaluation algorithm is then developed using the matrix analytic methods. In numerical examples, the optimal numbers of kanbans are determined by the proposed algorithm. The optimal numbers of kanbans are robust for the variations in production capacity distribution and demand distribution.

New approximate dynamic programming algorithms for large-scale undiscounted Markov decision processes and their application to optimize a production and distribution system

Undiscounted Markov decision processes (UMDPs) can formulate optimal stochastic control problems that minimize the expected total cost per period for various systems. We propose new approximate dynamic programming (ADP) algorithms for large-scale UMDPs that can solve the curses of dimensionality. These algorithms, called simulation-based modified policy iteration (SBMPI) algorithms, are extensions of the simulation-based modified policy iteration method (SBMPIM) (Ohno, 2011) for optimal control problems of multistage JIT-based production and distribution systems with stochastic demand and production capacity. The main new concepts of the SBMPI algorithms are that the simulation-based policy evaluation step of the SBMPIM is replaced by the partial policy evaluation step of the modified policy iteration method (MPIM) and that the algorithms starts from the expected total cost per period and relative value estimated by simulating the system under a reasonable initial policy.

Effective Estimation of Distribution Algorithm for Stochastic Job Shop Scheduling Problem

Abstract This paper propose an effective estimation of distribution algorithm (EDA), which solves the stochastic job-shop scheduling problem (S-JSP) with the uncertainty of processing time, to minimize the expected average makespan within a reasonable amount of calculation time. With the framework of proposed EDA, the probability model of operation sequence is estimated firstly. For sampling the processing time of each operation with the Monte Carlo methods, we use allocation method to decide the operation sequence then the expected makespan of each sampling is evaluated. Subsequently, updating mechanism of the probability models is proposed with the best solutions to obtain. Finally, for comparing with some existing algorithms by numerical experiments on the benchmark problems, we demonstrate the proposed effective estimation of distribution algorithm can obtain acceptable solution in the aspects of schedule quality and computational efficiency.

Optimal production sequencing problem to minimise line stoppage time in a mixed-model assembly line

Mixed-model assembly line is utilised to assemble many product variants on a single line in automobile and other industries. In just-in-time production system, the automation (Jidoka) allows workers to stop a conveyor line whenever they fail to complete their assembly operations within predetermined process times. Given this situation, it becomes important to determine the production sequence to minimise the total conveyor stoppage time in the mixed-model assembly line. In this article, we propose an explicit and complete formulation of the production sequencing problem for the mixed-model line, in which the objective function is to minimise the total line stoppage time. Some important properties are derived and by relaxing the integer constraints in the formulation, a branch and bound algorithm is developed. The performance of a commercial mathematical programming package is discussed by solving several numerical examples using the formulation.

A Hybrid EA for Reactive Flexible Job-shop Scheduling

Abstract In this paper, we consider a reactive flexible job-shop scheduling problem (rFJSP) under uncertainty environment. The most existing reactive scheduling methods are characterized by least commitment strategies such as real-time dispatching that create partial schedules based on local information. In rFJSP, two extensions of these dispatching strategies are to allow the system to select multiple machines assignment, and multiple operation process for each job. So, how to design an effective flexible rescheduling strategy is the key point of this paper. For solving this rFJSP, we propose a hybrid evolutionary algorithm (hEA) with combining genetic algorithm (GA) and particle swarm optimization (PSO). Finally, the experiments verify the effectiveness of proposed hEA, by comparing with different evolutionary approaches for several scale test problems of rFJSP.

Flexible Production Systems developed and utilized in DENSO CORPORATION and their Evaluation

As the global environment changes rapidly and the information technology develops quickly in recent years, requirements for the production become diverged widely and sophisticated highly. The flexible production systems developed in DENSO are evaluated by the performance function ‘Life Cycle Cost’ of the production volume and compared with conventional systems. The results show that the flexible production systems contribute to the management of the manufacturer.