To address the uncertainty of production environment in assembly job shop, in combination of the real-time feature of reinforcement learning, a dual Q-learning (D-Q) method is proposed to enhance the adaptability to environmental changes by self-learning for assembly job shop scheduling problem. On the basis of the objective function of minimising the total weighted earliness penalty and completion time cost, the top level Q-learning is focused on localised targets in order to find the dispatching policy which can minimise machine idleness and balance machine loads, and the bottom level Q-learning is focused on global targets in order to learn the optimal scheduling policy which can minimise the overall earliness of all jobs. Some theoretical results and simulation experiments indicate that the proposed algorithm achieves generally better results than the single Q-learning (S-Q) and other scheduling rules, under the arrival frequency of product with different conditions, and show good adaptive performance. Abbreviations: AFSSP, assembly flow shop scheduling problem; AJSSP, assembly job shop scheduling problem; RL, reinforcement learning; TASP, two-stage assembly scheduling problem

In a manufacturing system, reworking of the defectives and management of waste or scrap are important issues that call for immediate attention to meet the basic objectives and requirements for lean production system. This paper deals with a manufacturing process where finished goods are occasionally produced along with undesirable defective items and scrap. As the system is not always perfect, some scrap is produced during the manufacturing and/or rework processes. A rework process is necessary to correct the defective items into good finished products. The manufacturing process considered in this research deals with the process of rework that is performed ‘within’ the cycle. Also, three types of scrap detection scenarios are considered: (i) detection ‘before’ rework, (ii) detection ‘during’ rework, and (iii) detection ‘after’ rework. In order to avoid shortage of finished products due to scrap production, finished-product buffers have been considered to fulfil the customers’ timely demand. The required finished products for these buffers are ordered from different suppliers. Based on the observations in shop floors, a number of inventory models are developed for a single-stage production process wherein defective items are reworked, and scraps are detected and discarded during entire process. For the stated detection procedures, the total cost functions for a single-stage imperfect manufacturing system are developed with 100% inspection to find the optimum operational policy for manufacturing batch sizes and economic storage of inventories. Numerical examples are used to demonstrate these models, and an analysis is performed to study the sensitivity of different system parameters on several measures of performance.

We consider a manufacturing system which receives raw material from a supplier, processes it, and delivers it to the customer periodically. The system considered is imperfect and produces defectives at a constant rate. The finished product can only be delivered if the whole lot is quality-certified. Hence, defectives have to be reworked, and the whole lot quality-checked within the cycle. Three different scenarios are considered, viz. (a) a single lot of raw material for multiple lot of finished product and delivery of the product in multiple instalments, (b) a single lot of raw material for a multiple lot of finished product and delivery of the product in a single instalment, and (c) lot-for-lot and delivery of finished product in single instalment. A total cost equation is developed for each model, and the optimal ordering quantities are evaluated. The results found are encouraging and quite simple to use for practical purposes.