Learning to Dispatch Operations with Intentional Delay for Re-Entrant Multiple-Chip Product Assembly Lines

Abstract

As the demand for small devices with embedded flash memory increases, semiconductor manufacturers have been recently focusing on producing high-capacity multiple-chip products (MCPs). Due to the frequently re-entrant lots between the die attach (DA) and wire bonding (WB) assembly stages in MCP production, increased flow time and decreased resource utilization are unavoidable. In this paper, we propose a dispatcher based on artificial neural networks, which minimizes the flow time while maintaining high utilization of resources at the same time through exploiting the possible intentional delays on DA stage. Specifically, the proposed dispatcher learns the assignment preferences between available lots and DA resources based on assembly line data generated by using a simulator, then the proposed dispatcher performs lot dispatching decisions by considering the intentional delays. The numerical experiments were performed under various configurations of the MCP assembly lines, and the results show that the proposed dispatcher outperformed the existing methods.

A review of previous researches about machine learning theory in production management: Focusing on reinforcement learning

“…Existing studies on schedule planning using machine learning that have been conducted so far are as follows. Jae-seok Heo developed a decision-making model to increase operation rate in the semiconductor packaging process using a deep neural network (DNN) [10], and Cho Yong-cheol used an artificial neural network (Artificial Neural Network) to improve wafer processing in the semiconductor manufacturing process. An optimal scheduling method for movement paths was proposed [11], and Jinyoung Kim proposed a DQN packet scheduling algorithm in wireless networks using the DQN (Deep-Q-Network) algorithm [12].…”

An examination of the reinforcement learning methodology of machine learning to establish a schedule considering major KPIs such as delivery date in a high-tech manufacturing line

“…Existing studies on schedule planning using machine learning that have been conducted so far are as follows. Heo [7] developed a decision-making model to increase operation rate in the semiconductor packaging process using a deep neural network (DNN), and Cho [8] used an artificial neural network (Artificial Neural Network) to develop a decisionmaking model for the semiconductor manufacturing process. proposed an optimal scheduling method for the wafer's movement path, and Kim [9] proposed a DQN packet scheduling algorithm in a wireless network using the DQN (Deep-Q-Network) algorithm.…”

Intelligent Manufacturing Planning System Using Dispatch Rules: A Case Study in Roofing Manufacturing Industry

“…FIFO processes jobs in an orderly manner according to the first arrival of the customer's order. There were many researchers who used FIFO as the basic rules [34][35][36][37][38][39] due to its simple algorithm. It has no complexity in decision making [40].…”