Machine Learning Applications in Demand Forecasting and Order Fulfillment for Smart Manufacturing OSS

Shabrinath Motamary

Abstract: The demand-driven paradigm of the modern market environment imposes enormous challenges on enterprise production and logistics. Smart manufacturing based on cloud edge collaboration is changing the traditional manufacturing mode into on-demand order manufacturing mode, which requires enterprises to have a perfect demand forecasting model. Current cloud-edge architecture just focuses on how to guarantee system stability and computing resource composition, but seldom studies how to make full use of multi-source heterogeneous cloud-edge resources to provide competitive service level agreements for smart manufacturing. Therefore, in smart manufacturing, the supply chain sensing layer needs to build a multi-source heterogeneous time series forecasting mechanism. Among these tasks, demand forecasting based on time series has attracted widespread attention due to its wide application scenarios and challenging characteristics. Demand forecasting is a key business in a supply chain, and a variety of intelligent demand forecasting methods have been developed. In production management, forecasting the processing time of a specific job input is essential for dynamic scheduling in job shop systems. The scheduling model comprising the manufacturing attributes and characteristics of the order processing is built to forecast order completion date. This forecast mechanism plays a key role in order management and scheduling decision-making. Currently, the existing cloud-edge management systems are seldom designed for smart manufacturing and hardly predict multi-source heterogeneous time series. Therefore, the corresponding modeling methodology should be proposed to comprehensively consider diverse data characteristics (complexity, completeness, frequency, sampling rate, etc.). With the rapid development of the industrial internet, various types of demanding data are heavily accumulating in cloud-edge environments. However, how to effectively select and utilize such multi-source heterogeneous data to improve edge computing is still a challenging problem in future smart manufacturing. The cloud-edge collaborative computing architecture and technologies place strict requirements on the performance of multi-source heterogeneous data selection and processing methods. State-of-the-art cloud-edge computing methods often ignore the existence of diverse multi-source heterogeneous data. Hence, the key techniques regarding how to model the selection and processing of multi-source heterogeneous supply chain data should be studied, which may significantly change the cloud-edge resource management and scheduling algorithms.

Keywords: ML for Demand Forecasting, Smart Manufacturing OSS, Predictive Order Fulfillment, AI-Driven Supply Chain Optimization, Machine Learning in Production Planning, Demand Sensing Algorithms, Real-Time Inventory Forecasting, Intelligent Order Management, Data-Driven Manufacturing Decisions, Adaptive Forecasting Models, AI-Powered Supply Chain Analytics, ML-Based Inventory Optimization, Operational Support Systems in Manufacturing, Predictive Analytics for Manufacturing, Automated Demand Planning

How to Cite?: Shabrinath Motamary, "Machine Learning Applications in Demand Forecasting and Order Fulfillment for Smart Manufacturing OSS", Volume 10 Issue 12, December 2021, International Journal of Science and Research (IJSR), Pages: 1639-1653, https://www.ijsr.net/getabstract.php?paperid=MS2112143019, DOI: https://dx.doi.org/10.21275/MS2112143019