Optimal Modelling of Process Variations in Industry 4.0 Facility under Advanced P-Box uncertainty

Abstract

Industry 4.0 is the fourth and current working paradigm leaning on data exchange, data analysis and Internet of things for complete automation and optimization of manufacturing technologies. This I4.0 paradigm enables the possibility for optimal autonomous decision making and full autonomous operation of the facility. Since all manufacturing processes are inherently in-deterministic, optimal decision making will only work when provided with realistic uncertainty models. The authors have observed that the typical used prediction models are not accurate as they do not represent the in-deterministic nature of the processes in a realistic manner. The objective of this paper is to use the IoT- based logged data to design more relevant uncertainty models to optimize the industry 4.0 framework. For this purpose, p-box models are used to represent production variation as function of unknown in-deterministic parameters. Furthermore, by using regression learning on the database, we show that the uncertainty can be decreased. It is shown that classic stochastic models are not able to capture the full in-deterministic nature of the uncertainties. It makes therefore sense that autonomous decision making based on such stochastic models are unreliable and sub-optimal. Future research possibilities are discussed, based on the proposed uncertainty models, together with possible notions that will help in the development of reliable optimal decision making and more accurate adaptable Industry 4.0 framework.