Human and Machine Reliability Estimation in Discrete Simulations and Machine Learning for Industry 4.0 and 5.0

Currently, Industry 4.0 creates new opportunities for analyzing data on production processes and extracting knowledge from them. With the Internet of Things, data is continuously collected from machine sensors to analyze machine health. Thanks to artificial intelligence methods and discrete simulation, it is possible to process data and dynamically adjust the operating conditions of the production line to the expected time of failure-free operation of the machine or reliable work of an employee. Recently, machine learning techniques have been used to automatically adapt the production line to changes in a given production environment. The paper presents various methods of modeling actions, i.e., forecasting the failure-free operation time of a machine or the error-free working time of an employee. The possible actions the agent can perform, the possible prediction techniques that can be selected are presented. The time between failures is described by a log-normal distribution. The asymmetric lognormal distribution is much more flexible for practical modeling compared to the “perfectly” symmetric normal distribution. In practice, the asymmetric lognormal distribution, strongly shifted to the left, can be used to describe the decreasing time between failures due to human error, as well as the time between failures of a machine in the third phase of its life cycle, which decreases as the machine ages and its components wear out. The parameters of the distribution are estimated using the maximum-likelihood approach, theempirical moments approach, the renewal-theory approach, the empirical distribution function and the method based on coefficient of variation. Numerical examples of predicting failure-free operation times described by the log-normal distribution are presented. The results are compared assuming that failure-free times are described by exponential, normal and Weibull distributions. The results are also compared with an example of the simplest learning method.