The chapter examines ML methods that appear to be applied in implementing systems with intelligent behaviour. It depends on two workshops on learning in system of intelligent manufacturing, an intensive survey of the literature, and various commitments. Symbolic, sub-symbolic, and hybrid approaches, as well as their applications in manufacturing, are also discussed, as are hybrid solutions that attempt to combine the advantages of several methodologies. The advantages, inadequacies, and impediments of different creation methods are illustrated to decide suitable strategies for explicit circumstances.
TopIntroduction
Machine learning will unharness the subsequent wave of virtual disruption. Businesses are predicted to be prepared. Companies that were among the first to put it in place have reaped the benefits. Future technology consists of robotics and self-using deep learning, natural language processing, machine learning, computer vision, and so on. The next generation of machine learning apps is built around digitization. Industries that have embraced digitization are typically at the forefront of machine learning. They are also expected to be a growth driver. Changes in marketplace share and revenue are probably to be elevated via way of means of using machine learning. Several adjustments to the original approach to industrial automation are introduced by Industry 4.0. In this context, the Cyber-Physical System and Internet of Things technologies play a role in introducing intellectual automation and, thus implementing the notion of intelligent manufacturing, which is the main reason for smart products and best services (Kunst et al., 2019). Companies will encounter obstacles in a far more dynamic environment as a result of this unique approach. Many of these businesses are unprepared to deal with this new reality, in which a huge number of people do not always work together to boost productivity (Lee et al., 2014).
Predictive maintenance has several advantages, but it also has many drawbacks. On the one hand, productivity gains, system fault reduction (Chukwuekwe et al., n.d.), unexpected downtime minimization (Balogh et al., 2018), improved effectiveness in the utilization of human resources, financial(Schmidt & Wang, 2018), and optimization in maintenance intervention planning (Adhikari, 2018) are all benefits of Product data management. Machine Learning (ML) can be used for prognostics and failure prediction, such as estimating a machine's lifetime using a vast quantity of data to train an ML system (Balogh et al., 2018; Zhou & Tham, 2018), as well as diagnosing faults (Ansari et al., 2020; Bousdekis et al., 2019). Machine Learning (ML) research in areas directly related to design and manufacturing has not yet begun. It does, however, have a long history of success and failure, much like other disciplines of Artificial Intelligence (Lu, 1990) for a comprehensive assessment up to 1989. The adoption of a single ML approach or methodology into a current modeling and dynamic system for an explicit designing challenge has been the most typical feature of these R&D projects, with a few exceptions. Regardless, while learning is a universal process, machine learning in the context of engineering requires further investigation to completely appreciate the specifics. We hope to define and characterize “mutual human-machine learning” in future workplaces in this article. The key research difficulty is identifying how to characterize reciprocal learning when both human employees and intelligent computers of variable competence and intelligence participate in a shared endeavor. We investigate normal situations including humans and AI, just as human and machine skills in modern frameworks, to find an answer.
Figure 1.
Relationship of human-machine collaboration.