Intelligent Manufacturing Systems— A Tentative Forecast

Abstract

Manufacturing systems research and development started by linking extant modules (CAD, NC Programming, DNC, Automated Assembly, Scheduling) to each other. As systems have grown in size and complexity, these methods have become inefficient and research is now concentrated on developing synthesis methods for large-scale hierarchical system design. These methods, however, in the practical use of which the authors have gained some experience, are mostly deterministically based and assume either the predictability of all eventualities, or a reversion to human control when unforeseen (algorithmically unstipulated) situations occur.The scale and complexity of large manufacturing systems will grow so rapidly as to render both the above alternatives impracticable. Systems will have to possess an in-built resilience with respect to unforeseen situations, based on the ability to recognize trends and situations in real time from large amounts of data, to use these perceptions to solve the problem of finding an optimal recovery strategy, to carry out the required decision-making processes and finally to implement and check the decisions taken. Such an approach requires the development of novel, heterarchical manufacturing model definition and analysis methods, for which the present tools of Artificial Intelligence research hold promise, but are not yet adequate.