Neural Networks for Electronics Assembly

In August 1991 I had the opportunity of attending the annual general meeting of CIRP in Aix‐en‐Provence, France. CIRP is not all that well known generally. It is however a forum for the leading experts in manufacturing technology, and since its inception in 1951 it has evolved as the College of International Research in Production (CIRP). The annual general meeting of CIRP moves around the globe. The purpose of its existence is mainly to exchange and develop ideas in manufacturing technology with a view to providing a more scientific base for product and process design for manufacturing. Each year there are about 150 papers presented at the annual meeting and they form the Annals of CIRP which can be purchased or consulted by the public at large[l]. The papers selected by CIRP are generally of the highest quality, and reflect most advanced concepts in various areas of manufacturing engineering. The areas addressed by the scientific technical committees (STCs) of CIRP are: assembly, cutting, design, physical and chemical machine tools, optimization, dimensional metrology in quality assurance, and surfaces. I was invited by the STC on assembly to give an overview of the assembly research in Australia. This provided the opportunity for me to hear about the latest research on assembly. Of the 12 papers presented on assembly, including a keynote paper by Boothroyd and Alting on Design for Assembly and Disassembly, there were two which specifically dealt with the issue of quality control in electronic assembly, particularly via neural network data processing. With reference to the previous article by Pandya in Assembly Automation, Vol. 12 No. 4, the material which follows provides further insights into the neural network approach.