Neural knowledge helps gear designer’s decisions

Dr Daizhong Su of the department of Mechanical and Manufacturing Engineering at Nottingham Trent University has been working on the development of an intelligent hybrid computer system (IHS) that aims to computerise and speed up the lengthy gear design process.

The IHS blends artificial intelligence (AI), data processing and CAD/CAM/CAE into a single software environment for design specification, conceptual design and detailed design. Two AI techniques, a knowledge based system (KBS) and artificial neural network (ANN), have been utilised within the IHS to capture the design knowledge.

Distinguished from other computing techniques (such as CAD, numerical analysis, databases, etc.), the major functions of the KBS are symbolic reasoning and decision-making. It embodies both expert knowledge and expert inferencing means; the former is stored explicitly using a symbolic declarative language, and the latter consists of intelligent heuristic search and reasoning procedures.

The ANN is composed of many interconnected units (artificial neurones), each of which performs a weighted sum of its inputs. The units operate in parallel and arrange in the ways reminiscent of biological neural nets. The major functions of the ANNs are pattern-recognition and classification.

The neural networks have been utilised for two tasks: concept generation, and design data retrieval in detail design stage which requires the knowledge based on empirical data or interpretation by design experts.

ANNs obtain the knowledge via training using previously known results. The training data consists of two parts: the input data to the network and output data which are the known results corresponding to the input data. The neurones are connected via weights that are adapted during the training. On the completion of the training, the weights are saved into files.