Machine fault diagnostics using a transputer network

This paper describes the implementation of a machine-fault diagnostic system on a transputer network. The fault diagnostic system includes an artificial neural network and a fuzzy logic-based hypothesis and test program. The performance of this transputer network is compared with that of a 486 33 MHz personal computer. It is found that a single T800 25 MHz transputer is much slower than the 486 PC. However, the performance is improved substantially when transputers are implemented in parallel. With the hardware cost of transputers continuing to drop, it is economically feasible to introduce this technology to the shop floor, and to use it to perform many routine tasks.