Stress Calculation of Crankshaft Using Artificial Neural Network

A system that calculates the stress concentration factor of the crankpin fillet from six characteristic dimensions of the crankshaft was developed using an artificial neural network. The learning database was constructed based on the finite element analysis, and an “adaptive transfer function algorithm” was used for the learning calculations. The calculation errors of the stress concentration factors applied to crankshafts of small utility engines and outboard motors were found to be within -6.9 to +6.3% of the measured values. With this system, designers can calculate the stress concentrated at crankpin fillets precisely in a short time.