Static load estimation for magnetic alloy strips by using neural networks

Magnetoelastic (magnetic alloy) sensors have been used to monitor stress. Use of backpropagation (BP) type neural networks (NN) was proposed for estimation of the load by evaluating the magnetic response characteristics of the system. In the experiments, Metglas 2826MB magnetic alloy strip was located at the axis of two coils. An alternating magnetic field was then generated with the excitation coil by applying a sweep sine wave. The envelope of the signal of the detection coil was given to a BP type NN to estimate the load after the training process. The estimation errors of the training and test cases were less than 0.01% and 2.5% of the loading range respectively. The study indicated that BP type NN could be used for mapping and interpretation of the signals of the detection coil when magnetoelastic sensors are used.