Output Prediction of Helical Microfiber Temperature Sensors in Cycling Measurement by Deep Learning

Abstract The inconsistent response curve of delicate micro/nanofiber (MNF) sensors during cycling measurement is one the main factors which greatly limit their practical application. In this paper, we proposed a temperature sensor based on copper rod-supported helical microfiber (HMF). HMF exhibited different light intensity-temperature relationships in single-cycle measurements. Two neural networks, deep belief network (DBN) and backpropagation (BPNN), were employed respectively to predict sensing processes. input variables geometric parameters (the diameter, wrapped length, coiled turns, angle) output optical intensity under working root mean square error (RMSE) Pearson correlation coefficient ( R ) used evaluate predictive ability networks. DBN with two restricted Boltzmann machines (RBMs) provided best prediction results (RMSE heating process are 0.9705 °C 0.9969, while values RMSE cooling 0.786 6 0.997 7, respectively). obtained by optimal BPNN (five hidden layers, 10 neurons each layer, RMSE=1.126 °C, =0.995 7) slightly inferior those DBN. could accurately reliably operation, possibility for flexible application complex MNF wide range.