Application of artificial neural network to optimize sensor positions for accurate monitoring: an example with thermocouples in a crystal growth furnace

We propose to utilize artificial neural network (ANN) to optimize positions of a limited number of sensors for accurate monitoring, and demonstrate its effectiveness by a case study of four thermocouples in a directional solidification furnace. Our concept consists of choosing the positions with ANN that has the lowest loss from a multiplicity of ANNs, which were trained by the simulated temperature distributions along the outer crucible wall. Interestingly, the top ten ranks of accurate predictions contain positions around the crucible's bottom to suggest the importance of measuring temperatures carefully around high-temperature gradients that is the boundary between different materials.