| 摘要 |
<p>PURPOSE:To improve noise resistance, reduce the memory capacity, and allow high-speed processing by using a neural network when the state of a plant is diagnosed from the pattern of the process signal of the plant. CONSTITUTION:The transient pattern patterned 3 from the process signal inputted 2 from a plant 11 is compared with the load data of a neural network generated by a neural network calculation section 21 by the reference pattern learning in advance to estimate the abnormal position of the plant 11. An abnormal position information output section 5 retrieves the abnormality cause corresponding to the transient pattern and the apparatus name among the items assumed when the reference pattern is generated based on the estimated data and outputs them as the diagnostic information. The pattern comparison utilizing the characteristic of the neural network is performed. The data of the reference pattern are used in the form of the load data obtained by the learning of the neural network, and high-speed processing can be performed with a small memory capacity.</p> |
