| 摘要 |
<p>PURPOSE:To prevent the quality of water from deteriorating due to the rottenness of sludge and thereby improve durability significantly by making it possible to correct the non-uniformity of sludge flow distribution. CONSTITUTION:A stratified neural network 2 which controls an operation system 1 for operation initial sedimentation ponds 11, 12 enters a drawing flow amount A, an inflow amount C and an average drawn sludge concentration B as the initial value of each initial sedimentation pond 11, 12, then obtains control parameters for the rectification of an unbalance in sludge distribution, using sigmoid function as input/output function and thereby output the drawing time and drawing cycle time of each initial sedimentation pond 11, 12. In addition, an empirical rule is applied to an inflow amount, rainfall function and seasonal function in the stratified neural network 2, in order to allow the execution of learning at high speed. Further, the network 2 self-evaluates its output, the drawing amount A and the drawn sludge concentration B of each initial sedimentation pond 11, 12 to self-create learning data for supply.</p> |
