| 摘要 |
PURPOSE: To prevent such a situation that the past learning results are all wasted by checking every neural net at the time of restarting due to power abnormality or the like, and using an optimum neural net among usable neural nets for assignment. CONSTITUTION: A group control device 10 is provided with a CPU 11, and a learning neural net is updated by a learning means every completion of learning. An optimum elevator is judged by an assignment judging means 14 from the output pattern of an assigning neural net 12b, and an assignment signal is outputted. The group control device 10 is reset and restarted at the time of detecting abnormality by an abnormality generation monitoring means 15. At the time of restarting the group control device 10, the use propriety of neural nets 12a-12c is judged by a use propriety judging means 16. As a result, one of the neural nets judged to be usable is selected as the neural net 12c for assignment by a neural net selecting means 17 according to the specified priority. |
