摘要 |
PURPOSE: To save the handling time of preparing a program and to facilitate a counter measure to a secular change and a difference in environment by obtaining a successive correction for an optimum lighting time value with learning only with the rough setting of a parameter. CONSTITUTION: In a thermal fixing device 30 in an image forming device such as a copying machine, a means 20 calculates a lighting time at a prescribed lighting time interval by a fussy neural network, from the surface temp. of a thermal fixing roller 14 obtained by the calculation of the detection output of a temperature detecting means 11 by a means 18 and a variation in a temp. obtained from the surface temp. by a means 19 and executes the drive of the lighting and turning off of a heater 15 by a heater control circuit 17. Moreover, the obtained lighting time is inputted to a means 21 and the surface temp. after the lighting is predicted. Then, the result of the prediction and the actual result are compared by a means 22 and when the difference between them is larger than a prescribed value, for instance 5 deg.C, the means 20 adjusts the weight of a network. |