发明名称 |
Verfahren zur Erkennung von fehlerhaften Vorhersagen in einer neuromodellgestützten oder neuronalen Regelung |
摘要 |
The neural network process control system is configured with five neural process models Ä1-5Ü, each of which has four inputs for signals from different process parameter sensor Ä6-9Ü outputs. Each model has an output fed to a common weighting stage Ä10Ü that provides an output to a regulator Ä11Ü coupled to the process. The regulator can be a fuzzy controller. The process neural models are trained with data sets and the weighting module determines a standard variation value relative to an average value. This is compared with a defined limit and if exceeded an alarm Ä12Ü is activated. |
申请公布号 |
DE19530049(A1) |
申请公布日期 |
1997.02.27 |
申请号 |
DE1995130049 |
申请日期 |
1995.08.16 |
申请人 |
FROESE, THOMAS DIPL.-ING., 41334 NETTETAL, DE |
发明人 |
FROESE, THOMAS DIPL.-ING., 41334 NETTETAL, DE |
分类号 |
G05B13/02;G06F11/16;G06F11/22;G06N3/04;(IPC1-7):G05B13/04;G05B23/02;G06G7/60 |
主分类号 |
G05B13/02 |
代理机构 |
|
代理人 |
|
主权项 |
|
地址 |
|