Network type :
For now I've only tried NeuroShell 2's Beginner's System on the very first version of the input file since it is only a compatibility test.
This System is using a uses a three layer Backpropagation network.
Data :
The input file is using this format:
Red20, Red19, ..., Red2, Red1, RedNow, Blue20, Blue19, ..., Blue2, Blue1, BlueNow, AlarmType
93.4, 93.6, ..., 103.6, 100.8, 103.6, 92.9, 95.7, ..., 124.5, 129.6, 124.5, 0.25
112.1, 114.9, ..., 109.2, 107.4, 109.2, 124.8, 126.2, ..., 126.2, 123.4, 126.2, 0.25
129.5, 131.2, ..., 108.9, 110.8, 108.9, 112.8, 110.4, ..., 94.2, 92.8, 94.2, 0.25
138.9, 137.6, ..., 135.4, 131.0, 135.4, 134.2, 130.0, ..., 148.8, 148.2, 148.8, -0.25
146.7, 146.9, ..., 132.1, 130.4, 132.1, 135.4, 134.8, ..., 157.9, 156.7, 157.9, 0.25
The red part contains the temperatures on the upper thermocouple during the 20 seconds preceding the alarm. The blue part contains the temperatures on the lower thermocouple during the 20 seconds preceding the alarm. The last column is the alarm type when it's known.
The value are:
- 0.75: True alarm
- 0.25: Maybe an alarm
- -0.25: False alarm
- -0.75: Not an alarm (normal situation)
Conclusion :
There is still a lot of work...
|
