摘要
Forecast is very important for preventing and controlling the disaster of spontaneous combustion （sponcom）. Gaseous products of coal, such as carbon monoxide, ethylene, propane and hydrogen, are commonly used as indicators to reflect its status quo of sponcom in coal mines. Nevertheless, since the corresponding relationship between the temperature and the indicators is non-linear and can＇t be depicted with simple mathematical formula, it is very difficult to diagnose and forecast coal sponcom by monitoring indicator gases＇ distribution. A forward feeding 3-layer artificial neural network （ANN） model is employed to express the corresponding relation between temperature and index gases of coal sponcom more accurately. A large amount of data from programmed temperature oxidation experiments were employed to train the network to gain the connection strength between nerve cells and to accomplish the model. It proved in real coal productions that the ANN model can forecast coal sponcom accurately.
基金
Supported by the National Natural Science Foundation of China （10972178）
