用于煤自然发火期预测的神经网络模型和实验技术

根据煤的硫分、灰分以及煤自燃过程中的耗氧速率、CO和CO2产生率等随温度变化的序列值与煤自然发火期之间存在的密切对应关系，建立了前向多层人工神经网络模型，用已有的煤自然发火实验数据对网络进行训练，得到了神经元间的联结强度，从而准确地表征这种对应关系．设计了一套油浴程序升温实验装置，确定了实验试管的尺寸和实验条件，从而能够准确测定煤自燃在不同温度下的耗氧速率及气体产生率．将煤样油浴程序升温实验数据及煤质分析数据代入人工神经网络，可算出煤的自然发火期．与煤自然发火实验相比，该方法测定煤样的自然发火期用煤量减少了99％以上，实验耗时缩短了90％以上，二者测试结果的偏差小于3d．

New Experimental Technique for Determining Coal Self-Ignition Duration

An artificial neural network(ANN) model is set up to depict relation between the self-ignition duration and sulfur content, ash content,oxygen consumption rate,carbon monoxide as well as carbon dioxide generation rate of coal at different temperature of self heating process.The data from self-ignition modeling experiments are used for ANN training to obtain connection strength between nerve cells.An oil-bath programmed temperature experiment device is designed and the experimental condition and size of the test tube are determined for testing oxygen consumption and gases generation rate of coal during self-heating process precisely.The sulfur content,ash content and the data from the oil-bath experiment are taken as ANN input to evaluate the experiment self-ignition duration of coal.Compared with self-ignition modeling experiment,less than 1% of coal sample and 10% of time are required with error less than 3 d to test self-ignition duration of coal.