多传感器论文：基于小波神经网络多传感器融合技术的研究多传感器论文：基于小波神经网络多传感器融合技术的研究.doc

多传感器论文：基于小波神经网络多传感器融合技术的研究 【中文摘要】多传感器信息融合的实质是对不确定信息的处理,它需要能够处理不确定信息的数学方法。小波神经网络是小波变换理论与人工神经网络的相互结合的一种新的神经网络模型,它将小波变换良好的时频局域化性质和神经网络的自学习能力相互结合起来,具有较强的逼近能力和容错能力,在处理一些复杂的非线性、不确定系统等问题上表现出了优于传统神经网络的收敛速度、容错能力等,具有广泛的应用前景。本文利用信息融合将煤矿井下检测的多种信号采用经过遗传算法改进的小波神经网络进行融合处理。文章分析了多传感器信息融合、小波变换、人工神经网络、遗传算法的基本理论、运行机理以及各自的优缺点。通过对优缺点的分析,从而提取它们的优点并将它们融合起来组成一个新的融合算法,为工程实际应用提供理论基础。由于传统的BP网络Sigmoid函数自身的特性,实际上神经网络只是一个次优网络,有待于改进；BP算法采用误差函数梯度下降的方式进行迭代,收敛速度很慢,而且容易陷入局部极值的处境,并且具体的极值位置与权值的初始化数值有关。本文所采用的是遗传算法对小波网络中的参数进行优化学习,实现全局寻优,避免局部极值。本文针对煤矿瓦斯涌出量的监测问题进行了研究,设计了一种基于多传感器信息融合的模型,其融合算法采用小波神经网络和遗传算法,并用BP标准算法与它进行比较。仿真的结果表明该算法得到的监测误差精度有了很大提高,并且网络收敛的速度明显加快,监测效果很理想。 【英文摘要】Multi-sensor information fusion is the essence of dealing with uncertain information; uncertain information needs to be able to handle the mathematical method. Wavelet neural network is combined a wavelet transform theory with artificial neural networks. A new neural network model, which with a strong approximation ability combines the wavelet transform time-frequency localization good nature and self-learning ability of neural networks, and fault tolerance in dealing with complex non-linear, uncertain system is superior to other issues to show the convergence speed of traditional neural networks, fault tolerance, with a wide range of applications. In this paper, information fusion will be a variety of coal mine detection signal by wavelet neural network and genetic algorithm fusion.This paper analyzes the multi-sensor information fusion, wavelet transform, artificial neural networks, genetic algorithms the basic theory, operation mechanism, and their respective advantages and disadvantages. Characteristics of BP network the Sigmoid function itself, in fact, only a sub-optimal neural network need to be improved; BP algorithm uses the error function gradient descent iterative way, the convergence speed is very slow, and easily falls into local optimum, and the specific The extreme position and the initial value of the relevant weights. W