神经网络可拓融合技术在生物废气检测中的应用

将生物滴滤塔有机废气处理过程视为一个可控工业生态环境，以净化废气的流量、循环液流量、pH值、塔内温度等过程参数作为可拓物元，采用神经网络模型的可拓融合的方法对工业生态环境中微生物的反应过程的状态检测进行了研究。通过仿真和实验发现净化处理过程适合用人工神经网络（ANN）模型建模，而通过对遗传算法的改进能够提高其收敛速度。采用改进遗传算法（MGA）与LMBP算法相结合（MGA—LMBP），建立了神经网络模拟生物滴滤塔处理有机废气的过程可拓融合模型，利用已有的实验数据样本训练神经网络模型，取得了较好的效果。

APPLICATION OF INFORMATION FUSION TECHNOLOGY BASED ON NEURAL NETWORKS IN DETECTING THE TREATMENT PROCESS OF BIOLOGICAL WASTE GASES

Regarding the biochemical reactions in waste gas treatment process by a trickling bio-filter as a controllable industrial environment and taking the process parameters(the flux of purified waste gas,the flux of the cycle liquid,pH and temperature) as extendable units,the authors applied the extendable fusion method of neural networks to study the state detection of biochemical reaction processes of the microorganisms in the industrial environment.Simulation and experiment showed that Artificial Neural Networks are preferable for the establishment of the model for the purification treatment process and modification of genic algorithm(GA) can improve its convergent velocity.By adopting the hybrid algotithm(MGA-LMBP),which is composed of a modified GA(MGA) and a Levenberg-Marquardt BP algorithm(LMBP),an extendable fusion model was established for the simulation of the process of biological waste gas treatment in the trickling bio-filter by neural networks.Training the neural networks with the available experiment data gave satisfactory results.