针对复杂环境下农机设备的齿轮箱系统在故障诊断时存在易受现场噪声干扰和故障识别率低等问题,提出了一种基于改进的烟花算法和概率神经网络的齿轮箱智能故障诊断方法。为提高现有概率神经网络模式分类方法的性能,定义了一项样本相似度衡量指标以提高建模过程中训练样本的质量。将烟花算法与概率神经网络技术有机融合提出了一种改进的烟花算法-概率神经网络模式分类方法,利用烟花算法优化概率神经网络的平滑参数以确定网络参数的最优值,提高模式分类与识别精度。将改进的烟花算法-概率神经网络模式分类方法用于噪声环境下齿轮箱的故障诊断建模,构建故障特征参量与齿轮箱工作状况间的复杂非线性映射关系。应用结果表明,与基于BP神经网络、GABP(genetic algorithm back propagation)神经网络和概率神经网络的故障诊断模型相比,在不同程度噪声影响下烟花算法-概率神经网络模型均具有最高故障识别率。当噪声控制系数为0.01、0.02、0.04和0.06时,模型的故障识别率分别为100%、95.83%、93.33%和88.33%。该研究可为非线性复杂系统的故障诊断提供了一种可行的解决方案。
采用对硝基苯酚法建立爪哇正青霉生物转化合成8-异戊烯基柚皮素体系中O-脱甲基酶酶活的分光光度测定方法;研究生物催化合成8-异戊烯基柚皮素的过程;采用3种经典的细胞色素P450酶抑制剂对P450酶和生物转化反应之间的关系进行鉴定;研究O-脱甲基酶的酶学性质。研究结果表明,生物转化体系中O-脱甲基酶酶活力在第4天达到最大值,其最适反应温度60℃,最适p H 8.0。
In order to produce macromolecules by using the immobilized cells in microcapsules, a novel macro-porous NaCS-PDMDAAC capsules system was used to immobilize Candida valida in a semi-continuous culture mode for producing lipase. It was found that the encapsulation system had no negative effect on the growth and lipase productivity, and the lipase produced inside capsules could diffuse freely to the media outside capsules. The cultivation with the capsules was carried out for 15 batches and the lipase in each batch had almost the same activity. In comparison with free cell cultivation, the immobilized cell fermentation shortened obviously the period of glucose consumption from 24 hours to 8 hours and increased lipase productivity.