多重干扰下采煤机截割传动系统复合故障特征增强解耦及诊断方法研究

Extreme severe working conditions lead to frequent compound faults of shearer cutting transmission system, which seriously affects its production efficiency and even causes safety accidents. The multiple interferences generated by cutting coal and the weakness and coupling of the features of compound faults bring challenges to the compound faults diagnosis of shearer cutting transmission system. This project develops the research on feature enhanced decoupling and diagnosis method of compound faults of shearer cutting transmission system under multiple interferences, and its main research contents are as follows: A feature constrained – non-convex penalty regularized sparse representation method based on self-learning joint sparse dictionary is proposed to eliminate the strong noise and impact interference generated by cutting coal, and a tacholess order tracking model based on the adaptive estimation of instantaneous frequency is constructed, and the influence of the cutting speed fluctuation can be eliminated. Aiming at the weakness and coupling of the features of compound faults for cutting transmission system, an enhanced decoupling method of the feature signals of compound faults based on the blind separation of vibration sources sparse-lifting in hyperplane space is proposed, which can accurately separate the feature signals of multiple fault sources. And then, a feature quantitative fusion diagnosis model of multi-sensor based on self-adaptive neighborhood semi-supervised linear local tangent space arrangement is established, which ensures the accurate diagnosis of compound faults of shearer cutting transmission system. The research results provide the theoretical and technical support for the health monitoring, reliability improvement and efficient and safe operation of shearer.

极端恶劣工况导致采煤机截割传动系统复合故障频发，严重影响其高效生产，甚至引发安全事故。而采煤机截割煤岩产生的多重干扰及截割传动系统复合故障特征表现出的微弱性和耦合性，给截割传动系统复合故障精确诊断带来挑战。本项目开展多重干扰下采煤机截割传动系统复合故障特征增强解耦及诊断方法研究，主要包括：提出基于自学习联合稀疏字典的特征约束-非凸罚正则化稀疏表示方法，消除截割煤岩产生的强背景噪声和落煤冲击干扰；构建瞬时频率自适应准确估计的无键相阶次跟踪模型，消除截割转速波动影响；针对截割传动系统复合故障特征的微弱性和耦合性，提出基于超平面空间振源稀疏提升盲分离的复合故障特征信号增强解耦方法，准确分离多个故障振源特征信号；进而建立基于自适应邻域半监督线性局部切空间排列的多传感特征量化融合诊断模型，确保截割传动系统复合故障得到精确诊断。研究成果为采煤机的健康监测、可靠性提升及高效安全运行提供理论和技术保障。

针对极端恶劣工况导致采煤机截割传动系统复合故障频发，严重影响其高效生产，甚至引发安全事故的现状。本项目以工况多重干扰消除，故障特征增强提取和特征量化诊断为主线，开展了理论分析和实验研究。搭建了多重干扰下的传动系统故障模拟实验平台，具备了多重干扰下传动系统故障模拟实验条件；提出了基于IMRSVD和有效分量筛选的振动信号预处理方法，消除外界工况产生的强噪声背景干扰，有效提高振动信号质量；提出了基于同步挤压小波变换和转速脊线提取的传动系统变转速信号重构方法，实现了转速瞬时频率自适应准确估计，消除了传动系统转速波动对振动信号的干扰影响；建立了融入Fisher鉴别准则的标签一致稀疏表示模型，能够学习传动系统振动信号稀疏字典，实现传动系统振动信号的稀疏表示和故障诊断；提出了基于自适应最优最小熵反褶积的传动系统复合故障解耦提取模型，克服了复合故障特征所表现出的微弱性和耦合性，实现了复合故障特征的增强解耦；提出了基于融合型自动编码机的新型深度学习架构，保证了恶劣工况下故障特征提取过程的鲁棒性和适应性，实现了传动系统故障特征的自主自适应提取；建立了不同工况下基于降噪后CSCoh二维特征图和MSCNN的传动系统故障诊断模型，实现了传动系统故障特征的二维化表达和故障类型的精准诊断。研究成果有助于解决恶劣工况下传动系统振动信号预处理和故障诊断问题，为采煤机、掘进机、疏浚船及相关恶劣工况下大型机电装备传动系统故障的精准诊断提供理论支撑。