基于群体智能的柴油机性能评估与故障预测研究

The diesel engine, as the most widely used power machinery, plays an important role in all fields of national economy. The diesel engine detecting and diagnosing technique provides an effective means of forecasting faults in the early stage to ensure the safe and efficient operation of the engine. Big power diesel engines used for marine ships and locomotives are multi-cylinder engines with the similar structure, function and working conditions, which together provide torsion to the crank. The technical state of the cylinders reflects the health condition of the engine factually. Based on the diesel structure and operation character, the project proposes a novel diesel diagnosing method utilizing the cylinder performance discrepancy. The research includes: the diesel multi-cylinder is abstracted to ant swarm or bee swarm, then based on the Swarm Intelligence technology, the cylinder swarm operation character is compared to find the abnormal one with its relative operation parameters; on the bases multi-variable grey MGM model is adopted to forecast the transformation trend of diesel latency fault; Blending the cluster analysis results and diesel expert knowledge with the Bayesian Network, then the fault types and fault reasons will be got exactly. The resear project aims at providing novel diesel fault diagnosis method under the condition of fault sample deficiency, and the study results will help to improve the operation security and fuel economy for diesel power system.

柴油机作为最广泛应用的动力机械在国民经济各领域中发挥着重要作用，柴油机监测诊断技术是实现故障早期预报、确保柴油机安全、高效运行的有效手段。机车用、船用等大功率柴油机均为多缸柴油机，各气缸的结构和功能一致、工作条件相同，彼此协调共同向主轴输出扭矩，其技术状态能够切实反映整机的健康状况。根据柴油机特有的结构和运行特点，本研究提出利用故障所引起的气缸间性能差异实现柴油机监测诊断的新方法。研究内容包括：将柴油机"气缸群"抽象为"蚁群、蜂群"，基于群体智能聚类方法实现"气缸群"性能的自组织横向比较，发现异常气缸及相关运行参数；在此基础上，建立多变量灰色MGM模型，定量预测潜在故障的劣化趋势；采用贝叶斯网络融合"气缸群"聚类分析结果与领域专家知识，准确诊断柴油机故障类型及原因。本课题旨在探索故障样本数据缺乏条件下，实现柴油机故障预测的新途径，研究成果对提高柴油机动力系统的安全性、经济性具有重要意义。

柴油机作为一种重要的动力机械，广泛的应用在国民生产的各项领域中。柴油机能否安全稳定的运行直接影响整套设备的工作状态。如果柴油机发生故障，很有可能导致重大的人员伤亡和经济损失。车用、船舶用等大功率柴油机均为多缸柴油机，各气缸的结构和功能一致，工作条件相同，彼此独立运行，共同向主轴输出扭矩，“气缸的性能状态”能够切实反映整机的健康状况。根据柴油机所特有的结构和运行特点，本研究首先将柴油机内配置的多个结构和功能相同的气缸抽象成气缸“群体”，单个气缸则看作是独立“个体”，通过监测柴油机气缸的运行参数获得“个体”的属性特征，然后基于蚁群聚类算法（Ant Colony Clustering Algorithm）实现“个体”依据其属性特征的自组织横向比较，自动分离出性能异常或故障气缸。将离群分析方法引入到柴油机故障诊断研究中，提出通过计算“离群因子（CBGOF）”的方法，定量分析离群气缸相对于正常“气缸群”的性能退化程度。在此基础上，建立灰色预测GM模型，定量预测异常气缸“离群因子”的变化趋势，借此反映柴油机气缸性能的退化状态；采用贝叶斯网络融合“气缸群”聚类分析结果与领域专家知识，处理气缸故障与征兆间的复杂耦合关系，准确诊断柴油机故障类型及原因。本研究利用TBD234V12柴油机GT-power模型和上柴SC5D122D型柴油机实验台，模拟气缸供油不足、喷油角改变、排气门间隙劣变等柴油机故障模式，对所提出的算法进行了有效性验证。