基于支持向量机的组合分类方法及应用

为了解决采用神经网络、决策树作为弱分类器的AdaBoost组合分类存在的不足,进一步改善组合分类效果,提出采用支持向量机(SVM)作为弱分类器的一种新的组合分类诊断方法——AdaBoost-SVM。该方法没有采用一个固定的SVM的核参数,而是自适应调整SVM中的核参数,从而得到一组有效的SVM弱分类器。通过对基准数据库的测试及航空发动机故障样本的诊断,结果表明,所提AdaBoost-SVM方法较好地解决了现有的Ada-Boost组合分类方法中存在的弱分类器本身参数选取困难问题及训练轮数的合理选取问题,并具有更好的泛化性能,更适合对分散程度较大、聚类性较差的航空发动机故障样本进行分类。     

"问题联系式教学法"在诊断学教学中的应用

诊断学教学质量的提高，必须改变传统的教学模式，采用问题联系式教学方法，使师生互动，以培养学生的临床思维能力、分析问题和解决问题的能力为目的，从而激发学生的学习兴趣，调动学生的学习积极性。实践证明，问题联系式教学法是提高诊断学教学质量的有效途径。     

固体发动机失效树定量分析法

根据固体火箭发动机失效树定量分析中确定底事件失效概率存在的困难,提出了确定底事件失效概率的工程方法——征求意见评分法。通过对某型号发动机试样阶段结构可靠性实例分析,得出了较满意的结果。     

某型涡喷发动机的可靠性增长分析

针对某型涡喷发动机的一组含零故障的多台不同步截尾故障数据。给出了趋势检验、AMSAA模型的拟合优度检验及模型参数的极大似然估计方法。分析表明，该型发动机有显著的可靠性增长，且可用AMSAA模型拟合其故障数据，在考虑零故障后，其参数及MTBF的极大似然估计（MLE）更为可信。     

直接不交化方法在故障树定量分析中的应用

在布尔集合运算方法和不交化方法基础上，提出了单调关联故障树系统定量分析优化方法，通过故障树定量分析优化方法，可以方便地得到系统的可靠度、不可靠度、最小割集、最小路集、关键重要度等指标。     

非平稳信号的双重中值容错平滑算法

平滑是消除或削弱测量数据随机误差时常用的数据处理手段。文章建立了一组新的非平稳数据序列容错平滑算法，其采用两阶段两次滑动中值与滑动均值平滑组合，实现了数据简化与平滑的统一。理论分析和仿真计算结果证实，算法可有效避免斑点型异常数据的不利影响，有效削弱随机误差的影响，改进数据处理精度和数据处理结果可靠性。     

一个基于故障树模型的诊断系统

提出一种基于故障树模型的诊断方法，并基于该方法实现了一个验证性的诊断系统。基于故障树的层次诊断模型，提出了确定性推理方法和可能性推理方法及面向故障树的基于框架和广义规则的混合知识表示方法。最后在ＣＯＭ－ＰＡＱ４８６微机上，Ｗｉｎｄｏｗｓ环境下，用ｂｏｒｌａｎｄＣ＋＋实现了该诊断系统，并通过对一个电路系统实验台的诊断验证了方法的有效性。     

一种整体式天然气压缩机气阀故障诊断技术研究CSCD

整体式天然气压缩机在石油化工领域广泛应用,压缩机的正常工作对安全生产具有重要意义。为了有效诊断整体式压缩机气阀部件的故障,采用了Daubechies小波基函数对ZTY470型整体式天然气压缩机气阀振动信号的分解重构原理,详细介绍了以振动信号的能量为故障特征的故障诊断过程。对气阀阀片断裂和弹簧断裂的故障振动信号进行了小波包分析,并与正常工作时的信号进行了对比。实验表明该方法能准确的进行故障诊断和故障定位,证明了对整体式天然气压缩机气阀故障诊断的有效性。     

Multi-mode diagnosis of a gas turbine engine using an adaptive neuro-fuzzy system

Gas Turbine Engines (GTEs) are vastly used for generation of mechanical power in a wide range of applications from airplane propulsion systems to stationary power plants. The gas-path components of a GTE are exposed to harsh operating and ambient conditions, leading to several degradation mechanisms. Because GTE components are mostly inaccessible for direct measurements and their degradation levels must be inferred from the measurements of accessible parameters, it is a challenge to acquire reliable information on the degradation conditions of the parts in different fault modes. In this work, a data-driven fault detection and degradation estimation scheme is developed for GTE diagnostics based on an Adaptive Neuro-Fuzzy Inference System (ANFIS). To verify the performance and accuracy of the developed diagnostic framework on GTE data, an ensemble of measurable gas path parameters has been generated by a high-fidelity GTE model under (a) diverse ambient conditions and control settings, (b) every possible combination of degradation symptoms, and (c) a broad range of signal to noise ratios. The results prove the competency of the developed framework in fault diagnostics and reveal the sensitivity of diagnostic results to measurement noise for different degradation symptoms.     

Simultaneous state and actuator fault estimation for satellite attitude control systems

In this paper, a new nonlinear augmented observer is proposed and applied to satellite attitude control systems. The observer can estimate system state and actuator fault simultaneously. It can enhance the performances of rapidly-varying faults estimation. Only original system matrices are adopted in the parameter design. The considered faults can be unbounded, and the proposed augmented observer can estimate a large class of faults. Systems without disturbances and the fault whose finite times derivatives are zero piecewise are initially considered, followed by a discussion of a general situation where the system is subject to disturbances and the finite times derivatives of the faults are not null but bounded. For the considered nonlinear system, convergence conditions of the observer are provided and the stability analysis is performed using Lyapunov direct method. Then a feasible algorithm is explored to compute the observer parameters using linear matrix inequalities (LMIs). Finally, the effectiveness of the proposed approach is illustrated by considering an example of a closed-loop satellite attitude control system. The simulation results show satisfactory perfor-mance in estimating states and actuator faults. It also shows that multiple faults can be estimated successfully.