【作者】 徐红波；

【导师】 陈国华；

【作者基本信息】 华南理工大学 ， 化工过程机械， 2012， 博士

【摘要】 我国当前使用的桥式起重机大部分为20世纪80年代以前生产的，除少量退役报废外，大部分仍在继续使用，老化问题日益突出，超期使用依据不足，风险性较大。随着我国现代化建设事业的发展和改革的不断深入，生产应用中对桥式起重机的承载能力和可靠性提出了更高的要求。因此，如何科学评估起重机主梁结构的实际性态，是保障桥式起重机的安全使用，预防和控制重大事故，确保企业安全、高效生产的重要问题。同时，齿轮减速器因传动力矩大、工作环境恶劣等，导致其成为起重机最容易出现故障的部分，其运行状态往往直接影响到起重机是否正常工作。为了评估桥式起重机使用的安全可靠性，建立一套桥式起重机主梁及减速器的健康评估方法就显得尤为重要。论文重点开展下列研究工作：(1)桥式起重机主梁有限元模型修正技术的研究针对桥式起重机主梁实际结构与有限元计算模型之间存在的差异，研究了有限元模型修正的一般理论，提出了通过灵敏度分析来实现对桥式起重机有限元模型修正的方法。在结构有限元模型建模过程中，结合实测频率确定修正目标函数，对结构模型进行修正，以得出修正后结构模型的参数，最后根据结构模型修正所得到的参数，获得修正模型。修正后的有限元模型更能反映结构的实际动力特性，为主梁损伤评价方法的提出奠定了坚实的基础。(2)基于动力指纹的起重机主梁损伤分析为获得一种对主梁损伤较为敏感的动力指纹，着重研究了基于频率变化、基于振型变化、基于曲率模态、基于柔度差曲率和基于改进刚度指标五种动力指纹的特点，明确了基于动力指纹的主梁损伤分析步骤并将五种指纹用于桥式起重机主梁的损伤评定。通过将五种指标用于主梁损伤定位分析，发现曲率模态和改进刚度指标在结构前三阶的定位效果最好。为此，将曲率模态和改进刚度指标用于损伤程度识别，发现基于改进刚度指标指纹的程度识别能力较曲率模态指纹的程度识别能力更为敏感。说明改进刚度指标是一种优秀的损伤识别动力指纹。这为验证提出的主梁损伤评价方法提供了保证。(3)基于支持向量机的起重机主梁损伤评价通过对支持向量机理论和具体实现算法的分析，成功地将改进刚度指标(SVI)引入支持向量机(SVM)，确定了SVM用于损伤评价的最佳参数，实现了对桥式起重机主梁损伤的识别。为进一步提高评价精度与速率，对粒子群算法进行了改进，并用改进的算法对最小二乘支持向量机(LSSVM)的正则化参数和核函数参数进行了优化，优化后的方法被用到起重机损伤诊断中。结果证实该方法具有比传统分类算法更出色的分辨能力。(4)基于EMD与ARMA模型倒谱分析的桥式起重机减速器故障诊断简明阐述了桥式起重机减速器故障诊断的基本原理，并总结了齿轮和滚动轴承发生故障时出现的故障特征频率。针对桥式起重机减速器振动信号混杂的特点，提出了EMD与ARMA模型参数化倒双谱分析相结合的诊断方法。首先，对提取的实测截取信号进行AR模型延拓，抑制EMD分解的边界效应；然后，借助ARMA模型对延拓分解得到的各平稳信号序列建立精确模型，并对模型实施参数化切片倒谱运算，从而实现有效地提取故障特征信息的目的。(5)桥式起重机主梁及减速器的动力测试为获得用于主梁模型修正的实测频率数据和用于减速器诊断的实测故障信号，对桥式起重机主梁及减速器进行了动力学实验研究。通过对桥式起重机主梁金属骨架的模态测试，获得了主梁激振实验的振动衰减信号，并对测取的信号进行了传函分析，经参数识别方法确定了主梁振动的前三阶频率，为主梁的仿真模型修正提供了基础和支持。同时，通过对主梁起升机构的齿轮减速器的故障诊断测试，获取了在役减速器振动信号，从而为判断机器内部的故障原因及其故障的性质提供条件，为桥式起重机的实际性态作出客观判断与评价提供了依据。更多还原

【Abstract】 At present, most of overhead travlling cranes, used in our country, are produced beforethe1980s. They are servicing with the exception of that a few are scraped. Aging has becomean increasingly important issue for the serviced and the risk is becoming greater and greater.However, with the development of chinese construction industry and the deepening of thereform, higher request in practice is proposed for the bearing capacity and reliability ofoverhead travlling cranes. So, how to scientific evaluate the real state on girder of overheadtravlling crane is the key to avoid and control catastrophe failure in enterprise production.Meanwhile, gear reducer has a big transmission torque and poor working conditions, etc, itcauses reducer become the part of the most prone to go wrong. So reducer’s running statedirectly affects crane’s normal work. In order to assess the reliability of overhead travllingcrane so as to make scientific decision about normal use, downgraded or scraped and so on,dynamic evaluation is played in the thesis.The main research work is conducted as follows:(1) Research on the girder finite element model updating techniques.Targeting the difference between the finite element model and real girder structure, thetraditional theory of finite element model updating is analysed and a few kinds of methods,commonly used, are compared.Then the method, based on sensitivity analysis, is proposed toupdate the girder finite element model of overhead travlling crane. In the process of modeling,modification of the objective function is on the basis of the measured frequency. Aftermodifying the structural model, optimized calculating parameters are obtained, then themodel is updated. The updated finite element model reflected actual dynamic characteristicsin an even better fashion. It lays solid foundation for further analysis.(2) Damage analysis on modal dynamic fingerprint of the girder.The dynamic fingerprints, based on frequency variation, mode variation, curvature modevariation, flexibility variation curvature and stiffness variation index, are introduced and thecomputational method of dynamic fingerprints are expounded. These indexes are used on anoverhead travlling crane to compare damage identification capability. It is found that theperformance of curvature mode variation fingerprint and stiffness variation index fingerprintis the best when those dynamic fingerprints are used to locate damage. Then curvature modevariation fingerprint and stiffness variation index fingerprint are used to calculate damagelevels and the results show stiffness variation index fingerprint has a more accurateidentification capability. So, stiffness variation index is an excellent fingerprint. (3) Damage evaluation of crane’s girder based on support vector machine.The algorithm of support vector machine(SVM) and the principle of the kernel functionare analysed in theory. Then SVIs are input into SVM and appropriate kernel function isselected. With test, the best parameters are calculated to identify gieder damage of crane.Tofurther improve the identification accuracy and convergence rate, particle swam optimization(PSO) algorithm is introduced and improved. The improved PSO is used to optimize theparameters of least square support vector machine(LSSVM). After that, the method is used toidentify girder damage of overhead travlling crane.The result demonstrates IPSO-LSSVM hasbetter identifition ability than traditional classification algorithm.(4) Research on the application of EMD and ARMA bi-cepstrum fault diagnosismethod in gearbox of overhead traveling crane.Basic principle of fault diagnosis about crane gear reducer is concisely expounded andpossible characteristic frequency when gear and rolling bearing occurs fault is summarized.Targeting the time varying of signals measured on the surface of gear reducer, empirical modedecomposition(EMD) is occupied to modulate the signals; auto-regressive movingaverage(ARMA) method,which has higher accuracy and better applicable condition, is usedto establish model for the signal principal components of intrinsic mode functions from theEMD; then parametric bi-cepstrum estimation is implemented and accurate fault informationsare extracted. The method is applied in fault diagnosis of crane gear reducer, the resultsdemonstrate that: the presented method, combining EMD with parametric bi-cepstrum ofARMA model, can obtain the fault quefrency accurately to determine the actual conditions ofcrane gear reducer, and it is an effective health assessment method.(5) Dynamics test for girder and gearbox of overhead traveling craneThrough the modal test for the girder metal frame, fading signals of exciter test on girdermetal frame are collected. Frequency response analysis is conducted for the collected signalsand the informations of the first three modes are obtained with structural modal parameter id-entification method.This provides foundation and support for updating simulation model ofgirder metal frame.Otherwise, with fault diagnosis test on gear reducer of gieder lifting mech-anism, vibration signals are collected from reducer. This provides condition for deteminingthe failure causes or the fault nature.更多还原