摘要
安全阀是承压设备的超压保护装置,它的可靠工作对现代企业的安全运行起着至关重要的作用。在坚持“安全第一,预防为主”生产方针的前提下,建立和完善安全法规、管理体系在高危行业中就显得尤为重要。安全评价是安全管理中的关键技术,科学、系统地开展安全评价工作,对减少事故发生,减轻人员伤亡和财产损失具有重要的意义。本文主要对安全阀失效进行评价,以降低其运行风险,开展的主要工作归纳如下。
在实际使用过程中安全阀的失效涉及诸多因素且易受随机性因素的影响,具有模糊性和不确定性的特点,这些特点决定了系统状态的变化并不按照某一特定的规律或函数变化,整个系统是非线性动力学的。首先,本文将安全阀失效的相关影响因素作为因素集,把可能产生故障的元件作为评价集,采用模糊综合评价方法对安全阀的失效进行综合评价。然后,尝试着将模糊综合评价的分析思路应用到BP(Back Propagation)人工神经网络(Artificial Neural Network,ANN)模型中,利用MATLAB神经网络工具箱GUI,对样本数据进行仿真,接着采用训练好的网络对安全阀的现状进行评价。
在两种方法的比较中发现,模糊综合评价法在确定权重时,采用层次分析法和专家评判相结合,将定性问题转化为定量问题,可较为准确的反映因素集中各因素的权重分布情况。但由于因素之间的相互关联较为复杂,两两比较出来的权重带有一定的主观色彩,因此还需寻找另一种更为简便、科学的方法,使评价的结果更为客观、公正、合理。
BP神经网络的非线性动力学特性与安全阀失效的规律相耦合,克服了传统评价方法在解决非线性动力学问题上存在的缺陷,运用BP神经网络对安全阀的现状进行评价,得到的结果与实际检测结果一致,且快速、方便;通过对样本的不断训练,BP网络使权值收敛在一个稳定的范围,避免了主观因素的影响;同时BP人工神经网络还具有自适用性、自学习性、信息处理的并行性、结构的可塑性和良好的容错性等优点,因此将BP神经网络用于安全阀失效评价具有良好的评价效果和重要的实用价值。
Safety valve is the protection against overpressure device of compressive equipment, its reliable operation plays an important role for the modern enterprise. It is of great important to establish and perfect safety laws and regulations as well as management systems, so as to make sure that the production can go on safely. Persisting in "Firstly safety, Mainly prevention" under the premise of production polity. As the key technology of the safety management, it is significance to carry out safety assessment scientifically and systematically in order to reduce accidents, and casualties and property losses. Secondly, this paper focuses on the failure evaluation of safety valves, in order to reduce the risk of operation. The main work was summarized as follows.
The failure of safety valves is affected by many factors, and even many random factors easily, so there are characters of fuzziness and uncertainty. That determine the changes of system state do not according to a specific rules or function, but it is nonlinear dynamics. First, the model for the fuzzy comprehensive assessment of safety valve is proposed, in this model, the factors affecting the failure of safety valve is taken as the factor set; the elements of the safety valve, which are likely to fail, are taken as the assessment set. Secondly it was tried that analytical method of fuzzy comprehensive assessment is drawn into BP neural network model(NNS), this apply to failure analysis of safety valve. Based on fuzzy comprehensive assessment, BP neural network model of failure valuation was established. Then the sample data was emulated by using neural network toolbox in the MATLAB software-GUI. Lastly safety valve was valuated by using the trained network, its results are same as the practical results.
In comparison of the two method, we found that in the process of fuzzy comprehensive assessment, the weight of the element are decided according to analytic hierarchy process and expert evaluation, this method could relatively accurate the failure of safety valve. But because of the relationship between the factors are complex, there are certain subjective color in weight by using pairwise comparison. Therefore it is need to find another more simple and scientific method, this method will make the evaluation results becomes more objective, fair and reasonable.
BP Artificial neural network could compose new information processing system according to imitate human nervous system, its nonlinear dynamic is adapted to the rule of failure of safety valve, this overcomes the traditional evaluation method in solving nonlinear dynamics problems existing defects; Using BP neural network to evaluate the present status of safety valves, its results are consistent as the practical results, and overcome the subjective factors by; BP network makes weights converge to a stable range by using training sample; And there are the characters of the self-adaptability, self-study, concurrency of information processing, structural plasticity, good fault tolerance and so on. So application of ANN technique in safety evaluation has feasibility and strong adaptability.
引文
[1]赵铁锤,杨富,王如君等.安全评价(修订版).北京:煤炭工业出版社,2004,12-13
[2]张海波.建(构)筑物拆除爆破安全评价方法及应用研究:[广西大学硕士学位论文].南宁:广西大学,2007,2-3
[3]徐满贯.煤矿动态综合安全评价模式及应用研究:[西安科技大学博士学位论文].西安:西安科技大学,2006,4-5,16-18
[4]刘菊梅.陆上油气钻井作业安全评价方法的研究:[重庆大学硕士学位论文].重庆:重庆大学,2005,3-4
[5]李学桥,马莉.神经网络工程应用.重庆:重庆大学出版社,1996,2-3
[6]高隽.人工神经网络原理及仿真实例.北京:机械工业出版社,2003,17-18
[7]宋庆克,汪希龄,胡铁牛.多属性评价方法及发展评述.决策与决策支持系统,1997,4(7):128-138
[8]谢亚东,周国庆,周大鹏.压力容器的使用管理与安全运行.北京:原子能出版社,1995,157-158
[9]黄光禹.安全阀.上海:上海科技出版社,1982,11-12
[10]杨贵萍,白云峰,方静.安全阀常见故障分析及排除方法.农机使用与维修,2007(2):42-43
[11]王磊.安全阀常见故障分析及排除方法研究.吉林化工学院学报,2007,24(2):58-60
[12]王俊杰,朱国东.安全阀常见故障分析与排除方法.特种设备与安全,2005(6):37,40
[13]赵春岳,吉守军.浅谈弹簧式安全阀的选用和调试.阀门,2003(4):34-36
[14]马凌运.安全阀的正确选用与安装.化工劳动保护(安全技术与管理分册),1997,99(5):14-15
[15]魏新利,李慧萍,王自健.工业生产过程安全评价.北京:化工工业出版社,2005,12-13
[16]张智超.化工装置安全评价软件研究与实现:[大连理工大学硕士学位论文].南宁:广西大学,2007,2-3
[17]崔晓红.人工神经网络技术在安全评价中的应用:[西安建筑科技大学硕士学位论文].西安:西安建筑科技大学,2005,1,7-11,28-29
[18]蒋军成,郭振龙.工业装置安全卫生预评价方法(第二版).北京:化工工业 出版社,2004,15-16
[19]孙连捷.安全评价和安全评价的论述.劳动保护科学技术,1999,19(1):24-26
[20]施式亮.矿井安全非线性动力学评价模型及应用研究:[中南大学博士学位论文].长沙:中南大学,2001,19-23
[21]罗云,樊运晓,马晓春.安全健康新知丛书-风险分析与安全评价.北京:化学工业出版社,2004,144-152
[22]缪春生,赵建平.压力容器模糊分析方法的研究(一)—固有危险性评价和风险水平评价.压力容器,2005,22(3):1-5
[23]缪春生,赵建平.压力容器模糊分析方法的研究(二)—风险控制状态评价和风险可接受准则讨论.压力容器,2005,22(4):1-4
[24]龙勇.煤矿安全模糊综合评价理论与实践:[辽宁工程技术大学硕士学位论文].阜新:辽宁工程技术大学,2006,25-37
[25]吴丽萍.模糊综合评方法及其应用:[太原理工大学硕士学位论文].太原:太原理工大学,2006,36-43
[26]金承尧.石化装置在役安全阀的风险评价和风险管理:[南京工业大学硕士学位论文].南京:南京工业大学,2004,20-23
[27]API581.American Petroleum Institute:Risk-based Inspection Based Resource Document.First Edition,2000,5
[28]Chen Z,Huang G H,Chakma A.Hybrid Fuzzy-Stochastic Modeling Approach for Assessing.Journal of Environmental Engineering.2003(1):79-88
[29]Yin Y Y,Huang G H,Hipel K W.Fuzzy Relation Analysis for Multicriteria Water Resources Management.Journal of Water Resources Planning and Management.1999(1):41-47
[30]Wang Jia ding.The Analysis of loess Slop Instability Based on Fuzzy Information Method.BUSEFAL,1993:83-92
[31]彭祖赠,孙韫玉.模糊(Fuzzy)数学及其应用.武汉:武汉大学出版社,2002,147
[32]飞科科技产品研发中心.MATLAB6.5辅助神经网络分析与设计.北京:电子工业出版社,2003,9-11,64-69
[33]韩立群.人工神经网络教程.北京:北京邮电大学出版社,2007,14-15,58
[34]周开利,康耀红.神经网络模型及其MATLAB仿真程序设计.北京:清华大学出版社,2005,69
[35]Moller M F.A scaled conjugate gradient algorithm for fast supervised learning.Neural Networks,1993,6:525-533
[36]Charalambous C.Conjugate gradient algorithm for efficient training of artificial neural networks.IEEE Proceedings,1992,3(139):301-310
[37]Battiti R.First and second order methods for learning:Between steepest descent and Newton's method.Neural Computation,1992,2(4):141-166
[38]Foresee F D,Hagan M T.Gauss-Newton approximation to Bayesian regularization.Proceedings of the 1997 International Joint Conference on Networks,1997,1930-1935
[39]Ahmad S,Tesuro G.Scaling and Generalization in Neural Networks.Proc.of 1988Connectionist Models Summer School.Canada:1989
[40]Kohonen T.Self-Organization and Associative Memory.New York:Springer Verlag.1989
[41]Jacek M Z.Introduction to Artificial Neural Systems.St.Paul:West Publishing Company,1992
[42]马骥.基于MATLAB的BP神经网络在砂土液化评价中的应用:[南京大学硕士学位论文].南京:南京大学水文学及水资源(地质工程),2002,5-6
[43]闻新,周露,王丹力等.MATLAB神经网络应用设计.北京:科技出版社,2000,11-13