一種基于虛擬樣本的磨機負荷參數(shù)軟測量方法
【技術(shù)領(lǐng)域】
[0001] 本發(fā)明設(shè)及軟測量領(lǐng)域,具體設(shè)及一種基于虛擬樣本的磨機負荷參數(shù)軟測量方 法。
【背景技術(shù)】
[0002] 磨礦過程的優(yōu)化運行控制需要準確檢測磨機內(nèi)的負荷參數(shù)(參見文獻[1] P.Zhou,T.Y.Chai,Η.Wang,"Intelligentoptimal-settingcontrolforgrinding circuitsofmineralprocessing, "IEEETransactionsonAutomationScienceand lingineering, 6(2009)730-743.和文獻[2]T.Y.Qiai,"Operationaloptimization andfeedbackcontrolforcomplexindustrialprocesses,"ActaAutomatica Sinica,39 (2013) 1744-1757)。磨機內(nèi)部數(shù)W萬計的鋼球分層排列,不同層的鋼球?qū)δC內(nèi) 部物料和磨機筒體的沖擊力具有不同的強度和周期。通常測量得到的筒體振動信號是具有 不同時間尺度的多個子信號的混合。筒體振動是磨機振聲信號的主要來源。因此,運些機 械振動和振聲信號具有非穩(wěn)態(tài)和多組分特征。優(yōu)秀的領(lǐng)域?qū)<彝ㄟ^同時考慮多種運行工況 和多種來源信息可有效監(jiān)視磨機負荷狀態(tài)和部分磨機內(nèi)部的負荷參數(shù)。研究表明,人耳可 W從磨機振聲信號中分辨出有價值信息。事實上,人耳是一組自適應(yīng)帶通濾波器,人腦具有 多層認知結(jié)構(gòu)。領(lǐng)域?qū)<铱蓮亩嘣刺卣骱投喾N運行工況中提取有價值信息進行決策。領(lǐng)域 專家經(jīng)驗的差異和有限的精力難W保證磨機長期工作在優(yōu)化負荷狀態(tài)。針對運些情況,很 有必要模擬領(lǐng)域?qū)<业恼J知過程建立磨機負荷參數(shù)軟測量模型。
[0003] 在時域內(nèi),磨機筒體振動和振聲內(nèi)的有價值信息被隱含在寬帶隨機噪聲中(參見 ^南犬[3]Y. ,Zen邑,E.Forssber邑,"Monitoring邑rindin邑parametersbyvibrationsignal me曰surement-曰prim曰ry曰pplic曰tion, ''Miner曰IsEngineering, 1994, 7 (4) :495-501.)。 基于機械振動和振聲信號的磨機負荷參數(shù)建模需要關(guān)注3個子問題:多組分信號自適應(yīng)分 解、多源譜特征自適應(yīng)選擇、基于選擇多種運行工況的軟測量模型構(gòu)建。
[0004] 研究表明,信號處理可W簡化特征的選擇和提取過程(參見文獻[4] S.Shukla,S.Mishra,andB.Singh,"PowerQualityEventClassificationUnder NoisyConditionsUsingEMD-BasedDe-NoisingTechniques, "IEEETransactionon IndustrialInformatics, 10(2014) 1044-1054.)。磨機負荷參數(shù)與筒體振動和振聲信 號的功率譜密度(PSD)密切相關(guān)(參見文獻[5]J.Tang,L.J.Zhao,J.W.Zhou,Η.化e,T. Υ.Chai,"Experimentalanalysisofwetmillloadbasedonvibrationsignalsof 1 油orato巧-scaleballmillshell,"Mineralslingineering, 23(2010)720-730.),但運 些譜數(shù)據(jù)通常包含成千上萬的特征。很多維數(shù)約簡算法用于處理具有該特點的數(shù)據(jù)(參見 文南犬[6]J.Tan邑,T.Y.Chai,W.Yu,L.J.Zhao,"Modelingloadparametersofballmill ingrindingprocessb曰sedonselectiveensemblemulti-sensorinform曰tion,''IEEE TransactionsonAutomationScienceandEngineering, 10 (2013)726-740.)?;?互信息(MI)和偏最小二乘(PLS)的算法可W有效識別運些特征(參見文獻[6])。為 有效的融合運些頻譜特征,基于集成化S,選擇性集成(沈腳和核化S(KPLS)的軟測量 模型方法已有報道(參見文獻[7]J.Tang,T.Y.Chai,L.J.Zhao,W.化,比化e,"Soft sensorforparametersofmillloadbasedonmulti-spectralsegmentsPLS sub-modelsandon-lineadaptiveweightedfusionalgorithm,,,Neurocomputi ng, 78 (2012) 38-47.文獻[8]J.Tang,T.Y.Qiai,W·化,LJ.Zhao,"Featureextraction andselectionbasedonvibrationspectrumwithapplicationtoestimate theloadparametersofballmillingrindingprocess,,,ControlEngineering Practice, 20 (2012) 991-1004.)。但是,快速傅里葉變換(FFT)不適合于具有非穩(wěn)態(tài)特性 的機械振動和振聲信號的處理(參見文獻巧]Y.G.Lei,Z.J.He,Y.Y.Zi,"Applicationof theEEMDmethodtorotorfaultdiagnosisofrotatingmachinery,"Mechanical SystemsandSignalProcessing, 23 (2009) 1327-1338·)。離散小波變換、連續(xù)小波 變換(CWT)、小波包變換等時頻分析方法已經(jīng)被廣泛應(yīng)用于旋轉(zhuǎn)機械設(shè)備的故障診斷 (參見文南犬[10]G.K.Singh,S.A.S.AlKazzaz,"Isolationandidentificationof drybearingfaultsininductionmachineusingwavelettransform,,,Tribology International42(2009)849-861.;文獻[11]J.Cusido,LRomeral,J.A.Ortega,J. A. 民osero,andA.GarciaEspinosa,"Faultdetectionininductionmachinesusing powerspectraldensityinwaveletdecomposition, "IEEETrans.Ind.Electron.,v ol. 55,no. 2,pp. 633-643,Feb. 2008.文獻[12]M.Riera-Guasp,J.A.Antonino-Daviu,M. Pineda-Sanchez,民.Puche-Panadero,J.Perez-Cruz,"Ageneralapproachfor thetransientdetectionofslip-dependentfaultcomponentsbasedonthe discretewavelettransform,"IEEETrans.Ind.Electron.,55(2008)4167-4180.文 南犬[13]J.Seshadrinath,B.Singh,andB.K.Panigrahi,"VibrationAnalysisBased InterturnFaultDiagnosisinInductionMachines,"TransactiononIndustrial Informatics, 10(2014)340-350.文獻[14]P.K.Kankar,S.C.Sharma,S.P.Harsha,"Rolling elementbearingfaultdiagnosisusingautocorrelationandcontinuouswavelet transform,"JournalofVibrationandControl, 17 (201 ]_) 2081-2094.)D但這些方 法不能自適應(yīng)分解本文所面對的多組分信號,如面對任何具體實際問題必須為CWT選擇 合適的母小波。經(jīng)驗?zāi)B(tài)分解(EMD)技術(shù)通過自適應(yīng)分解獲取具有不同時間尺度的內(nèi) 稟模態(tài)函數(shù)(IMFs,也成為子信號)(參見文獻[15]N.E.Huang,Z.Shen,S.R.Long,"化6 empiricalmodedecompositionandtheHilbertspectrumfornon-linearandnon stationarytimeseriesanalysis, "Proc.民oyalSoc.LondonA, 454 (1998) 903-995.), 并且已經(jīng)被廣泛應(yīng)用于旋轉(zhuǎn)設(shè)備故障診斷(參見文獻[16]J.Faiz,V.化orbanian,and B.Μ.Ebrahimi, "EMD-BasedAnalysisofIndustrialInductionMotorsWith BrokenRotorBarsforIdentificationofOperatingPointatDifferentSupply Modes, "IEEETransactiononIndustrialInformatics, 10 (2014) 957-966.文南犬[17] Stuti.Shukla,S.Mishra,andBhimSingh,"PowerQualityEventClassificationUnder NoisyConditionsUsingEMD-BasedDe-NoisingTechniques,,,IEEETransactionon IndustrialInformatics, 10(2014) 1044-1054.文獻[1 引R.Y.Li,D.He,"Rotational machinehealthmonitoringandfaultdetectionusingEMD-basedacoustic emissionfe曰turequ曰ntific曰tion, "IEEETr曰ns曰ctiononInstrument曰tion曰nd Measurement, 61 (2012) 990-1001.)。文獻[19] (V.K.Rai,A.R.Mohanty,"Bearing faultdiagnosisusingFFTofintrinsicmodefunctionsinHiIbert-Huang transform,"MechanicalSystemsandSignalProcessing, 21 (2007)2607-2165.)和 文獻[20] (J.Tang,L.J.Zhao,H.化e,W.化,T.Y.Qia