Research On PSD Replication Of Electro-hydraulic Vibration Table Based On DSP

Vibration environment simulation has been broadly used in the fields of theaspects of aeronautics, astronautics, national defense, automobile industry and so on.Now the vibration environment simulation is an important means for measuring andincreasing the reliability of the products. By simulating the vibration environmentexactly that products will be exposed to in use, transform using the vibrationequipment and specified control strategies, faults of product design can be foundearly and provide the information for adjusting the design process.Upon the background of “985Project” Phase II to develop redundancy-drivenhydraulic vibration system of six degrees of freedom (DOF), which is developed byIEST (Institute of Electro-hydraulic Servo Simulation&Test System) of HarbinInstitute of Technology, the key control methods of power spectral density (PSD)replication through electro-hydraulic shaking table (EHST) has been theoreticallyand experimentally studied in this thesis.The shaking control system of EHST consists of the servo controller and thevibration controller. The servo controller is the control basic to realize lots ofvibration environment simulation with a digital control loop. Due to the lowfrequency band and small damp ratio of the hydraulic dynamic structure, the ThreeVariable Controller (TVC) is widely introduced in the servo system of EHST toextend the response band and improve the system stability. Given that thefoundation stiffness is similar to the hydraulic spring stiffness, the servo system ofEHST can be equivalent to2degree of freedom resonator system. In this case, thefunction of EHST would be difficult to be realized fully, and the performance ofTVC may be not achieved.The PSD replication of EHST mainly comprises the estimation of PSD, updateof driven PSD and generation of time-domain drive signal. In the literature, both thePSD estimation of control acceleration and the generation of time domain drivesignal can be accomplished by modern parameter methods. This paper takes theAuto-Regressive (AR) model as typical to describe the use of these numericalsimulation technologies in PSD replication. It is shown that these methods aredependent on the high precise Gaussian random signal. As it is difficult to get the stationary, ergodic and independent Gaussian White Noise, furthermore, the orderof these random processes must be very large to assure the precision of PSDreplication. Because of this, the loop time of PSD replication is increased. In viewof above consideration, this paper adopts the average period-gram method (Welchmethod) which is traditional non-parameter technology to estimate the PSD of theresponse acceleration signal of EHST. Meanwhile, sine-wave fitting is presented togenerate the time-domain drive signal in the flow of PSD replication. Systemimpedance is needed to modify the drive PSD, therefore, the system impedance isestimated more accurate, the loop time of PSD replication is less. To acquire thehigh rate of convergence of frequency response function (FRF) esitmaiton of EHST,subband adaptive system identification algorithm is proposed in this thesis. Thesimulation results show that estimation signal is divided into more bands, the rate ofconvergence of FRF estimation is higher. Cosine-modulated filter banks (CMFB)are time-domain filter banks, and the overlap between different bands is small. Thispaper takes the CMFB as the filter banks in subband adaptive system identificationalgorithm.The CMFB divide signal into identical bandwidth bands, the interior relationbetween wavelet transforms and filter banks has been unvealed recently. Signaldivided into different bandwidth bands by wavelet transforms is considered in thisthesis. Subband adaptive system identification algorithm based on wavelettransforms not only inherits the properties of that based on identical bandwidthbands signal division, but also has its own features, that is, the results of subbandadaptive system identification based on wavelet transforms is representation infrequency domain. As the development of wavelet theory, lots of wavelet basis areconstructed, which have different characteristics of orthogonality, symmetry andvanishing moment. The orthogonality and symmetry of wavelet basis play vital rolein the algorithm mentioned above, because they guarantee the signal divided intodifferent bands has little overlap and the subband filter banks have the linear phase.The constraint of domestic shking control techonology is that development ofdigital contoller is slower than the theoretical research. which has hardwaremultiplier, Havard bus structure and pipeline technique, is popular in the field ofsignal processing. Texas instruments’(TI) C2000series digital signal processors(DSPs) are used in this paper to construct the digital controller for PSD replication of EHST.The concept of distributed computing is adopted to develop the digitalcontroller, which is composed of algorithm computation unit and data acquireunit.The algorithm computation unit is in charge of the shaking control strategycomputation, and the data acquire unit is used to get the driven and feedback signalsof EHST. The dual port RAM (DPRAM) is used to communication between twofunctional units, and the handshake signal is generated by the interrupt arbitrationlogic of DPRAM. The monitor unit of signal processing platform expands DPRAMusing its ISA bus to share data with the DSPs system. The DSPs system softwaredevelopment with C language and assembly language, and intends to use the TI’sDSP library to improve processing speed. National instruments GUI languageLabview is adopted to develop the monitor software. Labview calls the dynamiclinking library (DLL) created by Windows Driver to read from or write to theDPRAM and display the results or adjust the control parameters.The servo system of EHST is realized by xPc Target Rapid Control Prototyping(RCP) technology, and the vibration controller is realized by the signal processingplatform designed before. The experimental results of PSD replication of EHSTverify the effectiveness of the approaches presented above and the signal processingplatform.