Measurement and Data Science.

Description

Nowadays, all of us enjoy the worldwide revival of measurement and data science caused by the revolution of sensory devices and the amazing data transmission, storage and processing capabilities available and embedded everywhere. Thanks to the unbelievable amount of recorded information and the theoretical results of measurement and data science, a great deal of newly developed products invade our surroundings and enable previously unconceivable smart services and support. This volume consists of a number of chapters covering the scientific results of researchers working in this field at the Department of Measurement and Information Systems of the Budapest University of Technology and Economics, Hungary. The book reports research results attained by carefully combining some of the classical theories of measurement and data processing. These new approaches and methods contribute to higher quality measurement design and measured data evaluation, and provide hints to find efficient implementations for instrumentation. Contents Preface Acknowledgements Chapter One 1.1 Introduction 1.2 Attributes of measurement processes 1.2.1 Observation in the case of noiseless system and observation models 1.2.2 Observation in the case of noisy system and observation models 1.2.3 Measurement processes based on observation models 1.2.4 Recursive evaluation of measurement processes based on observation models 1.2.5 Recursive estimations if the unknown quantity is a single value 1.2.6 Frequently used linear observation models 1.2.7 Evaluation of nonlinear observation models 1.2.8 Measurement processes using sliding-window methods 1.2.9 Summary: Recursive algorithms 1.3 Model-based signal representation and its recursive algorithms 1.3.1 Signal representation in signal spaces 1.3.2 Observers to compute signal parameters 1.3.3 Derivation of resonator-based structures 1.3.4 The resonator-based observer as universal signal processing structure 1.3.5 Signal synthesis using the resonator-based structure 1.3.6 Summary: Observer-based signal analysis and synthesis 1.4 Structural properties, aspects of implementation 1.4.1 Condition of boundedness in the case of resonator-based observers 1.4.2 Structural passivity and energy relations 1.4.3 Summary: Structural properties 1.5 Summary References Chapter Two 2.1 Introduction to Chapter 2 2.2 Adaptive Fourier Analysis 2.2.1 Introduction 2.2.2 Resonator-based observer 2.2.3 Algorithm of the AFA 2.2.3.1 Derivation of the algorithm 2.2.3.2 Fine tuning of the parameters 2.2.4 Convergence of the frequency estimator 2.2.4.1 Initial results and experiences 2.2.4.2 Block-adaptive Fourier analyser (BAFA) 2.2.5 Improvements 2.2.5.1 Adaptation for a prescribed time-frequency function 2.2.5.2 Adaptation to a decaying periodic signal 2.2.5.3 Adaptation in a wide frequency range 2.2.5.4 Further results 2.2.6 Summary 2.3 Spectral estimation in the case of data loss 2.3.1 Introduction 2.3.2 Estimation of the power spectral density function 2.3.3 Description of data loss 2.3.4 Spectral estimation using the resonator-based observer 2.3.5 FFT-based spectral estimation 2.3.5.1 The proposed procedure 2.3.5.2 Assessment of the procedure 2.3.5.3 Convergence of the proposed method 2.3.6 Frequency domain identification of data loss models 2.3.7 Summary 2.4 Active noise control 2.4.1 Introduction 2.4.2 The active noise control problem 2.4.3 Active control of periodic disturbances 2.4.4 Achievements related to periodic noise control 2.4.4.1 Identification of linear systems 2.4.4.2 Automatic offset compensation 2.4.4.3 Active nonlinear distortion reduction 2.4.5 Filtered error-filtered reference LMS algorithm 2.4.5.1 LMS-based noise control systems 2.4.5.2 Improvement of convergence speed 2.4.6 Wireless sensor network for active noise control 2.4.6.1 Synchronization 2.4.6.2 Overcoming the bandwidth constraint 2.4.6.3 Handling of data loss 2.4.6.4 Further results 2.4.7 Summary 2.5 Summary of Chapter 2 References Chapter Three 3.1 Introduction 3.2 Distortions of the signal path and possibilities for their compensation 3.3 Extension of finite bandwidth of linear systems 3.3.1 Deconvolution algorithms 3.3.1.1 Input error criterion 3.3.1.2 Output error criterion 3.3.1.3 Output error criterion + filtering 3.3.1.4 Iterative methods handling amplitude limits 3.3.1.5 Regularization 3.3.1.6 Signal model-based noise and inverse filtering 3.3.1.7 Inverse filtering based on a stochastic signal model 3.3.2 Automatic parameter optimization 3.3.3 Sampling jitter and its effect 3.3.4 Illustrations of application possibilities 3.3.4.1 Extension of the bandwidth of high-voltage dividers 3.3.4.2 Extension of the bandwidth of an accelerometer 3.3.4.3 Correction of images 3.4 Compensation of nonlinearities 3.4.1 Compensation of memoryless static nonlinearities in well-conditioned cases 3.4.2 Compensation of memoryless static nonlinearities in ill-conditioned cases—a model-based approach 3.4.3 Inverse filtering with learning systems 3.5 Sensor fusion 3.5.1 Extension of bandwidth by means of complementary filtering 3.5.2 Sensor fusion by means of Kalman filtering 3.6 Estimation of quantities that can be measured indirectly 3.6.1 Time-varying transfer function 3.6.2 Estimation of state variables that cannot be directly measured 3.6.3 An illustrative example 3.6.3.1 Parameter estimation of a permanent magnet synchronous motor 3.7 Application areas—results achieved at the department BME-MIT 3.7.1 Cost-effective measurement system using inverse filtering 3.7.2 Extending physical/technological barriers using inverse filtering methods 3.7.3 Complex sensors 3.7.4 Safety-critical systems 3.8 Summary References Chapter Four 4.1 Introduction 4.2 On linear system models and measurement design 4.2.1 Linear system models 4.2.2 Measurement setup 4.2.3 Measurement data 4.3 Estimating the frequency response function from measurements 4.4 Properties of the frequency transfer function estimate measured with periodic signals 4.4.1 Quality of the estimate—bias and variance 4.4.2 Variance reduction with averaging 4.5 Properties of the frequency transfer function estimate measured with random signals 4.5.1 Quality of the estimate—bias and variance 4.6 Estimating the frequency transfer matrix of a MIMO system 4.6.1 Optimizing input signals 4.7 Measuring frequency transfer characteristics in a closed loop 4.8 Selecting domain and excitation signals 4.9 Nonparametric identification in the frequency domain in the case of nonlinear systems 4.10 Modelling nonlinear effects 4.11 A wide range of input signals 4.12 The best linear approximation frequency characteristics 4.12.1 Theoretical principles 4.12.2 Model of nonlinear distortions 4.12.3 The variance of the best linear approximation-based nonparametric FRF estimate 4.12.4 The question of the frequency grid 4.12.5 Riemann-equivalent excitation signals 4.12.6 Relationship between stochastic and systematic nonlinear model errors 4.12.7 Measuring the best linear approximation 4.12.8 Best linear approximation measurement in a closed loop 4.13 The best linear approximation measurement—MISO systems 4.14 Best linear approximation FRF—application issues 4.14.1 Nonlinear models and the best linear approximation FRF 4.14.2 What is the BLA FRF good for? 4.14.3 The linear model alone 4.14.4 Indicator and estimator for nonlinear model structure, nonlinearity type, and nonlinear model degree 4.14.5 Initial values in nonlinear system identification References Chapter Five 5.1 Research background and objectives 5.2 Introduction to the field of reported investigations 5.2.1 Characterization of analogue-to-digital converters 5.2.2 Least squares (LS) sine-fitting 5.2.3 Sine wave histogram test for ADC characterization 5.2.4 Effects of improper frequency selection on the results of the histogram test 5.3 Verification of signal parameter settings for the sine wave histogram test 5.3.1 Estimation of sine wave parameters 5.3.2 Overdrive handling 5.3.3 Checking coherent sampling and relative prime conditions 5.3.4 Real measurement results 5.3.5 Summary of results 5.4 Numerical problems of sine-fitting algorithms 5.4.1 Some characteristics of floating-point arithmetic 5.4.2 Phase evaluation error 5.4.3 Conditioning of the system matrix 5.4.4 Summary of results 5.5 Maximum likelihood estimation 5.5.1 Attributes of maximum likelihood estimation 5.5.2 Application of ML estimation for ADC testing 5.5.3 The noise model 5.5.4 ML estimation of aperture jitter 5.5.5 Parameter space size reduction References Contributors