Residual Signal Research Articles

Robust fault estimation and isolation for a descriptor LPV system with disturbance

This paper proposes a new modus of fault estimation and isolation, which based on fault estimation filter using iterative leaning scheme combine with robust observer for descriptor linear parameter varying (D-LPV) systems with disturbance. According to the p actuators of the D-LPV multi-input multiple-output (MIMO) system, we designed p robust isolation observers and fault estimation filters. Firstly, assuming that the jth actuator fails, the corresponding jth robust isolation observer is designed by using the H∞ method, so that the residual signal is only sensitive to the jth actuator fault, but robust to disturbance and remaining faults. Next, by comparing the norm of residual signal with threshold of the designed filter, we can judge whether to start fault estimation filter. Then, iterative learning fault estimator use the filtered residual signal to modify the estimated value. Thus, this method of fault estimation and isolation is realized. Finally, the effectiveness and correctness of this method are verified in a simulation experiment.

Spatial basis functions based fault localisation for linear parabolic distributed parameter systems

Fault localisation for distributed parameter systems is as important as fault detection but is seldom discussed in the literature. The main reason is that an infinite number of sensors in the space are needed to construct a distributed residual signal, which is nearly impossible in practice. In this study, a fault detection and localisation filter which only uses a finite number of sensors is initiated based on an approximated ordinary differential equation model. Considering the limitations on computation resources for higher-order models in practice, a novel set of spatial basis functions is applied to the reduced-order fault detection and localisation filter design. Under certain conditions, the novel spatial basis functions obtain smaller state truncation error while the order is lower compared to the mostly used eigenfunctions of the spatial operator.

DFF-ResNet: An insect pest recognition model based on residual networks

Insect pest control is considered as a significant factor in the yield of commercial crops. Thus, to avoid economic losses, we need a valid method for insect pest recognition. In this paper, we proposed a feature fusion residual block to perform the insect pest recognition task. Based on the original residual block, we fused the feature from a previous layer between two 1×1 convolution layers in a residual signal branch to improve the capacity of the block. Furthermore, we explored the contribution of each residual group to the model performance. We found that adding the residual blocks of earlier residual groups promotes the model performance significantly, which improves the capacity of generalization of the model. By stacking the feature fusion residual block, we constructed the Deep Feature Fusion Residual Network (DFF-ResNet). To prove the validity and adaptivity of our approach, we constructed it with two common residual networks (Pre-ResNet and Wide Residual Network (WRN)) and validated these models on the Canadian Institute For Advanced Research (CIFAR) and Street View House Number (SVHN) benchmark datasets. The experimental results indicate that our models have a lower test error than those of baseline models. Then, we applied our models to recognize insect pests and obtained validity on the IP102 benchmark dataset. The experimental results show that our models outperform the original ResNet and other state-of-the-art methods.

Combining T2 measurements and crusher gradients into a single ASL sequence for comparison of the measurement of water transport across the blood-brain barrier.

PurposeArterial spin labeling can be used to assess the transition time of water molecules across the blood–brain barrier when combined with sequence modules, which allow a separation of intravascular from tissue signal. The bipolar gradient technique measures the intravascular fraction by removing flowing spins. The T2‐relaxation‐under‐spin‐tagging (TRUST) technique modulates the TE to differentiate between intravascular and extravascular spins based on T2. These modules were combined into a single time‐encoded pseudo‐continuous arterial spin labeling sequence to compare their mechanisms of action as well as their assessment of water transition across the blood–brain barrier.MethodsThis protocol was acquired on a scanner with 9 healthy volunteers who provided written, informed consent. The sequence consisted of a Hadamard‐encoded pseudo‐continuous arterial spin labeling module, followed by the TRUST module (effective TEs of 0, 40, and 80 ms) and bipolar flow‐crushing gradients (2, 4, and ∞ cm/s). An additional experiment was performed with TRUST and a 3D gradient and spin‐echo readout.ResultsGradients imperfectly canceled the intravascular signal, as evidenced by the presence of residual signal in the arteries at early postlabeling delays as well as the underestimation of the intravascular fraction as compared with the TRUST method. The TRUST module allowed us to detect the transport of water deeper into the vascular tree through changes in T2 than the used crusher gradients could, with their limited b‐value.ConclusionOf the implemented techniques, TRUST allowed us to follow intravascular signal deeper into the vascular tree than the approach with (relatively weak) crusher gradients when quantifying the transport time of water across the blood–brain barrier.

Replay Simulations with Personalized Metabolic Model for Treatment Design and Evaluation in Type 1 Diabetes.

The capacity to replay data collected in real life by people with type 1 diabetes mellitus (T1DM) would lead to individualized (vs population) assessment of treatment strategies to control blood glucose and possibly true personalization. Patek et al introduced such a technique, relying on regularized deconvolution of a population glucose homeostasis model to estimate a residual additive signal and reproduce the experimental data; therefore, allowing the subject-specific replay of what-if scenarios by altering the model inputs (eg, insulin). This early method was shown to have a limited domain of validity. We propose and test in silico a similar approach and extend the method applicability. A subject-specific model personalization of insulin sensitivity and meal-absorption parameters is performed. The University of Virginia (UVa)/Padova T1DM simulator is used to generate experimental scenarios and test the ability of the methodology to accurately reproduce changes in glucose concentration to alteration in meal and insulin inputs. Method performance is assessed by comparing true (UVa/Padova simulator) and replayed glucose traces, using the mean absolute relative difference (MARD) and the Clarke error grid analysis (CEGA). Model personalization led to a 9.08 and 6.07 decrease in MARD over a prior published method of replaying altered insulin scenarios for basal and bolus changes, respectively. Replay simulations achieved high accuracy, with MARD <10% and more than 95% of readings falling in the CEGA A-B zones for a wide range of interventions. In silico studies demonstrate that the proposed method for replay simulation is numerically and clinically valid over broad changes in scenario inputs, indicating possible use in treatment optimization.

Toward a Digital Twin: Time Series Prediction Based on a Hybrid Ensemble Empirical Mode Decomposition and BO-LSTM Neural Networks

Abstract Precise time series prediction serves as an important role in constructing a digital twin (DT). The various internal and external interferences result in highly nonlinear and stochastic time series. Although artificial neural networks (ANNs) are often used to forecast time series because of their strong self-learning and nonlinear fitting capabilities, it is a challenging and time-consuming task to obtain the optimal ANN architecture. This paper proposes a hybrid time series prediction model based on an ensemble empirical mode decomposition (EEMD), long short-term memory (LSTM) neural networks, and Bayesian optimization (BO). To improve the predictability of stochastic and nonstationary time series, the EEMD method is implemented to decompose the original time series into several components (each component is a single-frequency and stationary signal) and a residual signal. The decomposed signals are used to train the neural networks, in which the hyperparameters are fine-tuned by the BO algorithm. The following time series data are predicted by summating all the predictions of the decomposed signals based on the trained neural networks. To evaluate the performance of the proposed EEMD-BO-LSTM neural networks, this paper conducts two case studies (the wind speed prediction and the wave height prediction) and implements a comprehensive comparison between the proposed method and other approaches including the persistence model, autoregressive integrated moving average (ARIMA) model, LSTM neural networks, BO-LSTM neural networks, and EEMD-LSTM neural networks. The results show an improved prediction accuracy using the proposed method by multiple accuracy metrics.

Simultaneous mode, input and state estimation for switched linear stochastic systems

SummaryIn this paper, we propose a filtering algorithm for simultaneously estimating the mode, input and state of hidden mode switched linear stochastic systems with unknown inputs. Using a multiple‐model approach with a bank of linear input and state filters for each mode, our algorithm relies on the ability to find the most probable model as a mode estimate, which we show is possible with input and state filters by identifying a key property, that a particular residual signal we call generalized innovation is a Gaussian white noise. We also provide an asymptotic analysis for the proposed algorithm and provide sufficient conditions for asymptotically achieving convergence to the true model (consistency), or to the “closest” model according to an information‐theoretic measure (convergence). A simulation example of intention‐aware vehicles at an intersection is given to demonstrate the effectiveness of our approach.

Robust Metabolite Quantification from J-Compensated 2D 1H-13C-HSQC Experiments.

The spectral resolution of 2D H-C heteronuclear single quantum coherence (H-C-HSQC) nuclear magnetic resonance (NMR) spectra facilitates both metabolite identification and quantification in nuclear magnetic resonance-based metabolomics. However, quantification is complicated by variations in magnetization transfer, which among others originate mainly from scalar coupling differences. Methods that compensate for variation in scalar coupling include the generation of calibration factors for individual signals or the use of additional pulse sequence schemes such as quantitative HSQC (Q-HSQC) that suppress the JCH-dependence by modulating the polarization transfer delays of HSQC or, additionally, employ a pure-shift homodecoupling approach in the H dimension, such as Quantitative, Perfected and Pure Shifted HSQC (QUIPU-HSQC). To test the quantitative accuracy of these three methods, employing a 600 MHz NMR spectrometer equipped with a helium cooled cryoprobe, a Latin-square design that covered the physiological concentration ranges of 10 metabolites was used. The results show the suitability of all three methods for the quantification of highly abundant metabolites. However, the substantially increased residual water signal observed in QUIPU-HSQC spectra impeded the quantification of low abundant metabolites located near the residual water signal, thus limiting its utility in high-throughput metabolite fingerprinting studies.

Neural network-based fault detection for nonlinear networked systems with uncertain medium access constraint: Application to motor systems

The fault detection for a class of continuous-time nonlinear networked control systems with medium access constraint is concerned in this paper, where the occurring probability of transition from one sensor to another is allowed to be partially unknown and uncertain. First of all, a Markovian system approach is adopted to describe the access process of sensors, in which only one sensor is allowed to access the communication channel. A robust filter based residual generator is proposed to generate the residual signal such that it can be used to indicate whether the fault has occurred or not. The nonlinear term is approximated by a neural network, and the Lyapunov–Krasovskii functional is introduced to analyze the fault detection system, three sufficient conditions for the stochastic stability of fault detection error system are given, and the fault detection filter gains are calculated via solving some matrix inequalities. In the simulation, a second-order DC motor system is used to validate of the main results, which shows the effectiveness of the fault detector design.

V1 Projection Zone Signals in Human Macular Degeneration Depend on Task Despite Absence of Visual Stimulus

Identifying the plastic and stable components of the visual cortex after retinal loss is an important topic in visual neuroscience and neuro-ophthalmology.1-5 Humans with juvenile macular degeneration (JMD) show significant blood-oxygen-level-dependent (BOLD) responses in the primary visual area (V1) lesion projection zone (LPZ),6 despite the absence of the feedforward signals from the degenerated retina. Our previous study7 reported that V1 LPZ responds to full-field visual stimuli during the one-back task (OBT), not during passive viewing, suggesting the involvement of task-related feedback signals. Aiming to clarify whether visual inputs to the intact retina are necessary for the LPZ responses, here, we measured BOLD responses to tactile and auditory stimuli for both JMD patients and control participants with and without OBT. Participants were instructed to close their eyes during the experiment for the purpose of eliminating retinal inputs. Without OBT, no V1 responses were detected in both groups of participants. With OBT, to the contrary, both stimuli caused substantial V1 responses in JMD patients, but not controls. Furthermore, we also found that the task-dependent activity in V1 LPZ became less pronounced when JMD patients opened their eyes, suggesting that task-related feedback signals can be partially suppressed by residual feedforward signals. Modality-independent V1 LPZ responses only in the task condition suggest that V1 LPZ responses reflect task-related feedback signals rather than reorganized feedforward visual inputs.

Blade Rub-Impact Fault Identification Using Autoencoder-Based Nonlinear Function Approximation and a Deep Neural Network.

A blade rub-impact fault is one of the complex and frequently appearing faults in turbines. Due to their nonlinear and nonstationary nature, complex signal analysis techniques, which are expensive in terms of computation time, are required to extract valuable fault information from the vibration signals collected from rotor systems. In this work, a novel method for diagnosing the blade rub-impact faults of different severity levels is proposed. Specifically, the deep undercomplete denoising autoencoder is first used for estimating the nonlinear function of the system under normal operating conditions. Next, the residual signals obtained as the difference between the original signals and their estimates by the autoencoder are computed. Finally, these residual signals are used as inputs to a deep neural network to determine the current state of the rotor system. The experimental results demonstrate that the amplitudes of the residual signals reflect the changes in states of the rotor system and the fault severity levels. Furthermore, these residual signals in combination with the deep neural network demonstrated promising fault identification results when applied to a complex nonlinear fault, such as a blade-rubbing fault. To test the effectiveness of the proposed nonlinear-based fault diagnosis algorithm, this technique is compared with the autoregressive with external input Laguerre proportional-integral observer that is a linear-based fault diagnosis observation technique.

Denoising Method of the Φ-OTDR System Based on EMD-PCC

To filter out the phase noise of the Φ-OTDR system, a method based on empirical mode decomposition and Pearson correlation coefficient fusion (EMD-PCC) is proposed. First, the EMD-PCC method is simulated. The experimental results show that the SNR increases from 7.32 dB to 13.68 dB. Second, the beat signal detected by the Φ-OTDR system is demodulated by the I/Q quadrature demodulation method. Finally, the phase signal is decomposed by empirical mode decomposition to obtain the modal function and residual component. The Pearson correlation coefficient with the phase signal is calculated. Then, the threshold value is 0.4-1.0. The modal function within the threshold is superimposed with the residual component signal. In this paper, the experimental verification of PZT analog disturbance signal frequencies of 200 Hz, 300 Hz, 400 Hz and 500 Hz is performed. Then, wavelet denoising and EMD-soft denoising are compared with EMD-PCC denoising. Experimental results show that the proposed method can accurately restore the disturbance signal. The experiment of sawtooth disturbance signal frequencies of 400 Hz and 500 Hz verifies the applicability of the algorithm and shows that the method is suitable for other arbitrary forms of disturbance signals. The Φ-OTDR system has an important application in the field of oil and gas exploration. This method provides a good theoretical basis for the exploration field.

A novel spectrally selective fat saturation pulse design with robustness to B0 and B1 inhomogeneities: A demonstration on 3D T1-weighted breast MRI at 3 T

Spectrally selective fat saturation (FatSat) sequence is commonly used to suppress signal from adipose tissue. Conventional SINC-shaped pulses are sensitive to B0 off-resonance and B1+ offset. Uniform fat saturation with large spatial coverage is especially challenging for the body and breast MRI. The aim of this study is to develop spectrally selective FatSat pulses that offer more immunity to B0/B1+ field inhomogeneities than SINC pulses and evaluate them in bilateral breast imaging at 3T. Optimized composite pulses (OCP) were designed based on the optimal control theory with robustness to a targeted B0/ B1+ conditions. OCP pulses also allows flexible flip angles to meet different requirements. Comparisons with the vendor-provided SINC pulses were conducted by numerical simulation and in vivo scans using a 3D T1-weighted (T1w) gradient-echo (GRE) sequence with coverage of the whole-breast. Simulation revealed that OCP pulses yielded almost half of the transition band and much less sensitivity to B1+ inhomogeneity compared to SINC pulses with B0 off-resonance within ±200Hz and B1+ scale error within ±0.3 (P<0.001). Across five normal subjects, OCP FatSat pulses produced 25-41% lower residual fat signals (P<0.05) with 27-36% less spatial variation (P<0.05) than SINC. In contrast to conventional SINC-shaped pulses, the newly designed OCP FatSat pulses mitigated challenges of wide range of B0/ B1+ field inhomogeneities and achieved more uniform fat suppression in bilateral breast T1w imaging at 3T.

Fault diagnosis for semilinear distributed parameter systems with actuator/sensor based on iterative learning control

AbstractThis paper focuses on the fault diagnosis problem for a class of semilinear distributed parameter systems with actuator and sensor. To diagnose the actual fault, a fault estimator is designed by using an introduced virtual fault for this system. Based on the semigroup theory and λ‐norm, the residual signal is used in the PD‐type iterative learning control scheme to regulate the introduced virtual fault, making the virtual fault track the practical fault. Then, the goal of fault diagnosis is obtained. Finally, the effectiveness of the proposed method is tested on a Fisher equation and applied to a class of typical chemical industry processes.

Finite-horizon H ∞ filtering and fault isolation for a class of time-varying systems with sensor saturation

This paper is concerned with the finite-horizon filtering and fault isolation problems for a class of discrete time-varying systems subject to possible actuator fault and sensor saturation. A set of filters are designed and the residual signal is generated for a specific actuator fault. A residual matching approach is proposed to isolate different actuator failures once they are detected. The purpose of the addressed problem is to detect and isolate the fault such that the influence from the disturbances onto the errors satisfies the performance index. By using the stochastic analysis techniques and completing squares method, sufficient conditions are established for the existence of the desired filters, whose parameters can be iteratively calculated online by solving a series of recursive Riccati different equations (RDEs), in a given finite horizon. The effectiveness of the proposed method is demonstrated via a simulation example.

Identification of Cameroon's geological structures through a gravity separation and using seismic crustal models

Residual gravity data provide useful information for a geophysical interpretation of geological sources. Several techniques have been developed and applied to separate a different content in gravity data. Nevertheless, few methods actually incorporate factual geophysical or geological information in gravity data processing. Instead, most of these methods relay only on mathematical techniques that do not necessarily reflect the actual geological composition and problems inherent to gravity inversion, most notably a non-uniqueness of solutions to gravimetric inverse problems (i.e. infinity many geological configurations could theoretically provide the same gravity solution) as well as the fact that gravimetric inverse solutions are generally ill-posed problems that require employment of regularization techniques. To partially address this issue, we apply a gravimetric forward modelling based on available seismic crustal structure models to separate a residual gravity signal from observed gravity data in order to interpret more realistically detailed geological and tectonic features. This procedure involves the estimation of the uppermost mantle gravity contribution that is treated as a regional gravity data. The residual gravity data is then obtained by subtracting the regional gravity data from observed gravity data. In addition, observed gravity data are typically corrected for the gravitational contributions of topography and bathymetry (and possibly other known crustal density heterogeneities, such as sediments and density variations within the underlying consolidated crust). The application of these gravity corrections to observed (free-air) gravity data yields the Bouguer gravity data. We employ this forward modelling approach to the Bouguer gravity disturbances over Cameroon. Furthermore, we apply a regional-residual separation based on a spectral analysis and a robust polynomial fitting. The performance of these three techniques is compared in terms of their interpretational quality. The numerical analysis reveals that the residual gravity disturbance map obtained by applying the forward modelling technique resembles main known geological and tectonic configuration and closely reproduces even more detailed features. The results from using a spectral analysis could to some extent reproduce realistically tectonic and geological structures, except for more localized features otherwise seen in the forward modelling result. A robust polynomial fitting, on the other hand, fails to provide residual gravity disturbances that realistically reflect geological and tectonic configuration of the study area. The reason is that this method produces either only positive or only negative gravity disturbances, but not both. Moreover, this merely mathematical technique does not take into consideration available geological and geophysical information.

An optimal filter length selection method for MED based on autocorrelation energy and genetic algorithms

This paper proposed a method for exact selecting the optimal filter length of minimal entropy deconvolution (MED) to solve it recovering a single random pulse when the filter length is not improper. The energy ratio of autocorrelation between the filtered signal and the residual signal is adopted to measure the salience of periodic impulses. Then this index is used as an objective function of genetic algorithms (GA) to form an adaptive optimal selection method of filter length. The proposed method is verified by two different rolling bearing fault experiments. The results show that the proposed method reveals the periodic impulses successfully from the casing signals. Compared with other MED-based methods, the proposed method has better performance in detecting the weak fault signal.

Experimental Investigation of Spur Gear Tooth Crack Location and Depth Detection using Short-time Averaging Method and Statistical Indicators

Gear systems are one of the most functional power transmission systems in the industry. Crack is one of the common defects in gears which is caused by excessive loading, sudden impact and shortcomings in the gears construction. Initially, the crack will not result in structure collapse, but its growth can lead to irreparable damage. Therefore, detecting the crack and determining its location and depth are very important in this respect. In this paper, two encoders are used to obtain the spur gear pair transmission error speed. Moreover, the short-time averaging method (STAM) has been proposed thereby detecting the crack location and some statistical indicators have been used to estimate the crack depth in the spur gear tooth. For this purpose, a dynamical model in which mesh stiffness varying with time has been deployed to achieve the transmission error speed of the gear system. Additionally, a gear test rig including a single-stage gearbox, two encoders, and also an electronic board has been used. Encoders were installed on input and output shafts and the angular position of each shaft in time was saved in the computer using the electronic board. In addition, the transmission error speed was obtained by analyzing the received signals. Then, short-time averaging method was used to identify the crack location. Ultimately, some indicators such as ABS-max, FM0, Energy Ratio (ER) and Residual Signal Average were applied to the simulated results and experimental signals to fine the crack depth ratio. According to the results of this study, it seems safe to conclude that the STAM is a useful method in cracked tooth detection and the indicators have acceptable accuracy to find the crack depth ratio.

Zonotopic Fault Detection for Fuzzy Systems With Event-Triggered Transmission

This article deals with the zonotopic fault detection for fuzzy systems with event-triggered mechanism. An <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\ell _1/h_{\infty }$</tex-math></inline-formula> fault detector design method is deduced with considering the asynchronous premise variables caused by event-triggered transmission, such that the residual signal is sensitive to system failure while robust to amplitude-bounded disturbance, and noise. Then, a novel zonotopic residual evaluation process is constructed by converting the asynchronization of premise variables into system uncertainties. Different from most published works on event-triggered fault detection where constant thresholds are utilized for residual evaluation, zonotopic thresholds for residual signal are designed with taking the influences of disturbance, noise, asynchronous premise variables, and event-based transmission into consideration. Finally, a numerical example is utilized to verify the effective performance of the developed strategy.

Indoor Fingerprinting With Bimodal CSI Tensors: A Deep Residual Sharing Learning Approach

Wi-Fi-based indoor fingerprinting is attracting increasing interest in the research community due to the ubiquitous access in indoor environments. In this article, we propose ResLoc, a deep residual sharing learning-based system for indoor fingerprinting using bimodal channel state information (CSI) tensor data. The proposed ResLoc system employs CSI tensor data, including the angle of arrival and amplitude, collected from a small set of training locations with known coordinates to train the proposed dual-channel deep residual sharing learning model. The proposed new model extends the traditional deep residual learning model by incorporating two or more channels and let the channels exchange their residual signals after each residual block. Unlike prior deep-learning-based fingerprinting schemes, ResLoc only requires for training one group of weights for all the training locations. The proposed ResLoc system is implemented with commodity Wi-Fi devices and evaluated with extensive experiments in three representative indoor environments. The experimental results validate that the proposed ResLoc system can achieve high localization accuracy using a single Wi-Fi access point in indoor environments.