Wen Chenglin

Multi-level PCA and its application in fault diagnosis

The traditional principal component analysis (PCA) method divides the variable space into two parts: Principal subspace and Residual subspace by orthogonal decomposition. It has been widely used in fault detection process, but it is difficult to interpret the modes of the fault because of model compound effect, and the ability to distinguish the pattern which is no significant is affected. In industrial process, there may exist a larger fault cause deviation from the normal state of system and may exist security risks cause a larger fault, take different responses for different sizes of faults can reduce expenses, thus identify the fault size is extremely important. In this paper, we put forward a multi-level PCA method that the variable space is divided into several principal subspaces and a residual subspace to solve the problem of identify the size of the fault, and apply it to fault diagnosis. For different sizes of fault data, project them onto each subspace step by step, calculate indicators and compare with the control limits of normal subspaces. The method can not only find faults, but also can identify the fault size, according to the subspace in which the fault is detected. Simulation shows the effectiveness of the algorithm.

Fusion filtering method based on pseudo-measurement model library for multisensor systems with delay measurements

Information transferring process in sensor network will appear delay, out-of-sequence even dropout. How to make full use of such information is extremely important to improve the state estimation accuracy. In this paper, it presents a new method, which can effectively improve the estimation accuracy. Firstly, a local pseudo-measurement model library is established to deal with the measurement with several steps delay, then sent the local estimates to the fusion center; Secondly, in the fusion center, under the unbiased linear minimum variances weighting fused rule, the fusion Kalman filter based on a matrix weighted fusion algorithm processes the local estimates, thus obtain the global results. Simulations verify the effectiveness of this method.

A method of multi-surface evidence fusion slope stability analysis based on random set

The traditional methods of slope stability analysis are using cohesion and inner friction angle and other parameters to calculate the safety factor of stability. One of the most representative is the Sweden arc method which is most widely used in engineering. The problem lies in that the parameters being fixed value have not considered the existence of uncertainties of soil character. Different choice of the soil parameters often lead to different safety factor. The results tend to be more conservative, and there is no method to consider the comprehensive effect of multiple landslide. Based on the theory of random set and theory of evidence, as well as the Swedish arc method this paper gives a method of multi-surface evidence fusion slope stability analysis based on random set. First, we use two random sets to represent two uncertain parameters of the slope based on random set theory. By Extension Principles five sliding surface are mapped to five output random sets. Then use the Dempster combination rule to fuse five output random sets to get a comprehensive output. Following this transform it into pignistic cumulative distribution by TBM method to assess the stability of slope. Finally, we using the simulation comparing single surface results with multi-surface result, verifies the validity of the method.

Multi-sensor multi-OOSM distributed sequential fusion filtering

Multi-sensor fusion for OOSM system is still an open question in the field of wireless sensor network. In this paper, a new distributed sequential fusion algorithm for multi-sensor system with one-step-lag OOSMs is proposed in the sense of minimum trace of error covariance matrix. The main idea of this real time sequential fusion algorithm is that: firstly, an equivalent measurement “sampled” at the fusion time is obtained, once a measurement comes to the fusion center; then, a local estimation of the current state is obtained based on the equivalent measurement; furthermore, first round fusion is carried out for the second local estimate and the first local estimate to obtain the first local fusion when the second measurement comes; another round of fusion process is carried out for the third local estimate and the first local fusion estimation. Repeating this process until all the measurements are fused, sequentially. The final simulation illustrates its efficiency and validity of the proposed method.

The research of solve measurement dimension for inverse problem based on convex optimization

This paper studies how to transform vector to be estimated recovery problem into a convex optimization problem based on measure values. Restore ratio of vector to be estimated depends on the measured values dimension. So we can transform the problem of measurement dimension into the problem of calculating the tangent cone Gaussian width that induced by atomic norm. And the problem of solve Gaussian width mainly use dual structure. Eventually solving the dimension of the measurement depends on solving the problem of dual cone Gaussian width. Finally, this paper solves the sparse vector and the low-rank matrix through computer simulation software. Verify the validity of the number of dimensions of the measurements that determined.

Fault tolerant control for a class of nonuniformly sampled-data systems

This paper preliminarily investigates the problem of fault tolerant control for nonuniformly sampled-data systems. Firstly, based on the analysis that the adaptive fault diagnosis observer is not appropriate for sampled-data systems, a novel augmented fault estimation observer is developed to simultaneously estimate continuous-time state and fault by use of discrete-time nonuniformly sampled output. Using the on-line fault and state information, a sampled-data state feedback fault tolerant controller which is nonuniformly updated is designed to guarantee the stability of closed-loop system in presence of faults. Finally simulation results are presented to illustrate the efficiency of the proposed method.

Assessment of Surface Contamination of High Voltage Insulator Based on Evidence Theory

To overcome the bad performance of surface contamination assessment for the current high voltage insulator, which induced by a lot of uncertainties existing in measurement information and the thing between measurement and assessment result, a novel assessment strategy is proposed based on D-S evidence theory in this paper. Firstly, the D-S reasoning model is established to assess contamination extent on the basis of information or evidence. Secondly, the effective representation for uncertainties is solved by processing obscurely the corresponding measurement information. Afterwards, the uncertain reasoning for assessment result is finished based on fuzzified uncertain information by the proposed novel algorithm for the insulator contamination assessment. Finally, computer simulation example is illustrated to prove the validity of the proposed novel algorithm.

A new level measurement method based on acoustic standing wave

A new level measurement method based on acoustic standing wave is presented. Firstly, the system model about the frequencies of the acoustic standing waves is established and Kalman filter is used to preprocess the standing wave frequency. It will increase the accuracy of level measurement by reducing the influence of environment noise and measurement error, which are usually ignored by most of acoustic level measurement methods. Secondly, the new conversion technique can be given to calculate the level height with estimates of standing wave frequencies from Kalman filter. And it can overcome the strict limitation about frequency range in existing classical method. Moreover, the experiments show the proposed method effective.

Modeling range-only tracking in airborne ISAR and nonlinear filter

The main task of inverse synthetic aperture radar imaging is to use the echo data of some points that received from the scan lines during the observation periods to reconstruct the target images. However, since the target and airborne ISAR are all moving objects, some points may disappear from the radar antenna beams, which will affect imaging results. In order to improve image qualities, it is very important to track the target range by virtue of the collected information and accurately control the antennas of ISAR to face the target. Because of the relative motion between the radar and target, their relative distances are usually nonlinearly variable with time. In this paper, a nonlinear model of range-only system is constructed by using a method of chord length approximation, based on which, a particle filter algorithm is presented for real-time tracking to the relative range. Finally, a compare is made between the new algorithm and other filtering algorithms under the same system model. The simulation results show that the former have higher accuracy.

An approach of fault detection based on multi-mode

Conventional multi-scale principal component analysis (MSPCA) only detects fault, but it canpsilat detect fault types. For these problems, a method of fault detection based on multi-mode that incorporates MSPCA into adaptive resonance (ART) neural network is presented. Firstly, this method presents a wavelet transform for samples data, and principal component analysis can be used to analyze data at each scale. Then ART is used to classify reconstruction data. It can detect fault effectively, and ART2 can classify fault using wavelet denoising easily, it separates the fault successfully in the system. At last, it develops multi-mode fault detection in autocorrelation system application through computer simulation experiment. The theory and simulation experiments shows that this method is of wide application prospect.