Sensor faults mainly include: complete failure, fixed deviation, drift deviation and precision degradation.
Failure fault is the sudden failure of sensor measurement, and the measured value is always a constant;
Deviation fault mainly refers to a kind of fault that the measured value of sensor is different from the real value by a constant constant. The measurement with fault is parallel to the measurement without fault;
Drift fault is a kind of fault that the difference between the measured value and the true value of the sensor increases with time;
The decline of accuracy means that the measurement ability of the sensor becomes worse and the accuracy becomes lower. When the accuracy level decreases, the mean value of measurement does not change, but the variance of measurement changes.
Both fixed deviation fault and drift fault are not easy to find, which will cause a series of unpredictable problems in the process of fault occurrence, so that the control system can not function normally for a long time.
Fault classification of sensors
1. Classification by sensor failure degree
According to the degree of sensor fault, it can be divided into hard fault and soft fault.
Hard fault generally refers to the fault caused by structural damage, with large amplitude and sudden change; soft fault generally refers to the variation of characteristics, with small amplitude and slow change.
Hard fault is also called complete fault, when complete fault, the measured value does not change with the actual change, and always keeps a certain reading. This constant value is usually zero or the maximum reading. The measured value of the fault is approximately a horizontal straight line.
Soft faults include data deviation, drift, precision grade degradation and so on. Soft fault is relatively small and difficult to be found. Therefore, in a sense, the harm of soft fault is greater than that of hard fault, and its harm has gradually attracted people’s attention.
2. Classification according to the performance of fault
According to the performance of the fault, it can be divided into intermittent fault and permanent fault.
Intermittent faults are good and bad ; After permanent failure, it can not return to normal.
3. According to the process of fault occurrence and development
According to the process of fault occurrence and development, it can be divided into abrupt fault and gradual fault.
The change rate of sudden change fault signal is large, and that of slow change fault signal is small.
4. Classification by cause of failure
According to the fault causes, it can be divided into deviation fault, impact fault, open circuit fault, drift fault, short circuit fault, periodic interference, nonlinear dead time fault.
The causes of deviation fault are: bias current or bias voltage, etc;
The causes of impact fault are: random interference in power supply and ground wire, surge, spark discharge, burr in D / a converter, etc;
Causes of open circuit fault: signal line broken, chip pin not connected, etc;
The cause of drift fault: temperature, etc ;
Causes of short circuit fault: Bridge corrosion caused by pollution, line short circuit, etc;
Causes of periodic interference: power supply 50 Hz interference, etc;
The causes of nonlinear dead time fault are amplifier saturation, nonlinear link and so on.
In addition, from the perspective of modeling and simulation, it can be divided into multiplicative fault and pseudo fault. For bias fault, a constant or random small signal is added to the original signal; for impact interference, a pulse signal can be added to the original signal; for short circuit fault, the signal is close to zero; for open circuit fault, the signal is close to the maximum output value of the sensor; for short circuit fault, the signal is close to zero; For drift fault, the signal offsets the original signal at a certain rate; for periodic interference fault, the signal of a certain frequency is superimposed on the original signal.
Diagnosis method of sensor fault
From different perspectives, the classification of fault diagnosis methods is not the same. The fault diagnosis methods are divided into two parts: the method based on analytic mathematical model and the method independent of mathematical model.
1. The method based on analytic mathematical model
According to the different forms of residual, the methods based on analytical mathematical model can be further divided into parameter estimation method, state estimation method and equivalent space method.
The model-based fault diagnosis method is one of the earliest developed diagnosis methods, and it is also one of the most widely studied and applied diagnosis methods.
The advantages of the model are clear mechanism, simple structure, easy implementation, easy analysis and real-time diagnosis. It plays an important role in the field of fault diagnosis, and will still be the main research direction of sensor fault diagnosis in the future.
The disadvantages are large amount of calculation, complex system, modeling error, poor adaptability of the model, poor reliability, prone to false positives, missing positives and other phenomena, robustness of external disturbance, insensitive to noise and disturbance of the system.
At present, the research results of this diagnosis method are still mainly focused on linear systems, which is of great significance to the in-depth study of general fault diagnosis technology of nonlinear systems. At the same time, the robustness problem also has high research value.
2. Fault diagnosis method independent of mathematical model
At present, the control system is becoming more and more complex, because it is difficult to establish an accurate analytical mathematical model of the control system in practice, when there are modeling errors, the model-based fault diagnosis method will appear false positives, missing positives and other phenomena, so people attach great importance to the model-free fault diagnosis method.
The advantage of the method which does not depend on mathematical model is that it does not need the accurate model of the object and has strong adaptability. Its disadvantage is that its structure is complex and difficult to realize.
This fault diagnosis method independent of system model can be divided into data-driven fault diagnosis method, knowledge-based fault diagnosis method and discrete event based fault diagnosis method.
2.1 data driven approach
There are two kinds of data-driven methods: signal processing methods and statistical methods.
Some common fault diagnosis methods based on signal processing are: absolute value test and trend test, fault detection based on Kullback information criterion, fault detection based on adaptive sliding lattice filter, fault detection based on signal mode estimation, correlation analysis, wavelet analysis and information fusion.
2.2 knowledge based approach
Knowledge based fault diagnosis methods can be divided into symptom based fault diagnosis methods and qualitative model-based fault diagnosis methods.
2.3 discrete event based approach
The fault diagnosis method based on discrete event is a new fault diagnosis method developed in recent years. The basic idea is that the state of the discrete event model reflects not only the normal state, but also the fault state of the system.
With the progress of theoretical research and the continuous improvement of technical level, the research of sensor fault diagnosis will be more practical, and some practical problems will be gradually solved.
Editor in charge: CC