CN104238530A - Sensor fault diagnosis technology-based interlock alarm system - Google Patents

Abstract

The invention discloses a sensor fault diagnosis technology-based interlock alarm system. The sensor fault diagnosis technology-based interlock alarm system comprises a field device layer, a fault diagnosis layer and an interlock alarm layer, wherein the field device layer comprises a sensor arranged on an industrial object, a controller and an executer; the output end of the sensor is connected with the first input end of the controller, and the output end of the controller is used for controlling the industrial object through the executer; the fault diagnosis layer comprises a fault detection module, a fault diagnosis module and an active compensation fault-tolerant control module, and the interlock alarm layer comprises a production monitoring module and an alarm processing module. The product diagnoses the faults of a sensor in time during the working process of the sensor, thus interlock alarm can be carried out in time, and the working difficulty of searching and analysing the fault reasons of the sensors by technicists after the interlock alarm is also reduced; the product has important significance in guarantee on the quality of surfactant products, reduction for energy consumption, and prevention for safety accidents.

Description

A kind of interlock alarm system based on sensor fault diagnosis technology

Technical field

The invention discloses a kind of interlock alarm system based on sensor fault diagnosis technology, relate to the sensor fault diagnosis technical field in control system.

Background technology

The manufacturing of chemical products is the processes having suitable danger, and for the production run of surfactant device, the operations such as oxirane handling, storage, charging, sulfonating reaction all have suitable danger, and misoperation easily causes the accident.The production safety of assurance device must be carried out by control device safely and effectively.Present stage, enterprise generally adopted computing machine to come warning and the interlocking of control system, and speed and reliability promote greatly, and Comparatively speaking sensor fault becomes system and starts the main predisposing factors of reporting to the police, interlocking.Do not perform any action in the middle of interlock alarm system in the past detects controlled parameter process at sensor, only have and just carry out interlock alarm when controlled parameter reaches alarm limit, be also carrying out to the diagnosis of sensor fault afterwards.For the normal operation of Maintenance Table surface-active agent process units control system, in current interlock alarm system, the computer-controlled reliability of ubiquity improves, the problem of thing followed sensor reliability deficiency.

Summary of the invention

Technical matters to be solved by this invention is: for the defect of prior art, a kind of interlock alarm system based on sensor fault diagnosis technology is provided, abandon and adopt simple logic control to realize the mode of the overload alarm of controlled parameter in the past, adopt sensor fault diagnosis technology, before controlled parameter exceedes alarm limit, sensor experiences failure time, just can immediate analysis reason carry out corresponding compensation faults-tolerant control.

The present invention is for solving the problems of the technologies described above by the following technical solutions:

A kind of interlock alarm system based on sensor fault diagnosis technology, comprise scene equipment level, fault diagnosis layer and interlock alarm layer, wherein, described scene equipment level comprises the sensor be arranged on industrial object, also comprise controller and actuator, the output terminal of described sensor is connected with the first input end of controller, and the output terminal of controller controls industrial object through actuator;

Described fault diagnosis layer comprises fault detection module, fault diagnosis module and Active Compensation fault-tolerant control module, fault detection module detecting sensor fault exists, and the transmission frequency of fault-signal and fault mathematical model are saved in the knowledge base in fault diagnosis module, knowledge base carries out coupling diagnosis to fault-signal, compensate the compensating approach of faults-tolerant control storehouse realization to fault-signal by the historical failure in Active Compensation fault-tolerant control module, complete the Active Compensation faults-tolerant control for historical failure;

Described interlock alarm layer comprises production monitoring module and alert process module, when fault-signal current in fault diagnosis module and knowledge base fail to mate, judge that current fault-signal is as unknown failure, chain alarming layer starts production monitoring module monitors and becomes occurrence condition or start that alert process module sends warning, emergency power off is interlocked.

As present invention further optimization scheme, build fault detection module based on residual error envelope method of loci, concrete steps are as follows:

(1) formation of analyte sensors output signal, is described as the mathematical model of sensor fault:

$x=g(x,u)+\∂(x,u)---\left(1\right)$

$y\left(t\right)=\stackrel{\‾}{y}(x,u)+\mathrm{\γ}\left(t\right)+\mathrm{\ζ}(x,u,t)---\left(2\right)$

Wherein, x represents state vector; U represents input vector, and t represents time parameter; G (x, u) is mission nonlinear model; represent the uncertainty of model; Y (t) represents the actual output of sensor; represent the true output of measured signal; γ (t) represents the stochastic error produced in measuring process, and in normal distribution, mean value is 0; ζ (x, u, t) represents sensor fault function;

(2) residual error envelope method of loci design error failure detection module is adopted:

(201), adopt weight moving average filtering technique to sample to residual error, n sampled point before current sample time is chosen in sampling, obtains one group of sequence in a kth sampling instant:

wherein, for the residual error of a kth sampling instant;

Calculate the weighted moving average of above-mentioned sequence according to described weighted moving average, real-time update is carried out to new sampled value,

Wherein, η _ifor flexible strategy, n is sampled point number.

(202), adaptive threshold envelope track is set, the envelope track of residual error when namely sensor non-fault exports.

As present invention further optimization scheme, in step (202), described envelope track is by mission nonlinear model uncertain error scope the stochastic error γ (t) caused in linearized stability σ (t) and measuring process causes the threshold range of residual error to form along with time changing curve, and adaptive threshold envelope track Ψ (t) has its minimum and maximum border;

$\mathrm{\γ}\left(t\right)+\mathrm{\σ}\left(t\right)-\∂(x,u)\≤\mathrm{\Ψ}\left(t\right)\≤\mathrm{\γ}\left(t\right)+\mathrm{\σ}\left(t\right)-\∂(x,u)---\left(4\right)$

In a kth sampling instant, if the weighted moving average of residual error drop in the scope of track Ψ (t), judged that residual error is by mission nonlinear model uncertain error scope the stochastic error γ (t) that linearized stability σ (t) and measurement noises cause causes, and is now judged as that sensor is in normal work; Otherwise, in a kth sampling instant, if the weighted moving average of residual error outside the scope dropping on track Ψ (t), then think that residual error further comprises other compositions, be now judged as sensor failure.

As present invention further optimization scheme, sensor fault is classified as soft fault and hard fault according to the intensity of variation of signal amplitude after breaking down, pace of change by described historical failure feature database, amplitude changes less and is soft fault slowly, and amplitude changes greatly and is hard fault rapidly; Specifically comprise: deviation fault, drifting fault, precise decreasing fault and complete failure.

As present invention further optimization scheme,

(501) function representation of described deviation fault is:

ζ(x,u,t)＝K (5)

Wherein, K is constant;

(502) function representation of described drifting fault is:

ζ(x.u.t)＝K(t-t ₀) (6)

Wherein, K is constant, t ₀for fault initial time;

(503) function representation of described precise decreasing fault is:

ζ(x.u.t)～N(0,σ ₂ ²) (7)

Wherein, σ ₂ ²represent variance;

(504) function representation of described complete failure is:

Y (t)=V _max(or V _min) (8)

Wherein, V _max, V _minrepresent maximal value and the minimum value of instrument range respectively.

As present invention further optimization scheme, in the complete failure of described (504), fault has just started there is a forming process when occurring, and is expressed as:

Wherein, t ₀for fault initial time.

The present invention adopts above technical scheme compared with prior art, there is following technique effect: the present invention just can diagnose out the fault of sensor in time in the middle of the process of working sensor, so both can interlock alarm in time, also reduce that technician goes to search after interlock alarm again, the operational difficulties of analyte sensors failure cause simultaneously, to guarantee surfactant product quality, reduce energy consumption, prevent security incident and have great significance.

Accompanying drawing explanation

Fig. 1 is system architecture schematic diagram of the present invention.

Fig. 2 is the fault detection module based on residual error envelope method of loci in the present invention.

Fig. 3 is the failure function schematic diagram corresponding to sensor fault type different in the present invention,

Wherein: (a) deviation fault, (b) drifting fault, (c) precise decreasing fault, (d) complete failure.

Fig. 4 is in a specific embodiment of the present invention, the high temperature monitoring of oxirane intermediate tank and high liquid level monitoring schematic diagram,

Wherein: I1-nitrogen, I2-oxirane remove chuck from refrigeratory, I3-oxirane from tank car, I4-chilled water, I5-chilled water, O1-chilled water, O2-tail gas remove tail gas absorber, O3-oxirane removes ethoxylation device, V101-oxirane intermediate tank, P101-oxirane feeder pump, PV101-Feed Shut-Off valve, PV102-oxirane return line stop valve, PV103-ethylene oxide emissions pipeline stop valve, PV104-motorized valve, PV108-entrance stop valve, TT101A/B/C; TT102A/B/C; TT103A/B/C is respectively temperature parameter, LT101A; LT101B is respectively level parameter.

Fig. 5 is the interlocked control schematic diagram of specific embodiments of the invention ethylene oxide intermediate tank.

Embodiment

Be described below in detail embodiments of the present invention, the example of described embodiment is shown in the drawings, and wherein same or similar label represents same or similar element or has element that is identical or similar functions from start to finish.Being exemplary below by the embodiment be described with reference to the drawings, only for explaining the present invention, and can not limitation of the present invention being interpreted as.

Below in conjunction with accompanying drawing, technical scheme of the present invention is described in further detail:

System architecture schematic diagram of the present invention as shown in Figure 1, based on the interlock alarm system of sensor fault diagnosis technology, comprise scene equipment level, fault diagnosis layer and interlock alarm layer, wherein, described scene equipment level comprises the sensor be arranged on industrial object, also comprise controller and actuator, the output terminal of described sensor is connected with the first input end of controller, and the output terminal of controller controls industrial object through actuator; Described fault diagnosis layer comprises fault detection module, fault diagnosis module and Active Compensation fault-tolerant control module, fault detection module detecting sensor fault exists, and the transmission frequency of fault-signal and fault mathematical model are saved in the knowledge base in fault diagnosis module, knowledge base carries out coupling diagnosis to fault-signal, compensate the compensating approach of faults-tolerant control storehouse realization to fault-signal by the historical failure in Active Compensation fault-tolerant control module, complete the Active Compensation faults-tolerant control for historical failure; Described interlock alarm layer comprises production monitoring module and alert process module, when fault-signal current in fault diagnosis module and knowledge base fail to mate, judge that current fault-signal is as unknown failure, chain alarming layer starts production monitoring module monitors and becomes occurrence condition or start that alert process module sends warning, emergency power off is interlocked.

Scene equipment level: in the production run of surfactant, the sensor of scene equipment level once break down, produce the fault-signal departing from measured signal actual value and can be passed to controller part by control loop in the bamboo telegraph of whole level of factory scope.

Whether and analyze it fault diagnosis layer: whether traditional scene equipment level cannot failure judgement exist, just merely performs control algolithm and control strategy, therefore at the scene mechanical floor will increase fault diagnosis layer and detect out of order existence.If judge really there is sensor fault by fault detection module, so can collect those recurrent, that there is certain features fault-signals and study its occurrence frequency and the description of fault mathematical model, then the frequency size occurred according to its fault sets up knowledge base at fault diagnosis module successively, to treat to judge the detection signal in future.Even can compensate and correct fault-signal according to fault type further after knowledge base carries out coupling diagnosis to fault-signal, it is made to be returned to the time of day of measured signal, namely for the Active Compensation faults-tolerant control of historical failure, faults-tolerant control storehouse can be compensated by the historical failure in Active Compensation fault-tolerant control module and realize.

Interlock alarm layer: if fault-signal current in fault diagnosis module and knowledge base fail to mate, belong to non-empirical fault and unknown failure, interlock alarm layer will be entered for unknown failure system to process, by close supervision production status, exceed alarm limit once detection signal just to give a warning or emergency power off is interlocked, technician can change setting value by authority, carry out parameter tuning or amendment dependent instruction etc.

Judge whether sensor breaks down mainly by setting up the mathematical model of sensor fault, analyzes its failure function, the difference of contrast fault-signal and non-faulting signal, and detect that whether fault-signal exists with this.

The mathematical description of sensor fault: the formation of analyte sensors output signal, can be described as the mathematical model of sensor fault:

$x=g(x,u)+\∂(x,u)---\left(1\right)$

$y\left(t\right)=\stackrel{\‾}{y}(x,u)+\mathrm{\γ}\left(t\right)+\mathrm{\ζ}(x,u,t)---\left(2\right)$

Wherein, x represents state vector; U represents input vector; G (x, u) is mission nonlinear model; represent the uncertainty of model; Y (t) represents the actual output of sensor; represent the true output of measured signal; γ (t) represents the stochastic error produced in measuring process, and in normal distribution, mean value is 0; ζ (x, u, t) represents sensor fault function.

Residual error envelope method of loci:

Difference between the actual output of sensor and the true output of measured signal, namely be referred to as residual error.

Ideally, during sensor no-failure operation, residual error should be " 0 ", so for actual conditions be but sensor output signal residual error not only with sensor fault functional dependence, also with the uncertain error of system stochastic error γ (t), mission nonlinear model and the X factor such as linearized stability σ (t) is correlated with, even if it is also non-vanishing to result in the residual error when sensor no-failure operation, the foundation for the historical failure feature database of fault diagnosis module adds difficulty ^[5].In order to ensure the impact being not affected or less affected by these X factors, whether can not be that " 0 " is as judging whether sensor exists the standard of fault using residual error when design error failure detection module, but should to judge whether residual error is no more than a certain threshold range as judgment criteria, and this threshold range should comprise nonlinear model uncertain error the stochastic error γ (t) produced in linearized stability σ (t) and measuring process.Residual error envelope method of loci design error failure detection module can be adopted, as shown in Figure 2.

In order to accurate Calculation residual error, weight moving average filtering technique can be adopted to sample to residual error, (comprising the sampled point of current time) the n sampled point before at every turn all getting current sample time, so can obtain one group of sequence in a kth sampling instant calculate its weighted moving average, this weighted moving average upgrades all constantly there being new sampled value.

Wherein, η _ifor flexible strategy.

Adaptive threshold envelope track:

In order to effectively reject the impact that X factor causes residual error, adaptive threshold envelope track can be set, the envelope track of residual error when namely sensor non-fault exports.Envelope track is by mission nonlinear model uncertain error scope the minimum and maximum threshold range of the residual error that the stochastic error γ (t) caused in linearized stability σ (t) and measuring process causes is formed along with time changing curve, and adaptive threshold envelope track Ψ (t) has its minimum and maximum border.

$\mathrm{\γ}\left(t\right)+\mathrm{\σ}\left(t\right)-\∂(x,u)\≤\mathrm{\Ψ}\left(t\right)\≤\mathrm{\γ}\left(t\right)+\mathrm{\σ}\left(t\right)-\∂(x,u)---\left(4\right)$

In a kth sampling instant, if the weighted moving average of residual error drop in the scope of track Ψ (t), thought that residual error is primarily of mission nonlinear model uncertain error scope the stochastic error γ (t) that linearized stability σ (t) and measurement noises cause causes, and is now judged as that sensor is in normal work.Otherwise, in a kth sampling instant, if the weighted moving average of residual error outside the scope having dropped on track Ψ (t), just think that residual error is except further comprises other compositions by except above factors composition, is now judged as sensor failure.The fault detection module utilizing above-mentioned residual error envelope method of loci to design can greatly reduce the false alarm because X factor causes, and improves the reliability of interlock alarm system, reduces unnecessary economic loss.

If fault diagnosis module requires to judge whether current failure mates with historical failure, its fault type of immediate analysis, fault is assessed and decision-making, even can compensate and correct fault-signal according to fault type in Active Compensation fault-tolerant control module further ^[6].Then be judged as unknown failure if do not belonged to historical failure, system pays close attention to detection signal to transfinite by no and give a warning, and even emergency power off carries out interlocking guarantee system worked well.

In historical failure feature database, according to the intensity of variation of signal amplitude after breaking down, pace of change, sensor fault is classified as soft fault and hard fault, amplitude change is less and be soft fault slowly.Amplitude changes greatly and is hard fault rapidly.Can be divided in detail again: (a) soft fault-deviation fault; (b) soft fault-drifting fault; (c) soft fault-precise decreasing fault; (d) hard fault-complete failure Four types, as shown in Figure 3.

Soft fault-deviation fault: this type of fault referred to fault-signal and non-fault signal parallel to each other, differ a certain constant constant K between the two, its main cause producing this fault is the existence owing to having bias current or bias voltage, this constant undesired signal has been added on original signal.Its function representation is:

ζ(x,u,t)＝K (5)

Wherein, K is constant.

Soft fault-drifting fault: this type of fault refers to the existence due to temperature drift, and As time goes on the trend making fault-signal depart from non-fault signal constantly increase, and cause the class fault that the difference between two signals is also increasing, its representation is:

ζ(x.u.t)＝K(t-t ₀) (6)

Wherein, K is constant, t ₀for fault initial time.

Soft fault-precise decreasing fault: this type of fault refers to if calculated respectively to be had the mean value of fault-signal and non-fault signal to carry out result of calculation can not to change, but this does not represent that sensor fault does not exist, if calculate its variance further can find that the variance of fault-signal can be larger, illustrate that its signal fluctuation is comparatively violent, there is fault.Concrete representation is:

ζ(x.u.t)～N(0,σ ₂ ²)(7)

Wherein, σ ₂ ²represent variance.This class fault-signal external manifestation is not obvious, needs to calculate variance and just likely judges further, be therefore also difficult to find.

Hard fault-complete failure: this type of fault refers to that fault-signal becomes suddenly the constant class fault of a certain fixed value after output a period of time, its main reason is due to the sensor malfunctioning short trouble that causes and open fault suddenly, as signal wire broken string, chip pin breaks, circuit etch, short circuit etc.The fault-signal that short circuit causes is close to the minimum value of instrument range.The fault-signal that open circuit causes is close to the maximal value of instrument range.Such fault can represent to be become:

Y (t)=V _max(or V _min) (8)

Wherein, V _max, V _minrepresent maximal value and the minimum value of instrument range.

This kind of fault has just started there is a forming process when occurring, and can be expressed as:

Wherein, t ₀for fault initial time.

This type of fault is just difficult to go to revise once formation again, if can be detected when just occurring, plays a part key to preventing the extension of this type of fault.

If fault-signal judges not mate with historical failure storehouse through fault diagnosis module, then think that fault-signal is unknown failure type, interlock alarm layer enters into computer logic state of a control, by close supervision production status, transfinite once detection signal and just to give a warning or emergency power off is interlocked.

A specific embodiment of the present invention in surfactant production run for the high temperature monitoring of oxirane intermediate tank and high liquid level monitoring and interlocked control, as shown in Figure 4, Figure 5:

As oxirane intermediate tank V101 temperature (TT101A/B/C; TT102A/B/C; When TT103A/B/C) reaching 30 DEG C and 35 DEG C respectively, system display firsts and seconds high temperature alarm.When reaching 38 DEG C, system starts high temperature interlocking, system will close oxirane intermediate tank V101 Feed Shut-Off valve PV101, oxirane feeder pump P101 entrance stop valve PV108 and oxirane return line stop valve PV102 automatically, open ethylene oxide emissions pipeline stop valve PV103, motorized valve PV104 and tail gas absorber fire water back with water inlet line stop valve, after there is warning or interlocking, operating personnel should judge rapidly, in time, exactly, determine to cause the condition of reporting to the police or interlocking, and got rid of.

When oxirane intermediate tank V101 liquid level (LT101A, LT101B) reach 70%, DCS shows high liquid level warning and causes high level chain.Automatic closedown oxirane intermediate tank V101 Feed Shut-Off valve PV101.After there is this pressure warning interlocking, operating personnel should judge rapidly, in time, exactly, determine to cause the condition of reporting to the police or interlocking, and are got rid of.

The present invention devises fault detection module based on residual error envelope method of loci for the detection of fault-signal, and establish historical failure feature database for tracing trouble type to compensate faults-tolerant control accordingly according to the feature of each fault-signal, take computer logic to control for unknown failure simultaneously.Both can interlock alarm in time, also reduce that technician goes to search after interlock alarm again, the operational difficulties of analyte sensors failure cause.Not only there is theory significance, more have its actual application value, be applicable to the actual production process of surfactant, can wide popularization and application in the interlock alarm system of various automated production process.

By reference to the accompanying drawings embodiments of the present invention are explained in detail above, but the present invention is not limited to above-mentioned embodiment, in the ken that those of ordinary skill in the art possess, can also makes a variety of changes under the prerequisite not departing from present inventive concept.The above, it is only preferred embodiment of the present invention, not any pro forma restriction is done to the present invention, although the present invention discloses as above with preferred embodiment, but and be not used to limit the present invention, any those skilled in the art, do not departing within the scope of technical solution of the present invention, make a little change when the technology contents of above-mentioned announcement can be utilized or be modified to the Equivalent embodiments of equivalent variations, in every case be do not depart from technical solution of the present invention content, according to technical spirit of the present invention, within the spirit and principles in the present invention, to any simple amendment that above embodiment is done, equivalent replacement and improvement etc., within the protection domain all still belonging to technical solution of the present invention.

Claims (6)

1. based on an interlock alarm system for sensor fault diagnosis technology, it is characterized in that: comprise scene equipment level, fault diagnosis layer and interlock alarm layer, wherein,

Described scene equipment level comprises the sensor be arranged on industrial object, also comprises controller and actuator, and the output terminal of described sensor is connected with the first input end of controller, and the output terminal of controller controls industrial object through actuator;

Described fault diagnosis layer comprises fault detection module, fault diagnosis module and Active Compensation fault-tolerant control module, fault detection module detecting sensor fault exists, and the transmission frequency of fault-signal and fault mathematical model are saved in the knowledge base in fault diagnosis module, knowledge base carries out coupling diagnosis to fault-signal, compensate the compensating approach of faults-tolerant control storehouse realization to fault-signal by the historical failure in Active Compensation fault-tolerant control module, complete the Active Compensation faults-tolerant control for historical failure;

Described interlock alarm layer comprises production monitoring module and alert process module, when fault-signal current in fault diagnosis module and knowledge base fail to mate, judge that current fault-signal is as unknown failure, chain alarming layer starts production monitoring module monitors and becomes occurrence condition or start that alert process module sends warning, emergency power off is interlocked.

2. a kind of interlock alarm system based on sensor fault diagnosis technology as claimed in claim 1, is characterized in that, build fault detection module based on residual error envelope method of loci, concrete steps are as follows:

(1) formation of analyte sensors output signal, is described as the mathematical model of sensor fault:

$x=g(x,u)+\∂(x,u)---\left(1\right)$

$y\left(t\right)=\stackrel{\‾}{y}(x,u)+\mathrm{\γ}\left(t\right)+\mathrm{\ζ}(x,u,t)---\left(2\right)$

Wherein, x represents state vector; U represents input vector, and t represents time parameter; G (x, u) is mission nonlinear model; represent the uncertainty of model; Y (t) represents the actual output of sensor; represent the true output of measured signal; γ (t) represents the stochastic error produced in measuring process, and in normal distribution, mean value is 0; ζ (x, u, t) represents sensor fault function;

(2) residual error envelope method of loci design error failure detection module is adopted:

(201), adopt weight moving average filtering technique to sample to residual error, n sampled point before current sample time is chosen in sampling, obtains one group of sequence in a kth sampling instant:

wherein, for the residual error of a kth sampling instant;

Calculate the weighted moving average of above-mentioned sequence according to described weighted moving average, real-time update is carried out to new sampled value,

Wherein, η _ifor flexible strategy, n is sampled point number.

(202), adaptive threshold envelope track is set, the envelope track of residual error when namely sensor non-fault exports.

3. a kind of interlock alarm system based on sensor fault diagnosis technology as claimed in claim 1 or 2, it is characterized in that: in step (202), described envelope track is by mission nonlinear model uncertain error scope the stochastic error γ (t) caused in linearized stability σ (t) and measuring process causes the threshold range of residual error to form along with time changing curve, and adaptive threshold envelope track Ψ (t) has its minimum and maximum border;

$\mathrm{\γ}\left(t\right)+\mathrm{\σ}\left(t\right)-\∂(x,u)\≤\mathrm{\Ψ}\left(t\right)\≤\mathrm{\γ}\left(t\right)+\mathrm{\σ}\left(t\right)-\∂(x,u)---\left(4\right)$

In a kth sampling instant, if the weighted moving average of residual error drop in the scope of track Ψ (t), judged that residual error is by mission nonlinear model uncertain error scope the stochastic error γ (t) that linearized stability σ (t) and measurement noises cause causes, and is now judged as that sensor is in normal work; Otherwise, in a kth sampling instant, if the weighted moving average of residual error outside the scope dropping on track Ψ (t), then think that residual error further comprises other compositions, be now judged as sensor failure.

4. a kind of interlock alarm system based on sensor fault diagnosis technology as claimed in claim 1, it is characterized in that, sensor fault is classified as soft fault and hard fault according to the intensity of variation of signal amplitude after breaking down, pace of change by described historical failure feature database, amplitude changes less and is soft fault slowly, and amplitude changes greatly and is hard fault rapidly; Specifically comprise: deviation fault, drifting fault, precise decreasing fault and complete failure.

5. a kind of interlock alarm system based on sensor fault diagnosis technology as claimed in claim 4, is characterized in that:

(501) function representation of described deviation fault is:

ζ(x,u,t)＝K (5)

Wherein, K is constant;

(502) function representation of described drifting fault is:

ζ(x.u.t)＝K(t-t ₀) (6)

Wherein, K is constant, t ₀for fault initial time;

(503) function representation of described precise decreasing fault is:

ζ(x.u.t)～N(0,σ ₂ ²) (7)

Wherein, σ ₂ ²represent variance;

(504) function representation of described complete failure is:

Y (t)=V _max(or V _min) (8)

Wherein, V _max, V _minrepresent maximal value and the minimum value of instrument range respectively.

6. a kind of interlock alarm system based on sensor fault diagnosis technology as claimed in claim 5, it is characterized in that: in the complete failure of described (504), fault has just started there is a forming process when occurring, and is expressed as:

Wherein, t ₀for fault initial time.