CN118171155A - HMM-based equipment state monitoring and fault diagnosis method and system - Google Patents
HMM-based equipment state monitoring and fault diagnosis method and system Download PDFInfo
- Publication number
- CN118171155A CN118171155A CN202410247219.5A CN202410247219A CN118171155A CN 118171155 A CN118171155 A CN 118171155A CN 202410247219 A CN202410247219 A CN 202410247219A CN 118171155 A CN118171155 A CN 118171155A
- Authority
- CN
- China
- Prior art keywords
- state
- hmm
- fault
- lambda
- fault diagnosis
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 48
- 238000012544 monitoring process Methods 0.000 title claims abstract description 39
- 238000003745 diagnosis Methods 0.000 title claims description 36
- 238000007476 Maximum Likelihood Methods 0.000 claims abstract description 9
- 238000012549 training Methods 0.000 claims abstract description 8
- 239000011159 matrix material Substances 0.000 claims description 14
- 230000007704 transition Effects 0.000 claims description 7
- 230000006870 function Effects 0.000 claims description 5
- 238000003860 storage Methods 0.000 claims description 3
- 238000001514 detection method Methods 0.000 abstract description 5
- 230000005856 abnormality Effects 0.000 abstract description 3
- 238000010586 diagram Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 108091026890 Coding region Proteins 0.000 description 1
- 238000005452 bending Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 238000006073 displacement reaction Methods 0.000 description 1
- 230000005284 excitation Effects 0.000 description 1
- 238000009776 industrial production Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
Landscapes
- Testing And Monitoring For Control Systems (AREA)
Abstract
The invention relates to a method and a system for monitoring equipment state and diagnosing faults based on HMM, wherein the method comprises the following steps: finely classifying the state of the whole life cycle of the equipment from a normal state to various known fault states; for each normal or fault state of the device, determining all the characteristic parameters which can be observed in the state; training out the maximum probability P (O|lambda) of the observed sequence under the model by using a maximum likelihood estimation method for the observed sequence of the characteristic parameters under each state and an estimation model lambda= (pi, A, B); obtaining an unknown observation sequence O ', training out the maximum probability P (O ' |lambda) of the observation sequence HMM model observation sequence, sequentially calculating the KL distance of the probability density of P (O ' |lambda) and P (O|lambda) under each state, wherein the minimum distance is the KL distance between the unknown observation sequence and the corresponding state, and the minimum distance is the state between the newly obtained observation sequence and the corresponding state. The method is simple and feasible, has strong practicability, can realize early detection and alarm of fault abnormality, and avoids economic loss caused by post detection.
Description
Technical Field
The invention belongs to the technical field of industrial equipment state monitoring and fault diagnosis, and particularly relates to an equipment state monitoring and fault diagnosis method and system based on an HMM (Hidden Markov Model, HMM, hidden Markov model).
Background
Status monitoring and fault diagnosis of equipment have been important and difficult points in industrial production. Conventional sensors installed on equipment comprise temperature, pressure, flow, vibration and the like, production operators can see the change trend of the state of the equipment from the readings of the sensors, but only experienced experts can obtain the state degradation degree of the equipment and possible faults through comprehensive judgment, and a great deal of expertise and abundant experience are needed to make judgment.
Disclosure of Invention
The invention aims to provide a device state monitoring and fault diagnosis method and system based on an HMM (Hidden Markov Model, HMM, hidden Markov model) so as to solve the technical problems.
The invention provides an equipment state monitoring and fault diagnosis method based on an HMM, which comprises the following steps:
step 1, carefully classifying the state of the whole life cycle of equipment from a normal state to various known fault states;
Step 2, determining all the characteristic parameters which can be observed in each normal or fault state of the equipment in the step 1;
Step 3, training out the maximum probability P (O|lambda) of the observed sequence under the model by using a maximum likelihood estimation method through the observed sequence of the characteristic parameters and the estimated model lambda= (pi, A, B) under each state in the step 2; wherein pi is an initial state probability vector, A is a state transition probability matrix, and B is an observation value probability matrix;
And 4, obtaining an unknown observation sequence O ', training out the maximum probability P (O ' |lambda) of the observation sequence HMM model observation sequence by using the method of the step 3, and sequentially calculating the KL distances of the probability density P (O ' |lambda) and the P (O|lambda) in each state in the step 3, wherein the minimum distance is the state corresponding to the unknown observation sequence.
Further, in step 1, the device states are divided into a normal state, a fault 1 state, a fault 2 state, and a fault 3 state.
Further, the characteristic parameters in the step 2 include temperature, pressure, mechanical vibration, phase value, flow, power and valve opening.
Further, the model in step 3 is a hidden markov model, which is a probability model related to time sequence, and is described as follows:
n: the number of states of the Markov chain, and the state at the time t is q t epsilon {1,2, …, N };
o: any possible observed value corresponding to different states is recorded as o t at the time of t;
Pi: an initial state probability vector, pi= { pi 1,π2,…,πN},πi=P(q1 =i), 1.ltoreq.i.ltoreq.N;
a: a state transition probability matrix, a= { a ij},aij=P(qt+1=j|qt =i), 1.ltoreq.i, j.ltoreq.n;
B: observation probability matrix, b= { B j(k)},bj(k)=P(ot=vk|qt =j), 1.ltoreq.k.ltoreq.o.
Λ= (pi, a, B) is a discrete HMM.
For an observed sequence O under a certain known equipment state, given an estimation model λ= (pi, a, B), using a method of maximum likelihood parameter estimation, the maximum probability P (o|λ) of the observed sequence occurrence under the model λ= (pi, a, B) is trained.
Further, in step 4, after the unknown observation sequence O 'is obtained, the method of step 3 is used to train out the maximum probability P (O' |λ) of the HMM model of the observation sequence.
Further, the KL distance between two similar probability density functions q and p is defined as:
Sequentially calculating the KL distance of the probability density of P (O' |lambda) and P (O|lambda) in each state in the step 3, and comparing the sizes one by one; if the distance between the calculated result and the normal state KL is minimum, the unknown observation sequence corresponds to the normal state of the equipment, and if the distance between the calculated result and the equipment fault 1 state KL is minimum, the unknown observation sequence corresponds to the fault 1 state.
Further, the extent of deviation of the device from the normal state is reflected by the KL distance of the unknown observation sequence P (O' |λ) and the device normal state P (o|λ).
The invention also provides an equipment state monitoring and fault diagnosis based on the HMM, which comprises an equipment state monitoring and fault diagnosis module, wherein the equipment state monitoring and fault diagnosis module is used for executing the equipment state monitoring and fault diagnosis method based on the HMM according to any one of claims 1 to 7.
The present invention also provides a non-transitory computer readable storage medium storing computer instructions which, when executed by a processor, implement an HMM-based device state monitoring and fault diagnosis method as claimed in any one of claims 1-7.
The invention also provides an electronic device, comprising:
A memory and a processor, the memory and the processor being communicatively connected to each other, the memory having stored therein computer instructions, the processor executing the computer instructions to perform a HMM-based device status monitoring and fault diagnosis method of any of claims 1-7.
By means of the scheme, through the equipment state monitoring and fault diagnosis method and system based on the HMM, the method is simple and feasible, has strong practicability, can realize early fault abnormality discovery and alarm, and avoids economic loss caused by post detection.
The foregoing description is only an overview of the present invention, and is intended to provide a better understanding of the present invention, as it is embodied in the following description, with reference to the preferred embodiments of the present invention and the accompanying drawings.
Drawings
FIG. 1 is a step explanatory diagram of an HMM-based device status monitoring and fault diagnosis method of the present invention;
FIG. 2 is a schematic flow chart of an embodiment of a method for monitoring equipment status and diagnosing faults based on HMM according to the present invention;
FIG. 3 is a graph showing a comparison of KL distances of an embodiment of the HMM-based device state monitoring and fault diagnosis method of the present invention;
fig. 4 is a schematic diagram of an electronic device according to the present invention.
Detailed Description
The following describes in further detail the embodiments of the present invention with reference to the drawings and examples. The following examples are illustrative of the invention and are not intended to limit the scope of the invention.
Referring to fig. 1, the present embodiment provides a method for monitoring equipment status and diagnosing faults based on HMM, including the following steps:
Step S1, carefully classifying the state of the whole life cycle of the equipment from a normal state to various known fault states. Such as the device state may be classified as a normal state, a fault 1 state, a fault 2 state, a fault 3 state, and so on.
Step S2, for each normal or fault state of the device in step S1, determining all the characteristic parameters which can be observed in the state. The characteristic parameters include temperature, pressure, mechanical vibration, phase value, flow, power, valve opening and the like.
Step S3, training out the maximum probability P (O|lambda) of the observed sequence under the model by using a maximum likelihood estimation method through the observed sequence of the characteristic parameters and the estimated model lambda= (pi, A, B) under each state in the step S2; wherein pi is an initial state probability vector, A is a state transition probability matrix, and B is an observation value probability matrix.
And S4, obtaining an unknown observation sequence O ', training out the maximum probability P (O ' |lambda) of the observation sequence HMM model observation sequence by using the method of the step S3, and sequentially calculating the KL distances of the probability density P (O ' |lambda) and the P (O|lambda) in each state in the step S3, wherein the minimum distance is the state corresponding to the unknown observation sequence.
The method uses the hidden Markov model in industrial equipment state monitoring and fault diagnosis, can effectively identify known fault types and avoid accidents. The method is simple and feasible, has strong practicability, can realize early detection and alarm of fault abnormality, and avoids economic loss caused by post detection.
In this embodiment, the model in step S3 is a hidden markov model, which is a probability model related to time sequence, and is described as follows:
n: the number of states of the Markov chain, and the state at the time t is q t epsilon {1,2, …, N };
o: any possible observed value corresponding to different states is recorded as o t at the time of t;
Pi: an initial state probability vector, pi= { pi 1,π2,…,πN},πi=P(q1 =i), 1.ltoreq.i.ltoreq.N;
a: a state transition probability matrix, a= { a ij},aij=P(qt+1=j|qt =i), 1.ltoreq.i, j.ltoreq.n;
B: observation probability matrix (discrete HMM), b= { B j(k)},bj(k)=P(ot=vk|qt =j, 1.ltoreq.k.ltoreq.o.
Λ= (pi, a, B) is a discrete HMM.
For an observed sequence O under a certain known equipment state, given an estimation model λ= (pi, a, B), using a method of maximum likelihood parameter estimation, the maximum probability P (o|λ) of the observed sequence occurrence under the model λ= (pi, a, B) is trained.
In this embodiment, after the unknown observation sequence O 'is obtained in step S4, the method of step S3 is used to train out the maximum probability P (O' |λ) of the HMM model of the observation sequence.
The KL (Kullback-Leibler) distance between two similar probability density functions q and p is defined as:
Sequentially calculating the KL (Kullback-Leibler) distance of the probability density of P (O' |lambda) and P (O|lambda) in each state in the step S3, and comparing the magnitudes one by one; if the distance between the calculated result and the normal state KL is minimum, the unknown observation sequence corresponds to the normal state of the equipment, and if the distance between the calculated result and the equipment fault 1 state KL is minimum, the unknown observation sequence corresponds to the fault 1 state and the like. Meanwhile, the KL distance of the unknown observation sequence P (O' |λ) and the device normal state P (o|λ) reflects the degree of deviation of the device from the normal state.
The present invention will be described in further detail below with reference to fig. 2 to 3.
An industrial steam turbine is selected as a research object, and a steam turbine state monitoring and fault diagnosis method based on a Hidden Markov Model (HMM) comprises the following steps:
The first step: the state of the full life cycle of the steam turbine is finely classified from a normal state to various known fault states.
The state of the steam turbine is divided into a normal state, a mass unbalance fault, a dynamic and static friction fault, a rotor thermal bending fault, a misalignment fault, a structural resonance fault, a steam flow excitation fault and a rotor crack fault.
And a second step of: for each normal or fault condition of the turbine, all the characteristic parameters that can be observed in that condition are determined.
The characteristic parameters comprise main steam temperature, main steam pressure, main steam flow, mechanical vibration, vibration phase value, shaft seal air supply temperature, shaft displacement, expansion difference, active power and valve opening.
Classifying according to different characteristic parameters contained in different states, wherein the parameters are normally marked as O, the parameters are abnormally marked as X, and the classification is shown in the following table.
And a third step of: under each state of the steam turbine, a plurality of groups of observation sequences are selected to train out the maximum probability P (O|lambda) of the observation sequences under the model by using a maximum likelihood estimation method.
First, an estimation model λ= (pi, a, B) is preset, and the coding region of the model is divided into 3 parts. The first part is the initial state probability vector pi (i.e., pi region) of the feature parameters of the HMM, and the second part is the third part of the observation probability matrix B (i.e., B region) of the state transition matrix a (i.e., a region). Depending on the nature of the HMM type, there is a ij noteq 0 when j+.i. (i, j=1, 2,., N), and a NN =1. Meanwhile, each part satisfies the following conditions:
When P (o|λ) is maximum, the HMM maximum likelihood estimate can be given the parameter λ= (pi, a, B), where P (o|λ) is the fitness function. The logarithm of P (O|lambda) is conveniently taken for calculation. Taking the number of observation sequences as K, taking P (O (k) |lambda) as the observation probability of observing the sequence of the kth group, and taking the fitness function as follows:
The selection terminates the computation at the maximum evolution algebra.
Fourth step: and obtaining an observation sequence of an unknown state of the steam turbine, and repeating the method in the last step to train out the maximum probability P (O' |lambda) of the observation sequence of the sequence HMM model.
The distance between P (O' |lambda) and KL (Kullback-Leibler) of probability density of P (O|lambda) under each known state of the steam turbine is calculated in sequence, normalization processing is carried out on the vertical axis of the graph, and the minimum distance is the state corresponding to the newly acquired observation sequence, and the calculated example is shown in a reference figure 3.
The embodiment also provides equipment state monitoring and fault diagnosis based on the HMM, which comprises an equipment state monitoring and fault diagnosis module, wherein the equipment state monitoring and fault diagnosis module is used for executing the equipment state monitoring and fault diagnosis method based on the HMM.
The present embodiment also provides a non-transitory computer readable storage medium storing computer instructions that, when executed by a processor, implement the HMM-based device state monitoring and fault diagnosis method.
As shown in fig. 4, this embodiment further provides an electronic device, including:
The device comprises a memory 201 and a processor 202, wherein the memory 201 and the processor 202 are in communication connection, computer instructions are stored in the memory 201, and the processor 202 executes the computer instructions, so that the device state monitoring and fault diagnosis method based on the HMM is executed.
The above description is only of the preferred embodiments of the present invention and is not intended to limit the present invention, and it should be noted that it is possible for those skilled in the art to make several improvements and modifications without departing from the technical principle of the present invention, and these improvements and modifications should also be regarded as the protection scope of the present invention.
Claims (10)
1. An equipment state monitoring and fault diagnosis method based on HMM is characterized by comprising the following steps:
step 1, carefully classifying the state of the whole life cycle of equipment from a normal state to various known fault states;
Step 2, determining all the characteristic parameters which can be observed in each normal or fault state of the equipment in the step 1;
Step 3, training out the maximum probability P (O|lambda) of the observed sequence under the model by using a maximum likelihood estimation method through the observed sequence of the characteristic parameters and the estimated model lambda= (pi, A, B) under each state in the step 2; wherein pi is an initial state probability vector, A is a state transition probability matrix, and B is an observation value probability matrix;
And 4, obtaining an unknown observation sequence O ', training out the maximum probability P (O ' |lambda) of the observation sequence HMM model observation sequence by using the method of the step 3, and sequentially calculating the KL distances of the probability density P (O ' |lambda) and the P (O|lambda) in each state in the step 3, wherein the minimum distance is the state corresponding to the unknown observation sequence.
2. The HMM-based device state monitoring and fault diagnosis method according to claim 1, wherein the device state in step 1 is classified into a normal state, a fault 1 state, a fault 2 state, and a fault 3 state.
3. The HMM-based device status monitoring and fault diagnosis method of claim 1, wherein the characteristic parameters in step 2 include temperature, pressure, mechanical vibration, phase value, flow, power, valve opening.
4. The HMM-based device state monitoring and fault diagnosis method of claim 2, wherein the model in step 3 is a hidden markov model, which is a probabilistic model with respect to time sequences, described as follows:
n: the number of states of the Markov chain, and the state at the time t is q t epsilon {1,2, …, N };
o: any possible observed value corresponding to different states is recorded as o t at the time of t;
Pi: an initial state probability vector, pi= { pi 1,π2,…,πN},πi=P(q1 =i), 1.ltoreq.i.ltoreq.N;
a: a state transition probability matrix, a= { a ij},aij=P(qt+1=j|qt =i), 1.ltoreq.i, j.ltoreq.n;
B: observation probability matrix, b= { B j(k)},bj(k)=P(ot=vk|qt =j), 1.ltoreq.k.ltoreq.o.
Λ= (pi, a, B) is a discrete HMM.
For an observed sequence O under a certain known equipment state, given an estimation model λ= (pi, a, B), using a method of maximum likelihood parameter estimation, the maximum probability P (o|λ) of the observed sequence occurrence under the model λ= (pi, a, B) is trained.
5. The HMM-based equipment state monitoring and fault diagnosis method according to claim 1, wherein in step 4, after the unknown observation sequence O 'is obtained, the method of step 3 is used to train out the observation sequence HMM model with the maximum probability P (O' |λ).
6. The HMM-based device state monitoring and fault diagnosis method of claim 5, wherein the KL distance between two similar probability density functions q and p is defined as:
sequentially calculating the KL distance of the probability density of P (O' lambda) and P (O lambda) in each state in the step 3, and comparing the distances one by one; if the distance between the calculated result and the normal state KL is minimum, the unknown observation sequence corresponds to the normal state of the equipment, and if the distance between the calculated result and the equipment fault 1 state KL is minimum, the unknown observation sequence corresponds to the fault 1 state.
7. The HMM-based device state monitoring and fault diagnosis method according to claim 6, wherein the extent of device deviation from a normal state is reflected by the KL distance of the unknown observation sequence P (O' λ) and the device normal state P (O λ).
8. An HMM-based device state monitoring and fault diagnosis module, characterized by comprising a device state monitoring and fault diagnosis module for performing an HMM-based device state monitoring and fault diagnosis method according to any one of claims 1 to 7.
9. A non-transitory computer readable storage medium storing computer instructions which, when executed by a processor, implement an HMM-based device state monitoring and fault diagnosis method of any of claims 1-7.
10. An electronic device, comprising:
A memory and a processor, the memory and the processor being communicatively connected to each other, the memory having stored therein computer instructions, the processor executing the computer instructions to perform a HMM-based device status monitoring and fault diagnosis method of any of claims 1-7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202410247219.5A CN118171155A (en) | 2024-03-05 | 2024-03-05 | HMM-based equipment state monitoring and fault diagnosis method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202410247219.5A CN118171155A (en) | 2024-03-05 | 2024-03-05 | HMM-based equipment state monitoring and fault diagnosis method and system |
Publications (1)
Publication Number | Publication Date |
---|---|
CN118171155A true CN118171155A (en) | 2024-06-11 |
Family
ID=91346441
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202410247219.5A Pending CN118171155A (en) | 2024-03-05 | 2024-03-05 | HMM-based equipment state monitoring and fault diagnosis method and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN118171155A (en) |
-
2024
- 2024-03-05 CN CN202410247219.5A patent/CN118171155A/en active Pending
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109141881B (en) | Rotary machine health assessment method of deep self-coding network | |
KR101955305B1 (en) | Gas turbine sensor failure detection utilizing a sparse coding methodology | |
Filev et al. | An industrial strength novelty detection framework for autonomous equipment monitoring and diagnostics | |
US6587737B2 (en) | Method for the monitoring of a plant | |
CN115034248A (en) | Automatic diagnostic method, system and storage medium for equipment | |
CN108335021B (en) | Method for evaluating state reliability of wind turbine and maintenance decision optimization | |
CN111648992B (en) | Gas turbine compressor fault identification early warning method | |
CN112418306B (en) | Gas turbine compressor fault early warning method based on LSTM-SVM | |
CN112067335A (en) | Power plant blower fault early warning method based on multivariate state estimation | |
KR102206737B1 (en) | Method and system for fault diagnosis of pump-turbines based on machine learning | |
CN114645844B (en) | Method, computing device and computer medium for determining flow state of air compression station | |
CN111723925A (en) | Method, device, equipment and medium for fault diagnosis of on-road intelligent train air conditioning unit | |
CN114936520B (en) | Steam turbine fault early warning analysis method based on improved MSET | |
CN114738132B (en) | Supercharger rotation speed sensor fault diagnosis method and system based on real ship data | |
CN118171155A (en) | HMM-based equipment state monitoring and fault diagnosis method and system | |
CN112576454A (en) | Wind turbine generator main shaft temperature early warning method and device based on multi-dimensional early warning strategy | |
CN116383615A (en) | Multisource sensing information fusion and expansion method based on system state characterization | |
CN115526238A (en) | Induced draft fan fault early warning method based on all-condition convolution characteristic memory | |
Abid et al. | Adaptive data-driven approach for the remaining useful life estimation when few historical degradation sequences are available | |
US20180087489A1 (en) | Method for windmill farm monitoring | |
CN111914353B (en) | Rotor low-cycle fatigue loss detection method, device and computer equipment | |
Zarate et al. | Computation and monitoring of the deviations of gas turbine unmeasured parameters | |
KR100882887B1 (en) | A trend monitoring and diagnostic analysis method for a vacuum pump and a trend monitoring and diagnostic analysis system therefor and computer-readable storage media including a computer program which performs the mothod | |
Mahmoodian et al. | A life clustering framework for prognostics of gas turbine engines under limited data situations | |
Cody et al. | Trade-offs in metric learning for bearing fault diagnosis |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination |