CN115307943A - Method and system for detecting abnormal operation state of rotating mechanical equipment under multiple working conditions - Google Patents
Method and system for detecting abnormal operation state of rotating mechanical equipment under multiple working conditions Download PDFInfo
- Publication number
- CN115307943A CN115307943A CN202210927715.6A CN202210927715A CN115307943A CN 115307943 A CN115307943 A CN 115307943A CN 202210927715 A CN202210927715 A CN 202210927715A CN 115307943 A CN115307943 A CN 115307943A
- Authority
- CN
- China
- Prior art keywords
- data
- model
- support vector
- module
- mechanical equipment
- 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
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M99/00—Subject matter not provided for in other groups of this subclass
- G01M99/004—Testing the effects of speed or acceleration
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- General Engineering & Computer Science (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention discloses a method for detecting abnormal operation states of rotating mechanical equipment under multiple working conditions, which comprises the following steps: collecting and selecting vibration signals of the rotary mechanical equipment in a multi-working-condition normal running state, learning a description model, collecting the vibration signals of the rotary mechanical equipment under real-time monitoring and describing the model for abnormal detection; learning an improved support vector data description model from vibration signals of the rotating mechanical equipment in a multi-working-condition normal running state as a training process, and carrying out anomaly detection on real-time monitoring signals of the rotating mechanical equipment as a testing process; training process: extracting time domain characteristics and frequency domain characteristics of the subsequence, standardizing the characteristics, and learning the improved support vector data description model; and (3) testing process: and extracting time domain characteristics and frequency domain characteristics of the subsequence, and performing characteristic standardization on the extracted characteristics by adopting characteristic standardization parameters in a training process to obtain an abnormal detection result. The invention can detect the running state of the equipment and alarm the abnormal running state.
Description
Technical Field
The invention relates to the technical field of abnormal detection of running states of rotary mechanical equipment, in particular to a method and a system for detecting abnormal running states of rotary mechanical equipment under multiple working conditions.
Background
The rotary machine is a machine capable of performing a specific function by a rotational motion of a rotating member, and is widely used in the industrial field. However, the rotary machines are subject to various failure modes due to the possible severe working conditions and permanent operation, thus impairing the smooth development of the production task. Therefore, the abnormal operation state detection of the rotating mechanical equipment provides technical support for maintenance according to situations, and has important significance.
The operation of the rotating mechanical device will generate vibration signals in real time that imply information characterizing the state of the device. Most of the current anomaly detection technologies at home and abroad regard an anomaly detection task as a two-classification supervised learning problem or an unsupervised learning problem in methods, and train a model in a training set containing normal samples and abnormal samples by methods such as a support vector machine, a neural network, a cluster and an isolated forest, and then use the model to test data to be detected. Although the above unsupervised learning method has overcome the problems of high cost and difficulty in implementation due to the need of labeling the sample label by an expert, a more general and practical problem is that, in the task of detecting the abnormal operation state of many devices, the devices are mostly in the normal operation state, the abnormal operation state is terminated in time, the difficulty of fault injection is high, the cost is high, and therefore, the abnormal sample data is few or not available at all, so that the application of the conventional abnormal detection method is limited.
Disclosure of Invention
The invention aims to provide a method and a system for detecting abnormal running states of rotating mechanical equipment under multiple working conditions, so as to solve the problems.
The technical scheme adopted by the invention for solving the technical problem is as follows:
a method for detecting abnormal operation states of rotating mechanical equipment under multiple working conditions comprises the following steps:
collecting and selecting vibration signals of the rotary mechanical equipment in a multi-working-condition normal running state, learning an improved support vector data description model, collecting vibration signals of the rotary mechanical equipment under real-time monitoring, and performing anomaly detection through the established improved support vector data description model;
learning and improving a support vector data description model from vibration signals of the rotating mechanical equipment in a multi-working-condition normal running state as a training process, and carrying out anomaly detection on real-time monitoring signals of the rotating mechanical equipment as a testing process;
training process: truncating the normal state historical data of multiple working conditions by using a sliding window, dividing the historical data into subsequences with equal length, extracting time domain characteristics and frequency domain characteristics of the subsequences, and standardizing the characteristics; taking the normalized features as input data of an improved support vector data description model, and learning the improved support vector data description model;
the testing process comprises the following steps: cutting the vibration signal acquired in real time by using a sliding window to obtain a latest subsequence, extracting time domain characteristics and frequency domain characteristics of the subsequence, and performing characteristic standardization on the extracted characteristics by adopting characteristic standardization parameters in a training process; and taking the normalized features as input data of an improved support vector data description model to obtain an abnormal detection result of the model.
Further, the specific steps of the model training process for improving the description of the support vector data are as follows:
s1, establishing an optimal Gaussian mixture model of training data X, and dividing a sample into a certain data subset according to the maximum responsivity of the sample; determining a Gaussian mixture model component number K optimal to the training data X based on an optimal model selection criterion;
s2, for each data subset divided in the step S1, learning a combination coefficient of a basic kernel function of the data subset based on a central kernel alignment multi-kernel learning method, and constructing a multi-kernel function of the data subset;
and S3, for each data subset divided in the step S1, adopting the multi-kernel function constructed in the step S2, endowing fuzzy weight to each sample according to the probability density information of the Gaussian mixture model in the step S1, and learning a fuzzy multi-kernel support vector data description sub-model.
Further, in S2, for the data subset a k Memory for recording
y i Is sample x in the data subset i The corresponding label is used for model training only by samples in a normal state, so that y i ＝1，i＝1,2,...,|A k L, |; let K * ＝yy T For an NxN kernel matrix K, a corresponding centered kernel matrix
Wherein I is an nxn identity matrix, and l is a vector in which all the elements of nx1 are 1; the combination coefficient of the kernel function of the data subset is obtained by solving the following formula:
wherein eta m Is the combination coefficient of the mth basic kernel function, M is the number of basic kernel functions, K m Is the kernel matrix corresponding to the mth base kernel function, K mc Is the centralized kernel matrix corresponding to the mth base kernel function,
is a centred K * ，<·,·> F Is Frobenius inner product, | | · |. The purple F Is the Frobenius norm;
the constructed multi-core function of the data subset is
Wherein k is m (. Is) the mth base kernel function, x i And x j Are two samples of the kernel function to be solved in the data subset.
Further, in S2, for the data subset a k The learning of the fuzzy multi-core support vector data descriptor model is completed by solving the following formula:
wherein k is m (. Is) the mth base kernel function, C k Is a penalty factor that is a function of,
is a sample x i The weight of the blur of (a) is,
and
is a sample x i And x j A corresponding Lagrangian multiplier;
solving the optimal solution alpha of the above problem k Then, an important parameter calculation formula of the fuzzy multi-core support vector data description submodel is as follows:
wherein, sample x l Corresponding to
Is a non-zero value.
Further, the specific steps of the model test process for improving the description of the support vector data are as follows:
s1, sequentially using each learned fuzzy multi-core support vector data description sub-model to detect whether test data z is in a normal running state;
s2, when any fuzzy multi-core support vector data descriptor model detects that the test data is in a normal running state, judging that the running state of the test data is finally normal; and when all the fuzzy multi-core support vector data descriptor models detect that the test data are in abnormal operation states, judging that the operation states of the test data are finally abnormal.
Further, whether the test data z is in a normal operation state is detected through the following formula:
wherein k is m (. Is) the mth base kernel function, η m Is the combination coefficient of the basic kernel function, and M is the number of the basic kernel functions; if f (z) is less than or equal to R k Then, in the kth submodel, the number is testedIs detected as being in a normal operation state; if f (z) > R k Then in the kth submodel, the test data is detected as being in an abnormal operating state.
Further, the optimal model selection criterion adopted in the step S1 in the model training process described by the improved support vector data is a bayesian information criterion, and the optimal gaussian mixture model component K for the training data X is determined by the following formula:
where N is the number of samples of training data X, P (X) i Theta) is sample X in training data X i The global probability density estimate in the gaussian mixture model, θ, is a parameter of the gaussian mixture model.
A kind of rotating machinery apparatus multi-operating mode abnormal condition detection system, comprising: the system comprises an equipment data management module, a data preprocessing and feature extraction module, an abnormality detection core module, a database module and a human-computer interaction module;
the equipment data management module realizes real-time acquisition of vibration signal data by using the acceleration sensor; the device data management module realizes the storage management operation of real-time data and the reading management operation of historical data;
the data preprocessing and feature extraction module divides the original vibration signal data into subsequences with equal length by using a sliding window; the data preprocessing and feature extraction module extracts time domain features and frequency domain features of the subsequences; the data preprocessing and feature extraction module realizes the standardization of the extracted time domain features and frequency domain features;
the anomaly detection core module realizes model training by utilizing the time domain characteristics and the frequency domain characteristics of a standardized training set; the anomaly detection core module performs model test and detection result judgment on the time domain characteristics and the frequency domain characteristics of the standardized real-time acquired data to realize anomaly detection of the current running state of the equipment;
the database module realizes the storage of the acquired data; the database module realizes the storage of a time domain feature and a frequency domain feature extraction algorithm; the database module realizes the parameter storage of characteristic standardization after the truncation and characteristic extraction of the training set data by a sliding window; the database management module realizes model storage after the anomaly detection core module finishes model training;
the man-machine interaction module visualizes the monitoring data acquired in real time; the man-machine interaction module operates and controls selection of training data, selection of a feature extraction algorithm, selection of a feature standardization parameter, a model training process and selection of a detection model; and the man-machine interaction module displays the result of the abnormity detection in real time.
Further, the feature extraction algorithm in the database module feature extraction algorithm library selectable by the feature extraction submodule in the data preprocessing and feature extraction module includes: mean, root mean square value, peak, kurtosis, skewness, peak factor, impulse factor, form factor, and margin factor.
The invention has the beneficial effects that:
(1) The vibration signal data naturally generated when the rotary mechanical equipment runs in the normal state are fully utilized, the model training in the method does not need the data in the abnormal state or labels for the data, the data acquisition mode is simple, and the data acquisition cost is low;
(2) The method can keep higher abnormal detection accuracy rate, lower false alarm rate and false alarm rate under the conditions that normal state data are in multiple actual working conditions and each working condition is not distinguished and labeled;
(3) The method can keep higher abnormal detection accuracy rate and lower false alarm rate and false alarm rate under the condition that the normal state data is actually mixed into the abnormal state data unknowingly.
Drawings
FIG. 1 is a flow chart of the method of the present invention;
FIG. 2 is a schematic diagram of the system structure of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Referring to fig. 1, the invention discloses a method for detecting abnormal operating conditions of rotating mechanical equipment, comprising the following steps:
collecting and selecting vibration signals of the rotary mechanical equipment in a multi-working-condition normal operation state, learning an improved support vector data description model, collecting the vibration signals of the rotary mechanical equipment under real-time monitoring, and performing anomaly detection through the established improved support vector data description model;
learning an improved support vector data description model from vibration signals of the rotating mechanical equipment in a multi-working-condition normal running state as a training process, and carrying out anomaly detection on real-time monitoring signals of the rotating mechanical equipment as a testing process;
training process: intercepting the normal state historical data of multiple working conditions by using a sliding window, dividing the historical data into subsequences with equal length, extracting time domain characteristics and frequency domain characteristics of the subsequences and standardizing the characteristics; taking the normalized features as input data of an improved support vector data description model, and learning the improved support vector data description model;
the testing process comprises the following steps: truncating the vibration signal acquired in real time by using a sliding window to obtain a latest subsequence, extracting time domain characteristics and frequency domain characteristics of the subsequence, and performing characteristic standardization on the extracted characteristics by adopting characteristic standardization parameters in a training process; and taking the normalized features as input data of an improved support vector data description model to obtain an abnormal detection result of the model.
The further improvement of the invention is that the model training process described by the improved support vector data comprises the following specific steps:
s1, establishing an optimal Gaussian mixture model of training data X, and dividing a sample into a certain data subset according to the maximum responsivity of the sample; determining a Gaussian mixture model component number K optimal to the training data X based on an optimal model selection criterion;
s2, for each data subset divided in the step S1, learning a combination coefficient of a basic core function of the data subset based on a central core alignment multi-core learning method, and constructing a multi-core function of the data subset;
for data subset A k Memory for recording
y i Is sample x in the data subset i Corresponding labels, y, because the method only needs samples in normal state for model training i ＝1，i＝1,2,...,|A k L. Let K * ＝yy T For an NXN kernel matrix K, the corresponding centralized kernel matrix
Where I is an N × N identity matrix and l is a vector in which all the elements of N × 1 are 1. The combination coefficients of the basis kernel functions of the data subsets are solved by the following problems:
wherein eta is m Is the combination coefficient of the mth basic kernel function, M is the number of basic kernel functions, K m Is the kernel matrix corresponding to the mth base kernel function, K mc Is the centralized kernel matrix corresponding to the mth base kernel function,
is a centred K * ，<·,·> F Is Frobenius inner product, | | · |. The purple F Is the Frobenius norm;
the constructed multi-core function of the data subset is
Wherein k is m (. Is) the mth base kernel function, η m Is the combination coefficient of the basic kernel function, and M is the number of the basic kernel functions; x is a radical of a fluorine atom i And x j Are two samples of the kernel function to be solved in the data subset.
S3, for each data subset divided in the step S1, adopting the multi-kernel function constructed in the step S2, endowing each sample with fuzzy weight according to the probability density information of the Gaussian mixture model in the step S1, and learning a fuzzy multi-kernel support vector data description sub-model;
for data subset A k The learning of the fuzzy multi-core support vector data description submodel is completed by solving the following problems:
wherein k is m (. H) is the mth base kernel function, η m Is the combination coefficient of the kernel function, M is the number of kernel functions, C k Is a penalty factor that is a function of the time,
is a sample x i The weight of the blur of (a) is,
and
is a sample x i And x j A corresponding Lagrangian multiplier;
solving the optimal solution alpha of the above problem k Thereafter, the multi-core support vector data is blurredAn important parameter calculation formula for the descriptor model is as follows:
wherein k is m (. Is) the mth base kernel function, η m Is the combination coefficient of the basic kernel function, M is the number of the basic kernel functions, and sample x l Corresponding to
Is a non-zero value;
the further improvement of the invention is that the model test process described by the improved support vector data comprises the following specific steps:
s1, sequentially using each learned fuzzy multi-core support vector data description sub-model to detect whether test data z is in a normal operation state; more specifically, the following equation is calculated:
wherein k is m (. Is) the mth base kernel function, η m Is the combination coefficient of the base kernel function, and M is the number of base kernel functions. If f (z) is less than or equal to R k Then in the kth submodel, the test data is detected to be in a normal operation state; if f (z) > R k Then in the kth submodel, the test data is detected to be in an abnormal operation state;
s2, when any fuzzy multi-core support vector data descriptor model detects that the test data is in a normal running state, judging that the running state of the test data is finally normal; and when all the fuzzy multi-core support vector data descriptor models detect that the test data are in abnormal operation states, judging that the operation states of the test data are finally abnormal.
In order to further optimize the technical scheme, the optimal model selection criterion adopted in the step S1 in the model training process described by the improved support vector data is a Bayesian information criterion, and the optimal Gaussian mixture model component K for the training data X is determined by the following formula:
where N is the number of samples of training data X, P (X) i Theta) is sample X in training data X i The global probability density estimate in the gaussian mixture model, θ, is a parameter of the gaussian mixture model.
As shown in fig. 2, an embodiment of the present invention discloses a system for detecting an abnormal operating state of a rotating mechanical device under multiple operating conditions, including: the device comprises an equipment data management module, a data preprocessing and feature extraction module, an abnormality detection core module, a database module and a human-computer interaction module;
the equipment data management module comprises a real-time data acquisition submodule and a data management operation submodule; the real-time data acquisition submodule transmits vibration signal data acquired in real time to the data preprocessing and feature extraction module by using the acceleration sensor for preprocessing and feature extraction so as to be detected, transmits the vibration signal data acquired in real time to the data management operation submodule for data storage, and transmits the vibration signal data acquired in real time to the man-machine interaction module; the data management operation sub-module receives data transmitted by the real-time data acquisition sub-module and stores the data into a historical database of the database module on one hand, and reads historical data from the historical database of the database module and transmits the historical data to the data preprocessing and feature extraction module for preprocessing and feature extraction so as to train the model on the other hand; the data management operation sub-module receives historical data reading selection control from the man-machine interaction module;
the data preprocessing and feature extraction module comprises a sliding window truncation sub-module, a feature extraction sub-module and a feature standardization sub-module; the sliding window truncation sub-module receives data from the equipment data management module, divides the data into sub-sequence data with equal length by using a sliding window and transmits the sub-sequence data to the characteristic extraction sub-module; the characteristic extraction submodule receives the subsequence cut from the sliding window, reads a characteristic extraction algorithm library in the database module, extracts the time domain characteristic and the frequency domain characteristic of the subsequence, and transmits the time domain characteristic and the frequency domain characteristic to the characteristic standardization submodule; the feature extraction algorithm library receives feature extraction algorithm selection control from the human-computer interaction module; the characteristic standardization sub-module receives the characteristics extracted by the characteristic extraction sub-module, standardizes the characteristics in the training process, and stores related parameters adopted by characteristic standardization in a characteristic standardization parameter library of the database module; the characteristic standardization sub-module receives the characteristics extracted by the characteristic extraction sub-module, reads parameters in a database module characteristic standardization parameter library in the test process, and standardizes the received characteristics by adopting the read parameters; the characteristic standardization sub-module transmits the standardized characteristics to the abnormality detection core module; the characteristic standardization parameter library receives parameter selection control from the human-computer interaction module;
the anomaly detection core module comprises a model training submodule, a model testing submodule and a detection result judging submodule; in the training process, the model training submodule receives the features standardized by the data preprocessing and feature extraction module, learns and improves the support vector data description model, and stores the learned model into a model base of the database module; in the testing process, the model testing sub-module receives the features standardized by the data preprocessing and feature extraction module, reads a model base in the database module, receives model selection control from the human-computer interaction module, performs preliminary abnormal detection by using the selected model, and transmits a preliminary detection result to the detection result judging sub-module; the detection result judgment submodule receives the preliminary detection result from the model test submodule, gives out a final abnormal detection result according to the comprehensive judgment logic and transmits the final abnormal detection result to the man-machine interaction module; the comprehensive judgment logic is that if all the primary detection results are abnormal, the final abnormal detection result is abnormal, and if any one of the primary detection results is normal, the final abnormal detection result is normal;
the database module comprises a historical database, a characteristic standardized parameter library, a characteristic extraction algorithm library and a model library; the historical database receives data from the equipment data management module to realize the storage of real-time collected data; the historical database transmits historical data to the equipment data management module for model training; the feature extraction algorithm library transmits a feature extraction algorithm to the data preprocessing and feature extraction module; in the training process, the characteristic standardized parameter library receives the storage of the characteristic standardized parameters transmitted by the data preprocessing and characteristic extracting module; in the test process, the characteristic standardized parameter library transmits the characteristic standardized parameters to the data preprocessing and characteristic extracting module; the feature extraction algorithm library realizes the storage of time domain features and frequency domain feature extraction algorithms; the model library receives the learned model storage from the anomaly detection core module; the model library transmits the learned model to the abnormality detection core module for model test;
the man-machine interaction module comprises a monitoring signal visualization sub-module, an operation control interface sub-module and an abnormality detection result sub-module; the monitoring signal visualization submodule receives data from the equipment data management module to realize visualization of real-time acquisition of vibration signal data; the operation control interface sub-module operates and controls the equipment data management module to read and select a historical database in the database module; the operation control interface sub-module is used for carrying out operation data preprocessing and feature extraction module on a feature standardized parameter library and a feature extraction algorithm library in the database module to carry out parameter selection control and algorithm selection control; the operation control interface sub-module operation abnormity detection core module performs model selection control on a model library in the database module; the operation control interface sub-module is used for operating and controlling the training process and the testing process; and the abnormity detection submodule receives the final abnormity detection result from the abnormity detection core module and displays the final abnormity detection result.
In order to further optimize the technical scheme, the characteristic extraction algorithm in the database module characteristic extraction algorithm library selectable by the characteristic extraction submodule in the data preprocessing and characteristic extraction module comprises the following steps: calculating the average value, the root mean square value, the peak value, the kurtosis, the skewness, the peak factor, the pulse factor, the wave factor and the margin factor, wherein the formula is as follows:
(1) Mean value
(2) Root mean square value
(3) Peak value X p ＝max(|x(N)|)
(4) Kurtosis
(5) Deflection degree
(6) Crest factor
(7) Pulse factor
(8) Form factor
(9) Margin factor
Wherein x (N) = { x = 1 ,x 2 ,x 3 ,…,x N -is the vibration signal subsequence, N is the subsequence length, μ is the mean of all values of the subsequence, and x (N) - μ represents the mean of all values of the subsequence subtracted from each value in the vibration signal subsequence.
The vibration signal data naturally generated when the rotary mechanical equipment runs in a normal state are fully utilized, model training in the method does not need abnormal state data or label labeling on the data, the data acquisition mode is simple, and the data acquisition cost is low; the method can keep higher abnormal detection accuracy rate, lower false alarm rate and false alarm rate under the conditions that normal state data are in multiple actual working conditions and each working condition is not distinguished and labeled; the method can keep higher abnormal detection accuracy rate and lower false alarm rate and false alarm rate under the condition that the normal state data is actually mixed into the abnormal state data unknowingly.
The invention provides a method and a system for detecting the abnormal running state of the rotating mechanical equipment, which are not dependent on any abnormal running data aiming at the actual problem of the complete lack of the abnormal data, can adapt to the situation of multiple working conditions of normal state data used for model training, can adapt to the situation that some abnormal state data are unknowingly mixed in the normal state data used for model training, and have the advantages of high accuracy, low false alarm rate and low model training time complexity.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (9)
1. A method for detecting abnormal operation states of rotating mechanical equipment under multiple working conditions is characterized by comprising the following steps:
collecting and selecting vibration signals of the rotary mechanical equipment in a multi-working-condition normal running state, learning an improved support vector data description model, collecting vibration signals of the rotary mechanical equipment under real-time monitoring, and performing anomaly detection through the established improved support vector data description model;
learning an improved support vector data description model from vibration signals of the rotating mechanical equipment in a multi-working-condition normal running state as a training process, and carrying out anomaly detection on real-time monitoring signals of the rotating mechanical equipment as a testing process;
training process: intercepting the normal state historical data of multiple working conditions by using a sliding window, dividing the historical data into subsequences with equal length, extracting time domain characteristics and frequency domain characteristics of the subsequences and standardizing the characteristics; taking the normalized features as input data of an improved support vector data description model, and learning the improved support vector data description model;
the testing process comprises the following steps: cutting the vibration signal acquired in real time by using a sliding window to obtain a latest subsequence, extracting time domain characteristics and frequency domain characteristics of the subsequence, and performing characteristic standardization on the extracted characteristics by adopting characteristic standardization parameters in a training process; and taking the normalized features as input data of an improved support vector data description model to obtain an abnormal detection result of the model.
2. The method for detecting the abnormal operating state of the rotating mechanical equipment according to claim 1, wherein the specific steps of the model training process for improving the description of the support vector data are as follows:
s1, establishing an optimal Gaussian mixture model of training data X, and dividing a sample into a certain data subset according to the maximum responsivity of the sample; determining a Gaussian mixture model component number K optimal to the training data X based on an optimal model selection criterion;
s2, for each data subset divided in the step S1, learning a combination coefficient of a basic core function of the data subset based on a central core alignment multi-core learning method, and constructing a multi-core function of the data subset;
and S3, for each data subset divided in the step S1, adopting the multi-kernel function constructed in the step S2, endowing fuzzy weight to each sample according to the probability density information of the Gaussian mixture model in the step S1, and learning a fuzzy multi-kernel support vector data description sub-model.
3. The method according to claim 2, wherein in S2, for the data subset a, the abnormal operating conditions of the rotating mechanical equipment are detected k Memory for recording
y i Is sample x in the data subset i The corresponding label is used for model training only by samples in a normal state, so that y i ＝1，i＝1,2,…,|A k L, |; let K * ＝yy T For an NxN kernel matrix K, a corresponding centered kernel matrix
Wherein I is an nxn identity matrix, and l is a vector in which all the elements of nx1 are 1; the combination coefficient of the kernel function of the data subset is obtained by solving the following formula:
s.t.η m ≥0,
wherein eta m Is the combination coefficient of the mth basic kernel function, M is the number of basic kernel functions, K m Is the kernel matrix corresponding to the mth base kernel function, K mc Is a centralized kernel matrix corresponding to the mth base kernel function,
is a centralized K * ，<·,·> F Is Frobenius inner product, | | · |. The purple F Is the Frobenius norm;
the constructed multi-core function of the data subset is
Wherein k is m (,) is the mth base kernel function, x i And x j Are two samples of the kernel function to be solved in the data subset.
4. The method according to claim 3, wherein in S2, for the data subset A, the abnormal operation state detection method is applied to the rotating machinery equipment under multiple operating conditions k The learning of the fuzzy multi-core support vector data descriptor model is completed by solving the following formula:
wherein k is m (,) is the mth base kernel function, C k Is a penalty factor that is a function of the time,
is a sample x i The weight of the blur of (a) is,
and with
Is a sample x i And x j A corresponding Lagrangian multiplier;
solving the optimal solution alpha of the problem k Then, an important parameter calculation formula of the sub model of the fuzzy multi-core support vector data description is as follows:
wherein, sample x l Corresponding to
Is a non-zero value.
5. The rotating mechanical equipment multi-condition operation state anomaly detection method according to claim 4, wherein the model test process for improving support vector data description specifically comprises the following steps:
s1, sequentially using each learned fuzzy multi-core support vector data description sub-model to detect whether test data z is in a normal operation state;
s2, when any fuzzy multi-core support vector data description sub-model detects that the test data is in a normal operation state, judging that the operation state of the test data is finally normal; and when all the fuzzy multi-core support vector data descriptor models detect that the test data are in abnormal operation states, judging that the operation states of the test data are finally abnormal.
6. The abnormal detection method for the multi-condition operation state of the rotating mechanical equipment as claimed in claim 5, wherein whether the detection test data z is in the normal operation state is obtained by the following formula:
wherein k is m (. Is) the mth base kernel function, η m Is the combination coefficient of the kernel function, and M is the number of kernel functions; if f (z) is less than or equal to R k Then in the kth submodel, the test data is detected to be in a normal operation state; if f (z) > R k Then in the kth submodel, the test data is detected as being in an abnormal operating state.
7. The method for detecting the abnormal operating state of the rotating mechanical equipment under the multiple working conditions as claimed in claim 6, wherein the optimal model selection criterion adopted in the step S1 in the model training process described by the improved support vector data is a Bayesian information criterion, and the optimal Gaussian mixture model component K for the training data X is determined by the following formula:
where N is the number of samples of training data X, P (X) i Theta) is sample X in training data X i The global probability density estimate in the gaussian mixture model, θ, is a parameter of the gaussian mixture model.
8. A rotating machinery equipment multi-operating mode running state abnormity detection system is characterized by comprising: the system comprises an equipment data management module, a data preprocessing and feature extraction module, an abnormality detection core module, a database module and a human-computer interaction module;
the equipment data management module realizes real-time acquisition of vibration signal data by using the acceleration sensor; the device data management module realizes the storage management operation of real-time data and the reading management operation of historical data;
the data preprocessing and feature extraction module divides the original vibration signal data into equilong subsequences by using a sliding window; the data preprocessing and feature extraction module extracts time domain features and frequency domain features of the subsequences; the data preprocessing and feature extraction module realizes the standardization of the extracted time domain features and frequency domain features;
the anomaly detection core module realizes model training by utilizing the time domain characteristics and the frequency domain characteristics of a standardized training set; the anomaly detection core module performs model test and detection result judgment on the time domain characteristics and the frequency domain characteristics of the standardized real-time acquired data to realize anomaly detection of the current running state of the equipment;
the database module realizes the storage of the acquired data; the database module realizes the storage of a time domain feature and a frequency domain feature extraction algorithm; the database module realizes the parameter storage of characteristic standardization after the training set data is cut off through a sliding window and extracted; the database management module realizes model storage after the anomaly detection core module finishes model training;
the man-machine interaction module visualizes the monitoring data acquired in real time; the man-machine interaction module operates and controls selection of training data, selection of a feature extraction algorithm, selection of a feature standardization parameter, a model training process and selection of a detection model; and the man-machine interaction module displays the result of the abnormity detection in real time.
9. The system for detecting the abnormal operating state of the rotating mechanical equipment under the multiple operating conditions of claim 8, wherein the feature extraction algorithm in the database module feature extraction algorithm library selectable by the feature extraction submodule in the data preprocessing and feature extraction module comprises: mean, root mean square value, peak, kurtosis, skewness, peak factor, impulse factor, form factor, and margin factor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210927715.6A CN115307943A (en) | 2022-08-03 | 2022-08-03 | Method and system for detecting abnormal operation state of rotating mechanical equipment under multiple working conditions |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210927715.6A CN115307943A (en) | 2022-08-03 | 2022-08-03 | Method and system for detecting abnormal operation state of rotating mechanical equipment under multiple working conditions |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115307943A true CN115307943A (en) | 2022-11-08 |
Family
ID=83858125
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210927715.6A Pending CN115307943A (en) | 2022-08-03 | 2022-08-03 | Method and system for detecting abnormal operation state of rotating mechanical equipment under multiple working conditions |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115307943A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117056849A (en) * | 2023-10-12 | 2023-11-14 | 智能制造龙城实验室 | Unsupervised method and system for monitoring abnormal state of complex mechanical equipment |
| CN117195138A (en) * | 2023-11-07 | 2023-12-08 | 湖南展通科技集团有限公司 | Production equipment safety production management method based on artificial intelligence and related device |
-
2022
- 2022-08-03 CN CN202210927715.6A patent/CN115307943A/en active Pending
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117056849A (en) * | 2023-10-12 | 2023-11-14 | 智能制造龙城实验室 | Unsupervised method and system for monitoring abnormal state of complex mechanical equipment |
| CN117056849B (en) * | 2023-10-12 | 2024-02-02 | 智能制造龙城实验室 | Unsupervised method and system for monitoring abnormal state of complex mechanical equipment |
| CN117195138A (en) * | 2023-11-07 | 2023-12-08 | 湖南展通科技集团有限公司 | Production equipment safety production management method based on artificial intelligence and related device |
| CN117195138B (en) * | 2023-11-07 | 2024-02-20 | 湖南展通科技集团有限公司 | Production equipment safety production management method based on artificial intelligence and related device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109492193B (en) | Abnormal network data generation and prediction method based on deep machine learning model | |
| CN112527788B (en) | Method and device for detecting and cleaning abnormal value of transformer monitoring data | |
| CN115307943A (en) | Method and system for detecting abnormal operation state of rotating mechanical equipment under multiple working conditions | |
| US7539597B2 (en) | Diagnostic systems and methods for predictive condition monitoring | |
| EP2015186B1 (en) | Diagnostic systems and methods for predictive condition monitoring | |
| CN111353482A (en) | LSTM-based fatigue factor recessive anomaly detection and fault diagnosis method | |
| AU2002246994A1 (en) | Diagnostic systems and methods for predictive condition monitoring | |
| US20220004163A1 (en) | Apparatus for predicting equipment damage | |
| Mathew et al. | Regression kernel for prognostics with support vector machines | |
| CN114841250A (en) | Industrial system production abnormity detection and diagnosis method based on multi-dimensional sensing data | |
| CN110879586A (en) | Phase modulator fault diagnosis and state monitoring method and system | |
| CN116625683A (en) | Wind turbine generator system bearing fault identification method, system and device and electronic equipment | |
| WO2020230422A1 (en) | Abnormality diagnosis device and method | |
| CN116108371B (en) | Cloud service abnormity diagnosis method and system based on cascade abnormity generation network | |
| CN111934903A (en) | Docker container fault intelligent prediction method based on time sequence evolution genes | |
| CN117034157B (en) | Hydropower equipment fault identification method and system combining multimodal operation data | |
| CN117743969A (en) | Fault prediction method, model training method, device, electronic equipment and medium | |
| CN118094436A (en) | Special fault-sensitive self-supervision equipment abnormality detection method and system | |
| CN117540160A (en) | Knowledge constraint-based power equipment fault diagnosis method and related device | |
| CN115456142A (en) | Change point detection method based on comparison learning | |
| CN117972313A (en) | Equipment fault diagnosis system based on deep neural network | |
| Chen et al. | Fault diagnosis and Prediction Algorithm of Digital Distribution Network based on Deep Learning | |
| CN114693143A (en) | Health state evaluation method, system, equipment and medium of numerical control machine tool | |
| CN117992878A (en) | Abnormality detection method and system applied to power grid dispatching technical support system | |
| CN117055432A (en) | Electromechanical control equipment fault early warning system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |