CN116106045A - Device abnormality detection method and apparatus, and storage medium - Google Patents
Device abnormality detection method and apparatus, and storage medium Download PDFInfo
- Publication number
- CN116106045A CN116106045A CN202211707471.7A CN202211707471A CN116106045A CN 116106045 A CN116106045 A CN 116106045A CN 202211707471 A CN202211707471 A CN 202211707471A CN 116106045 A CN116106045 A CN 116106045A
- Authority
- CN
- China
- Prior art keywords
- sample
- information
- characteristic
- characteristic data
- abnormal
- 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
- 230000005856 abnormality Effects 0.000 title claims abstract description 70
- 238000001514 detection method Methods 0.000 title claims abstract description 14
- 230000002159 abnormal effect Effects 0.000 claims abstract description 74
- 238000000034 method Methods 0.000 claims abstract description 32
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 23
- 238000012216 screening Methods 0.000 claims abstract description 10
- 238000012544 monitoring process Methods 0.000 claims description 11
- 238000010276 construction Methods 0.000 claims description 8
- 238000001704 evaporation Methods 0.000 claims description 6
- 230000008020 evaporation Effects 0.000 claims description 6
- 230000008569 process Effects 0.000 claims description 6
- 229920006395 saturated elastomer Polymers 0.000 claims description 6
- 238000004590 computer program Methods 0.000 claims description 5
- 239000005457 ice water Substances 0.000 claims description 5
- 238000004458 analytical method Methods 0.000 claims description 3
- 239000000654 additive Substances 0.000 claims 1
- 230000000996 additive effect Effects 0.000 claims 1
- 238000004519 manufacturing process Methods 0.000 description 9
- 238000012549 training Methods 0.000 description 8
- 238000010586 diagram Methods 0.000 description 5
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 238000004891 communication Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 239000003826 tablet Substances 0.000 description 2
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000008092 positive effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
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
- G01M99/005—Testing of complete machines, e.g. washing-machines or mobile phones
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
A device abnormality detection method and apparatus, and a storage medium, the method comprising: acquiring a characteristic data set in a first period of equipment, determining m interval points based on a comparison result of a first time difference between acquisition time of adjacent characteristic data information and a set acquisition period, and dividing the characteristic data set into (|m+1|) subsets based on the m interval points, wherein m is more than or equal to 0; constructing a plurality of first sample sets for each sub-set based on the traceability cycle size n, inputting the sample subset of each first sample set into a convolutional neural network, and obtaining the convolutional neural network to output the prediction information corresponding to each sample subset; screening abnormal information in the prediction information corresponding to all sample subsets; explaining the abnormal reasons of the first sample set corresponding to the abnormal information based on the saproline additivity model interpreter, and obtaining an explanation result; and determining an abnormal point of the device according to the interpretation result. The problem that the equipment abnormality cause and the abnormality point position thereof cannot be found and positioned in time can be solved.
Description
Technical Field
The present disclosure relates to the field of device monitoring technologies, and in particular, to a device anomaly detection method and apparatus, and a storage medium.
Background
The problem of equipment abnormality in production equipment (such as equipment of ice machines, air compressors and the like) in factories can occur in the production process, and the equipment abnormality can seriously influence achievement of production targets, so that loss is caused to manufacturers.
In the production process, the production equipment is manually maintained regularly by a technician, and the occurrence of equipment abnormality cannot be effectively prevented. In some application scenarios, the data of the production equipment can be monitored through the monitoring model, and the abnormality detection is performed on the production equipment according to the monitored data. However, in the application process, when the monitored data reflect the abnormality of the equipment, the cause of the abnormality of the equipment cannot be found in time, and the abnormality point position of the equipment cannot be accurately positioned, so that the equipment failure rate becomes high.
Disclosure of Invention
The embodiment of the application provides a method and a device for detecting equipment abnormality and a storage medium, which can be classified according to time difference, construct training sets (X, Y) for similar data, directly splice and combine different types, and can effectively solve the time fault problem of time sequence data.
In a first aspect, an embodiment of the present application provides a method for detecting an equipment abnormality, where the method includes: acquiring a characteristic data set in a first period of equipment, wherein the characteristic data set comprises a plurality of characteristic data information which are ordered according to acquisition time, each characteristic data information comprises the acquisition time of the current characteristic data information and characteristic values of a plurality of characteristics, and the plurality of characteristics are a plurality of monitoring information when the equipment operates; determining m interval points based on a comparison result of a first time difference between the acquisition times of adjacent characteristic data information and a set acquisition period, and dividing the characteristic data set into (|m+1|) subsets based on the m interval points, wherein m is more than or equal to 0; constructing a plurality of first sample sets for each sub-set based on a traceability period size n, wherein each first sample set comprises a sample subset, the sample subset comprises n pieces of characteristic data information with continuous acquisition time, wherein a first time difference exists between any two adjacent characteristic data information, and the continuous acquisition time refers to that a second time difference between the first time difference and the set acquisition period is not larger than a time interval threshold;
inputting the sample subsets of each first sample set into a convolutional neural network, and acquiring prediction information corresponding to each sample subset output by the convolutional neural network; screening out abnormal information in the prediction information corresponding to all the sample subsets; performing abnormality cause interpretation on the first sample set corresponding to the abnormality information based on a saprolidine additivity model interpreter, and obtaining interpretation results; and determining an abnormal point of the equipment according to the interpretation result.
Further, the determining m interval points based on the comparison result of the first time difference between the acquisition times of the adjacent feature data information and the set acquisition period, and dividing the feature data set into (|m+1|) subsets based on the m interval points includes: traversing the characteristic data set, determining whether a first time difference between the acquisition times of adjacent characteristic data information is larger than the set acquisition period, if so, determining that interval points exist between the adjacent characteristic data information, determining m interval points based on a comparison result, and dividing the characteristic data set into (|m+1|) subsets based on the m interval points; if not, determining that no interval point exists between the adjacent characteristic data information.
Further, before constructing a number of first sample sets for each sub-set based on the trace back period size n, the method further comprises: determining the trace back period size n based on the first period and the acquisition period; wherein,
n represents the trace back period size, T represents the first period, and s represents the acquisition period.
Further, each first sample set further contains focal point data, and constructing a plurality of first sample sets for each sub-set based on the trace back period size n includes: starting with the first characteristic data information in the subset according to the sequence of the characteristic data information in the subset, and sequentially constructing a plurality of rounds of first sample sets by taking the characteristic data information as a starting element of the sample subset until all the subsets are constructed to complete the first sample sets; the construction process of the first sample set of one round comprises the following steps: starting from the initial element, acquiring n pieces of characteristic data information with continuous acquisition time as sample subsets of a current first sample set, and acquiring (n+1) th piece of characteristic data information as focal data of the current first sample set; and when the (n+1) th characteristic data information is the last characteristic data information in the subset from the current initial element, completing the construction of the first sample set of the last round of the current subset.
Further, the screening out abnormal information in the prediction information corresponding to all the sample subsets includes: comparing the prediction information corresponding to each sample subset with the focus data corresponding to the corresponding sample subset, and obtaining an anomaly score; and acquiring abnormal focus data, wherein the abnormal focus data is focus data with the abnormal score exceeding a score threshold.
Further, the score threshold is 3 times the standard deviation of all the anomaly scores.
Further, the saprolipram-based additivity model interpreter interprets the abnormal reasons of the first sample set corresponding to the abnormal information, and obtains interpretation results including: calculating characteristic errors between each characteristic value in each abnormal focus data and the corresponding characteristic value in the prediction information of the corresponding sample subset; calculating the total error corresponding to each piece of abnormal focus data, wherein the total error is the sum of the characteristic errors corresponding to each characteristic value in the abnormal focus data; sorting the characteristic errors of each characteristic value in each abnormal focus data from large to small based on the error values, and obtaining target characteristic information that the sum of the characteristic errors from large to small in the sorting exceeds a duty ratio threshold value by the total error; calculating a saprolipram value corresponding to each piece of target characteristic information based on a saprolipram additivity model interpreter; and determining an abnormality factor of the abnormal focus data based on the saprolimus value corresponding to each piece of target characteristic information in each piece of abnormal focus data.
Further, the determining the abnormality factor of the abnormal focus data based on the saprolimus value corresponding to each piece of target feature information in each piece of abnormal focus data includes: calculating a weighted average value based on the saprolimus value corresponding to each piece of target characteristic information in each piece of abnormal focus data; acquiring k target weighted averages with the front weighted average value of the weighted average value in the weighted average values corresponding to all abnormal focus data; and determining an abnormality factor of the abnormal focus data based on the saprolitic value corresponding to the k target weighted average values.
Further, determining the outlier of the device according to the interpretation result includes: after determining the abnormality factor of the abnormal focus data, highlighting the abnormality factor of the abnormal focus data and the saproliferation value of the abnormality factor; and acquiring an abnormality factor with a saprolipram value being a positive value from the abnormality factors, and determining that the abnormality factor with the saprolipram value being the positive value represents a current abnormality point of the equipment.
Further, when the device is an ice maker device, the plurality of monitoring information at the time of operation of the device includes at least one or more of: the ice machine equipment is characterized by comprising an ice water inlet temperature, an oil pressure difference, an oil tank outlet temperature, a motor power and a saturated evaporation temperature when in operation.
In a second aspect, an embodiment of the present application further provides an apparatus for detecting an abnormality of a device, where the apparatus includes: the device abnormality detection method comprises a processor and a memory, wherein the memory is used for storing at least one instruction, and the instruction is loaded and executed by the processor to realize the device abnormality detection method provided by the first aspect.
In a third aspect, an embodiment of the present application further provides an apparatus for detecting an abnormality of a device, where the apparatus includes:
the device comprises an acquisition module, a storage module and a storage module, wherein the acquisition module is used for acquiring a characteristic data set in a first period of equipment, the characteristic data set comprises a plurality of characteristic data information which are ordered according to acquisition time, each characteristic data information comprises the acquisition time of the current characteristic data information and characteristic values of a plurality of characteristics, and the plurality of characteristics are a plurality of monitoring information when the equipment operates;
the analysis module is used for executing the following steps: determining m interval points based on a comparison result of a first time difference between the acquisition times of adjacent characteristic data information and a set acquisition period, and dividing the characteristic data set into (|m+1|) subsets based on the m interval points, wherein m is more than or equal to 0; constructing a plurality of first sample sets for each sub-set based on a traceability period size n, wherein each first sample set comprises a sample subset, the sample subset comprises n pieces of characteristic data information with continuous acquisition time, wherein a first time difference exists between any two adjacent characteristic data information, and the continuous acquisition time refers to that a second time difference between the first time difference and the set acquisition period is not larger than a time interval threshold; inputting the sample subsets of each first sample set into a convolutional neural network, and acquiring prediction information corresponding to each sample subset output by the convolutional neural network; screening out abnormal information in the prediction information corresponding to all the sample subsets; performing abnormality cause interpretation on the first sample set corresponding to the abnormality information based on a saprolidine additivity model interpreter, and obtaining interpretation results; and determining the abnormal point of the equipment according to the interpretation result.
In a fourth aspect, embodiments of the present application further provide a computer-readable non-volatile storage medium having stored thereon a computer program that, when executed by a processor, implements the device abnormality detection method provided in the first aspect.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief description will be given below of the drawings that are needed in the embodiments or the prior art descriptions, it being obvious that the drawings in the following description are some embodiments of the present application, and that other drawings may be obtained according to these drawings without inventive effort to a person skilled in the art.
Fig. 1 is a flow chart of a method for detecting device abnormality according to an embodiment of the present application;
FIG. 2 is a schematic view of a feature data set provided in one embodiment of the present application;
FIG. 3 is a sample subset training schematic provided in one embodiment of the present application;
FIG. 4 is a schematic diagram of an anomaly factor for an anomaly of a device provided in one embodiment of the present application;
fig. 5 is a schematic structural diagram of an apparatus for detecting an abnormality of a device according to an embodiment of the present application.
Detailed Description
For the purposes of making the objects, technical solutions and advantages of the embodiments of the present application more clear, the technical solutions of the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is apparent that the described embodiments are some embodiments of the present application, but not all embodiments. All other embodiments, which can be made by one of ordinary skill in the art based on the embodiments herein without making any inventive effort, are intended to be within the scope of the present application.
Fig. 1 is a flow chart of a method for detecting device abnormality according to an embodiment of the present application.
Referring to fig. 1, the detection method may include:
step 101: a feature data set is acquired for a first period of time of the device.
In one embodiment, the acquisition of the characteristic data set for the first period of time of the device may be a data set that acquires device operational data for one or more target devices within 30 minutes. The target device may be, for example, an ice maker and an air compressor (hereinafter, simply referred to as an air compressor).
In one embodiment, the feature data set contains a plurality of feature data information ordered by acquisition time. Fig. 2 is a schematic diagram of a feature data set according to an embodiment of the present application. Referring to fig. 2, a plurality of pieces of acquired feature data information are taken as time series samples for training: x0, x1, x2, x3, x4, x5, x6, x7, x8, x9, x10, x11, x12, x13 …. The plurality of feature data information may be acquired according to a preset acquisition period, as shown in fig. 2, where the preset acquisition period is 3min, and thus one feature data information may be acquired once every 3 min. For example, the first data acquired after the feature data information is acquired is taken as x0, the first data is acquired again after 3 minutes, the next data after x0 is acquired is taken as x1, and so on until the acquisition is completed.
In one embodiment, each feature data information includes a collection time of the current feature data information and feature values of a plurality of features, wherein the plurality of features are a plurality of monitoring information when the device is operated. By way of example, several types of monitoring information at the device runtime may include one or more of the following: ice water inlet temperature, oil pressure difference, oil tank outlet temperature, motor power and saturated evaporation temperature when the ice machine equipment operates.
Step 102: m interval points are determined based on a comparison result of a first time difference between acquisition times of adjacent feature data information and a set acquisition period, and the feature data set is divided into (|m+1|) subsets based on the m interval points.
In the process of collecting the characteristic data information, a time fault problem may occur. Referring to fig. 2, the acquisition times of two adjacent feature data information shown in the dashed boxes are "02:06" and "03:39", respectively, and the time difference between the acquisition times of the two adjacent feature data information is greater than the acquisition period (3 min), so that it can be determined that a time fault exists between the two adjacent feature data information.
In order to reduce the influence of the time faults on the data accuracy, in one embodiment, the feature data set can be divided based on the time faults, and then the divided feature data set can be correspondingly processed so as to reduce the influence of the time faults on the data accuracy.
In one embodiment, m interval points may be determined based on a comparison result of a first time difference between acquisition times of adjacent feature data information and a set acquisition period, and the feature data set may be divided into (|m+1|) subsets based on the m interval points. Specifically, the feature data set may be traversed, whether a first time difference between acquisition times of adjacent feature data information is greater than a set acquisition period is determined, if so, it is determined that there are interval points between adjacent feature data information, that is, 1 interval point is determined every time it is determined that the first time difference between acquisition times of a pair of adjacent feature data information is greater than the set acquisition period, m interval points may be determined after traversing the feature data set, and then the feature data set may be divided into (|m+1|) subsets based on the m interval points; if not, determining that no interval point exists between the adjacent characteristic data information. Illustratively, after traversing the feature data set, 1 interval point is determined, i.e., m=1, and the feature data set may be divided into (|1+1|) subsets, i.e., the feature data set may be divided into 2 subsets based on the interval point. Referring to fig. 2, the feature data set only includes 1 time slice (i.e., 1 interval point), that is, the interval point between the acquisition times "02:06" and "03:39" of two adjacent feature data information shown in the dashed line frame, so that the feature data set may be divided into a first subset and a second subset, where the first feature data information of the feature data set is used as initial data of the first subset, and the feature data information corresponding to the acquisition time "02:06" is used as cut-off data of the first subset. And taking the characteristic data information corresponding to the acquisition time '03:39' as initial data of the second subset, and taking the original cut-off data in the characteristic data set as cut-off data of the second subset. The division of the feature data set is completed in the above manner. It should be noted that if there is no temporal fault in the feature data set, that is, m=0, the feature data set is divided into (|0+1|) subsets, in other words, the feature data set does not need to be divided.
After the feature data set is divided in the above manner, the feature data set after division (including the condition of no division) can be subjected to corresponding data merging processing, so that the influence of time fault on the data accuracy is reduced. Specifically, this can be achieved by the following step 103.
Step 103: a number of first sample sets are constructed for each sub-set based on the trace back period size n.
In one embodiment, the following steps may also be performed before step 103 is performed:
determining the trace back period size n based on the first period and the acquisition period;
wherein ,
n denotes the trace back period size (lockback_s ize), T denotes the first period, and s denotes the acquisition period.
For example, the first period is 30min, the acquisition period is 3min, and the trace back period size can be calculated to be 10, i.e., n=10.
After determining the trace back period size n, step 103 is performed, i.e. a number of first sample sets are constructed for each sub-set based on the trace back period size n. In one embodiment, each first sample set further comprises focal point data, constructing a number of first sample sets for each sub-set based on the trace back period size n comprises: starting with the first characteristic data information in the subset according to the sequence of the characteristic data information in the subset, and sequentially constructing a plurality of rounds of first sample sets by taking the characteristic data information as a starting element of the sample subset until all the subsets are constructed to complete the first sample sets; the construction process of the first sample set of one round comprises the following steps: starting from the initial element, acquiring n pieces of characteristic data information with continuous acquisition time as sample subsets of a current first sample set, and acquiring (n+1) th piece of characteristic data information as focal data of the current first sample set; and when the (n+1) th characteristic data information is the last characteristic data information in the subset from the current initial element, completing the construction of the first sample set of the last round of the current subset. Wherein, the continuous acquisition time refers to that the first time difference between any two adjacent characteristic data information and the second time difference between the set acquisition period are not greater than the time interval threshold. For example, if the trace back period size n=10, taking the first characteristic data information (X0) in the first subset as the current initial element, the first 10 pieces of characteristic data information (X0, X1, X2, X3, X4, X5, X6, X7, X8, X9) are obtained as the sample subset X of the current first sample set 0 Acquiring the current (10+1) th characteristic data information, namely (x 10) as the focal number of the current first sample setAccording to Y 0 Thereby obtaining a current first sample set (X 0 ，Y 0 ). After a first sample set is acquired, a bit is followed backward according to the acquisition sequence of the characteristic data information to be used as a current initial element, if the initial element of the previous round is (X0), the initial element of the current round (i.e. the current initial element) is (X1), and then the acquired characteristic data information with continuous acquisition time of the current first 10 times is (X1, X2, X3, X4, X5, X6, X7, X8, X9 and X10), namely a sample subset X of the current first sample set 1 = (x 1, x2, x3, x4, x5, x6, x7, x8, x9, x 10), and correspondingly, the current (10+1) th feature data information is (x 11), i.e. the focal data Y of the first sample set 1 = (X11), i.e. the current first set of samples (X 1 ，Y 1 ). And so on, until the current (n+1) th characteristic data information is the last characteristic data information in the subset, completing the construction of the first sample set of the current subset until all the subset construction is completed. Wherein a first time difference exists between any two adjacent characteristic data information in 10 characteristic data information which are continuous in acquisition time in any one sample subset, and a second time difference between the first time difference and a set acquisition period is not larger than a time interval threshold. Wherein, in sample subset X 1 For example, the period of acquisition is set to 3min, the acquisition time of any two adjacent pieces of characteristic data information has a first time difference Δt1, and the second time difference Δt2 between each first time difference Δt1 and the set acquisition period (3 min) is not greater than a time interval threshold, which may be 0. It should be noted that, in step 102, the feature data set is divided into (|m+1|) subsets, each subset constructs a plurality of first sample sets, in a scene where the set collection period is 3min, a maximum value of a first time difference Δt1 existing between any two adjacent feature data information in the sample subset of each first sample set is 3min, and then a difference value (i.e., a second time difference) between the maximum value (3 min) of the first time difference Δt1 and the set collection period (3 min) is 0, and then a time interval threshold may be set to be 0.
Step 104: inputting the sample subsets of each first sample set into a convolutional neural network, and obtaining the convolutional neural network to output prediction information corresponding to each sample subset.
All the acquired first sample sets may be trained by constructing the first sample sets of all the subsets by means of step 103 described above. In one embodiment, the deep cnn (deep learning convolutional neural network) model may be used to self-encode training all the first sample sets. In the self-coding training, a sample subset of each first sample set may be input into a convolutional neural network, and prediction information corresponding to each sample subset may be output by the convolutional neural network. Fig. 3 is a sample subset training schematic provided in one embodiment of the present application. Referring to FIG. 3, a sample subset of all the first sample sets obtained may be input into a convolutional neural network, illustratively, h first sample sets are obtained by step 103, and further in self-encoding training, the sample subset [ X ] of the h first sample sets may be obtained 0 、X 1 、X 2 …X h-1 ，X h ]The convolutional neural network is input, and the prediction information [ Y 'corresponding to each sample subset can be predicted through the convolutional neural network' 0 、Y' 1 、Y' 2 …Y' h-1 ，Y' h ]。
Step 105: and screening out abnormal information in the prediction information corresponding to all the sample subsets.
In one embodiment, after predicting the prediction information corresponding to all the sample subsets through the convolutional neural network, the abnormal information in the prediction information corresponding to all the sample subsets may be further screened out. In one embodiment, the screening manner of the anomaly information may be that the prediction information corresponding to each sample subset is compared with the focus data corresponding to the corresponding sample subset, and an anomaly score is obtained, so that the anomaly focus data may be obtained, where the anomaly focus data is focus data whose anomaly score exceeds a score threshold. Wherein, for each first sample set, the first sample set comprises a sample subset and focus data, and the sample subset comprises the operation of the device for a previous period of timeThe data and the focus data are actual operation data of the equipment at the subsequent time point. The deep learning convolutional neural network can be used for predicting the prediction information of the sample subset (the operation data of the equipment in the previous period of time), namely, the prediction operation data of the equipment in the subsequent time point is calculated. Further, the focus data (actual operation data) may be compared with the prediction need information (predicted operation data) to obtain abnormal focus data. Referring to fig. 3, the anomaly information filtering may be prediction information [ Y 'corresponding to each sample subset' 0 、Y' 1 、Y' 2 …Y' h-1 ，Y' h ]And corresponding focus data [ Y ] 0 、Y 1 、Y 2 …Y h-1 ，Y h ]The comparison is performed and an anomaly score (loss) is obtained. After the anomaly score is obtained, the anomaly score may be screened, and in an embodiment, focal data with the anomaly score exceeding the score threshold may be screened, and then the screened focal data with the anomaly score exceeding the score threshold may be regarded as the anomaly focal data. The score threshold may be 3 standard deviations of all anomaly scores in one embodiment.
After determining the abnormal focal data, the abnormal focal data may be interpreted as an abnormal cause, which may be specifically implemented through step 106.
Step 106: and performing abnormality cause interpretation on the first sample set corresponding to the abnormality information based on the saprolidine additivity model interpreter, and obtaining interpretation results.
In one embodiment, the specific steps of explaining the cause of the abnormality may include:
calculating a characteristic error between each characteristic value in each abnormal focus data and a corresponding characteristic value in the prediction information of the corresponding sample subset; wherein, the calculation formula of the characteristic error between each characteristic value and the corresponding characteristic value in the prediction information of the corresponding sample subset is |Y' (i，j) -|Y (i，j) |，j＝0,1，…，h。
Calculating the total error corresponding to each piece of abnormal focus data, wherein the total error is the sum of characteristic errors corresponding to each characteristic value in the abnormal focus data;
sorting the characteristic errors of each characteristic value in each abnormal focus data from large to small based on the error values, and obtaining target characteristic information that the total error of the sum of the characteristic errors from large to small exceeds a duty ratio threshold in the sorting; in one embodiment, the duty cycle threshold may be 80%, i.e., target feature information is obtained that has a total duty cycle error in the order of more than 80% from the sum of the large to small feature errors. Where attention may be paid to the corresponding feature in which the error in each feature value is relatively large, exemplary, several features of the device when operating include: ice water inlet temperature, oil pressure difference, oil tank outlet temperature, motor power and saturated evaporation temperature when the ice machine equipment operates. In an application scene with the duty ratio threshold value of 80%, if the error value corresponding to the oil tank water outlet temperature and the saturated evaporation temperature accounts for 82% of the total error, the oil tank water outlet temperature and the saturated evaporation temperature of the equipment can be paid attention to preferentially, so that larger abnormal points in the running process of the equipment can be solved preferentially, and the basic stability of the running of the equipment is ensured.
Calculating a saprolipram value (shape value) corresponding to each piece of target feature information based on a saprolipram additivity model interpreter (shape interpreter); in one embodiment, the saproli value dimension for each target feature information is (lockback_s ize=10, features=h).
And determining an abnormality factor of the abnormal focus data based on the saprolimus value corresponding to each piece of target characteristic information in each piece of abnormal focus data. In one embodiment, determining an anomaly factor for the anomaly focus data based on the saprolitic value includes:
calculating a weighted average value based on the saprolitic value corresponding to each target feature information in each abnormal focus data;
acquiring the weighted average value of the k targets before weighted average in the weighted average values corresponding to all abnormal focus data; in one embodiment, k=50, i.e., the first 50 target weighted averages can be weighted averaged.
And determining an abnormality factor of the abnormal focus data based on the saprolitic value corresponding to the k target weighted average values.
And determining an abnormal factor of the abnormal focus data to obtain an interpretation result.
Fig. 4 is a schematic diagram of an anomaly factor of an anomaly of a device according to an embodiment of the present application.
Referring to fig. 4, ecw_tt27 represents the variable ECW before 27 minutes, and the box with diagonal lines indicates that the shape value of the feature is positive, and the result has a positive effect; the vertical bar indicates that the shape value is negative, with negative effects on the result. The positive influencing factors that can be obtained to cause possible anomalies in the results are: LCDW (ice water inlet temperature), KW (motor power), OI L_PD (oil pressure difference) and CTRL_PNT (oil tank outlet temperature) at different times in the past, and the influence factors should be immediately checked in the next step, so that equipment abnormality, production influence and enterprise loss are avoided.
Step 107: and determining the abnormal point of the equipment according to the interpretation result.
After the interpretation result is obtained, the outlier of the device may be determined according to the interpretation result, and in an embodiment, the determining the outlier may include:
after determining the abnormality factor of the abnormal focus data, highlighting the abnormality factor of the abnormal focus data and the saproliferation value of the abnormality factor;
and acquiring an abnormality factor with a saprolipram value being a positive value from the abnormality factors, and determining that the abnormality factor with the saprolipram value being a positive value represents a current abnormality point of the device.
Through the steps 101-107, the problem of time faults existing between adjacent characteristic data information is solved without adopting any data filling method, the self-coding training result of the deep CNN model is combined, the reasons of equipment abnormality are found in time based on a shape interpreter, the abnormal point positions of the equipment are accurately positioned, the abnormal reasons are found in time when the equipment is abnormal, the equipment fault occurrence rate is reduced, and the production efficiency is improved.
Fig. 5 is a schematic structural diagram of an apparatus for detecting an abnormality of a device according to an embodiment of the present application. Referring to fig. 5, the apparatus may include a processor 501 and a memory 502, where the memory 502 is configured to store at least one instruction, and the instruction when loaded and executed by the processor 501 implements the device abnormality detection method provided in any embodiment of the present application.
Another embodiment of the present application further provides an apparatus for detecting an abnormality of a device, which may include:
the device comprises an acquisition module, a storage module and a control module, wherein the acquisition module is used for acquiring a characteristic data set in a first period of equipment, the characteristic data set comprises a plurality of characteristic data information which are ordered according to acquisition time, each characteristic data information comprises the acquisition time of the current characteristic data information and characteristic values of a plurality of characteristics, and the plurality of characteristics are a plurality of monitoring information when the equipment operates;
the analysis module is used for executing the following steps:
determining m interval points based on a comparison result of a first time difference between acquisition times of adjacent characteristic data information and a set acquisition period, and dividing a characteristic data set into (|m+1|) subsets based on the m interval points, wherein m is more than or equal to 0;
constructing a plurality of first sample sets for each sub-set based on a traceability period size n, wherein each first sample set comprises a sample subset, and the sample subset comprises n pieces of characteristic data information with continuous acquisition time, wherein a first time difference exists between any two adjacent characteristic data information, and the continuous acquisition time means that a second time difference between the first time difference and a set acquisition period is not larger than a time interval threshold;
inputting a sample subset of each first sample set into a convolutional neural network, and acquiring the convolutional neural network to output prediction information corresponding to each sample subset;
screening abnormal information in the prediction information corresponding to all sample subsets;
explaining the abnormal reasons of the first sample set corresponding to the abnormal information based on the saproline additivity model interpreter, and obtaining an explanation result;
and determining the abnormal point of the equipment according to the interpretation result.
Another embodiment of the present application also provides a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the device abnormality detection method provided by any of the embodiments of the present application.
Another embodiment of the present application also provides a computer program product, including a computer program or an instruction, which when executed by a processor, implements the device abnormality detection method provided in any of the embodiments of the present application.
It should be noted that the terminals according to the embodiments of the present application may include, but are not limited to, a personal Computer (Persona l Computer, PC), a personal digital assistant (Persona l Digita l Ass i stant, PDA), a wireless handheld device, a tablet Computer (Tab let Computer), a mobile phone, an MP3 player, an MP4 player, and the like.
It may be understood that the application may be an application program (nat iveApp) installed on the terminal, or may also be a web page program (webApp) of a browser on the terminal, which is not limited in this embodiment of the present application.
It will be clear to those skilled in the art that, for convenience and brevity of description, specific working procedures of the above-described systems, apparatuses and units may refer to corresponding procedures in the foregoing method embodiments, which are not repeated herein.
In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods may be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative, e.g., the division of the elements is merely a logical function division, and there may be additional divisions when actually implemented, e.g., multiple elements or components may be combined or integrated into another system, or some features may be omitted or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be an indirect coupling or communication connection via some interfaces, devices or units, which may be in electrical, mechanical or other form.
The units described as separate units may or may not be physically separate, and units shown as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
In addition, each functional unit in each embodiment of the present application may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit. The integrated units may be implemented in hardware or in hardware plus software functional units.
The integrated units implemented in the form of software functional units described above may be stored in a computer readable storage medium. The software functional unit is stored in a storage medium, and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) or a Processor (Processor) to perform part of the steps of the methods described in the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-only Memory (ROM), a random access Memory (Random Access Memory, RAM), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
The foregoing description of the preferred embodiments of the present invention is not intended to limit the invention to the precise form disclosed, and any modifications, equivalents, improvements and alternatives falling within the spirit and principles of the present invention are intended to be included within the scope of the present invention.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present application, and not for limiting the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some or all of the technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present application.
Claims (13)
1. A method for detecting device anomalies, the method comprising:
acquiring a characteristic data set in a first period of equipment, wherein the characteristic data set comprises a plurality of characteristic data information which are ordered according to acquisition time, each characteristic data information comprises the acquisition time of the current characteristic data information and characteristic values of a plurality of characteristics, and the plurality of characteristics are a plurality of monitoring information when the equipment operates;
determining m interval points based on a comparison result of a first time difference between the acquisition times of adjacent characteristic data information and a set acquisition period, and dividing the characteristic data set into (|m+1|) subsets based on the m interval points, wherein m is more than or equal to 0;
constructing a plurality of first sample sets for each sub-set based on a traceability period size n, wherein each first sample set comprises a sample subset, the sample subset comprises n pieces of characteristic data information with continuous acquisition time, wherein the first time difference exists between any two adjacent characteristic data information, and the continuous acquisition time refers to that the second time difference between the first time difference and the set acquisition period is not larger than a time interval threshold;
inputting the sample subsets of each first sample set into a convolutional neural network, and acquiring prediction information corresponding to each sample subset output by the convolutional neural network;
screening out abnormal information in the prediction information corresponding to all the sample subsets;
performing abnormality cause interpretation on the first sample set corresponding to the abnormality information based on a saprolidine additivity model interpreter, and obtaining interpretation results; and
and determining the abnormal point of the equipment according to the interpretation result.
2. The method of claim 1, wherein the determining m interval points based on the comparison of the first time difference between the acquisition times of adjacent feature data information and a set acquisition period, and dividing the feature data set into (|m+1|) subsets based on the m interval points comprises:
traversing the characteristic data set, determining whether a first time difference between the acquisition times of adjacent characteristic data information is larger than the set acquisition period, if so, determining that interval points exist between the adjacent characteristic data information, determining m interval points based on a comparison result, and dividing the characteristic data set into (|m+1|) subsets based on the m interval points; if not, determining that no interval point exists between the adjacent characteristic data information.
3. The method of claim 1, further comprising, prior to constructing a number of first sample sets for each subset based on the trace back period size n:
determining the trace back period size n based on the first period and the acquisition period;
wherein ,
n represents the trace back period size, T represents the first period, and s represents the acquisition period.
4. A method according to claim 1 or 3, wherein each first sample set further comprises focal data, and wherein constructing a number of first sample sets for each sub-set based on the trace back period size n comprises:
starting with the first characteristic data information in the subset according to the sequence of the characteristic data information in the subset, and sequentially constructing a plurality of rounds of first sample sets by taking the characteristic data information as a starting element of the sample subset until all the subsets are constructed to complete the first sample sets;
the construction process of the first sample set of one round comprises the following steps:
starting from the initial element, acquiring n pieces of characteristic data information with continuous acquisition time as sample subsets of a current first sample set, and acquiring (n+1) th piece of characteristic data information as focal data of the current first sample set;
and when the (n+1) th characteristic data information is the last characteristic data information in the subset from the current initial element, completing the construction of the first sample set of the last round of the current subset.
5. The method of claim 1, wherein the screening out abnormal information in the prediction information corresponding to all the sample subsets comprises:
comparing the prediction information corresponding to each sample subset with the focus data corresponding to the corresponding sample subset, and obtaining an anomaly score; and
and acquiring abnormal focus data, wherein the abnormal focus data is focus data with the abnormal score exceeding a score threshold.
6. The method of claim 5, wherein the score threshold is 3 standard deviations of all of the anomaly scores.
7. The method of claim 5, wherein the performing, by the saprolily additive model interpreter, the anomaly cause interpretation on the first sample set corresponding to the anomaly information, and obtaining the interpretation result includes:
calculating characteristic errors between each characteristic value in each abnormal focus data and the corresponding characteristic value in the prediction information of the corresponding sample subset;
calculating the total error corresponding to each piece of abnormal focus data, wherein the total error is the sum of the characteristic errors corresponding to each characteristic value in the abnormal focus data;
sorting the characteristic errors of each characteristic value in each abnormal focus data from large to small based on the error values, and obtaining target characteristic information that the sum of the characteristic errors from large to small in the sorting exceeds a duty ratio threshold value by the total error;
calculating a saprolipram value corresponding to each piece of target characteristic information based on a saprolipram additivity model interpreter;
and determining an abnormality factor of the abnormal focus data based on the saprolimus value corresponding to each piece of target characteristic information in each piece of abnormal focus data.
8. The method of claim 7, wherein determining the anomaly factor for the anomaly focus data based on the saprolimus value corresponding to each of the target feature information in each of the anomaly focus data comprises:
calculating a weighted average value based on the saprolimus value corresponding to each piece of target characteristic information in each piece of abnormal focus data;
acquiring k target weighted averages with the front weighted average value of the weighted average value in the weighted average values corresponding to all abnormal focus data;
and determining an abnormality factor of the abnormal focus data based on the saprolitic value corresponding to the k target weighted average values.
9. The method according to claim 7 or 8, wherein determining an outlier of the device based on the interpretation result comprises:
after determining the abnormality factor of the abnormal focus data, highlighting the abnormality factor of the abnormal focus data and the saproliferation value of the abnormality factor;
and acquiring an abnormality factor with a saprolipram value being a positive value from the abnormality factors, and determining that the abnormality factor with the saprolipram value being the positive value represents a current abnormality point of the equipment.
10. The method of any of claims 1-8, wherein when the device is an ice maker device, the plurality of monitoring information for the device's operation includes at least one or more of: the ice machine equipment is characterized by comprising an ice water inlet temperature, an oil pressure difference, an oil tank outlet temperature, a motor power and a saturated evaporation temperature when in operation.
11. An apparatus abnormality detection device, characterized by comprising:
a processor and a memory for storing at least one instruction which when loaded and executed by the processor implements the method of any of claims 1-10.
12. An apparatus abnormality detection device, characterized by comprising:
the device comprises an acquisition module, a storage module and a storage module, wherein the acquisition module is used for acquiring a characteristic data set in a first period of equipment, the characteristic data set comprises a plurality of characteristic data information which are ordered according to acquisition time, each characteristic data information comprises the acquisition time of the current characteristic data information and characteristic values of a plurality of characteristics, and the plurality of characteristics are a plurality of monitoring information when the equipment operates;
the analysis module is used for executing the following steps:
determining m interval points based on a comparison result of a first time difference between the acquisition times of adjacent characteristic data information and a set acquisition period, and dividing the characteristic data set into (|m+1|) subsets based on the m interval points, wherein m is more than or equal to 0;
constructing a plurality of first sample sets for each sub-set based on a traceability period size n, wherein each first sample set comprises a sample subset, the sample subset comprises n pieces of characteristic data information with continuous acquisition time, wherein the first time difference exists between any two adjacent characteristic data information, and the continuous acquisition time refers to that the second time difference between the first time difference and the set acquisition period is not larger than a time interval threshold;
inputting the sample subsets of each first sample set into a convolutional neural network, and acquiring prediction information corresponding to each sample subset output by the convolutional neural network;
screening out abnormal information in the prediction information corresponding to all the sample subsets;
performing abnormality cause interpretation on the first sample set corresponding to the abnormality information based on a saprolidine additivity model interpreter, and obtaining interpretation results;
and determining the abnormal point of the equipment according to the interpretation result.
13. A computer readable non-transitory storage medium having stored thereon a computer program, which when executed by a processor, implements the method according to any of claims 1-10.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211707471.7A CN116106045A (en) | 2022-12-29 | 2022-12-29 | Device abnormality detection method and apparatus, and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211707471.7A CN116106045A (en) | 2022-12-29 | 2022-12-29 | Device abnormality detection method and apparatus, and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116106045A true CN116106045A (en) | 2023-05-12 |
Family
ID=86263129
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211707471.7A Pending CN116106045A (en) | 2022-12-29 | 2022-12-29 | Device abnormality detection method and apparatus, and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116106045A (en) |
-
2022
- 2022-12-29 CN CN202211707471.7A patent/CN116106045A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20200064822A1 (en) | Unsupervised anomaly detection, diagnosis, and correction in multivariate time series data | |
| CN111064614A (en) | Fault root cause positioning method, device, equipment and storage medium | |
| CN113098723A (en) | Fault root cause positioning method and device, storage medium and equipment | |
| CN113568774B (en) | Multi-dimensional time sequence data real-time abnormality detection method using unsupervised deep neural network | |
| CN112883929B (en) | On-line video abnormal behavior detection model training and abnormal detection method and system | |
| WO2023169274A1 (en) | Data processing method and device, and storage medium and processor | |
| CN109643087B (en) | System and method for selecting controllable parameters for operational safety of a device | |
| CN115796708B (en) | Big data intelligent quality inspection method, system and medium for engineering construction | |
| CN116739829A (en) | Big data-based power data analysis method, system and medium | |
| CN111680727A (en) | Naive Bayes-based fault diagnosis method, diagnosis device, and storage medium | |
| CN116823233B (en) | User data processing method and system based on full-period operation and maintenance | |
| CN113123955B (en) | Plunger pump abnormity detection method and device, storage medium and electronic equipment | |
| CN116106045A (en) | Device abnormality detection method and apparatus, and storage medium | |
| CN113593074B (en) | Method and device for generating monitoring video | |
| CN116993537A (en) | Power load abnormality detection method and system based on serial GRU (generic routing unit) self-encoder | |
| CN116435995A (en) | Time series processing method, computer readable storage medium and electronic device | |
| CN116804957A (en) | System monitoring method and device | |
| CN115809160A (en) | Data processing method, electronic device, storage medium, and program product | |
| CN111563078A (en) | Data quality detection method and device based on time sequence data and storage device | |
| CN110516659A (en) | The recognition methods of ball-screw catagen phase, device, equipment and storage medium | |
| CN113051700A (en) | Equipment reliability monitoring method and device | |
| CN114257490B (en) | Wireless network performance alarm analysis method and device | |
| CN116304763B (en) | Power data pre-analysis method, system, equipment and medium | |
| CN112671791B (en) | Method, device, computer equipment and readable storage medium for detecting entity abnormity | |
| CN117669803A (en) | Equipment fault prediction method, equipment fault prediction device and nonvolatile storage medium |
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 |