Alarm method and device based on time scale

Abstract

The invention relates to an alarm method and device based on a time scale. The method comprises the following steps: acquiring vibration data of target equipment; determining prediction sample data of the target equipment from the vibration data according to the time scale; inputting the prediction sample data into the support vector machine model, and outputting distribution characteristics corresponding to the prediction sample data; performing anomaly detection on the distribution characteristics, and determining an anomaly index of the target equipment; and when the abnormal index exceeds the alarm threshold value, executing an early warning strategy. A specific type of support vector machine is optimized by applying an increment and decrement learning strategy, the burden of data processing can be reduced, the calculation efficiency can be improved, real-time online monitoring is further supported, and the dynamic change of a monitoring parameter distribution area is accurately tracked by using a time scale; the comprehensive data acquisition is realized, and the alarm accuracy is improved.

Description

Technical Field

[0001] The present application relates to the field of security detection, and in particular to an alarm method and device based on a time scale. Background Art

[0002] With the rapid development of industrial Internet and intelligent manufacturing, the safety, stability and reliability of complex electromechanical equipment have received increasing attention. These devices usually work in complex environments and changeable working conditions, and their monitoring parameters are affected by many factors, such as equipment status, operating history and working conditions. Therefore, for the health monitoring of complex electromechanical equipment, the needs of automatic real-time data processing, adapting to changes in working conditions, and reducing false alarms and missed alarms have become particularly important.

[0003] However, when the equipment status remains stable, it is impossible to accurately study the slight changes in the monitoring parameters; once the equipment status becomes abnormal, the probability distribution area of ​​the monitoring parameters will change accordingly. Faced with such changes in the distribution area of ​​the monitoring parameters, it is often impossible to accurately predict and warn.

[0004] Therefore, there is an urgent need to develop an alarm method and device based on a time scale to solve one or more of the above-mentioned problems. Summary of the invention

[0005] In view of this, in order to solve the above technical problems or part of the technical problems, the embodiments of the present invention provide an alarm method and device based on a time scale.

[0006] In a first aspect, the present application provides an alarm method based on a time scale, the method comprising:

[0007] Get vibration data of target device;

[0008] Determining predicted sample data of the target device from the vibration data according to a time scale;

[0009] Inputting the predicted sample data into a support vector machine model, and outputting distribution features corresponding to the predicted sample data;

[0010] Performing anomaly detection on the distribution feature to determine an abnormality index of the target device;

[0011] When the abnormal indicator exceeds the alarm threshold, the early warning strategy is executed.

[0012] In a possible implementation, determining the predicted sample data of the target device from the vibration data according to a time scale includes:

[0013] Use the time scale to determine the target period corresponding to the forecast moment;

[0014] Determining a data sequence corresponding to a target period from the vibration data;

[0015] The data sequence is used as prediction sample data of the target device.

[0016] In a possible implementation, the vibration data includes vibration sub-data at different vibration moments;

[0017] The step of determining a data sequence corresponding to a target time period from the vibration data comprises:

[0018] Creating a corresponding time tag according to the vibration moment, and marking the time tag into the corresponding vibration sub-data;

[0019] Determine the start time and end time corresponding to the target time period;

[0020] By using the time tag, a vibration sub-data sequence corresponding to the start time to the end time is determined from the vibration data as a data sequence corresponding to the target time period.

[0021] In a possible implementation, the method further includes:

[0022] Obtain historical vibration data of target equipment;

[0023] Performing data preprocessing on the historical vibration data to obtain a sample support vector;

[0024] Using the sample support vector to establish a vector machine model as an initial vector machine model;

[0025] The initial vector machine model is updated to obtain a support vector machine model.

[0026] In a possible implementation manner, performing data preprocessing on the historical vibration data to obtain a sample support vector includes:

[0027] Perform vector extraction on historical vibration data to obtain an initial support vector set;

[0028] Selecting sample support vectors from the initial support vector set according to features of the initial support vectors in the initial support vector set.

[0029] In a possible implementation manner, the updating of the initial vector machine model to obtain a support vector machine model includes:

[0030] Reselecting support vectors from the initial support vector set according to the features of the initial support vectors in the initial support vector set;

[0031] The re-selected support vector is used as a new sample support vector to execute the step of establishing a vector machine model using the sample support vector to obtain a new vector machine model, and the vector machine model is added to the support vector machine model set;

[0032] A support vector machine model is selected from the support vector machine model set.

[0033] In a possible implementation, it further includes:

[0034] Determining whether there are unselected support vectors in the initial support vector set;

[0035] In the case that there are unselected support vectors in the initial support vector set, the step of reselecting sample support vectors is continued.

[0036] In a possible implementation, before reselecting support vectors from the initial support vector set according to the features of the initial support vectors in the initial support vector set, the method further includes:

[0037] Inputting the historical vibration data into a vector machine model, and outputting distribution features corresponding to the historical vibration data;

[0038] Obtaining a state parameter of the target device according to the distribution characteristics;

[0039] Obtaining a parameter difference between the state parameter and the historical state parameter recorded in the state tag of the historical vibration data;

[0040] When the parameter difference does not satisfy a preset difference rule, the sample support vector is deleted from the initial support vector set.

[0041] In a second aspect, the present application provides an alarm device based on a time scale, the device comprising:

[0042] A first acquisition module, used to acquire vibration data of a target device;

[0043] A sample determination module, configured to determine predicted sample data of the target device from the vibration data according to a time scale;

[0044] An output module, used to input the predicted sample data into a support vector machine model and output distribution features corresponding to the predicted sample data;

[0045] An indicator determination module, used to perform anomaly detection on the distribution feature and determine an abnormal indicator of the target device;

[0046] The execution module is used to execute the early warning strategy when the abnormal indicator exceeds the alarm threshold.

[0047] In a possible implementation, the device further includes:

[0048] A second acquisition module is used to acquire historical vibration data of the target device;

[0049] A preprocessing module, used for performing data preprocessing on the historical vibration data to obtain a sample support vector;

[0050] A building module is used to build a vector machine model using the sample support vector as an initial vector machine model;

[0051] The updating module is used to update the initial vector machine model to obtain a support vector machine model.

[0052] The above-mentioned technical scheme provided by the embodiment of the present application has the following advantages over the prior art: the method provided by the embodiment of the present application optimizes a specific type of support vector machine by applying an incremental and decremental learning strategy, which can reduce the burden of data processing and improve computing efficiency, thereby supporting real-time online monitoring; in addition, the time scale is used to accurately track the dynamic changes of the distribution area of ​​the monitoring parameters; not only comprehensive data collection is achieved, but also the accuracy of the alarm is improved, ensuring that complex electromechanical equipment can still be monitored in real time and its status can be accurately warned under complex working environments and changeable working conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0054] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0055] One or more embodiments are exemplarily described by pictures in the corresponding drawings, and these exemplified descriptions do not constitute limitations on the embodiments. Elements with the same reference numerals in the drawings represent similar elements, and unless otherwise stated, the figures in the drawings do not constitute proportional limitations.

[0056] Figure 1 A schematic diagram of a time scale-based alarm method provided in an embodiment of the present application;

[0057] Figure 2 A schematic diagram of the process steps of the support vector machine model alarm provided in the embodiment of the present application;

[0058] Figure 3 A schematic diagram of the process steps of a time scale-based alarm method provided in an embodiment of the present application;

[0059] Figure 4 A schematic flow chart of another time scale-based alarm method provided in an embodiment of the present application;

[0060] Figure 5 A schematic diagram of the geometric description of the support vector machine provided in the embodiment of the present application;

[0061] Figure 6 A schematic diagram of the steps for establishing a support vector machine model provided in an embodiment of the present application;

[0062] Figure 7a A schematic diagram of support vector increment provided in an embodiment of the present application;

[0063] Figure 7b A schematic diagram of support vector reduction provided in an embodiment of the present application;

[0064] Figure 8 A schematic diagram of a normal and stable operation state of a device provided in an embodiment of the present application;

[0065] Fig. 9 A schematic diagram of a normal and stable operation state of a device containing sparse abnormal data provided in an embodiment of the present application;

[0066] Fig.10 A schematic diagram of a device mutation state provided in an embodiment of the present application;

[0067] Fig.11 A schematic diagram of a device slowly changing state provided in an embodiment of the present application;

[0068] Fig.12 A schematic diagram of the structure of a time scale-based alarm device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0069] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0070] The disclosure below provides many different embodiments or examples to implement different structures of the present invention. In order to simplify the disclosure of the present invention, the parts and settings of specific examples are described below. Of course, they are only examples, and the purpose is not to limit the present invention. In addition, the present invention can repeat reference numbers and / or letters in different examples. This repetition is for the purpose of simplification and clarity, and does not itself indicate the relationship between the various embodiments and / or settings discussed.

[0071] In order to solve the problem that the equipment monitoring methods in the prior art often rely on regular inspections and fixed threshold settings, there are fixed alarm threshold settings, which are difficult to adapt to changes in the operating status of different equipment, and cannot achieve real-time monitoring of equipment status and comprehensive data collection, resulting in insufficient timely and accurate fault warnings, and ultimately unable to meet the needs of equipment variability and complexity in industrial production. Although the details of the changes in the monitoring parameters cannot be accurately studied, their distribution areas must be stable in a probabilistic sense. Once the state quantity changes abnormally, the distribution area of ​​the monitoring parameters will inevitably change accordingly. In the face of changes in the distribution area of ​​the monitoring parameters, the current dynamic monitoring and other methods cannot accurately make predictions and warnings. After obtaining vibration data, this application selects predicted sample data and then inputs the sample data into a vector machine model, outputs the distribution characteristics of the sample data, thereby realizing abnormal detection of the equipment, and executing a warning strategy when the abnormal index exceeds the alarm threshold; by applying an incremental reduction learning strategy to optimize a specific type of support vector machine, the burden of data processing can be reduced and computing efficiency can be improved, thereby supporting real-time online monitoring. In addition, the time scale is used to accurately track the dynamic changes of the distribution area of ​​the monitoring parameters; not only comprehensive data collection is achieved, but also the accuracy of the alarm is improved, ensuring real-time monitoring of complex electromechanical equipment and accurate warning of its status under complex working environments and changing working conditions.

[0072] Figure 1 A flowchart of a time-scale-based alarm method provided in an embodiment of the present application is shown in FIG. Figure 1 As shown, the method specifically includes:

[0073] S101, obtaining vibration data of a target device;

[0074] In this embodiment, the vibration condition of the equipment is monitored in real time by a vibration sensor pre-installed on the target equipment; then, the vibration signal collected by the sensor is transmitted to the central processing unit through the data acquisition system, and various signal processing techniques such as Fourier transform and wavelet transform are used to convert the time domain signal into a frequency domain signal in order to better analyze the vibration characteristics of the equipment. By analyzing the vibration signal, the abnormal vibration mode in the operation of the equipment is identified, and potential faults and problems are discovered in time, so that corresponding maintenance measures can be taken to ensure the normal operation of the equipment and extend its service life.

[0075] S102, determining predicted sample data of the target device from the vibration data according to a time scale;

[0076] The time scale is a standard or unit for dividing and defining time. It is used to measure and record the time range or interval of an event or phenomenon. It plays a vital role in the field of data processing and analysis. Through accurate time scales, the operating status of electromechanical equipment can be tracked and analyzed more carefully, so as to achieve rapid response and accurate prediction of equipment status changes. In this device, the application of time scale is not limited to the data collection stage, but also runs through the entire monitoring and alarm process, ensuring the timeliness and accuracy of data processing. For example, when analyzing vibration data, the time scale can help determine the sampling frequency of the data, which in turn affects the efficiency and effectiveness of signal processing and feature extraction. The time scale is conducive to understanding and organizing data, making it more meaningful and interpretable.

[0077] In this embodiment, when extracting the prediction sample data of the target device, the vibration data needs to be divided according to a predetermined time scale, including steps such as determining the sampling frequency of the data, selecting an appropriate time window, and defining the time series of the data. For example, the vibration data can be divided according to time scales such as hours, days, weeks, or months, so as to better observe and analyze the operating status of the device in different time periods.

[0078] The process of extracting data using time scale is also the process of screening and organizing the original data. Specifically, key information related to the target equipment is extracted from a large amount of vibration data, such as the vibration frequency, amplitude and vibration mode of the equipment; using time scale to determine sample data not only improves the efficiency and accuracy of data processing, but also provides a solid foundation for subsequent predictive analysis. By comparing and analyzing data at different time scales, we can have a deeper understanding of the operating laws and changing trends of the equipment, so as to promptly discover potential problems and take corresponding maintenance measures.

[0079] S103, inputting the predicted sample data into a support vector machine model, and outputting distribution features corresponding to the predicted sample data;

[0080] The vector machine model has powerful classification and regression capabilities, can process high-dimensional data and effectively extract features; it is widely used in the field of machine learning, especially in text classification, image recognition, and bioinformatics; the vector machine model finds the best classification hyperplane by maximizing the boundaries between different categories, thereby achieving effective data segmentation, which can not only improve classification accuracy, but also avoid the problem of overfitting to a certain extent; it has good generalization ability and can perform well on unknown data.

[0081] In this embodiment, the prediction sample data obtained according to the time scale is input into the support vector machine model, and the distribution characteristics corresponding to these prediction sample data are obtained through processing and analysis by the indicator vector machine model. Specifically, the support vector machine model will classify and extract features of the input prediction sample data according to the classification boundaries and decision functions learned during the training process, thereby revealing the distribution laws and characteristics of these data in the high-dimensional space, so that the internal structure of the prediction sample data provides an important reference for subsequent data analysis and decision-making.

[0082] S104, performing anomaly detection on the distribution feature to determine an abnormality index of the target device;

[0083] In this embodiment, the distribution characteristics are analyzed and evaluated in detail to detect anomalies, thereby determining whether the target device has abnormal indicators. Specifically, relevant data of the target device is first collected, and the distribution characteristics of the predicted sample data are deeply statistically analyzed. Then, advanced algorithms and models are used to systematically detect these distribution characteristics to identify any abnormal indicators that deviate from the normal range, which can effectively identify potential problems of the target device, so that corresponding measures can be taken to repair and optimize it.

[0084] S105: When the abnormal indicator exceeds the alarm threshold, execute the early warning strategy.

[0085] In this embodiment, when the abnormal index exceeds the alarm threshold, the early warning strategy is executed. For example, when the vibration data generated during the operation of the equipment exceeds the preset alarm threshold, the system will automatically trigger and execute a series of early warning strategies.

[0086] Specifically, an alarm threshold for vibration intensity is set. For example, when the vibration intensity exceeds a certain value (such as 5g), it is considered abnormal. Once it is determined through analysis that the equipment vibration data reaches or exceeds this alarm threshold, the system will immediately initiate an early warning strategy. The specific early warning strategy can be to send an alarm message to relevant personnel, record abnormal data for subsequent analysis, or initiate the equipment's self-protection mechanism, such as automatic shutdown to avoid possible further damage. By setting the alarm threshold, possible problems in equipment operation can be discovered and dealt with in a timely manner to ensure stable operation of the equipment.

[0087] The time scale-based alarm method provided in the embodiment of the present application first collects the vibration data of the target device, analyzes these data according to the time series, and identifies the sample data set for prediction; then, inputs these predicted sample data into the support vector machine model to extract the corresponding distribution characteristics; further, an abnormal detection algorithm is executed on these distribution characteristics to determine the abnormal indicators of the target device; once these abnormal indicators exceed the preset alarm threshold, the early warning mechanism will be immediately activated, and the burden of data processing and the calculation efficiency can be reduced by applying the incremental reduction learning strategy to optimize the support vector machine of a specific type, thereby supporting real-time online monitoring. In addition, the time scale is also used to accurately track the dynamic changes of the distribution area of ​​the monitoring parameters; not only comprehensive data collection is achieved, but also the accuracy of the alarm is improved, ensuring real-time monitoring of complex electromechanical equipment and accurate early warning of its status in the case of complex working environment and changing working conditions.

[0088] like Figures 2 to 3 As shown in FIG. 1 , the support vector machine is combined with the sliding window time scale analysis to obtain the process of the device status dynamic adaptive early warning method based on the analysis time scale as follows:

[0089] Step 1: Through the online algorithm of support vector machine, the target sample set {x k-m+1 ,…,x k-1 ,x k The distribution D of} in the feature space F k ;

[0090] Step 2: When k+1 time comes, new monitoring data x is collected k+1 , calculate I k+1 , compared with the alarm threshold η, to determine whether to alarm;

[0091] Step 3: x k+1 Add the target sample set and change x k-m+1 Delete from the target sample set and calculate the new distribution D online k+1 ;

[0092] Step 4: k←k+1, repeat steps 2 to 3.

[0093] Among them, the anomaly index is a relative index of spatial distance. Setting a threshold η for it determines the distance distribution area D of the new data. k At what relative level does an alarm need to be triggered?

[0094] Assume that the target sample set X at time k is k Satisfies the normal distribution N(μ k ,υ k ), then I k+1 ≥η is equivalent to:

[0095]

[0096] Among them, C(συ k ,ν) is a boundary threshold of the normal distribution.

[0097] The dynamic adaptive early warning method of equipment status based on the analysis time scale can better meet the dynamic conditions. Specifically, with the continuous updating of monitoring data, the distribution area is continuously estimated at the analysis scale, so it can adapt to the changes in equipment status. By monitoring the operation data of the equipment in real time, the early warning parameters are dynamically adjusted to ensure the sensitivity and accuracy of the early warning system. In addition, personalized equipment monitoring parameters are used as target samples to solve personalized alarm problems; the specific operating characteristics and historical data of each device are fully considered, so as to achieve more accurate and targeted early warning; through the above alarm method, the early warning system can detect potential faults and anomalies in time according to the actual operation status of the equipment, so as to take measures in advance to avoid the occurrence of equipment failures and improve the reliability and safety of the equipment.

[0098] When the state of the equipment remains stable, although the details of the changes in the monitoring parameters cannot be accurately studied, their distribution area must be stable in a probabilistic sense. Once the state quantity changes abnormally, the distribution area of ​​the monitoring parameters will inevitably change accordingly. This application establishes an analysis time scale on the monitoring parameters through a sliding window, performs kernel-based statistical analysis on data fragments within a time scale, and estimates their distribution area in the feature space. This process is performed continuously to obtain distribution area information that is dynamically updated in chronological order.

[0099] In an optional solution of the embodiment of the present invention, determining the predicted sample data of the target device from the vibration data according to the time scale includes:

[0100] Use the time scale to determine the target period corresponding to the forecast moment;

[0101] In this embodiment, the specific target period corresponding to the prediction moment is determined from the time axis by the time scale; specifically, the time scale determines the time period corresponding to the prediction moment on the time axis;

[0102] For example, in order to ensure that the status prediction of the target equipment can accurately reflect its actual operation, it is necessary to evaluate the performance of the equipment in the past week; at this time, the time scale is set to a time span of one week. By analyzing the data within this week, a more comprehensive equipment status evaluation can be obtained, and the corresponding target period at the prediction moment is defined as the time range of one week before the prediction moment, ensuring the accuracy and reliability of the prediction results. Through a detailed analysis of the equipment status in the previous week, the status of the equipment at the prediction moment is predicted, thereby providing strong data support for the maintenance and operation of the equipment.

[0103] Determining a data sequence corresponding to a target period from the vibration data;

[0104] In this embodiment, after determining the target time period corresponding to the prediction moment, the vibration data is first preprocessed, including filtering, denoising and other steps to ensure the accuracy and reliability of the data; then, the data is matched with the target time period using a timestamp or other time marker to extract the vibration data sequence within the time period; thereby obtaining a vibration data sequence that is completely corresponding to the target time period and representative, providing a solid foundation for further data analysis and processing.

[0105] The data sequence is used as prediction sample data of the target device.

[0106] In this embodiment, the data sequence is used as the prediction sample data of the target device so that more accurate predictions and decisions can be made in subsequent analysis and processing; the data sequence is then used as input for training and verification through a support vector machine model to generate reliable prediction results, better understand the behavior pattern and operating status of the target device, and thus optimize the performance of the device and improve its reliability.

[0107] In addition, since the support vector machine is based on the theory of structural risk minimization, it can effectively coordinate the contradiction between the complexity and generalizability of the learning machine. At the same time, the introduction of the parameter v makes a compromise between the volume of the distribution area and the number of sample points it contains. Therefore, it can not only avoid the influence of a few abnormal points on the estimation results, but also does not require special processing of abnormal data, thereby ensuring the continuity of dynamic estimation.

[0108] For example, there are sample points x1, x2, ..., x from the same distribution l ∈R n , l is the number of samples, α i ≥0 is the Lagrange coefficient, K(x i ,x j )=<φ(x i ),φ(x j )>=K i,j is the kernel function.

[0109] When the new monitoring parameter x k+1 When it arrives, determine whether it belongs to the distribution area determined by the target sample in the sliding window at time k, and quantify the expression x k+1 The distance from this distribution, the decision equation is as follows:

[0110]

[0111] According to the above decision equation, if Then x k+1 Belongs to the distribution at time k; x k+1 It is located on the distribution boundary; x k+1 is outside the distribution, The smaller it is, the farther it is from the k-time distribution.

[0112] In order to conform to common practice, The negative logarithm of is used as an outlier indicator:

[0113]

[0114] When I k+1 <0,x k+1 Belongs to the distribution at time k; I k+1 =0, it is on the distribution boundary; I k+1 >0 is outside the distribution.

[0115] In an optional solution of the embodiment of the present invention, the vibration data includes vibration sub-data at different vibration moments;

[0116] The step of determining a data sequence corresponding to a target time period from the vibration data comprises:

[0117] Creating a corresponding time tag according to the vibration moment, and marking the time tag into the corresponding vibration sub-data;

[0118] Determine the start time and end time corresponding to the target time period;

[0119] By using the time tag, a vibration sub-data sequence corresponding to the start time to the end time is determined from the vibration data as a data sequence corresponding to the target time period.

[0120] In this embodiment, before obtaining the data sequence corresponding to the target period, firstly, the corresponding time tags are created according to the time when the vibration occurs, and these time tags are accurately marked in the corresponding vibration sub-data; it is more convenient to identify and track the specific time when each vibration event occurs. Then, the start and end times of the target period are determined, and the vibration sub-data sequence occurring between the start and end times of the target period is screened out from the entire vibration data set through the created time tags, as the input data for studying and analyzing the vibration situation in the target period.

[0121] Figure 4 A flowchart of another time-scale-based alarm method provided in an embodiment of the present application is shown in FIG. Figure 4 As shown, the method specifically includes:

[0122] S401, obtaining historical vibration data of a target device;

[0123] In this embodiment, in order to accurately predict the state of the target device, the historical vibration data of the device in the past period of time is first obtained. The selected historical vibration data includes the vibration characteristics of the device under different operating conditions, so as to help obtain the operating rules of the device. By collecting these historical vibration data, the vibration conditions of the device in normal operation, abnormal conditions and when a fault occurs are recorded. Subsequently, these historical vibration data are used to establish a support vector model (SVM), which will be specifically used for the state prediction of the target device.

[0124] S402, performing data preprocessing on the historical vibration data to obtain a sample support vector;

[0125] In this embodiment, when processing the historical vibration data, a series of data preprocessing steps are first required, such as data cleaning, feature selection, etc., to ensure the quality and availability of the data. Then, the corresponding sample support vector is found through a training algorithm using the preprocessed data.

[0126] S403, using the sample support vector to establish a vector machine model as an initial vector machine model;

[0127] The specific process of using the sample support vector to establish a vector machine model and use it as the initial vector machine model

[0128] In this embodiment, the sample support vector decides the data points on the boundary; when building the model, the support vector will be used as the input of the model and trained through the vector machine algorithm to obtain a model that can accurately classify data as the initial vector machine model.

[0129] S404: Update the initial vector machine model to obtain a support vector machine model.

[0130] In this embodiment, after the initial vector machine model is established, the initial vector machine model will be verified based on the historical vibration data. Specifically, based on the gap between the model output state and the actual state, it is judged whether the support vector of the current sample needs to be updated. The purpose is to identify those sample points that have a greater impact on the model performance, that is, the support vectors, by comparing the difference between the model prediction results and the actual results; and according to the gap between the output results and the true values, the sample support vectors will be adjusted and optimized to improve the initial vector machine model, improve the accuracy of the model, and obtain a more efficient and reliable support vector machine model.

[0131] The time-scale-based alarm method provided in the embodiment of the present application uses the historical vibration data of the target device to establish a support vector machine model that is compatible with the target device, and in the process of creating the support vector machine model, the difference between the output state and the actual state of the target device recorded in the historical vibration data is used to update the support vector model, and finally an accurate and reliable support vector machine model is obtained, so that the model can predict the future state of the device based on the current vibration characteristics, thereby discovering potential faults or abnormal conditions in advance. Fault prevention of the equipment is achieved, and the reliability of the equipment is significantly improved.

[0132] like Figure 5 As shown, the support vector machine is used to find the optimal distribution area D in the feature space F, which can be geometrically described by the figure below. To simplify the problem, it is assumed that a normalized kernel function is selected to satisfy the condition K(x,x)=1, such as the radial basis kernel function. Satisfying this condition means that the norm of all kernel functions is equal to 1, and the sample is projected into a unit hypersphere F that may be infinite in dimension. ω and ρ define a hyperplane W, which is orthogonal to ω and has a distance from the center of the sphere equal to ρ / ||ω|. D is the part separated by W that is far from the center of the hypersphere. The purpose of optimization is to find the maximum value of ρ / ||ω| under given conditions, when the volume of D is the smallest.

[0133] exist Figure 5 The decision function f(x) of the non-support vector set (denoted as NSV) i )>0, it is located within the distribution area D, and the decision function f(x i )≤0, it is on or outside the boundary of D; f(x i )=0 is called the marginal support vector set, denoted as MSV, and its subscript set is S={s1,s2,…,s ls}, ls is the number of support vectors; f(x i )<0 is called the non-marginal support vector set, denoted as NMSV, and its corresponding ξ i>0, in order to ensure the compactness of the target class distribution area, the samples in this set are considered to be outliers outside the distribution area. Through such division, the target sample set consists of three subsets, X = MSV ∪ NMSV ∪ NSV. The parameter v is both the upper bound of the non-boundary support vector and the lower bound of the support vector. When training the same sample set X, adjusting this parameter can transform the points in NSV, MSV and NMSV. The smaller v is, the more points NSV contains and the fewer points NMSV contains; the larger v is, the fewer points NSV contains and the more points NMSV contains.

[0134] Figure 6 A schematic diagram of the steps for establishing a support vector machine model provided in the embodiment of the present application is as follows: Figure 6 As shown in Figure 2, the process of establishing the support vector machine model is as follows:

[0135] First, the historical data is preprocessed to extract the parameters required to initialize the model. Then, based on the status information of the equipment in the historical data, the corresponding sample labels are constructed. During the model training process, the difference between the true value and the predicted value of each sample is calculated. By analyzing these errors, the current performance of the model is evaluated and adjusted accordingly.

[0136] Specifically, if the error of a vector sample is large, a new sample vector is reselected to improve the accuracy of the model. This process is iterated continuously until all samples in the data stream are selected, and the final optimized support vector machine model is obtained.

[0137] In an optional solution of the embodiment of the present invention, performing data preprocessing on the historical vibration data to obtain a sample support vector includes:

[0138] Perform vector extraction on historical vibration data to obtain an initial support vector set;

[0139] Selecting sample support vectors from the initial support vector set according to features of the initial support vectors in the initial support vector set.

[0140] In this embodiment, by performing detailed analysis and processing on historical vibration data, key feature information of the historical vibration data is extracted, and the historical vibration data is converted into vector form according to the key feature information to obtain a set of initial support vectors; then vector selection is performed step by step in chronological order, and sample support vectors are selected from the initial support vector set.

[0141] In an optional solution of the embodiment of the present invention, the updating process of the initial vector machine model to obtain the support vector machine model includes:

[0142] Reselecting support vectors from the initial support vector set according to the features of the initial support vectors in the initial support vector set;

[0143] The re-selected support vector is used as a new sample support vector to execute the step of establishing a vector machine model using the sample support vector to obtain a new vector machine model, and the vector machine model is added to the support vector machine model set;

[0144] A support vector machine model is selected from the support vector machine model set.

[0145] In this embodiment, after selecting sample support vectors from the initial support vector set to establish an initial vector machine model, support vectors are further selected from the initial support vector set based on the feature information possessed by each initial support vector in the initial support vector set, and the new support vectors are used as sample support vectors. The step of using the sample support vectors to establish the vector machine model is performed again to update the initial vector machine model. In this process, the new vector machine models are recorded in the support vector machine model set.

[0146] Specifically, when continuing to select new support vectors from the initial support vector set, those elements that are suitable as support vectors can be re-selected from the initial set according to the chronological order; this ensures that all vectors in the set have a chance to be selected, and each vector is only selected once to avoid duplication; at the same time, this also helps to maintain the representativeness and diversity of the support vectors.

[0147] In an optional solution of the embodiment of the present invention, the method further includes:

[0148] Determining whether there are unselected support vectors in the initial support vector set;

[0149] In the case that there are unselected support vectors in the initial support vector set, the step of reselecting sample support vectors is continued.

[0150] In this embodiment, after the above-mentioned initial vector machine model is updated to add the new vector machine model to the support vector machine model set, it is also necessary to determine whether all support vectors in the initial support vector set are selected to construct the support vector machine model, to ensure that all support vectors have the opportunity to participate in the model construction process; if there are unselected support vectors, the selection process needs to continue until all support vectors are considered, and the use of a cyclic iteration method ensures that the model can make full use of the information in the initial support vector set, thereby improving the accuracy and generalization ability of the model.

[0151] The above are the implementation steps of this embodiment. Next, the specific application of the sliding window to the support vector is discussed in detail.

[0152] The core operation of the support vector machine is to solve the quadratic programming problem. Assume that the sliding window length is m. If the monitoring parameter slides once, the quadratic programming problem is solved again. The single calculation amount is o(m 3 ), which cannot meet the needs of real-time alarm. This application proposes a support vector machine algorithm improved by incremental and decremental learning methods, as follows:

[0153] In the continuous estimation process, the monitoring parameters in the current sliding window are used as target samples. The addition of new data is called the incremental process, and the deletion of old data is called the decrement process. The total number of samples in the sliding window is kept as m. When the data slides, the incremental process is processed first, and then the decrement process is processed. By adjusting the parameter α i and g i , to ensure that the KKT condition still holds in the target sample set.

[0154] Let new data x c is added to the target sample set, α c is the corresponding Lagrange coefficient. c After adding, α c The adjustment amount is Δα c , then the differential form of the equation for the KKT condition is as follows:

[0155]

[0156]

[0157] For the convenience of analysis, we first assume that non-support vectors, boundary support vectors, and non-boundary support vectors will not transform into each other. i ∈MSV,g i ≡0, that is, Δg i =0, only α i Changing, we can get the following matrix operation:

[0158]

[0159] in K M is the kernel matrix of the boundary support vector MSV. Decomposing it yields:

[0160] Δρ=-βΔα c ;

[0161] Δα j =-β j Δα c j∈S;

[0162] In the formula R=-Q M-1 ; for x i ∈NMSV and x i ∈NSV,α i Keep 0 and No change, that is, Δα i = 0, only g i Change, from the differential form formula we can get:

[0163] Δg i =γ i Δα c ;

[0164]

[0165] Thus, Δα is determined c The relationship between Δρ and Δα c and the boundary support vector set parameter α i The relationship between the changes and Δα c corresponds to non-boundary support vectors and non-support vectors g i , and describes the situation where non-support vectors, boundary support vectors, and non-boundary support vectors do not transform. c The impact of changes on other target sample parameters.

[0166] In practical situations, Δα c This may lead to data conversion between the three data sets. In order to reduce the complexity of the calculation, α c It needs to change gradually according to the process of data set transformation until the equilibrium state, which refers to x c It is determined to be one of the non-support vectors, boundary support vectors or non-boundary support vectors, and all target sample sets meet the KKT condition.

[0167] like Figure 7a As shown, assuming that there are sample points x1, x2, …, x from the same distribution l ∈R n , l is the number of samples. Let the new sample x c The incremental operation process added to the target sample set is as follows:

[0168] Step 1: Calculate g from the following c , if g c >0, indicating that x c ∈NSV (non-support vector set, denoted as NSV), the whole process ends, g c The calculation formula is as follows:

[0169]

[0170] Among them, ρ is the boundary threshold, v is the compromise parameter 0≤v≤1, K(x i ,x j )=<φ(x i ),φ(x j )>=K i,j is the kernel function, α i ≥0 is the Lagrange coefficient.

[0171] Step 2: If g c ≤0, according to the following Figure 7a , calculate Δα when all data sets change c , find where Δα c Minimal conversion operations;

[0172] Step 3: Adjust ρ and α j j∈S, adjust

[0173] That is: Δρ=-βΔα c , Δα j =-β j Δα c j∈S,Δg i =γ i Δα c ;

[0174] Step 4: Adjust each data set;

[0175] Step 5: If x c If the boundary support vector set or non-boundary support vector set is entered, stop; otherwise, jump to step 2.

[0176] When the boundary support vector set MSV changes, R also changes accordingly. In order to avoid repeated matrix inversion operations, it is processed according to the Woodbury formula. d When joining MSV:

[0177]

[0178] In the formula When the sample x in MSV o When removing from the collection:

[0179] R ij ←R ij -R oo -1 R io R oj , where i,j∈S,i,j≠o.

[0180] like Figure 7b As shown, unlike the incremental process, when the old sample x cWhen x is deleted from the target sample set, if c If it is a non-support vector, it can be deleted directly. Otherwise, it is necessary to gradually adjust the parameters of other target samples until x c Entering the non-support vector set, the corresponding reduction operation process is as follows:

[0181] Step 1: If α c =0, indicating that x c ∈NSV, the whole process ends;

[0182] Step 2: If α c >0, according to Figure 7b , calculate Δα when all data sets change c , find where Δα c Minimal conversion operations;

[0183] Step 3: Adjust ρ and α j j∈S, adjust

[0184] Step 4: Adjust each data set;

[0185] Step 5: If x c If it enters the non-support vector set, stop, otherwise jump to step 2.

[0186] In addition, in the process of mutual transformation between non-support vectors, boundary support vectors and non-boundary support vectors, it is possible that the boundary support vector set is empty. According to the differential form formula 2 It can be seen that due to Δα c It can only be equal to 0, and other sample parameters cannot be adjusted based on it, causing the algorithm to be unable to continue. However, according to the differential form formula 1 It can be seen that at this time Δg i =-Δρ, so use -Δρ instead of Δα c , we get γ i =1, according to Δg i =γ i Δα c Update the parameter g of the current target sample i , the algorithm can continue.

[0187] The core idea of ​​the online support vector machine algorithm is to ensure that the KKT condition is met on the current target sample set through local adjustment; when a new sample is introduced and an old sample is removed, the number of target samples in the sliding window remains m, so α i The upper bound of remains unchanged at 1 / vm, and the form of the dual problem remains unchanged; if the sliding step size s>1, it can be decomposed into s single-input and single-output processes for processing, and the obtained solution is still the exact quadratic programming global optimal solution.

[0188] In addition, the support vector machine online algorithm makes full use of the calculation results of the previous moment, recursively calculates the distribution area of ​​the current target sample, and reduces the amount of calculation each time the data is updated. When the added or deleted sample points are non-support vector points, the distribution of the data in the sliding window at two adjacent moments does not change, and only g c The calculation result of the calculation formula is used to judge, that is, all operations are completed. Therefore, for a data set with stable distribution, except for the first sliding window that requires more operations, the amount of calculation at subsequent moments only depends on the number of support vectors, that is, a kernel function operation is performed on each support vector.

[0189] When boundary support vectors or non-boundary support vectors appear in the new data, or the deleted sample points are boundary support vectors or non-boundary support vectors, a finite number of analytical calculations are required to ensure that the target samples in the sliding window maintain the KKT condition; since each distribution change will generate analytical calculations, under the same number of data points, the observation sequence with continuously changing distribution has the largest amount of calculation.

[0190] Exemplarily, in order to verify the computational efficiency of the online support vector machine algorithm and the influence of data distribution on computational efficiency, a two-dimensional data set is constructed. The data set consists of random points that obey two two-dimensional Gaussian functions, where μ1 = [0.2, 0.2], σ1 = [0.2, 0.2], μ2 = [-0.2, -0.2], σ2 = [0.2, 0.2], and the order of appearance of data points is divided into two types: random arrangement and arrangement by horizontal axis. The standard support vector machine is used as the comparison object, with the sliding window length m = 50, the sliding step length s = 1, ν = 0.1, and the kernel parameter σ = 0.33. The two algorithms use the same parameters to continuously slide and estimate the data distribution area. The computing computer CPU is 2.4G, and Matlab is selected as the software platform: The computing time when the data points are randomly arranged is shown in Table 1, and the computing time when the data points are arranged by horizontal axis is shown in Table 2:

[0191] Table 1

[0192]

[0193] Table 2

[0194]

[0195] Through analysis, it can be found that under two different data arrangements, the time used by the online algorithm is less than the calculation time of the standard support vector machine. This is because the online algorithm only adjusts some parameters when the sample window slides, while the standard algorithm needs to recalculate the distribution of 50 samples in the feature space. The order of data points has no effect on the standard algorithm, but has a greater impact on the online algorithm. When the data set is arranged in the order of the horizontal axis, the distribution of the target samples in the sliding window is constantly changing, and the calculation time is about twice as long as when it is randomly distributed. The time in the first two tables is averaged, and the calculation effect of the first sliding window is removed. The following conclusions are obtained: For a two-dimensional data set and under the current parameter setting conditions, the standard algorithm takes an average of 0.018 seconds for each sliding of the sample window, and the algorithm takes an average of 0.005 seconds when the samples are randomly arranged, and an average of 0.009 seconds when the samples are arranged sequentially. Since the single calculation amount of the standard algorithm is o(m 3 ), and the single calculation amount of the online algorithm has nothing to do with the sliding window length. Therefore, when the window length increases, the efficiency gap between the online algorithm and the original algorithm will further increase.

[0196] In addition, the efficiency of the algorithm is also reflected in its dynamic management of support vectors. Under the sliding window mechanism, only some data points are retained, and these data points are the points that have the greatest impact on the model in the current window. In this way, the algorithm can reduce unnecessary calculations and focus on those data points that have a significant impact on the current decision boundary. This not only improves the algorithm's computational efficiency, but also reduces the impact of outdated data on the model, thereby improving the generalization ability of the model to a certain extent.

[0197] In practical applications, the performance of the algorithm is also affected by the sliding window size and the sliding step size. The window size determines the amount of data that the algorithm can process, while the step size affects the frequency of algorithm updates. Choosing the right window size and step size is crucial to balancing computational efficiency and model accuracy. If the window is too small, it may lead to frequent model updates and increase the computational burden. If the window is too large, it may not be able to capture changes in data distribution in a timely manner, affecting the accuracy of the model.

[0198] In order to further optimize the performance of the online support vector machine algorithm, it is possible to consider introducing an adaptive mechanism to automatically adjust the size and step size of the sliding window according to the dynamic changes in the data distribution. For example, when a significant change in the data distribution is detected, the algorithm can reduce the sliding step size to increase the response speed of the model; on the contrary, if the data distribution is relatively stable, the step size can be appropriately increased to reduce the amount of calculation.

[0199] The time scale-based alarm method provided in the embodiment of the present application, combined with the online support vector machine algorithm, provides an efficient and adaptable solution, which is suitable for scenarios requiring real-time monitoring and rapid response. By dynamically adjusting model parameters and optimizing the calculation process, it can effectively process large-scale data streams and maintain the accuracy and stability of the model when the data distribution changes.

[0200] In an optional solution of the embodiment of the present invention, before reselecting support vectors from the initial support vector set according to the features of the initial support vectors in the initial support vector set, the method further includes:

[0201] Inputting the historical vibration data into a vector machine model, and outputting distribution features corresponding to the historical vibration data;

[0202] Obtaining a state parameter of the target device according to the distribution characteristics;

[0203] Obtaining a parameter difference between the state parameter and the historical state parameter recorded in the state tag of the historical vibration data;

[0204] When the parameter difference does not satisfy a preset difference rule, the sample support vector is deleted from the initial support vector set.

[0205] In this embodiment, the historical vibration data mentioned above are input into a vector machine model. After being processed by the model, the distribution characteristics corresponding to these historical vibration data can be obtained. Then, based on the distribution characteristics, the state parameters of the target device are further obtained. The parameter differences between these state parameters and the historical state parameters recorded in the historical vibration data are further obtained. If the parameter differences do not meet the preset difference rules, those data used as sample support vectors are deleted from the initial support vector set.

[0206] Based on the above content, a specific embodiment of the alarm method based on the time scale is proposed, which describes how to implement the alarm mechanism specifically through the time scale.

[0207] Specifically, in this embodiment, in order to evaluate the alarm effect of the dynamic monitoring method on the abnormal status of the equipment, the following four data distribution conditions are processed: stable distribution, sparse singular points in stability, sudden change in distribution area and slow change in distribution area to simulate four situations that occur during the operation of the equipment: normal and stable operation, a stable operation process containing sparse abnormal data, sudden state change and slow state degradation.

[0208] The main research goal of support vector machine is the distribution of high-dimensional space, so it has unique advantages in the learning problem of high-dimensional data, but for intuitive description, one-dimensional simulated data is used for analysis. The point where the decision equation is equal to zero is the boundary support vector point. The boundary support vector at the moment determines the distribution boundary of the data in the sliding window at that moment. Therefore, in the one-dimensional case, the boundary can be directly represented; select the sliding window length m = 50, the sliding step size s = 1, the parameter v = 0.1, and select the Gaussian function as the kernel function, and the kernel parameter σ = 0.3.

[0209] Figures 8 to 11 The boundary support vectors and anomaly indicators calculated at different times are shown in four cases. At time k, the data in the sliding window is {x k-m+1 ,…,x k-1 ,x k}, calculate the distribution D k , the boundary support vector NSV and the data points are plotted on the left side of the diagram, and the abnormal index I k+1 Plotted on the right side of the schematic.

[0210] like Figure 8 As shown in the figure, when the equipment runs smoothly, the state quantity remains stable and the distribution of the monitoring parameters is stable. Therefore, the boundary support vectors are roughly the same at each moment, and the abnormal indicators fluctuate only in a very small area.

[0211] In two cases, a stable distribution data set with sparse singular points may be obtained. One is that wild points appear during the monitoring process, and the other is that there is an adjustment process between the two data collection points, and the equipment status can be restored in time. When a singular point appears in the monitoring data at time k+1, due to the distribution D at time k, k does not contain such a point, so I k+1 Increase, continue to calculate D k+1 , x k+1 Enter the non-boundary support vector set and are outside the distribution area. At the next moment, if no new abnormal point appears, the point continues to stay in the non-boundary support vector set until it is deleted from the sliding window.

[0212] like Fig. 9 As shown in the figure, sparse outliers only cause changes in the anomaly index at the corresponding moment, and do not affect the recognition of other data points by the boundary support vector and the distribution area, so it has good robustness.

[0213] like Fig.10 As shown in Figure 2, when the device state changes suddenly, the distribution area of ​​the characteristic data changes accordingly. At the critical point, I k+1 Rapidly increases, and subsequent data points continue to fall on D kIn addition, the data points that enter the non-boundary support vector set gradually move to the boundary support vector set, changing the distribution boundary. As the time window slides, the distribution area stabilizes again and corresponds to the new state. Similarly, the anomaly index stabilizes again after passing through an area of ​​drastic changes; compared with the data set with sparse singular points, the out-of-limit anomaly index is in a concentrated area, and the fluctuation range of the anomaly index also changes significantly after the state changes.

[0214] like Fig.11 As shown in the figure, when the equipment status gradually deteriorates, the distribution area of ​​the characteristic data changes slowly. Although some data may exceed the distribution range at the previous moment, these deviations are usually small. In addition, the speed at which the subsequent data changes the distribution of the target sample in the sliding window is faster than the evolution speed of the equipment status itself. Therefore, the abnormal indicators remain within a limited range and do not change significantly.

[0215] The support vector machine is committed to finding the optimal balance between the compactness of data distribution and the inclusion of a sufficient number of samples, which results in some training data points being excluded from the target category; it is this balance strategy that ensures that the dynamic adaptive early warning method of equipment status based on the analysis time scale will not be disturbed by a few abnormal points while remaining sensitive to abnormal conditions.

[0216] In addition, this method can continuously adapt to changes in data and estimate and characterize changes in state quantities through distribution areas with strong generalization capabilities. The established anomaly indicators are not only of great significance in theory, but also simplify the complexity of threshold setting in practical applications and can achieve satisfactory alarm effects.

[0217] In summary, the dynamic adaptive early warning method of equipment status based on the analysis time scale can accurately predict and warn the status of complex electromechanical equipment under complex working environments and changeable working conditions.

[0218] Fig.12 A schematic diagram of a time scale-based alarm device provided in an embodiment of the present application is shown in FIG. Fig.12 As shown, the device specifically includes:

[0219] The first acquisition module 1201 is used to acquire vibration data of a target device;

[0220] A sample determination module 1202, configured to determine predicted sample data of the target device from the vibration data according to a time scale;

[0221] An output module 1203 is used to input the predicted sample data into a support vector machine model and output distribution features corresponding to the predicted sample data;

[0222] An indicator determination module 1204 is used to perform anomaly detection on the distribution feature and determine an abnormal indicator of the target device;

[0223] The execution module 1205 is used to execute the early warning strategy when the abnormal indicator exceeds the alarm threshold.

[0224] In a possible implementation, the sample determination module 1202 is further used to determine a target time period corresponding to a prediction moment using a time scale; determine a data sequence corresponding to the target time period from the vibration data; and use the data sequence as prediction sample data for the target device.

[0225] In one possible implementation, the sample determination module 1202 is further used to create a corresponding time tag according to the vibration moment, and mark the time tag into the corresponding vibration sub-data; determine the start moment and end moment corresponding to the target time period; and use the time tag to determine the vibration sub-data sequence corresponding to the start moment to the end moment from the vibration data as the data sequence corresponding to the target time period.

[0226] In a possible implementation manner, the device further includes:

[0227] The second acquisition module 1206 is used to acquire historical vibration data of the target device;

[0228] A preprocessing module 1207 is used to perform data preprocessing on the historical vibration data to obtain a sample support vector;

[0229] Establishing module 1208, used to establish a vector machine model using the sample support vector as an initial vector machine model;

[0230] The updating module 1209 is used to update the initial vector machine model to obtain a support vector machine model.

[0231] In a possible implementation manner, the preprocessing module 1207 is further configured to perform vector extraction on the historical vibration data to obtain an initial support vector set; and select sample support vectors from the initial support vector set according to features of the initial support vectors in the initial support vector set.

[0232] In one possible implementation, the preprocessing module 1207 is also used to reselect support vectors from the initial support vector set based on the characteristics of the initial support vectors in the initial support vector set; use the reselected support vectors as new sample support vectors to execute the step of establishing a vector machine model using the sample support vectors to obtain a new vector machine model, and add the vector machine model to the support vector machine model set; and select a support vector machine model from the support vector machine model set.

[0233] In one possible implementation, the device further includes a judgment module 1210 (not shown in the figure) for judging whether there are unselected support vectors in the initial support vector set; if there are unselected support vectors in the initial support vector set, the step of reselecting sample support vectors is continued.

[0234] In one possible implementation, the preprocessing module 1207 is also used to input the historical vibration data into a vector machine model, and output distribution characteristics corresponding to the historical vibration data; obtain state parameters of the target device according to the distribution characteristics; obtain parameter differences between the state parameters and the historical state parameters recorded in the state labels of the historical vibration data; and delete the sample support vector from the initial support vector set if the parameter difference does not meet a preset difference rule.

[0235] The time scale based alarm device provided in this embodiment can be as follows: Fig.12 The time scale based alarm device shown in can be implemented as follows Figure 1-11 All steps of time-scale-based alarming in Figure 1-11 The technical effect of the time scale-based alarm is shown in the figure. For details, please refer to Figure 1-11 For the sake of brevity, the relevant description is not repeated here.

[0236] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0237] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a general hardware platform, and of course, by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the relevant technology can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0238] It should be understood that the terms used in the text are only for the purpose of describing specific example embodiments, and are not intended to be limiting. Unless the context clearly indicates otherwise, the singular forms "one", "an" and "said" as used in the text may also be meant to include plural forms. The terms "include", "comprise", "contain", and "have" are inclusive, and therefore specify the existence of stated features, steps, operations, elements and / or parts, but do not exclude the existence or addition of one or more other features, steps, operations, elements, parts, and / or combinations thereof. The method steps, processes, and operations described herein are not interpreted as necessarily requiring them to be performed in the specific order described or illustrated, unless the execution order is clearly indicated. It should also be understood that additional or alternative steps may be used.

[0239] The foregoing is merely a specific embodiment of the present invention, which enables those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A time scale-based alarm method, characterized in that: include: Get vibration data of target device; Determining predicted sample data of the target device from the vibration data according to a time scale; Inputting the predicted sample data into a support vector machine model, and outputting distribution features corresponding to the predicted sample data; Performing anomaly detection on the distribution feature to determine an abnormality index of the target device; When the abnormal indicator exceeds the alarm threshold, the early warning strategy is executed.

2. The method according to claim 1, characterized in that The determining the predicted sample data of the target device from the vibration data according to the time scale includes: Use the time scale to determine the target period corresponding to the forecast moment; Determining a data sequence corresponding to a target period from the vibration data; The data sequence is used as prediction sample data of the target device.

3. The method according to claim 2, characterized in that The vibration data includes vibration sub-data at different vibration moments; The step of determining a data sequence corresponding to a target time period from the vibration data comprises: Creating a corresponding time tag according to the vibration moment, and marking the time tag into the corresponding vibration sub-data; Determine the start time and end time corresponding to the target time period; By using the time tag, a vibration sub-data sequence corresponding to the start time to the end time is determined from the vibration data as a data sequence corresponding to the target time period.

4. The method according to claim 1, characterized in that: The method further comprises: Obtain historical vibration data of target equipment; Performing data preprocessing on the historical vibration data to obtain a sample support vector; Using the sample support vector to establish a vector machine model as an initial vector machine model; The initial vector machine model is updated to obtain a support vector machine model.

5. The method according to claim 4, characterized in that The performing data preprocessing on the historical vibration data to obtain a sample support vector includes: Perform vector extraction on historical vibration data to obtain an initial support vector set; Selecting sample support vectors from the initial support vector set according to features of the initial support vectors in the initial support vector set.

6. The method according to claim 5, characterized in that The updating process of the initial vector machine model to obtain a support vector machine model includes: Reselecting support vectors from the initial support vector set according to the features of the initial support vectors in the initial support vector set; The re-selected support vector is used as a new sample support vector to execute the step of establishing a vector machine model using the sample support vector to obtain a new vector machine model, and the vector machine model is added to the support vector machine model set; A support vector machine model is selected from the support vector machine model set.

7. The method according to claim 5, characterized in that The method further comprises: Determining whether there are unselected support vectors in the initial support vector set; In the case that there are unselected support vectors in the initial support vector set, the step of reselecting sample support vectors is continued.

8. The method according to claim 6, characterized in that Before reselecting support vectors from the initial support vector set according to the features of the initial support vectors in the initial support vector set, the method further includes: Inputting the historical vibration data into a vector machine model, and outputting distribution features corresponding to the historical vibration data; Obtaining a state parameter of the target device according to the distribution characteristics; Obtaining a parameter difference between the state parameter and the historical state parameter recorded in the state tag of the historical vibration data; When the parameter difference does not satisfy a preset difference rule, the sample support vector is deleted from the initial support vector set.

9. An alarm device based on time scale, characterized in that: include: A first acquisition module, used to acquire vibration data of a target device; A sample determination module, configured to determine predicted sample data of the target device from the vibration data according to a time scale; An output module, used to input the predicted sample data into a support vector machine model and output distribution features corresponding to the predicted sample data; An indicator determination module, used to perform anomaly detection on the distribution feature and determine an abnormal indicator of the target device; The execution module is used to execute the early warning strategy when the abnormal indicator exceeds the alarm threshold.

10. The device according to claim 9, characterized in that The device also includes: A second acquisition module is used to acquire historical vibration data of the target device; A preprocessing module, used for performing data preprocessing on the historical vibration data to obtain a sample support vector; A building module is used to build a vector machine model using the sample support vector as an initial vector machine model; The updating module is used to update the initial vector machine model to obtain a support vector machine model.

Images