Mechanical seal failure early warning method, device, electronic equipment and program product

Abstract

The present invention relates to the field of fluid sealing technology, and in particular to a mechanical seal fault warning method, device, electronic equipment and program product, comprising: calculating the historical pressure average value and the historical pressure standard deviation of the third-level sealing channel of the mechanical seal system of the kernel main pump in a time interval, and calculating the first difference between the historical pressure average value and the pressure average value after the warning time, determining the first difference corresponding to a "healthy" training set or a "dangerous" training set according to the historical pressure standard deviation and a first preset standard deviation, and constructing a first kernel density estimation model; when a new pressure value is received, calculating the current pressure average value and the current pressure standard deviation in the time interval, and calculating a first danger distance based on the current pressure average value and a preset pressure danger threshold; determining a training sample based on the current pressure standard deviation and the first preset standard deviation, substituting the first danger distance and the training sample into the first kernel density estimation model to obtain a failure probability, and performing a fault warning according to the failure probability.

Description

Technical Field

[0001] The present invention relates to the technical field of fluid sealing, and in particular to a mechanical sealing failure early warning method, device, electronic equipment and program product. Background Art

[0002] The safe operation and maintenance of nuclear power plants are of great importance. The main pump mechanical seal is one of the key components of nuclear power plants. It not only needs to ensure low leakage or no leakage operation standards, but also effectively reduce the wear between friction pairs to extend the service life of the equipment. With the development of industrial automation and intelligence, the state monitoring and fault warning technology of mechanical seals has gradually become a research hotspot. Although some monitoring technologies have been applied to the state monitoring of mechanical seals, the existing warning strategies mostly rely on thresholds set by technicians based on experience. This method has obvious limitations when processing real-time sensor monitoring data.

[0003] With the advancement of sensor technology and the enhancement of data collection capabilities, the amount of monitoring data for mechanical seals has gradually increased. Traditional early warning methods based on empirical thresholds have been unable to adapt to current needs. These methods often fail to fully utilize monitoring data and have difficulty capturing subtle changes in faults, resulting in insufficient accuracy and response speed of early warning systems, which cannot meet the modern industry's pursuit of high reliability and high efficiency. Summary of the invention

[0004] The present invention provides a mechanical seal fault early warning method, device, electronic equipment and program product to solve the problems that the traditional early warning method based on empirical threshold cannot make full use of monitoring data and has difficulty in capturing subtle changes in the occurrence of faults, resulting in insufficient accuracy and response speed of the early warning system and unable to meet the modern industry's pursuit of high reliability and high efficiency.

[0005] A first aspect of the present invention provides a mechanical seal fault warning method, comprising the following steps: collecting historical pressure values ​​of a third-level sealing channel of a nuclear main pump mechanical seal system within a preset time interval; calculating a historical pressure average value and a historical pressure standard deviation within a preset time interval based on the historical pressure values, and calculating a first difference between the historical pressure average value and the pressure average value after the warning time, and determining a "healthy" training sample or a "dangerous" training sample corresponding to the first difference based on a comparison result of the historical pressure standard deviation and the first preset standard deviation, and constructing a first kernel density estimation model based on the "healthy" training samples and the "dangerous" training samples; when a monitoring system receives a new pressure value during actual operation, calculating a current pressure average value and a current pressure standard deviation within a preset time interval, and calculating a first danger distance based on the current pressure average value and a preset pressure danger threshold; determining a target training sample corresponding to the first kernel density estimation model based on a comparison result of the current pressure standard deviation and the first preset standard deviation, substituting the first danger distance and the corresponding target training sample into the first kernel density estimation model, obtaining the failure probability of the nuclear main pump mechanical seal system, and performing a fault warning when the absolute value of the failure probability is greater than a preset abnormal threshold.

[0006] Optionally, the preset failure probability calculation formula is:

[0007] ;

[0008] in, is the density value, is the failure probability, Dangerous distance.

[0009] Optionally, the first kernel density estimation model is:

[0010]

[0011] in, is the kernel function, is the density value, is the bandwidth, is the number of training samples, is the feature vector to be predicted, is the first feature vectors.

[0012] Optionally, the above-mentioned mechanical seal failure warning method also includes: collecting historical monitoring data of the multi-sensor channels of the nuclear main pump mechanical seal system in the preset time interval; using a preset dimensionality reduction method to reduce the dimensionality of the multi-dimensional features in the historical monitoring data to obtain one-dimensional features, and calculating the historical one-dimensional feature mean and the historical one-dimensional feature standard deviation within the preset time interval based on the one-dimensional features, and calculating the second difference between the historical one-dimensional feature mean and the one-dimensional feature mean after the preset time interval, and determining the "healthy" training sample or the "dangerous" training sample corresponding to the second difference according to the comparison result of the one-dimensional feature standard deviation and the second preset standard deviation, based on the "healthy" training sample and The "dangerous" training samples are used to construct a second kernel density estimation model; when the monitoring system receives a new one-dimensional feature during actual operation, the current one-dimensional feature mean and the current one-dimensional feature standard deviation within a preset time interval are calculated, and a second dangerous distance is calculated based on the current one-dimensional feature mean and the preset mean dangerous threshold; based on the comparison result of the current one-dimensional feature standard deviation and the second preset standard deviation, the target training samples corresponding to the second kernel density estimation model are determined, the second dangerous distance and the corresponding target training samples are substituted into the second kernel density estimation model to obtain the failure probability of the nuclear main pump mechanical sealing system, and a fault warning is issued when the absolute value of the failure probability is greater than the preset abnormal threshold.

[0013] Optionally, after collecting the monitoring data of the multi-sensor channels of the nuclear main pump mechanical seal system in the preset time interval, it also includes: calculating the historical signal average value and the historical signal average standard deviation within the preset time interval based on the monitoring data of the multi-sensor channels, and calculating the third difference between the historical signal average value and the signal average value after the warning time, and determining the "healthy" training sample or "dangerous" training sample corresponding to the third difference according to the comparison result of the historical signal average standard deviation and the third preset standard deviation, and constructing a third kernel density estimation model based on the "healthy" training samples and the "dangerous" training samples; when the monitoring system receives a new signal average value during actual operation, calculate the current signal average value and the current signal average standard deviation within the preset time interval, and calculate the third danger distance based on the current pressure average value and the preset average danger threshold; determine the target training sample corresponding to the third kernel density estimation model based on the comparison result of the current signal average standard deviation and the third preset standard deviation, substitute the third danger distance and the corresponding target training sample into the third kernel density estimation model, obtain the failure probability of the nuclear main pump mechanical seal system, and perform a fault warning when the absolute value of the failure probability is greater than the preset abnormal threshold.

[0014] Optionally, the third kernel density estimation model is:

[0015] ;

[0016] in, is a symmetric window width matrix, To solve the determinant of the window width matrix H, is the kernel function, is the feature vector to be predicted, For the training set i feature vectors, for The first feature, for The first feature, for The dth feature in for The dth feature in is the bandwidth of the first dimension feature, is the bandwidth of the d-th dimension feature.

[0017] A second aspect of the present invention provides a mechanical seal failure warning device, including: a first acquisition module, used to collect historical pressure values ​​of the third-stage sealing channel of the core main pump mechanical seal system in a preset time interval; a first training module, used to calculate the historical pressure average value and the historical pressure standard deviation within the preset time interval based on the historical pressure values, and calculate the first difference between the historical pressure average value and the pressure average value after the warning time, and determine the "healthy" training sample or the "dangerous" training sample corresponding to the first difference based on the comparison result of the historical pressure standard deviation and the first preset standard deviation, and construct a first core training sample based on the "healthy" training sample and the "dangerous" training sample. Density estimation model; a first calculation module, used to calculate the current pressure average value and the current pressure standard deviation within a preset time interval when the monitoring system receives a new pressure value during actual operation, and calculate the first danger distance based on the current pressure average value and the preset pressure danger threshold; a first warning module, used to determine the target training sample corresponding to the first kernel density estimation model based on the comparison result of the current pressure standard deviation and the first preset standard deviation, substitute the first danger distance and the corresponding target training sample into the first kernel density estimation model, obtain the failure probability of the nuclear main pump mechanical sealing system, and issue a fault warning when the absolute value of the failure probability is greater than the preset abnormal threshold.

[0018] Optionally, the preset failure probability calculation formula is:

[0019] ;

[0020] in, is the density value, is the failure probability, Dangerous distance.

[0021] Optionally, the first kernel density estimation model is:

[0022]

[0023] in, is the kernel function, is the density value, is the bandwidth, is the number of training samples, is the feature vector to be predicted, is the first feature vectors.

[0024] Optionally, the above-mentioned mechanical seal failure warning device also includes: a second acquisition module, which is used to collect historical monitoring data of the multi-sensor channels of the nuclear main pump mechanical seal system in the preset time interval; a second training module, which is used to use a preset dimensionality reduction method to reduce the dimension of the multi-dimensional features in the historical monitoring data to obtain one-dimensional features, and calculate the historical one-dimensional feature mean and the historical one-dimensional feature standard deviation within the preset time interval based on the one-dimensional features, and calculate the second difference between the historical one-dimensional feature mean and the one-dimensional feature mean after the preset time interval, and determine the "healthy" training sample or the "dangerous" training sample corresponding to the second difference according to the comparison result of the one-dimensional feature standard deviation and the second preset standard deviation, based on the "healthy" training sample and The "dangerous" training samples are used to construct a second kernel density estimation model; the second calculation module is used to calculate the current one-dimensional feature mean and the current one-dimensional feature standard deviation within a preset time interval when the monitoring system receives a new one-dimensional feature during actual operation, and calculate the second danger distance based on the current one-dimensional feature mean and the preset mean danger threshold; the second warning module is used to determine the target training samples corresponding to the second kernel density estimation model based on the comparison result of the current one-dimensional feature standard deviation and the second preset standard deviation, substitute the second danger distance and the corresponding target training samples into the second kernel density estimation model, obtain the failure probability of the nuclear main pump mechanical sealing system, and issue a fault warning when the absolute value of the failure probability is greater than the preset abnormal threshold.

[0025] Optionally, after collecting the monitoring data of the multi-sensor channels of the nuclear main pump mechanical seal system in the preset time interval, the second early warning module is also used to calculate the historical signal average value and the standard deviation of the historical signal average value within the preset time interval based on the monitoring data of the multi-sensor channels, and calculate the third difference between the historical signal average value and the signal average value after the early warning time, and determine the "healthy" training sample or the "dangerous" training sample corresponding to the third difference according to the comparison result of the historical signal average value standard deviation and the third preset standard deviation, and construct a third kernel tight seal based on the "healthy" training samples and the "dangerous" training samples. degree estimation model; when the monitoring system receives a new signal average value during actual operation, the current signal average value and the standard deviation of the current signal average value within a preset time interval are calculated, and a third danger distance is calculated based on the current pressure average value and the preset average value danger threshold; based on the comparison result of the current signal average value standard deviation and the third preset standard deviation, the target training sample corresponding to the third kernel density estimation model is determined, the third danger distance and the corresponding target training sample are substituted into the third kernel density estimation model to obtain the failure probability of the nuclear main pump mechanical sealing system, and a fault warning is issued when the absolute value of the failure probability is greater than the preset abnormal threshold.

[0026] Optionally, the target kernel density estimation model is:

[0027] ;

[0028] in, is a symmetric window width matrix, To solve the determinant of the window width matrix H, is the kernel function, is the feature vector to be predicted, For the training set i feature vectors, for The first feature, for The first feature, for The dth feature in for The dth feature in is the bandwidth of the first dimension feature, is the bandwidth of the d-th dimension feature.

[0029] A third aspect of the present invention provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the mechanical seal failure warning method as described in the above embodiment.

[0030] A fourth aspect of the present invention provides a computer program product, including a computer program, which is executed by a processor to implement the mechanical seal failure early warning method as described in the above embodiment.

[0031] In the above implementation, historical pressure values ​​of the third-stage sealing channel of the nuclear main pump mechanical seal system within a preset time interval are collected, and the historical pressure average value and the historical pressure standard deviation within the preset time interval are calculated based on the historical pressure values, and the first difference between the historical pressure average value and the pressure average value after the warning time is calculated, and the "healthy" training sample or the "dangerous" training sample corresponding to the first difference is determined based on the comparison result of the historical pressure standard deviation and the first preset standard deviation, and a first kernel density estimation model is constructed based on the "healthy" training samples and the "dangerous" training samples. When the monitoring system receives a new pressure value during actual operation, the current pressure average value and the current pressure standard deviation within the preset time interval are calculated, and the first danger distance is calculated based on the current pressure average value and the preset pressure danger threshold, and the target training sample corresponding to the first kernel density estimation model is determined based on the comparison result of the current pressure standard deviation and the first preset standard deviation, and the first danger distance and the corresponding target training sample are substituted into the first kernel density estimation model to obtain the failure probability of the nuclear main pump mechanical seal system, and a fault warning is performed when the absolute value of the failure probability is greater than the preset abnormal threshold. As a result, the problems of traditional early warning methods based on empirical thresholds that cannot make full use of monitoring data and have difficulty in capturing subtle changes in the occurrence of faults, resulting in insufficient accuracy and response speed of the early warning system and inability to meet the modern industry's pursuit of high reliability and high efficiency, are solved. This method can more accurately capture changes in the operating status of mechanical seals, has higher early warning accuracy, can greatly improve the reliability of mechanical seals, reduce unexpected downtime, and reduce maintenance costs, which is of great practical significance for ensuring the continuity and stability of industrial production.

[0032] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0034] Figure 1 This is a flow chart of a mechanical seal failure early warning method provided according to an embodiment of the present invention;

[0035] Figure 2 A schematic diagram of a mechanical seal structure of a nuclear main pump according to an embodiment of the present invention;

[0036] Figure 3A flowchart of early warning of a single-channel mechanical seal failure according to an embodiment of the present invention;

[0037] Figure 4 A schematic diagram of a No. 1 pump failure warning result according to an embodiment of the present invention;

[0038] Figure 5 A schematic diagram of a No. 3 pump failure warning result according to an embodiment of the present invention;

[0039] Figure 6 is an exemplary diagram of a mechanical seal failure early warning device according to an embodiment of the present invention;

[0040] Figure 7 FIG. 4 is a schematic diagram of a structure of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0041] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and should not be construed as limiting the present invention.

[0042] The mechanical seal fault warning method, device, electronic device and program product of the embodiment of the present invention are described below with reference to the accompanying drawings. In view of the problem that the traditional warning method based on empirical threshold mentioned in the above background technology cannot make full use of monitoring data and has difficulty in capturing subtle changes in the occurrence of faults, resulting in insufficient accuracy and response speed of the warning system, and cannot meet the pursuit of high reliability and high efficiency in modern industry, the present invention provides a mechanical seal fault warning method, in which the historical pressure values ​​of the third-stage sealing channel of the nuclear main pump mechanical seal system in a preset time interval are collected, and the historical pressure average value and the historical pressure standard deviation in the preset time interval are calculated based on the historical pressure values, and the first difference between the historical pressure average value and the pressure average value after the warning time is calculated, and the first difference corresponding to the comparison result of the historical pressure standard deviation and the first preset standard deviation is determined. The "healthy" training samples or "dangerous" training samples of the monitoring system are used to construct the first kernel density estimation model based on the "healthy" training samples and the "dangerous" training samples. When the monitoring system receives a new pressure value in actual operation, the current pressure average value and the current pressure standard deviation within the preset time interval are calculated, and the first dangerous distance is calculated based on the current pressure average value and the preset pressure dangerous threshold. The target training sample corresponding to the first kernel density estimation model is determined based on the comparison result of the current pressure standard deviation and the first preset standard deviation. The first dangerous distance and the corresponding target training sample are substituted into the first kernel density estimation model to obtain the failure probability of the nuclear main pump mechanical seal system, and a fault warning is issued when the absolute value of the failure probability is greater than the preset abnormal threshold. In this way, the traditional warning method based on the empirical threshold cannot make full use of the monitoring data, and it is difficult to capture the subtle changes in the occurrence of faults, resulting in insufficient accuracy and response speed of the warning system, and cannot meet the pursuit of high reliability and high efficiency in modern industry.

[0043] In order to overcome the problem that the traditional early warning method based on empirical threshold cannot make full use of monitoring data and cannot capture subtle changes in the occurrence of faults, resulting in insufficient accuracy and response speed of the early warning system, and cannot meet the limitations of modern industry's pursuit of high reliability and high efficiency, it is necessary to develop a more advanced fault early warning method that can deeply mine and analyze monitoring data, accurately identify the operating status of mechanical seals, and issue fault warnings in a timely manner. The mechanical seal fault early warning method based on non-parametric probability density estimation came into being in this context. By applying non-parametric statistical principles, this method can effectively process data of various distribution forms without making any prior assumptions about the distribution form of the data, thereby more accurately capturing the changes in the operating status of mechanical seals and providing the possibility of achieving accurate early warning.

[0044] In addition, the proposed method also meets the urgent needs of the current industrial field for intelligent and automated systems, and is expected to be widely used in multiple industries such as petroleum, chemical, electric power, and pharmaceuticals, providing strong technical support for the health management and fault prevention of mechanical seals. Through this innovative early warning method, the reliability of mechanical seals can be greatly improved, unexpected downtime can be reduced, and maintenance costs can be reduced, which has important practical significance for ensuring the continuity and stability of industrial production.

[0045] Specifically, Figure 1 A schematic flow chart of a mechanical seal failure early warning method provided by an embodiment of the present invention.

[0046] Assume that the equipment's acquisition system collects data from multiple sensors online according to the sampling rate and sets the following signal processing rules:

[0047] (1) Set a feature extraction interval duration (Determined according to the analysis needs, typically 5 minutes. Too long will cause the method to respond slowly, too short may cause unstable feature extraction and increase the amount of calculation due to too many intervals), divide the intervals according to the length of time, preprocess the signal in each time period to extract feature vectors or determine that the interval data is invalid. The content of preprocessing and the correspondence between the signal and the feature vector may be complex, the following examples are all possible:

[0048] The signal of a single channel is averaged over the time period and used as a dimension of the feature vector;

[0049] The average value and standard deviation of a single channel signal in a time period are taken as the two dimensions of the feature vector;

[0050] Compute the relationship between multi-channel signals (sum, difference, etc.) as a dimension of the feature vector;

[0051] A channel signal is filtered and processed according to its characteristics before calculating the average value.

[0052] After implementing empirical mode decomposition on a channel signal according to its characteristics, steps such as calculating the average value are performed on each component;

[0053] For known changes caused by humans and not caused by faults, they can be marked as invalid data, and this interval will not be included in subsequent processing;

[0054] No. The signal feature vector of the valid interval Wei Jiwei , let the number of dimensions be .

[0055] (2) The system monitors and processes the real-time signals online. When the duration of an interval ends, its feature vector will be analyzed to determine each feature The distance from the threshold in a statistical sense.

[0056] The following steps further describe the specific method:

[0057] (3) The method presupposes a minimum sample size , maximum sample size and update model length cycle After collecting the initial After the feature vectors are reached, they are used to train the model. When they are added to the training set (check whether the maximum sample size is exceeded at this time), If the maximum sample size is exceeded , then discard the oldest sample to make the sample size equal to ), and update the model. , then this strategy degenerates to updating the model for each newly added signal feature vector; the update of some types of models may take a long time, in which case Set to a large number so that the model does not need to be updated frequently.

[0058] The model is trained based on a certain number of existing feature vectors, and then new feature vectors can be input into the model for recognition to obtain a return value. The return value expresses the extent to which the newly input feature vector is likely to be normal (or abnormal) based on the existing feature vectors used for training.

[0059] (4) If the model has been established, Input model to get return value .like Threshold exceeded , an alarm is issued. You can also set multiple thresholds to trigger alarms of different levels. "Exceed" can be greater than or less than (including greater than or equal to / less than or equal to), or a combination of the two.

[0060] Specifically, the mechanical seal failure early warning method includes the following steps:

[0061] like Figure 1 As shown, in step S101, historical pressure values ​​of the third-stage sealing channel of the nuclear main pump mechanical sealing system in a preset time interval are collected.

[0062] In step S102, the historical pressure average value and the historical pressure standard deviation within the preset time interval are calculated based on the historical pressure value, and the first difference between the historical pressure average value and the pressure average value after the warning time is calculated, and the "healthy" class training sample or the "dangerous" class training sample corresponding to the first difference is determined according to the comparison result of the historical pressure standard deviation and the first preset standard deviation, and a first kernel density estimation model is constructed based on the "healthy" class training samples and the "dangerous" class training samples;

[0063] In step S103, when the monitoring system receives a new pressure value during actual operation, it calculates the current pressure average value and the current pressure standard deviation within a preset time interval, and calculates the first danger distance based on the current pressure average value and the preset pressure danger threshold.

[0064] In step S104, based on the comparison result between the current pressure standard deviation and the first preset standard deviation, the target training sample corresponding to the first kernel density estimation model is determined, the first danger distance and the corresponding target training sample are substituted into the first kernel density estimation model to obtain the failure probability of the nuclear main pump mechanical sealing system, and a fault warning is issued when the absolute value of the failure probability is greater than the preset abnormal threshold.

[0065] Among them, the preset pressure danger threshold and the preset abnormal threshold can be thresholds set in advance by the user, can be thresholds obtained through a limited number of experiments, or can be thresholds obtained through a limited number of computer simulations. No specific limitation is made here. Relevant technical personnel in this field can also set the preset warning threshold to multiple thresholds to trigger alarms of different levels.

[0066] Optionally, in some embodiments, the preset failure probability calculation formula is:

[0067] ;

[0068] in, is the density value, is the failure probability, Dangerous distance.

[0069] Optionally, in some embodiments, the first kernel density estimation model is:

[0070]

[0071] in, is the kernel function, is the density value, is the bandwidth, is the number of training samples, is the feature vector to be predicted, is the first feature vectors.

[0072] For a multi-stage seal, the following table "Monitoring Data" column and Figure 2 The data shown were monitored online.

[0073] Table 1

[0074]

[0075] There are three ways to use the multi-channel signal features in the above table. The first is to use a signal feature that directly corresponds to the fault for analysis; the second is to reduce the dimension of all features to form one-dimensional data for early warning; the third is to directly use all signals and use multidimensional non-parametric probability estimation methods for early warning. These three methods are slightly different in the process of data processing and model building.

[0076] As a first embodiment, Figure 3 As shown, the embodiment of the present invention uses the pressure before the third-stage seal as a fault warning feature. Under this condition, the probability of failure is easy to calculate. Moreover, since the typical form of nuclear main pump shaft seal failure is the failure of the third-stage seal, the failure of the third-stage seal is closely related to the change of the front pressure, so using the pressure before the third-stage seal as a judgment feature has engineering significance. This embodiment is only used as a method description and is still valid for other features.

[0077] After analysis, it is found that when the third-stage mechanical seal fails, the pressure in front of the third-stage seal will fluctuate and decrease. When the seal is operating normally, the pressure in front of the third-stage seal fluctuates very little and remains stable at 5.4Mpa. Based on this phenomenon, execute steps (1)-(4).

[0078] (1) In the historical monitoring database, take the historical pressure average value and historical pressure standard deviation of the three-level sealed pressure sensor channel within the preset time interval, and calculate the first difference with the pressure average value after the warning time. For example: if the preset time interval is 5 minutes and the warning time is 7 days, then the pressure average value within every 5 minutes is subtracted from the pressure average value within 5 minutes after 7 days to obtain the first difference and add it to the database. Set the first preset standard deviation , the pressure standard deviation within the preset time interval is greater than the first preset standard deviation The corresponding first difference is classified as a "healthy" training sample, and the pressure standard deviation within the preset time interval is less than The corresponding first difference is classified as a "dangerous" training sample, and a first kernel density estimation model is constructed based on the "healthy" training samples and the "dangerous" training samples. It should be noted that there can be multiple standard deviation thresholds, and the pressure drop data is divided into multiple training data sets according to needs.

[0079] (2) For the newly collected current pressure value of the third-stage sealing channel of the nuclear main pump mechanical seal system, intercept the latest preset time interval length and extract the current pressure average value and the current pressure standard deviation , compare the current pressure standard deviation with the first preset pressure standard deviation Compare and determine the target training sample corresponding to the first kernel density estimation model, and set the preset pressure danger threshold In engineering, 4.5Mpa is usually taken, and the first danger distance is obtained by subtracting the latest average pressure from the preset pressure danger threshold. d :

[0080]

[0081] (3) Based on the target training sample and the first danger distance in step (2), the density value is obtained by substituting them into the first kernel density estimation model, and the probability index of the newly collected data in this distribution is given by using the preset failure probability calculation formula. , which is the failure probability. Specifically:

[0082] According to the judgment in (2), use the corresponding "healthy" class training samples or "dangerous" class training samples ( The density value is calculated for the pressure drop value of the mean pressure of the i-th time interval after the warning time. In the target kernel density estimation model, is the kernel function, typically, the following formula can be taken, namely the Gaussian kernel function:

[0083]

[0084] Next, calculate the failure probability based on the density value:

[0085]

[0086] (4) When the failure probability has a return value, if the absolute value of the failure probability , then it prompts an exception.

[0087] The embodiment of the present invention tests the full cycle failure data of the seals of pumps 1 and 3 of a certain unit. The test results are as follows: Figure 4 and Figure 5 As shown, it can be found that the failure probability value begins to rise sharply in the early stage of the third-level sealing pressure drop, which can provide an early warning of sealing failure.

[0088] Optionally, in some implementations, the above-mentioned mechanical seal failure warning method also includes: collecting historical monitoring data of multiple sensor channels of the nuclear main pump mechanical seal system in a preset time interval; using a preset dimensionality reduction method to reduce the dimensionality of the multi-dimensional features in the historical monitoring data to obtain one-dimensional features, and calculating the historical one-dimensional feature mean and the historical one-dimensional feature standard deviation within the preset time interval based on the one-dimensional features, and calculating the second difference between the historical one-dimensional feature mean and the one-dimensional feature mean after the preset time interval, and determining the "healthy" class training sample or the "dangerous" class training sample corresponding to the second difference based on the comparison result of the one-dimensional feature standard deviation and the second preset standard deviation, based on the "healthy" class training sample. The second kernel density estimation model is constructed based on the training samples and the "dangerous" training samples; when the monitoring system receives a new one-dimensional feature in actual operation, the current one-dimensional feature mean and the current one-dimensional feature standard deviation within the preset time interval are calculated, and the second danger distance is calculated based on the current one-dimensional feature mean and the preset mean danger threshold; based on the comparison result of the current one-dimensional feature standard deviation and the second preset standard deviation, the target training samples corresponding to the second kernel density estimation model are determined, the second danger distance and the corresponding target training samples are substituted into the second kernel density estimation model to obtain the failure probability of the nuclear main pump mechanical sealing system, and a fault warning is issued when the absolute value of the failure probability is greater than the preset abnormal threshold.

[0089] As a second embodiment, for the monitoring data in Table 1, if you want to use all of them and still use the one-dimensional signal processing method, you can follow the following steps when processing the data:

[0090] In the historical monitoring database, the average value of all sensor channels related to the fault within the preset time interval is taken, and the data is reduced in dimension using an unsupervised learning method to obtain a one-dimensional feature, and the historical one-dimensional feature mean and the historical one-dimensional feature standard deviation within the preset time interval are calculated based on the one-dimensional feature, and the second difference between the historical one-dimensional feature mean and the one-dimensional feature mean after the preset time interval is calculated. For example: if the time interval is 5 minutes, and the unsupervised learning method is principal component analysis or autoencoder, the average value feature of the multi-channel sensor is reduced in dimension to a one-dimensional feature, and the historical one-dimensional feature mean and the historical one-dimensional feature standard deviation are added to the historical monitoring database.

[0091] Set the second preset standard deviation , the historical one-dimensional feature standard deviation within the preset time interval is greater than the second preset standard deviation The corresponding second difference is classified as a "healthy" training sample, and the one-dimensional feature standard deviation within the preset time interval is less than the second preset standard deviation. The corresponding second difference is classified as a "dangerous" training sample, and a second kernel density estimation model is constructed based on the "healthy" training sample and the "dangerous" training sample. It should be noted that there can be multiple standard deviation thresholds, and the one-dimensional feature data is divided into multiple training data sets according to requirements.

[0092] Furthermore, when the monitoring system receives a new one-dimensional feature during actual operation, it calculates the current one-dimensional feature mean and the current one-dimensional feature standard deviation within a preset time interval, and calculates the second danger distance based on the current one-dimensional feature mean and the preset mean danger threshold.

[0093] Based on the comparison result of the current one-dimensional feature standard deviation and the second preset standard deviation, the target training samples corresponding to the second kernel density estimation model are determined, and the second danger distance and the corresponding target training samples are substituted into the second kernel density estimation model, where the formula of the second kernel density estimation model is the formula of the first kernel density estimation model, and the failure probability of the mechanical sealing system of the nuclear main pump is obtained, and a fault warning is issued when the absolute value of the failure probability is greater than the preset abnormal threshold.

[0094] Optionally, in some embodiments, after collecting the monitoring data of the multi-sensor channels of the nuclear main pump mechanical seal system in the preset time interval, it also includes: after collecting the monitoring data of the multi-sensor channels of the nuclear main pump mechanical seal system in the preset time interval, it also includes: calculating the historical signal average value and the standard deviation of the historical signal average value within the preset time interval based on the monitoring data of the multi-sensor channels, and calculating the third difference between the historical signal average value and the signal average value after the warning time, and determining the "healthy" training sample or the "dangerous" training sample corresponding to the third difference according to the comparison result of the historical signal average value standard deviation and the third preset standard deviation, based on the "healthy" training sample The third kernel density estimation model is constructed based on the training samples or "dangerous" training samples; when the monitoring system receives a new signal average value during actual operation, the current signal average value and the standard deviation of the current signal average value within the preset time interval are calculated, and the third danger distance is calculated based on the current pressure average value and the preset average danger threshold; based on the comparison result of the current signal average value standard deviation and the third preset standard deviation, the target training sample corresponding to the third kernel density estimation model is determined, the third danger distance and the corresponding target training sample are substituted into the third kernel density estimation model to obtain the failure probability of the nuclear main pump mechanical sealing system, and a fault warning is issued when the absolute value of the failure probability is greater than the preset abnormal threshold.

[0095] Optionally, in some embodiments, the third kernel density estimation model is:

[0096] ;

[0097] in, is a symmetric window width matrix, To solve the determinant of the window width matrix H, is the kernel function, is the feature vector to be predicted, For the training set i feature vectors, for The first feature, for The first feature, for The dth feature in for The dth feature in is the bandwidth of the first dimension feature, is the bandwidth of the d-th dimension feature.

[0098] As a third embodiment, if all the monitoring data in Table 1 are to be used, the steps correspond to the following (1) to (4):

[0099] (1) In the historical monitoring database, take the historical signal average value and the historical signal average standard deviation within the preset time interval for all sensor channels. And calculate the change value of the average value of all signals within the preset time interval after the preset warning time. For example: if the preset time interval is 5 minutes and the preset warning time is 7 days, then the average value of all signals within every 5 minutes is subtracted from the average value of all signals within 5 minutes after 7 days to obtain the third difference of the signal average value and add it to the database. Set the third preset standard deviation of each signal As long as the standard deviation of the signal average value of a certain signal within the preset time interval is greater than the third preset standard deviation , then the third difference of the average value of the corresponding signal within the preset warning time is classified as a "healthy" training sample, and the standard deviation within the preset time interval is less than the third preset standard deviation The corresponding third difference of the signal average value is classified as a "dangerous" training sample, and a third kernel density estimation model is constructed based on the "healthy" training sample and the "dangerous" training sample, including a "healthy" kernel density estimation model and a "dangerous" kernel density estimation model. It should be noted that there can be multiple standard deviation thresholds, and the change data of the signal average value is divided into multiple training data sets according to requirements.

[0100] (2) For newly collected multi-channel sensor data, intercept the latest preset time interval length and extract the current signal average value of each signal and the standard deviation of the current signal mean , compare the current signal average standard deviation with the third preset standard deviation Compare and determine the target kernel density estimation model in the corresponding third kernel density estimation model. Set the preset average value danger threshold The third danger distance is obtained by subtracting the current signal average value from the preset average danger threshold value :

[0101]

[0102] Finally all Constitute the danger distance vector D.

[0103] (3) Further, the target training sample and the third danger distance corresponding to the third kernel density estimation model are used to give the probability index of the current signal average value in this distribution , which is the failure probability. Specifically:

[0104] According to step (2), the corresponding target training samples are determined, i.e., the "healthy" training samples or the "dangerous" training samples. ( For the i The change in the signal mean value of a time interval after the preset warning time). There are d-dimensional features in total. The density value is calculated and the failure probability is calculated based on the density value:

[0105]

[0106] (4) Set a preset abnormal threshold and judge when the failure probability has a return value If the absolute value of the failure probability is greater than the preset abnormal threshold, a fault warning is issued.

[0107] In summary, the beneficial effects of the present invention include:

[0108] First, the method framework can be universally adapted to different signal types. Online monitoring of mechanical seals, especially its application in engineering, is a recent development, so its standardization is extremely low, and there is no fixed form for the type of measured physical quantity, number of channels, and distribution of values ​​(range, distribution type). Based on the model form mentioned in the present invention, it can be applied under different sensor forms or distributions.

[0109] Second, in situations where the amount of data is large, it is possible to fully utilize the monitoring data of multiple channels or the data with more dense sampling in a single channel, which has higher accuracy than the method of setting thresholds based on a single measured value.

[0110] Third, the mechanical seal of the nuclear main pump is related to the safe operation and maintenance and economic benefits of the entire nuclear power plant. Condition monitoring and fault warning can prevent major accidents and are of great significance to social security.

[0111] According to the mechanical seal fault warning method proposed in an embodiment of the present invention, historical pressure values ​​of the third-level sealing channel of the nuclear main pump mechanical seal system in a preset time interval are collected, and the historical pressure average value and the historical pressure standard deviation in the preset time interval are calculated based on the historical pressure values, and the first difference between the historical pressure average value and the pressure average value after the warning time is calculated, and the "healthy" training sample or the "dangerous" training sample corresponding to the first difference is determined according to the comparison result of the historical pressure standard deviation and the first preset standard deviation, and a first kernel density estimation model is constructed based on the "healthy" training samples and the "dangerous" training samples. When the monitoring system receives a new pressure value in actual operation, the current pressure average value and the current pressure standard deviation in the preset time interval are calculated, and the first danger distance is calculated based on the current pressure average value and the preset pressure danger threshold, and the target training sample corresponding to the first kernel density estimation model is determined based on the comparison result of the current pressure standard deviation and the first preset standard deviation, and the first danger distance and the corresponding target training sample are substituted into the first kernel density estimation model to obtain the failure probability of the nuclear main pump mechanical seal system, and a fault warning is performed when the absolute value of the failure probability is greater than the preset abnormal threshold. As a result, the problems of traditional early warning methods based on empirical thresholds that cannot make full use of monitoring data and have difficulty in capturing subtle changes in the occurrence of faults, resulting in insufficient accuracy and response speed of the early warning system and inability to meet the modern industry's pursuit of high reliability and high efficiency, are solved. This method can more accurately capture changes in the operating status of mechanical seals, has higher early warning accuracy, can greatly improve the reliability of mechanical seals, reduce unexpected downtime, and reduce maintenance costs, which is of great practical significance for ensuring the continuity and stability of industrial production.

[0112] Next, the mechanical seal failure warning device according to the embodiment of the present invention will be described with reference to the accompanying drawings.

[0113] Figure 6 It is a block diagram of a mechanical seal failure warning device according to an embodiment of the present invention.

[0114] like Figure 6 As shown, the mechanical seal failure warning device 10 includes: a first acquisition module 100 , a first training module 200 , a first calculation module 300 and a first warning module 400 .

[0115] Among them, the first acquisition module 100 is used to collect the historical pressure values ​​of the third-stage sealing channel of the nuclear main pump mechanical seal system in a preset time interval; the first training module 200 is used to calculate the historical pressure average value and the historical pressure standard deviation within the preset time interval based on the historical pressure values, and calculate the first difference between the historical pressure average value and the pressure average value after the warning time, and determine the "healthy" training sample or the "dangerous" training sample corresponding to the first difference based on the comparison result of the historical pressure standard deviation and the first preset standard deviation, and construct a first kernel density estimation model based on the "healthy" training sample and the "dangerous" training sample; the first calculation Module 300 is used to calculate the current pressure average value and the current pressure standard deviation within a preset time interval when the monitoring system receives a new pressure value during actual operation, and calculate the first danger distance based on the current pressure average value and the preset pressure danger threshold; the first warning module 400 is used to determine the target training sample corresponding to the first kernel density estimation model based on the comparison result of the current pressure standard deviation and the first preset standard deviation, substitute the first danger distance and the corresponding target training sample into the first kernel density estimation model, obtain the failure probability of the nuclear main pump mechanical sealing system, and issue a fault warning when the absolute value of the failure probability is greater than the preset abnormal threshold.

[0116] Optionally, in some embodiments, the preset failure probability calculation formula is:

[0117] ;

[0118] in, is the density value, is the failure probability, Dangerous distance.

[0119] Optionally, in some embodiments, the first kernel density estimation model is:

[0120]

[0121] in, is the kernel function, is the density value, is the bandwidth, is the number of training samples, is the feature vector to be predicted, is the first feature vectors.

[0122] Optionally, in some embodiments, the above-mentioned mechanical seal failure warning device 10 further includes: a second acquisition module, which is used to collect historical monitoring data of multiple sensor channels of the nuclear main pump mechanical seal system in a preset time interval; a second training module, which is used to use a preset dimensionality reduction method to reduce the dimensionality of the multi-dimensional features in the historical monitoring data to obtain one-dimensional features, and calculate the historical one-dimensional feature mean and the historical one-dimensional feature standard deviation within the preset time interval based on the one-dimensional features, and calculate the second difference between the historical one-dimensional feature mean and the one-dimensional feature mean after the preset time interval, and determine the "healthy" training sample or the "dangerous" training sample corresponding to the second difference according to the comparison result of the one-dimensional feature standard deviation and the second preset standard deviation, based on the "healthy" training sample or the "dangerous" training sample. The second kernel density estimation model is constructed by using the “one-dimensional” class training samples and the “dangerous” class training samples; the second calculation module is used to calculate the current one-dimensional feature mean and the current one-dimensional feature standard deviation within a preset time interval when the monitoring system receives a new one-dimensional feature in actual operation, and calculate the second danger distance based on the current one-dimensional feature mean and the preset mean danger threshold; the second early warning module is used to determine the target training samples corresponding to the second kernel density estimation model based on the comparison result of the current one-dimensional feature standard deviation and the second preset standard deviation, substitute the second danger distance and the corresponding target training samples into the second kernel density estimation model, obtain the failure probability of the nuclear main pump mechanical sealing system, and issue a fault warning when the absolute value of the failure probability is greater than the preset abnormal threshold.

[0123] Optionally, in some embodiments, after collecting the monitoring data of multiple sensor channels of the nuclear main pump mechanical seal system in a preset time interval, the second early warning module is also used to calculate the historical signal average value and the standard deviation of the historical signal average value within the preset time interval based on the monitoring data of the multiple sensor channels, and calculate the third difference between the historical signal average value and the signal average value after the warning time, and determine the "healthy" training sample or "dangerous" training sample corresponding to the third difference based on the comparison result of the historical signal average value standard deviation and the third preset standard deviation, and construct a third kernel density estimation model based on the "healthy" training samples and the "dangerous" training samples; when the monitoring system receives a new signal average value during actual operation, calculate the current signal average value and the standard deviation of the current signal average value within the preset time interval, and calculate the third danger distance based on the current pressure average value and the preset average danger threshold; determine the target training sample corresponding to the third kernel density estimation model based on the comparison result of the current signal average value standard deviation and the third preset standard deviation, substitute the third danger distance and the corresponding target training sample into the third kernel density estimation model, obtain the failure probability of the nuclear main pump mechanical seal system, and perform a fault warning when the absolute value of the failure probability is greater than the preset abnormal threshold.

[0124] Optionally, in some embodiments, the third kernel density estimation model is:

[0125] ;

[0126] in, is a symmetric window width matrix, To solve the determinant of the window width matrix H, is the kernel function, is the feature vector to be predicted, For the training set i feature vectors, for The first feature, for The first feature, for The dth feature in for The dth feature in is the bandwidth of the first dimension feature, is the bandwidth of the d-th dimension feature.

[0127] It should be noted that the above explanation of the embodiment of the mechanical seal failure warning method is also applicable to the mechanical seal failure warning device of this embodiment, and will not be repeated here.

[0128] According to the mechanical seal fault warning device proposed in an embodiment of the present invention, historical pressure values ​​of the third-stage sealing channel of the nuclear main pump mechanical seal system in a preset time interval are collected, and the historical pressure average value and the historical pressure standard deviation in the preset time interval are calculated based on the historical pressure values, and the first difference between the historical pressure average value and the pressure average value after the warning time is calculated, and the "healthy" training sample or the "dangerous" training sample corresponding to the first difference is determined according to the comparison result of the historical pressure standard deviation and the first preset standard deviation, and a first kernel density estimation model is constructed based on the "healthy" training samples and the "dangerous" training samples. When the monitoring system receives a new pressure value in actual operation, the current pressure average value and the current pressure standard deviation in the preset time interval are calculated, and the first danger distance is calculated based on the current pressure average value and the preset pressure danger threshold, and the target training sample corresponding to the first kernel density estimation model is determined based on the comparison result of the current pressure standard deviation and the first preset standard deviation, and the first danger distance and the corresponding target training sample are substituted into the first kernel density estimation model to obtain the failure probability of the nuclear main pump mechanical seal system, and a fault warning is performed when the absolute value of the failure probability is greater than the preset abnormal threshold. As a result, the problems of traditional early warning methods based on empirical thresholds that cannot make full use of monitoring data and have difficulty in capturing subtle changes in the occurrence of faults, resulting in insufficient accuracy and response speed of the early warning system and inability to meet the modern industry's pursuit of high reliability and high efficiency, are solved. This method can more accurately capture changes in the operating status of mechanical seals, has higher early warning accuracy, can greatly improve the reliability of mechanical seals, reduce unexpected downtime, and reduce maintenance costs, which is of great practical significance for ensuring the continuity and stability of industrial production.

[0129] Figure 7 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. The electronic device may include:

[0130] A memory 701 , a processor 702 , and a computer program stored in the memory 701 and executable on the processor 702 .

[0131] When the processor 702 executes the program, the mechanical seal failure early warning method provided in the above embodiment is implemented.

[0132] Furthermore, the electronic device further comprises:

[0133] The communication interface 703 is used for communication between the memory 701 and the processor 702 .

[0134] The memory 701 is used to store computer programs that can be executed on the processor 702 .

[0135] The memory 701 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0136] If the memory 701, the processor 702 and the communication interface 703 are implemented independently, the communication interface 703, the memory 701 and the processor 702 can be connected to each other through a bus and communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 7 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0137] Optionally, in a specific implementation, if the memory 701, the processor 702 and the communication interface 703 are integrated on a chip, the memory 701, the processor 702 and the communication interface 703 can communicate with each other through an internal interface.

[0138] The processor 702 may be a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present invention.

[0139] An embodiment of the present invention further provides a computer program product, including a computer program, wherein the computer program is executed by a processor to implement the mechanical seal failure early warning method as described in the above embodiment.

[0140] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or N embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0141] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "N" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0142] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present invention includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present invention belong.

[0143] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be specifically implemented in any computer program product for use with an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purposes of this specification, a "computer program product" can be any device that can contain, store, communicate, propagate or transmit a program for use with an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples of computer program products (a non-exhaustive list) include the following: an electrical connection with one or N wirings (electronic device), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). Furthermore, the computer program product may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example by optically scanning the paper or other medium and then editing, interpreting or, if necessary, processing in another suitable manner, and then storing it in a computer memory.

[0144] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above embodiment, the N steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0145] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer program product, which, when executed, includes one or a combination of the steps of the method embodiment.

[0146] In addition, each functional unit in each embodiment of the present invention may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer program product.

[0147] The computer program product mentioned above may be a read-only memory, a disk or an optical disk, etc. Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limiting the present invention. A person of ordinary skill in the art may change, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. A mechanical seal failure early warning method, characterized in that: The following steps are involved: Collect the historical pressure values ​​of the third-stage sealing channel of the nuclear main pump mechanical sealing system in the preset time interval; Based on the historical pressure values, a historical pressure average value and a historical pressure standard deviation within a preset time interval are calculated, and a first difference between the historical pressure average value and the pressure average value after the warning time is calculated, and a "healthy" training sample or a "dangerous" training sample corresponding to the first difference is determined according to a comparison result of the historical pressure standard deviation and the first preset standard deviation, and a first kernel density estimation model is constructed based on the "healthy" training samples and the "dangerous" training samples; When the monitoring system receives a new pressure value during actual operation, it calculates the current pressure average value and the current pressure standard deviation within a preset time interval, and calculates the first danger distance based on the current pressure average value and the preset pressure danger threshold; Based on the comparison result of the current pressure standard deviation and the first preset standard deviation, the target training sample corresponding to the first kernel density estimation model is determined, and the first danger distance and the corresponding target training sample are substituted into the first kernel density estimation model to obtain the failure probability of the nuclear main pump mechanical sealing system, and a fault warning is issued when the absolute value of the failure probability is greater than the preset abnormal threshold.

2. The mechanical seal failure early warning method according to claim 1, characterized in that: The preset failure probability calculation formula is: ； in, is the density value, is the failure probability, Dangerous distance.

3. The mechanical seal failure early warning method according to claim 1, characterized in that: The first kernel density estimation model is: ； in, is the kernel function, is the density value, is the bandwidth, is the number of training samples, is the feature vector to be predicted, is the first feature vectors.

4. The mechanical seal failure early warning method according to claim 1, characterized in that: Also includes: Collect historical monitoring data of the multi-sensor channel of the nuclear main pump mechanical seal system in the preset time interval; Using a preset dimensionality reduction method, the multi-dimensional features in the historical monitoring data are reduced in dimension to obtain a one-dimensional feature, and the historical one-dimensional feature mean and the historical one-dimensional feature standard deviation within a preset time interval are calculated based on the one-dimensional feature, and a second difference between the historical one-dimensional feature mean and the one-dimensional feature mean after the preset time interval is calculated, and the "healthy" class training sample or the "dangerous" class training sample corresponding to the second difference is determined based on the comparison result of the one-dimensional feature standard deviation and the second preset standard deviation, and a second kernel density estimation model is constructed based on the "healthy" class training samples and the "dangerous" class training samples; When the monitoring system receives a new one-dimensional feature during actual operation, it calculates the current one-dimensional feature mean and the current one-dimensional feature standard deviation within a preset time interval, and calculates a second danger distance based on the current one-dimensional feature mean and a preset mean danger threshold; Based on the comparison result between the current one-dimensional feature standard deviation and the second preset standard deviation, the target training sample corresponding to the second kernel density estimation model is determined, and the second danger distance and the corresponding target training sample are substituted into the second kernel density estimation model to obtain the failure probability of the nuclear main pump mechanical sealing system, and a fault warning is issued when the absolute value of the failure probability is greater than the preset abnormal threshold.

5. The mechanical seal failure early warning method according to claim 4, characterized in that: After collecting the monitoring data of the multi-sensor channel of the nuclear main pump mechanical seal system in the preset time interval, it also includes: Based on the monitoring data of the multi-sensor channel, a historical signal average value and a standard deviation of the historical signal average value within a preset time interval are calculated, and a third difference between the historical signal average value and the signal average value after the warning time is calculated, and according to a comparison result of the historical signal average value standard deviation and the third preset standard deviation, a "healthy" training sample or a "dangerous" training sample corresponding to the third difference is determined, and a third kernel density estimation model is constructed based on the "healthy" training samples and the "dangerous" training samples; When the monitoring system receives a new signal average value during actual operation, the current signal average value and the standard deviation of the current signal average value within a preset time interval are calculated, and a third danger distance is calculated based on the current pressure average value and the preset average value danger threshold value; Based on the comparison result of the current signal average standard deviation and the third preset standard deviation, the target training sample corresponding to the third kernel density estimation model is determined, and the third danger distance and the corresponding target training sample are substituted into the third kernel density estimation model to obtain the failure probability of the nuclear main pump mechanical sealing system, and a fault warning is issued when the absolute value of the failure probability is greater than the preset abnormal threshold.

6. The mechanical seal failure early warning method according to claim 5, characterized in that: The third kernel density estimation model is: ； in, is a symmetric window width matrix, To solve the determinant of the window width matrix H, is the kernel function, is the feature vector to be predicted, For the training set i feature vectors, for The first feature, for The first feature, for The dth feature in for The dth feature in is the bandwidth of the first dimension feature, is the bandwidth of the d-th dimension feature, is the number of training samples.

7. A mechanical seal failure early warning device, characterized in that: include: The first acquisition module is used to collect the historical pressure value of the third-stage sealing channel of the nuclear main pump mechanical sealing system during a preset time interval; A first training module is used to calculate a historical pressure average value and a historical pressure standard deviation within a preset time interval based on the historical pressure value, and calculate a first difference between the historical pressure average value and the pressure average value after the warning time, and determine a "healthy" training sample or a "dangerous" training sample corresponding to the first difference according to a comparison result of the historical pressure standard deviation and the first preset standard deviation, and construct a first kernel density estimation model based on the "healthy" training samples and the "dangerous" training samples; A first calculation module, configured to calculate a current pressure average value and a current pressure standard deviation within a preset time interval when the monitoring system receives a new pressure value during actual operation, and calculate a first danger distance based on the current pressure average value and a preset pressure danger threshold; The first warning module is used to determine the target training sample corresponding to the first kernel density estimation model based on the comparison result of the current pressure standard deviation and the first preset standard deviation, substitute the first danger distance and the corresponding target training sample into the first kernel density estimation model, obtain the failure probability of the nuclear main pump mechanical sealing system, and issue a fault warning when the absolute value of the failure probability is greater than a preset abnormal threshold.

8. The mechanical seal failure warning device according to claim 7, characterized in that: The preset failure probability calculation formula is: ； in, is the density value, is the failure probability, Dangerous distance.

9. An electronic device, characterized in that: It comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the mechanical seal failure early warning method as claimed in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the mechanical seal failure early warning method as described in any one of claims 1 to 6 is implemented.

Images