Bearing temperature monitoring method in starting process of large fluid machinery

Abstract

The invention provides a bearing temperature monitoring method in the starting process of a large fluid machine, and the method comprises the steps: obtaining a state vector, and constructing a time window vector through a sliding time window; by calculating the distance between the time window vectors, uniformly distributed time window vectors are selected for model construction; carrying out vector distance analysis on the state vector of the monitoring time period to obtain the output of the model; and constructing a bearing temperature early warning vector based on the reconstruction output at the continuous monitoring moment, and realizing bearing temperature monitoring. According to the invention, the temperature of the bearing in the starting process of the large-scale fluid machinery can be effectively monitored, early warning is carried out in time when a fault risk exists, and the method has high accuracy and sensitivity.

Description

TECHNICAL FIELD

[0001] The application relates to a bearing temperature monitoring method in a starting process of a large fluid machine, in particular to a bearing temperature monitoring method for a starting process of a large fluid machine based on a sliding time window operation, vector distance analysis and early warning logic construction. BACKGROUND

[0002] Large fluid machines such as pumps, compressors and fans are key equipment in the industrial field, and the bearing temperature of the large fluid machines is an important parameter reflecting the running state of the equipment. In the starting process, the bearing bears a large friction and impact load as the mechanical parts accelerate from a stationary state to a rated speed, and the temperature changes significantly. If the bearing temperature abnormally rises, it may lead to lubrication failure, material wear and even equipment damage, thereby causing shutdown accidents and affecting production safety and efficiency.

[0003] At present, the bearing temperature monitoring method mainly relies on the traditional threshold alarm method. This kind of method is usually based on a fixed threshold or simple trend analysis, and an alarm is triggered when the bearing temperature exceeds the preset threshold. However, the traditional method has the following limitations: 1) the fixed threshold method cannot effectively capture the subtle changes of the bearing temperature, which may cause early warning delay or false alarm; 2) the starting process of the large fluid machine is affected by many factors (such as load change, environmental temperature, etc.), and the traditional method is difficult to dynamically adapt to the temperature changes under different working conditions; 3) the traditional method usually only focuses on the temperature value at a single time point, and ignores the change rule of the bearing temperature in the time sequence, and cannot comprehensively reflect the equipment state.

[0004] Therefore, it is urgent to develop a new monitoring method which can effectively solve the above limitations and improve the accuracy and sensitivity of the bearing temperature monitoring in the starting process of the large fluid machine, and provide strong support for the safe operation of the industrial equipment. SUMMARY

[0005] The purpose of the application is to provide a bearing temperature monitoring method for a starting process of a large fluid machine based on a sliding time window operation, vector distance analysis and early warning logic construction, which aims to solve some limitations in the prior art and improve the accuracy and sensitivity of the bearing temperature monitoring in the starting process of the fluid machine.

[0006] In order to achieve the above purpose, the technical scheme provided by the application is as follows: A bearing temperature monitoring method for a starting process of a large fluid machine, the temperature monitoring method is based on a sliding time window operation, vector distance analysis and early warning logic construction, and comprises the following steps: S1, collecting bearing temperature data of the fluid machine in the starting process to obtain a plurality of state vectors. The state vectors under the safe working condition are taken to form a model database of the monitoring method; S2, the state vector in the model database is operated by a sliding time window to obtain a plurality of time window vectors; S3, distances between the time window vectors are calculated, uniformly distributed time window vectors are selected to form a memory matrix, and a distance matrix of the memory matrix is obtained, and the model is constructed; S4, during the monitoring period, the state vector corresponding to the bearing temperature of the fluid machine starting process is operated by a time window to obtain a monitoring time window vector as an input of the model; the distance between the monitoring time window vector and the time window vector in the memory matrix is calculated to obtain a distance vector; and the reconstruction output of the model is obtained by analyzing the vector distance; S5, after obtaining the reconstruction outputs of a plurality of continuous monitoring time window vectors, the bearing temperature of the large fluid machine starting process is monitored according to the bearing temperature early warning vector and the monitoring early warning logic.

[0007] As preferred, in step S1, the following steps are specifically included: S101, during the starting process of the fluid machine, the bearing temperature data is collected; the bearing temperature data collected each time is constructed into a state vector as follows: ， Among them, indicates the sampling time index, indicates the number of bearing temperature measuring points, indicates the state vector at the time, respectively indicates the data of the th bearing temperature measuring point at the time ; S102, the state vector under the safe starting condition of the fluid machine is included in the model database of the monitoring method, and is used for constructing the model.

[0008] As preferred, in step S2, the following steps are specifically included: S201, the state vector in the model database is operated by a time window to obtain a time window vector as follows: , Among them, indicates the length of the time window, indicates the time window vector, respectively indicates the time window data of the th bearing temperature measuring point at the time

[0009] , ， Among them, ; S202. Slide the execution of step S201 on the time axis to obtain multiple time window vectors, denoted as... subscript This indicates the number of time window vectors obtained in this step.

[0010] Preferably, step S3 specifically includes the following steps: S301. Calculate the distance between any two time window vectors, as follows: , in, , represents the time index of the time window vector. , Represents the two time window vectors involved in the distance calculation. , They represent the first Each bearing temperature measuring point is at and Time window data corresponding to a given moment The Euclidean norm of a vector; S302. By maximizing the minimum distance operation, the time window vectors obtained in step S202 are filtered, and the memory matrix is ​​constructed as follows: ， in, Represents the memory matrix, This indicates the number of time window vectors obtained after filtering. The mapping representing the filtering satisfies and set The model is ; S303. Construct the distance matrix corresponding to the memory matrix. The distance matrix is The square matrix is ​​as follows: , Among them, index .

[0011] To further explain, in step S302, let The set of time window vectors is represented as We need to select from them. A subset is formed by time window vectors. The method for maximizing the minimum distance requires: , The operation is to maximize the minimum distance, ensure that the selected time window vector is as uniform as possible, avoid too dense or overlap, that is, too small distance.

[0012] As preferred, in step S4, specifically includes the following steps: S401, in the starting process of the fluid machinery in the monitoring period, collecting bearing temperature data; the monitored bearing temperature data is constructed into a state vector as follows: ， Wherein, represents the sampling time index in the monitoring period, represents the state vector at the time, respectively represents the data of the th bearing temperature measuring point at the time ; S402, repeat step S401 in the monitoring period, based on the obtained multiple state vectors, get the monitoring time window vector as the input of the model through the time window operation as follows: , Wherein, represents the monitoring time window vector, respectively represents the time window data corresponding to the th bearing temperature measuring point at the time , as follows: ， S403, calculate the distance between the model input and the time window vector in the memory matrix , to form the distance vector as follows: , Wherein, represents the distance vector, represents the distance between the monitoring time window vector and the time window vector , specifically as follows: , S404, based on the vector distance analysis, the model outputs the monitoring result of the monitoring time window vector as follows: ， Wherein, represents the reconstruction result output by the model for , the scale coefficient makes the sum of the elements of is 1.

[0013] Further, in step S404, let Therefore, the monitoring result can be rewritten as follows: ， The monitoring result output by the model is a weighted sum of the time window vectors in the memory matrix, and the weight vector reflects the proportion of each time window vector in reconstructing the monitoring result.

[0014] As a preferred, in step S5, specifically includes the following steps: S501, extracting the monitoring result output by the model at the time t from the memory matrix. The reconstructed state vector at the time t is as follows: ， wherein, represents the first state vector reconstructed at the time t output by the model, represents the first reconstructed value of the data of the first monitoring point at the time t output by the model; S502, repeating step S4 to extract the reconstructed state vector at the time t, and obtaining a total of reconstructed state vectors of the model at the time t ; S503, the bearing temperature high warning vector and the low warning vector are as follows: , wherein represents the index; S504, monitoring the th bearing temperature monitoring point of the fluid machine at the time t, setting the bearing temperature monitoring deviation value, and the rules are as follows: , wherein represents the index; , wherein, , respectively represent the th element value of the warning vector ; the deviation value indicates that the temperature deviates from the safe interval ​​​​​According to the degree of deviation, a pre-warning threshold is set for the deviation value, the pre-warning is set according to the degree of deviation, and multi-level pre-warning is implemented. S505, steps S4 and S5 of steps S501, S502, S503 and S504 are performed on the remaining time in the monitoring period, the monitoring result of the monitoring period is obtained, and the monitoring of the bearing temperature of the large fluid machine in the starting process is realized.

[0015] Further, since the length of the time window is Therefore, the reconstruction result of the model output of the monitoring time window vector contains information of the continuous time, that is, the reconstruction result of a certain monitoring time appears in In step S502, step S4 is repeated for multiple times, that is, in order to collect the reconstruction state vectors of the model at In step S503, the upper limit and the lower limit of the reconstruction state vectors are used to divide the pre-warning vector.

[0016] Compared with the prior art, the present application has the following beneficial effects: The present application significantly improves the accuracy and sensitivity of the bearing temperature monitoring of the large fluid machine in the starting process by using the sliding time window operation, vector distance analysis and pre-warning logic construction. The sliding time window dynamically captures the temperature change rule, the vector distance analysis quantifies the data similarity, the adaptive pre-warning logic enhances the monitoring adaptability, and the device state is comprehensively reflected. The method reduces the false alarm rate and the missed alarm rate, realizes early warning, provides valuable time for equipment maintenance, effectively prevents equipment failure, and has high precision and high sensitivity. The present application provides reliable support for the safety monitoring of industrial large fluid machines. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 The working flowchart of the large fluid machine starting process bearing temperature monitoring method provided by the present application is shown in the figure; Figure 2 The working flowchart of the time window vector screening method by the maximum minimum distance method in step S3 is shown in the figure; Figure 3 The safety interval diagram of the temperature measurement point data and the model output in the bearing temperature monitoring task of the present application in the safe starting condition is shown in the figure; Figure 4 The safety interval diagram of the temperature measurement point data and the model output in the bearing temperature monitoring task of the present application in the safe starting condition is shown in the figure; Figure 5 The safety interval diagram of the temperature measurement point data and the model output in the bearing temperature monitoring task of the present application in the safe starting condition is shown in the figure; Figure 6 Fig. 2 is a schematic diagram of an abnormal temperature monitoring deviation value and a multi-stage early warning interval in a bearing temperature monitoring task in a failure risk starting condition of an embodiment of the present application. DETAILED DESCRIPTION

[0018] In order to make the above objectives, characteristics and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the drawings. However, the present application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without departing from the connotation of the present application, so the present application is not limited by the specific embodiments disclosed below.

[0019] The object selected in the embodiment is an auxiliary feed water pump of a domestic nuclear power plant, as shown in Fig. 1. Figure 1 Fig. 2 shows a workflow diagram of the bearing temperature monitoring method for the starting process of the auxiliary feed water pump in the embodiment based on the sliding time window operation, vector distance analysis and early warning logic construction. The steps in the embodiment will be introduced below.

[0020] S1, collect bearing temperature data of the starting process of the auxiliary feed water pump to obtain a plurality of state vectors. The state vectors in a safe working condition are taken to form a model database of the monitoring method. Specifically, the following steps are included: S101, collect bearing temperature data in the starting process of the auxiliary feed water pump; and form a state vector from the bearing temperature data collected each time, as follows: ， wherein, denotes the sampling time index, denotes the number of bearing temperature measurement points, denotes the state vector at the time, denotes the data of the i-th bearing temperature measurement point at the time.

[0021] In the embodiment, the number of bearing temperature measurement points of the auxiliary feed water pump is .

[0022] S102, include the state vectors in the safe starting condition of the auxiliary feed water pump into the model database of the monitoring method for constructing the model.

[0023] Step S2, use the sliding time window operation on the state vectors in the model database to obtain a plurality of time window vectors. Specifically, the following steps are included: ​​​S201, a time window operation is performed on the state vector in the model database, and in this embodiment, the length of the time window is 10 time points, and the time window vector is obtained as follows: , wherein, represents the time window vector, represents the time window data corresponding to the th bearing temperature measurement point at the th time point, as follows: ， wherein, .

[0024] S202, step S201 is executed by sliding on the time axis, and in this embodiment, 2000 time window vectors are obtained, denoted as .

[0025] Step S3, the distance between the time window vectors is calculated, the uniformly distributed time window vectors are selected to form a memory matrix, and the distance matrix of the memory matrix is obtained, and the construction of the model is completed. Specifically, the following steps are included: S301, the distance between any two time window vectors is calculated, as follows: , wherein, , represents the time point index of the time window vector, , represents the two time window vectors participating in the distance operation, , represents the time window data corresponding to the th bearing temperature measurement point at the th and th time point, represents the Euclidean norm of the vector.

[0026] S302, in this embodiment, through the maximum minimum distance operation, 1000 time window vectors obtained in step S202 are screened and form a memory matrix as follows: ， wherein, represents the memory matrix, represents the mapping of the screening operation, satisfying and the modulus of the set is .

[0027] The workflow diagram for screening the time window vector by the maximum minimum distance method is shown in Figure 2 .

[0028] S303. Construct the distance matrix corresponding to the memory matrix. The distance matrix is The square matrix is ​​as follows: , Among them, index .

[0029] Step S4: During the monitoring period, perform a time window operation on the state vector corresponding to the bearing temperature during the auxiliary feedwater pump start-up process to obtain the monitoring time window vector, which serves as the input to the model; calculate the distance between the monitoring time window vector and the time window vector in the memory matrix to obtain the distance vector; and obtain the reconstructed output of the model through vector distance analysis. Specifically, this includes the following steps: S401. During the start-up process of the auxiliary feedwater pump in the monitoring period, collect bearing temperature data; construct a state vector from the monitored bearing temperature data, as follows: ， in, This indicates the index of the sampling time within the monitoring period. express The state vector at time t, They represent the first Each bearing temperature measuring point is at Data at any given moment.

[0030] S402. Repeat step S401 during the monitoring period. Based on the obtained multiple state vectors, use a time window operation to obtain a monitoring time window vector as the input to the model, as follows: , in, Represents the monitoring time window vector. They represent the first Each bearing temperature measuring point The time window data corresponding to each moment is as follows: ， S403, Input of Calculation Model With memory matrix The distances between the vectors within the intermediate time windows are represented by the following distance vectors: , in, Represents the distance vector. Represents the monitoring time window vector With time window vector The distance is as follows: , S404. Based on vector distance analysis, the model outputs a monitoring time window vector. The monitoring results are as follows: ， in, The representation model is for The output reconstruction result, scale coefficient Make The sum of the elements is 1.

[0031] Step S5: After obtaining the reconstructed output of multiple continuous monitoring time window vectors, the bearing temperature is monitored during the start-up process of large fluid machinery based on the bearing temperature early warning vector and monitoring and early warning logic. Specifically, this includes the following steps: S501, Results from Model Output extract The reconstructed state vector at time step is as follows: ， in, This indicates the model output in The first state vector reconstructed at time step 1. These represent the first output of the model, respectively. Each measuring point is at The first reconstructed value of the time-based data.

[0032] S502, Time Setting Repeat step S4 to extract. The reconstructed state vectors at time t = ... 10 reconstructed state vectors at time 10 .

[0033] S503, High Bearing Temperature Warning Vector and low warning vector as follows: , in Indicates an index.

[0034] S504, Regarding the auxiliary water supply pump The first moment Temperature measurement points of each bearing To perform monitoring, set the bearing temperature monitoring deviation value according to the following rules: , in, , These represent the early warning vectors. , the first element value; the deviation value indicates the degree of temperature deviation from the safety interval , a pre-warning threshold value can be set for the deviation value, and the pre-warning is graded according to the degree of deviation to implement multi-level pre-warning.

[0035] In the embodiment, the pre-warning rules are as follows: , S505, steps S4 and S501, S502, S503, and S504 are performed for the remaining time points in the monitoring period to obtain the monitoring result of the monitoring period, and the monitoring of the bearing temperature of the auxiliary feed water pump in the start-up process in the nuclear power plant is realized.

[0036] In the embodiment, under the safe start-up working condition of the auxiliary feed water pump, the interval result of the monitoring of a certain measuring point in the monitoring task of the bearing temperature is as shown in Figure 3 , and the deviation value of the monitoring of the measuring point is as shown in Figure 4 ; under the start-up condition with a risk of failure of the auxiliary feed water pump, the interval result of the monitoring of a certain measuring point in the monitoring task of the bearing temperature is as shown in Figure 5 , and the deviation value of the monitoring of the measuring point is as shown in Figure 6 .

[0037] In the embodiment, the monitoring result is statistically analyzed: the false alarm rate of the monitoring and pre-warning of the bearing temperature in the start-up process under the safe working condition is 0, which reflects the accuracy of the application; for the start-up process with a risk of failure, the monitoring and pre-warning rate of the bearing temperature is 411s earlier than that by the traditional threshold method on average, which reflects the sensitivity of the application, and the failure omission rate is 0. It can be seen that the application can dynamically capture the change rule of the bearing temperature in the time sequence by introducing the sliding time window operation; combined with the vector distance analysis, the similarity or difference between the data is quantified, so as to improve the accuracy of the monitoring and avoid the equipment failure caused by the failure to discover the slight temperature change in time; the safety region output by the model combined with the pre-warning threshold value can dynamically adapt to the temperature change under different working conditions, and the problem of poor adaptability of the traditional fixed threshold method under different working conditions is avoided.

[0038] The above-described embodiment is only a preferred scheme of the application, and is not intended to limit the application. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the application. Therefore, any technical scheme obtained by equivalent replacement or equivalent transformation falls within the protection scope of the application.

Claims

1. A method for monitoring bearing temperature during the start-up process of large fluid machinery, characterized in that, The temperature monitoring method is based on sliding time window operation, vector distance analysis, and early warning logic, and includes the following steps: S1. Collect bearing temperature data during the start-up process of fluid machinery to obtain multiple state vectors, and take the state vectors under safe operating conditions to form a model database for the monitoring method; S2. Apply a sliding time window operation to the state vectors in the model database to obtain multiple time window vectors; S3. Calculate the distance between time window vectors, select uniformly distributed time window vectors to form a memory matrix, and obtain the distance matrix of the memory matrix to complete the model construction; S4. During the monitoring period, the state vector corresponding to the bearing temperature during the start-up process of the fluid machinery is subjected to time window operation to obtain the monitoring time window vector, which is used as the input of the model. Calculate the distance between the monitoring time window vector and the time window vector in the memory matrix to obtain the distance vector; obtain the model reconstruction output through vector distance analysis; S5. After obtaining the reconstructed output of multiple continuous monitoring time window vectors, the bearing temperature is monitored during the start-up process of large fluid machinery based on the bearing temperature early warning vector and the monitoring and early warning logic.

2. The monitoring method according to claim 1, characterized in that, Step S1 specifically includes the following steps: S101. During the start-up process of the fluid machinery, bearing temperature data is collected; the bearing temperature data collected each time is used to construct a state vector, as follows: ， in, Indicates the sampling time index. This indicates the number of measuring points for bearing temperature. express The state vector at time t, They represent the first Each bearing temperature measuring point Data at any given time; S102. The state vector under the safe start-up condition of fluid machinery is incorporated into the model database of the monitoring method for model construction.

3. The monitoring method according to claim 1, characterized in that, Step S2 specifically includes the following steps: S201. Apply a time window operation to the state vectors in the model database to obtain the following time window vectors: ， in, This indicates the length of the time window. Represents the time window vector. They represent the first Each bearing temperature measuring point The time window data corresponding to each moment is as follows: ， in, ; S202. Slide the execution of step S201 on the time axis to obtain multiple time window vectors, denoted as... subscript This indicates the number of time window vectors obtained in this step.

4. The monitoring method according to claim 3, characterized in that, Step S3 specifically includes the following steps: S301. Calculate the distance between any two time window vectors, as follows: ， in, , represents the time index of the time window vector. , Represents the two time window vectors involved in the distance calculation. , They represent the first Each bearing temperature measuring point and Time window data corresponding to a given moment The Euclidean norm of a vector; S302. By maximizing the minimum distance operation, the time window vectors obtained in step S202 are filtered, and the memory matrix is ​​constructed as follows: ， in, Represents the memory matrix, This indicates the number of time window vectors obtained after filtering. The mapping of the filter satisfies and set The model is ; S303. Construct the distance matrix corresponding to the memory matrix. The distance matrix is The square matrix is ​​as follows: ， Among them, index .

5. The monitoring method according to claim 4, characterized in that, Step S4 specifically includes the following steps: S401. During the startup process of the fluid machinery in the monitoring period, collect bearing temperature data; construct a state vector from the monitored bearing temperature data, as follows: ， in, This indicates the index of the sampling time within the monitoring period. express The state vector at time t, They represent the first Each bearing temperature measuring point Data at any given time; S402. Repeat step S401 during the monitoring period. Based on the obtained multiple state vectors, use a time window operation to obtain a monitoring time window vector as the input to the model, as follows: ， in, Represents the monitoring time window vector. They represent the first Each bearing temperature measuring point The time window data corresponding to each moment is as follows: ， S403, Input of Calculation Model With memory matrix The distances between the vectors within the intermediate time windows are represented by the following distance vectors: ， in, Represents the distance vector. Represents the monitoring time window vector With time window vector The distance is as follows: ， S404. Based on vector distance analysis, the model outputs a monitoring time window vector. The monitoring results are as follows: ， in, The representation model is for The output reconstruction result, scale coefficient Make The sum of the elements is 1.

6. The monitoring method based on claim 5, characterized in that, Step S5 specifically includes the following steps: S501, Results from Model Output extract The reconstructed state vector at time step is as follows: ， in, This indicates the model output in The first state vector reconstructed at time step 1. The first output of the model represents the... Each measuring point is at The first reconstructed value of the time-series data; S502, Time Setting Repeat step S4 to extract. The reconstructed state vectors at time t = ... Moment Reconstructed state vectors ; S503, High Bearing Temperature Warning Vector and low warning vector as follows: ， in Indicates an index; S504, for fluid machinery The first moment Temperature measurement points of each bearing To perform monitoring, set the bearing temperature monitoring deviation value according to the following rules: ， in, , These represent the early warning vectors. , The Each element value; the deviation value indicates the temperature. Deviating from the safe range The degree of deviation is assessed, and warning thresholds are set for the deviation values. Warnings are set according to the degree of deviation, and multi-level warnings are implemented. S505. For the remaining times during the monitoring period, execute steps S501, S502, S503, and S504 of steps S4 and S5 to obtain the monitoring results for the monitoring period, thereby realizing the monitoring of bearing temperature during the start-up process of fluid machinery.

Images