CN114913302B - Rotary joint life prediction system and method based on multi-sensor fusion - Google Patents

Abstract

The invention discloses a rotary joint life prediction system and method based on multi-sensor fusion, belonging to the field of industrial equipment life prediction. Divide the rotary joint into a three-dimensional grid in space, and set the data collection frequency for multiple sensors at the same time, collect the index data of the rotary joint according to the data collection frequency, establish a database, and then establish a life prediction model for the rotary joint to predict each sub-grid Finally, take the minimum value of the service life of all sub-grids as the overall service life of the rotary joint. The invention can not only understand the overall situation of the rotary joint, but also observe local details, and obtain more comprehensive information; reflect the actual situation of the rotary joint from multiple angles, and use various index data to help improve the accuracy of prediction , to achieve accurate evaluation and prediction of the state of the rotary joint from multiple dimensions, and the prediction results are accurate and reliable, which is conducive to reducing the probability of failure.

Description

Rotary joint life prediction system and method based on multi-sensor fusion

technical field

The invention relates to the field of life prediction of industrial equipment, and in particular to a device and method for life prediction of a rotary joint.

Background technique

The rotary joint is a closed rotary connector that rotates 360° to convey the medium. The role of the rotary joint is to input the liquid from the side of the pipeline into the rotating or reciprocating equipment, and then discharge it from the sealing device for the connection. The application fields of rotary joints cover almost every processing and manufacturing industry. Such as: metallurgy, machine tools, power generation, petroleum, rubber, plastics, textiles, printing and dyeing, pharmaceuticals, cigarettes, papermaking, food, feed processing, etc.

With the rapid development of the economy, the use of industrial equipment is increasing day by day, and the resulting problems are becoming more and more obvious. The longer the industrial equipment is used, the mechanical structure inside the equipment will be damaged to varying degrees, resulting in continuous decline in performance. Damage to the inside of the rotary joint will lead to a decrease in sealing performance, and there is a risk of leakage of the conveyed medium, and even the high temperature and high pressure medium will cause the equipment to burst. Conventional rotary joint overhaul cannot meet the increasing demands. However, the fault detection of rotary joints is mostly confirmed through factory acceptance, first inspection, calibration and other methods, but it is impossible to confirm whether the actual operating status of the rotary joint is healthy or not, and whether there are safety hazards or equipment hidden dangers during the calibration cycle. It affects the stability and reliability of the rotary joint as a conveying device. Whether the actual operating state of the rotary joint is healthy or not, and the safety or hidden dangers of the equipment during the inspection period are all reflected in the life of the rotary joint. Therefore, a reasonable plan to predict the service life of the rotary joint is proposed to avoid the failure in industrial production caused by the loss of the internal mechanical structure of the rotary joint in advance. In actual use and in most patents, little consideration is given to accurately predicting the service life of rotary joints, and remedial measures are often taken after a failure occurs. Therefore, the present invention takes this situation into consideration, improves the traditional method, can predict the service life of the rotary joint in time, better understand the situation of the rotary joint, and allow the enterprise to make a reasonable production plan.

Contents of the invention

The object of the present invention is to provide a life prediction system and method for a rotary joint based on multi-sensor fusion to address the shortcomings of the prior art.

Technical scheme: the technical scheme adopted by the present invention to solve the problem is:

A life prediction system for rotary joints based on multi-sensor fusion, including a grid division module, a multi-sensor data acquisition module, a database module, a model building module, a life prediction module and a result analysis module, the grid division module implements the rotary joint Carry out three-dimensional grid-like division in space, and label the coordinates of each sub-grid; the multi-sensor data acquisition module sets the data acquisition frequency for multi-sensors, and performs the rotary joint index data acquisition according to the data acquisition frequency, and rotates The joint index data is stored in the database module; the model building module calls all the rotary joint index data in the database module and performs training in it to obtain the relationship between the service life and other indexes, thereby obtaining the trained rotary joint life prediction model , and stored in the life prediction module; the life prediction module saves the trained rotary joint life prediction model, and inputs the index data of the rotary joint to be predicted into the life prediction module, and the service life of each rotary joint subgrid can be obtained. The result analysis module sorts the service life of all sub-grids, and selects the one with the shortest service life among all sub-grids as the overall service life of the rotary joint.

Further, the sensors include a speed sensor for collecting rotational speed S, a torque sensor for collecting torque N, a timer for collecting service life H, a vibration sensor for collecting vibration signal V, a temperature sensor for collecting medium temperature T, and a pressure sensor for collecting medium pressure P .

The invention also discloses a method for predicting the life of a rotary joint based on multi-sensor fusion, which includes the following steps:

S1: Divide the rotary joint into a three-dimensional grid in space, and label the coordinates of each divided sub-grid;

S2: Collect and save the index data of the rotary joint;

S3: Train all the collected rotary joint index data to obtain the relationship between the service life and other indicators, so as to obtain the trained rotary joint life prediction model;

S4: Input the index data of the rotary joint to be predicted into the life prediction model to obtain the service life of each rotary joint sub-grid, sort the service life of all sub-grids, and select the one with the shortest service life among all sub-grids as the rotary joint overall service life.

Further, in step S1, the division range of the three-dimensional grid includes at least the main part of the rotary joint, and the three-dimensional cube grid is divided with the length of one tenth of the diameter of the main part of the rotary joint as the side length, and each sub-grid is divided into rows Label with column coordinates.

Further, in step S2, the collected index data includes rotational speed S, torque N, service life H, vibration signal V, medium temperature T and medium pressure P, and the collection frequency of each index data is 0.1s.

Further, the collected vibration signal V is converted into the decibel value of the sound signal D by the following formula.

为放大系数；D为声音分贝平均值，单位为dB；a由振动传感器获取。In the formula, a is the overall average value of the vibration acceleration;

is the amplification factor; D is the average sound decibel, in dB; a is obtained by the vibration sensor.

Further, the service life H in the rotary joint index data is used as the output prediction value, and the remaining indicators are used as the input value of the prediction model, and the data is continuously recorded until 10 hours after the rotary joint is completely damaged.

Further, in step S3, construct the rotary joint life prediction model expression, the specific expression is:

H _{t, i} = f{S _{t, i} , N _t , g _i (d) D _{t, i} , g _i (d) T _{t, i} , P _{t, i} }

In the formula, t is a certain moment; i is the number of sub-grids, which are labeled in the order of rows; H _{t, i} is the service life of the i-th sub-grid at time t; N _t is the service life of the rotary joint to be tested at time t Torque, the torque value of all sub-grids at time t is the same; f is the corresponding law of input and output index data; S _{t, i} is the speed of the i-th sub-grid at time t; D _{t, i} is the i-th sub-grid at time t T _{t, i} is the temperature value of the i-th sub-grid at time t; P _{t, i} is the pressure value of the i-th sub-grid at time t; g _i (d) is the correlation coefficient function that changes with distance, When i is the sub-grid where the sensor monitoring point is located, g _i (d) is 1, and the rest of the sub-grids conform to g(d) distribution, and g(d) is an exponential decay function; d is the distance between two sub-grids distance.

Further, the formula for calculating the distance d between two sub-grids is specifically:

In the formula, p and q represent any two sub-grids in 3-dimensional space; p _j and q _j represent the corresponding components of p and q in 3-dimensional space, respectively.

Further, the service life H _t,i of each sub-grid is predicted from the corresponding rule f and the input index data of each sub-grid.

Beneficial effect: compared with the prior art, the present invention has the following advantages:

(1) Unlike traditional rotary joints that use sensors for overall monitoring, the present invention divides the main part of the rotary joint into three-dimensional grids in space, divides them into several sub-grids, and evaluates the state of each sub-grid Prediction can not only understand the overall situation of the rotary joint, but also observe the local details, and obtain more comprehensive information.

(2) The collected index data includes indexes obtained by five different sensors, which can reflect the actual situation of the rotary joint from multiple angles, and the use of various index data can help improve the accuracy of prediction.

(3) The index data of the rotary joint is collected by multiple sensors and the life prediction model is established. Taking into account the changes of the index of different sub-grids with the distance, a predictive regressor based on multi-Bayesian estimation is given to predict each The service life of the subgrid, taking the minimum service life as the overall service life of the rotary joint is beneficial to reduce the probability of failure.

Description of drawings

Fig. 1 is the block diagram of the rotary joint life prediction system based on multi-sensor fusion of the present invention;

Fig. 2 is the sensor schematic diagram of the rotary joint life prediction system based on multi-sensor fusion of the present invention;

Fig. 3 is a flow chart of the life prediction method of a rotary joint based on multi-sensor fusion in the present invention.

Detailed ways

The present invention will be further illustrated below in conjunction with the accompanying drawings and specific embodiments. This embodiment is implemented on the premise of the technical solution of the present invention. It should be understood that these embodiments are only used to illustrate the present invention and are not intended to limit the scope of the present invention.

As shown in Figure 1, the present invention discloses a life prediction system for rotary joints based on multi-sensor fusion, including a grid division module 1, a multi-sensor data acquisition module 2, a database module 3, a model building module 4, a life prediction module 5 and Results analysis module6.

Among them, the grid division module 1 realizes three-dimensional grid division of the rotary joint in space, and labels the coordinates of each sub-grid.

The multi-sensor data acquisition module 2 sets the data acquisition frequency for the multi-sensor, collects the index data of the rotary joint according to the data acquisition frequency, and stores the index data of the rotary joint into the database module 3; as shown in Figure 2, the sensor includes acquisition speed S Speed sensor, torque sensor to collect torque N, timer to collect service life H, vibration sensor to collect vibration signal V, temperature sensor to collect medium temperature T, and pressure sensor to collect medium pressure P.

The model building module 4 retrieves all the index data of the rotary joint in the database module 3 and conducts training in it to obtain the relationship between the service life and other indexes, so as to obtain a well-trained rotary joint life prediction model.

The life prediction module 5 saves the trained rotary joint life prediction model, and inputs the index data of the rotary joint to be predicted into the life prediction module 5 to obtain the service life of each rotary joint subgrid.

The result analysis module 6 sorts the service life of all sub-grids, and selects the one with the shortest service life among all sub-grids as the overall service life of the rotary joint.

The invention also discloses a life prediction method for rotary joints based on multi-sensor fusion. Based on the realization of the above-mentioned system, the three-dimensional grid-like division of the rotary joint in space is carried out, and at the same time, the data collection frequency is set for multiple sensors, and the data collection frequency is carried out according to the data collection frequency. Collect the index data of the rotary joint, establish the database, and then establish the life prediction model of the rotary joint to predict the service life of each sub-grid, and finally take the minimum value of the service life of all sub-grids as the overall service life of the rotary joint. As shown in Figure 3, it specifically includes the following steps:

Step S1: Use the mesh division module 1 to divide the rotary joint into a three-dimensional grid in space, and label the coordinates of each sub-grid.

The division range should at least include the main part of the rotary joint, which can ensure that most of the space of the rotary joint can be predicted. The three-dimensional cube grid is divided with the length of one tenth of the diameter of the main part of the rotary joint as the side length. Grids are numbered by row and column coordinates. The smaller the side length of the sub-grid, the better the final prediction effect, and the clearer the situation of each part of the rotary joint is known.

Step S2: Use the multi-sensor data acquisition module 2 to set the data acquisition frequency for the multi-sensors, collect the index data of the rotary joint according to the data acquisition frequency, and store the index data of the rotary joint in the database module 3 .

In the multi-sensor data acquisition module 2, the collection index and collection frequency are set as follows: speed S every 0.1s, torque N every 0.1s, service life H every 1s, vibration signal V every 0.1s, medium temperature T every 0.1s , Medium pressure P every 0.1s. The collected vibration signal V is converted into the decibel value of the sound signal D by the following formula;

为放大系数；D为声音分贝平均值，单位为dB；a由振动传感器获取。In the formula, a is the overall average value of the vibration acceleration;

is the amplification factor; D is the average sound decibel, in dB; a is obtained by the vibration sensor.

In the database module 3, the service life H in the index data of the rotary joint is used as the output prediction value, and the rest of the indexes are used as the input value of the prediction model. Damaged index data is beneficial to the fitting of the subsequent regression predictor, making the subsequent prediction structure more accurate.

Step S3: Use the model building module 4 to retrieve all the rotary joint index data in the database module 3 and perform training in it to obtain the relationship between the service life and other indicators, so as to obtain the trained rotary joint life prediction model and save it in the life prediction In module 5;

Construct the life prediction expression of the rotary joint, the specific expression is:

H _{t, i} = f{S _{t, i} , N _t , g _i (d) D _{t, i} , g _i (d) T _{t, i} , P _{t, i} }

In the formula, t is a certain moment; i is the number of sub-grids, which are labeled in the order of rows; H _{t, i} is the service life of the i-th sub-grid at time t; N _t is the service life of the rotary joint to be tested at time t Torque, the torque value of all sub-grids at time t is the same; f is the corresponding law of input and output index data; S _{t, i} is the speed of the i-th sub-grid at time t; D _{t, i} is the i-th sub-grid at time t T _{t, i} is the temperature value of the i-th sub-grid at time t; P _{t, i} is the pressure value of the i-th sub-grid at time t; g _i (d) is the correlation coefficient function that changes with distance, When i is the sub-grid where the sensor monitoring point is located, g _i (d) is 1, and the rest of the sub-grids conform to g(d) distribution, and g(d) is an exponential decay function; d is the distance between two sub-grids distance.

The distance d between two sub-grids is obtained using the Euclidean distance formula, and the specific calculation formula is:

In the formula, p and q represent any two sub-grids in 3-dimensional space; p _j and q _j represent the corresponding components of p and q in 3-dimensional space, respectively.

The service life H _t,i of each sub-grid is predicted by the corresponding law f and the input index data of each sub-grid. The corresponding rule f of the input and output index data is generated by the following multi-Bayesian regression predictor. The predictor can predict the required index data according to the input index data. The predicted result is the actual physical quantity. The multi-Bayesian regression predictor is specific The expression is:

为旋转接头输出预测指标数据；P(y|x₁，...，x₅)为在x₁，...，x₅均发生的情况下y发生的概率；P(y)为y发生的概率。In the formula, x ₁ ,..., x ₅ are the input index data of the rotary joint; y is the output index data to be predicted of the rotary joint;

Output predictor data for rotary joints; P(y|x ₁ ,...,x ₅ ) is the probability of y occurrence when x ₁ ,...,x ₅ all occur; P(y) is y occurrence The probability.

Step S4: Input the index data of the rotary joint to be predicted into the life prediction module 5, and the service life of each rotary joint sub-grid can be obtained, and the result analysis module 6 sorts the service life of all sub-grids, and screens out all sub-grids The shortest service life is regarded as the overall service life of the rotary joint, and the specific expression is:

In the formula, H _t,i is the predicted service life of the i-th sub-grid at time t; H _{t min} is the minimum service life of it at time t.

The difference between the present invention and the traditional rotary joint using sensors for overall monitoring is that the method divides the main part of the rotary joint into a three-dimensional grid in space, divides it into several sub-grids, and evaluates and predicts the state of each sub-grid . The collected indicator data includes indicators acquired by five different sensors, which can reflect the actual situation of the rotary joint from multiple angles. The index data of the rotary joint is collected by multiple sensors and the life prediction model is established. Taking into account the change of the index of different sub-grids with the distance, a predictive regressor based on multi-Bayesian estimation is given to predict each sub-grid service life. The invention can timely predict the service life of the rotary joint, better understand the situation of the rotary joint, and allow enterprises to make reasonable production plans.

The specific implementation described above is only a preferred embodiment of the present invention, and is not used to limit the implementation of the present invention and the scope of the claims. All equivalent changes and modifications made according to the content of the patent protection scope of the present invention should be included in the Within the scope of the patent application of the present invention.

Claims (9)

1. A rotary joint life prediction system based on multi-sensor fusion is characterized in that: comprises a grid dividing module (1), a multi-sensor data acquisition module (2), a database module (3), a model building module (4), a life prediction module (5) and a result analysis module (6),

the grid dividing module (1) is used for realizing three-dimensional grid division of the rotary joint in space and marking coordinates of each sub-grid;

the multi-sensor data acquisition module (2) sets data acquisition frequency for the multi-sensor, acquires rotary joint index data according to the data acquisition frequency, and stores the rotary joint index data into the database module (3);

the model building module (4) is used for retrieving and training index data of all the rotary joints in the database module (3) to obtain the relation between the service life and other indexes, so as to obtain a trained rotary joint life prediction model; the specific expression of the rotary joint life prediction model is as follows:

H _t,i ＝f{S _t,i ,N _t ,g _i (d)·D _t,i ,g _i (d)·T _t,i ,P _t,i }

wherein t is a certain time; i is the number of sub-grids, and the sub-grids are marked according to the sequence of rows; h _t,i The service life of the ith sub-grid t moment; n (N) _t The torque value of all the sub-grids at the moment t is the same as the torque of the rotary joint to be tested at the moment t; f is a corresponding rule of input and output index data; s is S _t,i The rotation speed at the moment t of the ith sub-grid; d (D) _t,i Sound decibels at the time t of the ith sub-grid; t (T) _t,i The temperature value at the moment t of the ith sub-grid; p (P) _t,i The pressure value at the moment t of the ith sub-grid; g _i (d) G when i is the subgrid where the monitoring point of the sensor is located as the correlation coefficient function changing along with the distance _i (d) 1, the rest sub-grids conform to the distribution of g (d), and g (d) is an exponential decay function; d is the distance between two sub-grids;

the life prediction module (5) stores a trained life prediction model of the rotary joint, the rotary joint index data to be predicted is input into the life prediction module (5) to obtain the service life of each rotary joint sub-grid,

and the result analysis module (6) sorts the service lives of all the sub-grids, and screens out the service life which is shortest in all the sub-grids and is used as the whole service life of the rotary joint.

2. A rotary joint life prediction system based on multi-sensor fusion according to claim 1, wherein: the sensor comprises a speed sensor for collecting the rotating speed S, a torque sensor for collecting the torque N, a timer for collecting the service life H, a vibration sensor for collecting the vibration signal V, a temperature sensor for collecting the medium temperature T and a pressure sensor for collecting the medium pressure P.

3. The rotary joint service life prediction method based on multi-sensor fusion is characterized by comprising the following steps of:

s1: dividing the rotary joint into three-dimensional grids in space, and marking the coordinates of each divided sub-grid;

s2: acquiring index data of the rotary joint and storing the index data;

s3: training all acquired rotary joint index data to obtain the relation between the service life and other indexes, thereby obtaining a trained rotary joint life prediction model; constructing a rotary joint life prediction model expression, wherein the specific expression is as follows:

H _t,i ＝f{S _t,i ,N _t ,g _i (d)·D _t,i ,g _i (d)·T _t,i ,P _t,i }

wherein t is a certain time; i is the number of sub-grids, and the sub-grids are marked according to the sequence of rows; h _t,i The service life of the ith sub-grid t moment; n (N) _t The torque value of all the sub-grids at the moment t is the same as the torque of the rotary joint to be tested at the moment t; f is a corresponding rule of input and output index data; s is S _t,i The rotation speed at the moment t of the ith sub-grid; d (D) _t,i Sound decibels at the time t of the ith sub-grid; t (T) _t,i The temperature value at the moment t of the ith sub-grid; p (P) _t,i The pressure value at the moment t of the ith sub-grid; g _i (d) G when i is the subgrid where the monitoring point of the sensor is located as the correlation coefficient function changing along with the distance _i (d) 1, the rest sub-grids conform to the distribution of g (d), and g (d) is an exponential decay function; d is the distance between two sub-grids;

s4: inputting the rotary joint index data to be predicted into a life prediction model to obtain the service life of each rotary joint sub-grid, sequencing the service lives of all the sub-grids, and screening out the shortest service life of all the sub-grids as the whole service life of the rotary joint.

4. A method for predicting the lifetime of a rotary joint based on multi-sensor fusion according to claim 3, wherein: in step S1, the dividing range of the three-dimensional grid at least includes a main body portion of the rotary joint, the three-dimensional cube grid is divided by taking the length of one tenth of the diameter of the main body portion of the rotary joint as the side length, and each sub-grid is marked according to row and column coordinates.

5. A method for predicting the lifetime of a rotary joint based on multi-sensor fusion according to claim 3, wherein: in step S2, the collected index data includes a rotation speed S, a torque N, a service life H, a vibration signal V, a medium temperature T and a medium pressure P, and the collection frequency of each index data is 0.1S.

6. The rotary joint life prediction method based on multi-sensor fusion according to claim 5, wherein: the collected vibration signal V is converted into a decibel value of the sound signal D by the following formula

Wherein a is the overall average value of vibration acceleration;

is the amplification factor; d is the average value of sound decibels, and the unit is dB; a is acquired by a vibration sensor.

7. A method for predicting the lifetime of a rotary joint based on multi-sensor fusion according to claim 3, wherein: and taking the service life H in the rotary joint index data as an output predicted value, taking the rest indexes as predicted model input values, and continuously recording the data until the rotary joint is completely damaged for 10 hours.

8. A method for predicting the lifetime of a rotary joint based on multi-sensor fusion according to claim 3, wherein: the calculation formula of the distance d between the two sub-grids is specifically as follows:

wherein p and q represent any two sub-grids in 3-dimensional space; p is p _j And q _j Representing the corresponding components of p and q in three-dimensional space, respectively.

9. A method for predicting the lifetime of a rotary joint based on multi-sensor fusion according to claim 3, wherein: predicting the service life H of each sub-grid by the corresponding rule f and the input index data of each sub-grid _t,i 。

Images