A risk assessment method for pumps based on decision tree

Abstract

A risk assessment method for a pump based on a decision tree includes collecting pressure data, identifying the maximum pump pressure establishment process, sending data to the cloud, training a decision tree model, and comparing and evaluating theoretical time with actual time. The risk assessment method for a pump based on a decision tree of the present invention is easy to monitor. Since the pump pressure is monitored and then the entire maximum pump pressure establishment process is extracted, compared to the method of monitoring vibration, this method has no additional noise signal interference results; the overall performance of the pump can be preliminarily judged by analyzing the pump pressure establishment process. When the pump performance is reduced due to wear, it can also be monitored by monitoring the pump pressure establishment process, thereby performing an overall risk assessment of the hydraulic pump.

Description

Technical Field

[0001] The invention relates to the technical field of equipment risk assessment, and in particular to a risk assessment method for a pump based on a decision tree. Background Art

[0002] As the power core of the hydraulic system, the importance of hydraulic pumps is self-evident and they are widely used in various occasions. However, the risk assessment of hydraulic pumps is relatively complicated. Usually, hydraulic pumps often show signs of oil leakage, component damage, excessive noise, etc. at the beginning of a failure. As the failure continues to intensify, the hydraulic pump often causes a pressure drop due to the failure, which eventually leads to the hydraulic pump not being able to work properly, and even causing serious safety accidents. Therefore, it is extremely important to evaluate the health status of the hydraulic pump, judge the operating status of the hydraulic pump, and prevent possible failures.

[0003] After searching the literature and patents of the prior art, it is found that the common methods for evaluating the health status of plunger pumps are as follows:

[0004] By monitoring the vibration signals of the hydraulic pump during operation, a prediction model is established for the frequency domain signals of the hydraulic pump vibration under a large number of historical conditions through the convolutional neural network method, and then the deep learning model is fine-tuned on a small amount of target hydraulic pump data using the idea of ​​transfer learning.

[0005] By monitoring the outlet pressure of the hydraulic pump and decomposing it using the empirical mode decomposition method, a group of intrinsic mode functions are obtained and their characteristics are extracted; combined with the typical time domain characteristics of the original signal, the characteristic vector of the signal is finally constructed.

[0006] Since the hydraulic pump produces strong noise during operation, when the hydraulic pump is in the early stage of failure, the vibration signal generated by the failure is easily covered by the noise. In addition, in the early stage of failure, the hydraulic pump has only a small amount of leakage and can still reach the set pressure. Summary of the invention

[0007] The technical problem to be solved by the present invention is to overcome the deficiencies of the above-mentioned prior art and provide a risk assessment method for pumps based on a decision tree which is simple to operate and has good effects.

[0008] The present invention is implemented by the following technical solution: a risk assessment method for a pump based on a decision tree, comprising the following steps:

[0009] S1. Collecting the pressure data of the hydraulic pump at the edge end, and the engine speed, engine torque percentage and pilot valve pressure data at the same time;

[0010] S2, automatically identifying the maximum pump pressure establishment process of the hydraulic pump at the edge end;

[0011] S3, sending the data of the process of establishing the maximum pump pressure to the cloud;

[0012] S4, training a decision tree model using data from the process of establishing the maximum pump pressure;

[0013] S5, loading the decision tree model to the edge;

[0014] S6. Compare the expected pressure buildup time obtained by the decision tree model with the actual pressure buildup time to roughly evaluate the performance of the hydraulic pump.

[0015] The method further comprises: in step S1, the pressure data of the hydraulic pump and the pilot valve pressure are collected through a pressure transmitter, and the engine speed and the engine torque percentage are obtained from the CAN signal of the engine.

[0016] In step S2, the maximum pump pressure establishment process is performed by judging the pilot pressure of the pump, and includes the following steps:

[0017] T1, set the time when the pilot pressure starts to change from the static pressure to time point A;

[0018] T2, set the time when the pilot pressure reaches near the maximum pressure as time point B;

[0019] T3, determine whether the pilot pressure between time point A and time point B increases monotonically;

[0020] T4, judging whether it is the process of establishing the maximum pump pressure, judging whether the main pump pressure at time point A is the static pressure value, and judging whether the main pump pressure at time point B is the maximum pump pressure;

[0021] T5. Determine the maximum pump pressure establishment process;

[0022] T6. Calculate the average value of the engine speed and the torque percentage data between time point A and time point B as the engine speed value and the torque percentage value, and take the time difference between time point A and time point B as the time value.

[0023] The engine speed value, the torque percentage value and the corresponding maximum pump pressure establishment time obtained in step T6 are brought into the decision tree model for training.

[0024] The decision tree model in step S4 is divided into two layers, the first layer is the interval division of the torque percentage, and the second layer is the interval division of the engine speed.

[0025] In step S4, the actual data is brought into the decision tree model to find the maximum pump pressure build-up time within the torsion percentage interval and the engine speed interval where the actual data is located.

[0026] In step S6, the calculated theoretical maximum pump pressure build-up time is compared with the actual maximum pump pressure build-up time, and the difference between the two is used as a measurement parameter.

[0027] The deviation of the maximum pump pressure buildup time in the same interval is set to 60 points, and the score = 60 + (maximum deviation - actual deviation) / maximum deviation * 40.

[0028] The present invention has the following advantages: the decision tree-based pump risk assessment method of the present invention is easy to monitor, because the pump pressure is monitored and then the entire maximum pump pressure establishment process is extracted. Compared with the method of monitoring vibration, this method has no additional noise signal interference results; the pump pressure establishment process of the pump can be analyzed to make a preliminary judgment on the overall performance of the pump. When the performance of the pump is reduced due to wear, it can also be monitored by monitoring the pump pressure establishment process, thereby performing an overall risk assessment of the hydraulic pump. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 is a flow chart of the present invention;

[0030] Figure 2 It is the technical roadmap of the present invention;

[0031] Figure 3 It is a schematic diagram of the content of data collection of the present invention;

[0032] Figure 4 is a schematic diagram of a sample setup of the present invention;

[0033] Figure 5 It is a schematic diagram of the identification of the preprocessing of the original signal and the process of establishing the maximum pump pressure of the present invention;

[0034] Figure 6 It is a training diagram of the engine speed of the present invention;

[0035] Figure 7 It is a schematic diagram of the 100% twist training of the present invention;

[0036] Figure 8 It is a schematic diagram of the processing process of theoretical values ​​and actual values ​​of the present invention. Implementation

[0037] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0038] like Figures 1 to 8 A risk assessment method for a pump based on a decision tree is shown, comprising the following steps:

[0039] S1. Collect the pressure data of the hydraulic pump at the edge end, and the engine speed, engine torque percentage and pilot valve pressure data at the same time; the pressure data of the hydraulic pump and the pilot valve pressure data are collected through the pressure transmitter, and the engine speed and engine torque percentage are obtained from the engine's CAN signal.

[0040] S2, automatically identifying the maximum pump pressure establishment process of the hydraulic pump at the edge end;

[0041] The process of establishing the maximum pump pressure can be judged by the pilot pressure of the pump. In a single action, when the pilot valve is opened, the main valve core also starts to move. After the main valve core is opened, it means that the process of establishing the maximum pump pressure begins. Finally, after the pressure of the main pump tends to stabilize, the entire process of establishing the maximum pump pressure is completed. The reason why the pressure of the pilot valve is used as a sign of the establishment of the maximum pump pressure is that there are many hydraulic components in the main pump circuit, and the pressure of the main pump is very high. Even in a stable state, its pressure fluctuations will be large, and it is difficult to use this pressure as an indicator to divide the pressure range; while the pressure value of the pilot valve circuit is very small, and there are no other hydraulic components to interfere, and there is almost no change in its pressure in a stable state. In actual experiments, the pilot pressure has been at 0.2bar when there is no action. When the maximum pump pressure is reached, the pilot pressure reaches 41.2bar-41.3bar.

[0042] The specific steps include:

[0043] T1, set the time when the pilot pressure starts to change from the static pressure to time point A; the static pressure is generally 0.2 bar;

[0044] T2, set the time point B when the pilot pressure reaches the maximum pressure; the maximum pressure is generally 41.2 bar;

[0045] T3, determine whether the pilot pressure between time point A and time point B is monotonically increasing; that is, the pilot pressure value of row N+1 is greater than the pilot pressure value of row N;

[0046] T4, determine whether it is the process of establishing the maximum pump pressure, determine whether the main pump pressure at time point A is a static pressure (with some fluctuations) value, and determine whether the main pump pressure at time point B is the maximum pump pressure (with some fluctuations);

[0047] T5, determine the maximum pump pressure establishment process; if the pilot pressure between time point A and time point B is not monotonically increasing, then this process is not the maximum pump pressure establishment process; when the pump pressure at time point A is not the static pressure of the pump, or the pump pressure at time point B is not the maximum pressure of the pump, then this process is not the maximum pump pressure establishment process;

[0048] T6. When the requirement of step T5 for the maximum pump pressure establishment time is met, the average value of the engine speed and torque percentage data between time point A and time point B is calculated as the engine speed value and torque percentage value, and the time difference between time point A and time point B is used as the time value; the obtained engine speed value, torque percentage value and the corresponding maximum pump pressure establishment time are brought into the decision tree model for training.

[0049] S3. Send the data of the maximum pump pressure establishment process to the cloud.

[0050] The data during the process of establishing the maximum pump pressure include, but are not limited to, engine speed, engine torque percentage, pressure, time corresponding to pressure, etc.

[0051] The collected data can be connected to the collector, and the collector will transmit the data to TBOX. TBOX has a built-in tb-gateway, which sends the collected data to the TB platform through the TB gateway.

[0052] S4. A decision tree model is trained using data from the process of establishing the maximum pump pressure to obtain the theoretical maximum pump pressure establishment time; the decision tree model is divided into two layers, the first layer is the interval division of the torque percentage, and the second layer is the interval division of the engine speed.

[0053] In the first layer, the torsion percentage and the corresponding maximum pump pressure build-up time are brought into the model, and the torsion percentage is divided into intervals to obtain the torsion percentages in different intervals. The engine speeds in different torsion percentage intervals and the corresponding maximum pump pressure build-up time are brought into the second layer model to obtain different engine speed intervals.

[0054] The maximum pump pressure build-up time within the same torque percentage interval and the same engine speed interval is calculated to obtain the range of the maximum pump pressure build-up time within the torque percentage interval and the engine speed interval.

[0055] Bring the actual data into the decision tree model, find the maximum pump pressure buildup time within the torque percentage interval and engine speed interval where the actual data is located, compare the actual time with the maximum pump pressure buildup time interval to see whether it is within the maximum pump pressure buildup time interval. A theoretical value can be obtained through methods such as median, mode, and mean.

[0056] S5, loading the decision tree model to the edge;

[0057] S6. Compare the theoretical maximum pump pressure build-up time with the existing actual measured maximum pump pressure build-up time to conduct an overall risk assessment of the pump.

[0058] The expected pressure buildup time obtained by the decision tree model is compared with the actual pressure buildup time to roughly evaluate the performance of the hydraulic pump. The evaluation method is to compare the calculated theoretical maximum pump pressure buildup time with the actual maximum pump pressure buildup time. The difference between the two is used as a measurement parameter. The deviation of the maximum pump pressure buildup time in the same interval is set to 60 points. The score = 60 + (maximum deviation - actual deviation) / maximum deviation * 40. This score can roughly make a simple preliminary evaluation of the pump.

[0059] The decision tree-based pump risk assessment method of the present invention performs a preliminary risk assessment on the pump by monitoring the maximum pump pressure establishment time. When the pump is severely worn or fails, it will cause leakage of the hydraulic oil in the pump. However, when there is a slight leakage, the pump can still reach the maximum pressure, but at the same time, the time to rise to the maximum pressure will be longer, so a preliminary assessment of the pump can be performed by judging the process of establishing the maximum pump pressure. By analyzing the maximum pump pressure establishment process, calculating the time for establishing the maximum pump pressure under various working conditions, comparing the actual maximum pump pressure establishment time, and conducting a risk assessment on the hydraulic pump, the operating status of the hydraulic pump can be judged, thereby preventing possible failures, which is extremely important.

[0060] Finally, it should be noted that the above description is only a preferred example of the present invention and is not intended to limit the present invention. Although the present invention is described in detail with reference to the above embodiments, those skilled in the art can still modify the technical solutions described in the above embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A risk assessment method for a pump based on a decision tree, characterized in that: The steps include: S1. Collecting the pressure data of the hydraulic pump at the edge end, and the engine speed, engine torque percentage and pilot valve pressure data at the same time; S2, automatically identifying the maximum pump pressure establishment process of the hydraulic pump at the edge end; S3, sending the data of the process of establishing the maximum pump pressure to the cloud; S4. A decision tree model is trained using the data from the process of establishing the maximum pump pressure. The decision tree model is divided into two layers, the first layer is the interval division of the torque percentage, and the second layer is the interval division of the engine speed. The actual data is brought into the decision tree model to find the maximum pump pressure establishment time within the torque percentage interval and the engine speed interval where the actual data is located; S5, loading the decision tree model to the edge; S6. Compare the theoretical maximum pump pressure build-up time obtained by the decision tree model with the actual measured maximum pump pressure build-up time to evaluate the performance of the hydraulic pump; In step S2, the maximum pump pressure establishment process is performed by judging the pilot pressure of the pump, and includes the following steps: T1, set the time when the pilot pressure starts to change from the static pressure to time point A; T2, set the time when the pilot pressure reaches near the maximum pressure as time point B; T3, determine whether the pilot pressure between time point A and time point B increases monotonically; T4, judging whether it is the process of establishing the maximum pump pressure, judging whether the main pump pressure at time point A is the static pressure value, and judging whether the main pump pressure at time point B is the maximum pump pressure; T5, determine the maximum pump pressure establishment process; if the pilot pressure between time point A and time point B is not monotonically increasing, then this process is not the maximum pump pressure establishment process; when the pump pressure at time point A is not the static pressure of the pump, or the pump pressure at time point B is not the maximum pressure of the pump, then this process is not the maximum pump pressure establishment process; T6, calculate the average value of the engine speed and the torsion percentage data between time point A and time point B as the engine speed value and the torsion percentage value, and take the time difference between time point A and time point B as the time value; The engine speed value, the torque percentage value and the corresponding maximum pump pressure establishment time obtained in step T6 are brought into the decision tree model for training.

2. A risk assessment method for a pump based on a decision tree as claimed in claim 1, characterized in that: In step S1, the pressure data of the hydraulic pump and the pilot valve pressure are collected through the pressure transmitter, and the engine speed and the engine torque percentage are obtained from the CAN signal of the engine.

3. A risk assessment method for a pump based on a decision tree as claimed in claim 1, characterized in that: In step S6, the calculated theoretical maximum pump pressure build-up time is compared with the actual maximum pump pressure build-up time, and the difference between the two is used as a measurement parameter.

4. A risk assessment method for a pump based on a decision tree as claimed in claim 3, characterized in that: The deviation of the maximum pump pressure buildup time in the same interval is set to 60 points, and the score = 60 + (maximum deviation - actual deviation) / maximum deviation * 40.

Images