A method for handling valve fault characteristics in reciprocating compressors based on logarithmic dynamometer diagrams

Abstract

This invention provides a method for processing valve fault features in reciprocating compressors based on logarithmic dynamometer diagrams. Addressing the difficulty in finding suitable fault features to reflect valve faults under varying operating conditions, this method establishes a mapping between the compression and expansion processes of the logarithmic dynamometer diagram and operating parameters using a BP neural network. Based on this, a predicted logarithmic dynamometer diagram is obtained. By comparing the difference between the actual and predicted logarithmic dynamometer diagrams, valve faults are reflected. The method includes the following modules: 1) a data preprocessing module; 2) a logarithmic dynamometer diagram prediction module; and 3) a difference comparison module. This method reduces the dimensionality of the two-dimensional dynamometer diagram features (pressure and volume) to one dimension, significantly improving computational efficiency in fault diagnosis research. It also boasts advantages such as low hardware investment costs and high practicality.

Description

Technical Field

[0001] This invention belongs to the field of reciprocating compressor valve fault diagnosis technology, and more specifically, relates to a method for processing reciprocating compressor valve fault features based on logarithmic dynamometer diagrams. Background Technology

[0002] Reciprocating compressors are widely used in petrochemical and other fields due to their high compression efficiency and wide pressure range. The valve is a core component of a reciprocating compressor, controlling the intake and exhaust of gas. Frequent opening and closing of the valve, coupled with airflow impacts and collisions with the valve seat, makes it one of the components with the highest failure rate in reciprocating compressors. Valve failure can lead to pressure ratio imbalance, reduced discharge volume, and decreased efficiency. In severe cases, broken valve plates or springs can fall into the cylinder, potentially causing cylinder scoring and rendering the unit unusable. Therefore, timely and effective detection of valve failures is crucial for ensuring the safe and long-term operation of reciprocating compressors. The accuracy and practicality of valve failure diagnosis largely depend on the selection and processing of valve failure signals.

[0003] Currently, the most widely used valve fault signals are valve cover vibration signals and cylinder dynamic pressure signals. The principle behind vibration monitoring for fault diagnosis is that the valve disc generates impact vibrations during opening and closing. When a fault occurs, the vibration characteristics change. Because this monitoring method is non-invasive, and the vibration sensor is mounted on the valve cover, the vibration signal from the valve is attenuated when it reaches the vibration probe mounted on the valve cover. Furthermore, the compressor itself experiences significant vibration, resulting in strong non-stationarity and a low signal-to-noise ratio in the vibration signal at the valve cover. Compared with vibration monitoring, the use of cylinder dynamic pressure signal for valve fault diagnosis has the characteristics of low signal interference, low investment cost and sensitive fault response. The usual practice is to associate the pressure signal with the operation process of the reciprocating compressor to generate the reciprocating compressor indicator diagram, and analyze the valve fault from the reciprocating compressor indicator diagram. However, there are still some limitations: (1) Analyzing the valve fault based on the indicator diagram depends on the indicator diagram under normal operating conditions for comparison. However, for industrial reciprocating compressors, the operating conditions are time-varying and gradual, which makes it difficult for the normal indicator diagram used for reference to adapt to all operating conditions. On the other hand, due to the gradual change of operating conditions, the amount of normal indicator diagram data extracted from all operating conditions will be huge and difficult to apply to industrial sites. (2) Different faults of the air valve have different effects on the four stages of expansion, intake, compression and exhaust of the indicator diagram. For example, when the exhaust valve leaks, the expansion and compression process curves shift to the right compared to the normal indicator diagram, while when the intake valve leaks, the expansion and compression process curves shift to the left. Most fault analysis based on the indicator diagram is similar to the above qualitative description and lacks quantitative indicators, making it difficult to quantitatively extract fault characteristics for computer-aided analysis and diagnosis.

[0004] In the article "Fault Diagnosis of Reciprocating Compressor Valves Based on Geometric Properties of Dynamometer Diagram and Neural Network" published in the January 2018 issue of "Compressor Technology", a method for extracting features from dynamometer diagrams based on the geometric properties of planar graphics was proposed. This method uses the geometric property parameters of the dynamometer diagram as feature vectors to diagnose valve faults. Under certain compressor operating conditions, this method has a high fault identification rate. However, since changes in operating conditions will also cause changes in the geometric properties of the dynamometer diagram, it will be difficult to distinguish between changes in dynamometer diagram features caused by faults and changes caused by changes in operating conditions. Therefore, this method is not applicable under varying operating conditions. Summary of the Invention

[0005] To overcome the shortcomings of existing technologies, this invention provides a method for processing fault features of reciprocating compressor valves based on logarithmic dynamometer diagrams. The purpose is to establish a normal logarithmic dynamometer diagram for reference in a logarithmic coordinate system for the dynamometer diagrams of normal or fault states under various operating conditions within the compressor's operating range. Based on this, the difference between the dynamometer diagram to be measured and the reference dynamometer diagram is quantitatively obtained to obtain a feature expression with high cohesion and high inter-class variability. Using this feature as input to the valve fault diagnosis model can significantly improve the fault diagnosis effect.

[0006] The technical solution adopted in this invention is as follows: A method for handling fault characteristics of reciprocating compressor valves based on logarithmic dynamometer diagrams, comprising the following steps:

[0007] S100: Select N operating conditions within the operating range of the reciprocating compressor, collect cylinder dynamic pressure signal, flywheel key phase signal, and operating condition parameter signals (intake temperature and pressure signal, exhaust temperature and pressure signal, cylinder wall temperature signal) for M cycles of stable operation of the compressor under each operating condition, generate M indicator diagrams under each operating condition, and record the average value of the corresponding operating condition parameter signals.

[0008] S200: Take the natural logarithm of the horizontal and vertical coordinates of the dynamometer charts collected under each working condition to obtain the corresponding logarithmic dynamometer chart;

[0009] S300: Draw the intake and exhaust pressure lines in logarithmic coordinates on the logarithmic dynamometer diagram. These pressure lines will intersect the compression and expansion segments of the logarithmic dynamometer diagram at four points. Use linear interpolation to insert the four intersection points into the logarithmic dynamometer diagram. Then, connect the two points closest to the start and end points of compression to form a new compression segment, and connect the two points closest to the start and end points of expansion to form a new expansion segment. Finally, calculate the slopes of the two straight lines, which are respectively the compression exponent k. c With expansion process exponent k e ;

[0010] S400: The compression process index and expansion process index calculated from each indicator diagram of each working condition, along with the collected working condition parameters, are taken as a set of data, resulting in M ​​sets of data for N working conditions. The M sets of data for each working condition are randomly divided into training set and test set according to the ratio a:b, where a>b.

[0011] S500: The operating parameters (intake temperature and pressure, exhaust temperature and pressure, cylinder wall temperature) and compression and expansion indices in the training set are standardized and the standardization criteria are saved. The operating parameters in the test set are also standardized according to the criteria.

[0012] S600: Establish a three-layer BP neural network with standardized operating parameters as input nodes, t hidden layer nodes, and standardized compression and expansion process indices as output nodes respectively. Use training set data as training samples to train the model and obtain the process index prediction model.

[0013] S700: On the test set, the standardized operating parameters are input into the process index prediction model to obtain the predicted standardized compression and expansion process indices. After de-standardization using the standardization criteria of the training set, the predicted compression and expansion process indices are obtained. The process indices are used as the slopes of the compression and expansion process lines, respectively. The compression start point of the new compression segment and the expansion start point of the new expansion segment in step S300 are used as the starting points of the compression and expansion process lines to initially obtain the predicted compression and expansion process lines. The linear equations of the process lines are then obtained. The logarithmic volume values ​​on the original compression and expansion segments are substituted into the linear equations to obtain the logarithmic pressure values ​​on the predicted process lines. The compression and expansion process lines terminate at the exhaust process line and the intake process line, respectively, to obtain the final predicted expansion and compression process lines. The intake and exhaust processes are simplified to intake and exhaust pressures, respectively, to finally obtain the predicted logarithmic dynamometer diagram of the test sample.

[0014] S800: On the test set, firstly, a simplified original logarithmic dynamometer diagram is obtained using the method in step S300. Then, the natural logarithms of the two pressures under the same volume natural logarithm of the original logarithmic dynamometer diagram and the predicted logarithmic dynamometer diagram are subtracted. All differences are then tiled and standardized in one dimension according to the expansion, intake, compression and exhaust processes to obtain a data sequence reflecting the working state of the valve.

[0015] The present invention has the following advantages: The present invention uses cylinder dynamic pressure as a fault signal to reflect the fault characteristics of the air valve. Only a single dynamic pressure sensor and key phase sensor are needed for each compression cylinder. Compared with the vibration analysis method that installs vibration sensors on each air valve, the hardware investment cost is lower and the practicality is stronger.

[0016] Furthermore, the compressor operation prediction model obtained based on multi-condition data training can well adapt to valve fault extraction under variable operating conditions. By comparing the logarithmic dynamometer diagram generated by the machine learning model with the actual logarithmic dynamometer diagram, the abnormal changes in the dynamometer diagram caused by valve faults can be fully reflected, and normal and abnormal data can be fully distinguished. Moreover, this processing method has the characteristic of reducing the dynamometer diagram features of two-dimensional attributes of pressure and volume to one dimension, which can greatly improve the efficiency of computational processing in fault diagnosis research. Attached Figure Description

[0017] Figure 1 This is a flowchart of the method of the present invention;

[0018] Figure 2 It is a logarithmic indicator diagram of a reciprocating compressor;

[0019] Figure 3 It is the actual logarithmic dynamometer diagram and the predicted logarithmic dynamometer diagram;

[0020] Figure 4 These are the indicator curves for different operating conditions under normal conditions of the air valve;

[0021] Figure 5 The indicator curves of the air valve under different fault conditions under the same operating conditions;

[0022] Figure 6 This is a waveform diagram showing the difference in logarithmic dynamometer readings under different operating conditions under normal conditions.

[0023] Figure 7 The waveform diagram showing the difference between the logarithmic dynamometer diagram under normal and various fault conditions; Detailed Implementation

[0024] The dynamometer diagram is an important fault diagnosis diagram reflecting the working status of valves in a reciprocating compressor. Changes in its shape correspond to different valve fault types. Since changes in compressor operating conditions also cause changes in the shape of the dynamometer diagram, directly using the dynamometer diagram for fault diagnosis under varying operating conditions can easily lead to false alarms. This invention provides a method for processing valve fault features in reciprocating compressors based on logarithmic dynamometer diagrams. By using this method to process the dynamometer diagram, the influence of changes in operating conditions can be eliminated, enabling the extraction of valve fault features under varying operating conditions.

[0025] See Figure 1 This is a flowchart of the method of the present invention. The method for processing the fault characteristics of a reciprocating compressor valve based on a logarithmic dynamometer diagram can be summarized into three modules: data preprocessing, logarithmic dynamometer diagram prediction, and difference comparison.

[0026] In the data preprocessing module, a logarithmic dynamometer is obtained by logarithmizing the dynamometer diagram. The process is then divided using intake and exhaust pressures. The start and end points of the compression and expansion processes are obtained in a logarithmic coordinate system. The slopes of the lines connecting the start and end points of the two processes are calculated to obtain the compression and expansion process exponents. Based on this, a dataset is constructed using operating parameters, and data standardization is performed. Specifically, the steps include:

[0027] Step S100: Select N operating conditions within the operating range of the reciprocating compressor. Use a dynamic pressure sensor and a proximity switch sensor to collect the cylinder dynamic pressure signal and flywheel keyway signal during the stable operation of the compressor under each operating condition. Use a temperature sensor and a pressure sensor to collect operating condition parameter signals: intake temperature and pressure signal, exhaust temperature and pressure signal, and cylinder wall temperature signal. The switching point of the proximity switch sensor is the piston position when the instantaneous cylinder volume is at its minimum. Collect data for M operating cycles for each operating condition, and collect K data points for each operating cycle. Obtain the reciprocating compressor indicator diagram with cylinder dynamic pressure as the vertical axis and cylinder volume as the horizontal axis. At the same time, record the average value of the operating condition parameters in each operating cycle: intake temperature and pressure, exhaust temperature and pressure, and cylinder wall temperature.

[0028] The M indicator diagrams for the N operating conditions mentioned in this step are represented as follows:

[0029]

[0030] Among them, D i This represents all dynamometer diagram data for the i-th working condition out of N working conditions; [·] m This represents the data of the m-th indicator diagram among M indicator diagrams; (p k V k ) represents a data point in the indicator diagram; p k V represents the dynamic pressure of the cylinder; k The instantaneous volume of the cylinder is expressed by the following formula:

[0031] V k =x k ·π·D 2 / 4

[0032] Where, x k D represents the instantaneous piston displacement; D represents the piston diameter.

[0033] Furthermore, the formula for calculating the instantaneous piston displacement is:

[0034]

[0035] Where r represents the crank radius; L represents the connecting rod length; α kα represents the instantaneous crank angle. k = k × 360 / K.

[0036] Step S200: Take the natural logarithm of the horizontal and vertical coordinates of the dynamometer diagrams generated under each working condition in step S100 to obtain the corresponding logarithmic dynamometer diagram;

[0037] In step S200, the specific formulas for calculating the cylinder dynamic pressure and instantaneous cylinder volume using the natural logarithm are as follows:

[0038] p k =lnp k

[0039] V k '=lnV k

[0040] Where, p k ' indicates the logarithmic pressure value; V k ' represents the logarithmic volume value.

[0041] S300: Draw the intake and exhaust pressure lines in logarithmic coordinates on the logarithmic dynamometer diagram. These pressure lines will intersect the compression and expansion sections of the logarithmic dynamometer diagram at points 1, 2, 3, and 4. Figure 2 As shown, Figure 2 A specific example of a logarithmic indicator diagram for a reciprocating compressor is provided. Linear interpolation is used to insert the four intersection points into the logarithmic indicator diagram. Then, points 1 and 2 are taken as the start and end points of compression, and points 3 and 4 as the start and end points of expansion. Connecting points 1 and 2 forms the new compression segment, and connecting points 3 and 4 forms the new expansion segment. These new compression and expansion segments are then connected to the intake and exhaust segments of the original logarithmic indicator diagram to obtain a simplified original logarithmic indicator diagram. Finally, the slopes of the two straight lines are calculated, and these slopes represent the compression process exponent k, respectively. c With expansion process exponent k e ;

[0042] In step S300, the formula for calculating the instantaneous cylinder volume using linear interpolation in logarithmic coordinates is as follows:

[0043]

[0044] Where, p a+1 With p a It is the known logarithmic value of the cylinder dynamic pressure; V a+1 With V a It is the known logarithmic value of the instantaneous cylinder volume; p add This is the logarithmic value of the cylinder dynamic pressure at the intersection point. Let the intersection point 1, which is closest to the compression start point, be (p). add1 V add1The intersection point 2, which is close to the end of the compression, is (p) add2 V add2 The intersection point 3, which is close to the starting point of expansion, is (p) add3 V add3 The intersection point 4, which is close to the end of the expansion, is (p) add4 V add4 ).

[0045] Furthermore, in step S300, the compression process index k c and the expansion process index k e The calculation formula is:

[0046]

[0047] S400: The compression process index and expansion process index calculated from each indicator diagram of each working condition, along with the collected working condition parameters, are taken as a set of data, resulting in M ​​sets of data for N working conditions. The M sets of data for each working condition are then randomly divided into training set and test set according to the ratio a:b (a>b).

[0048] S500: Standardize the operating parameters (intake temperature and pressure, exhaust temperature and pressure, cylinder wall temperature) and compression and expansion indices in the training set, save the standardization criteria, and standardize the operating parameters in the test set according to the criteria.

[0049] The changes in the compression and expansion processes of a reciprocating compressor are affected by the valve malfunction status and operating conditions. By establishing the relationship between the compression and expansion process indices and operating parameters, the impact of operating condition changes can be understood, thus highlighting only valve malfunctions. Therefore, a machine learning method is adopted. Based on the data processed by the data preprocessing module, a mapping between operating parameters and process indices is established through data training, so as to achieve the goal of predicting the logarithmic dynamometer diagram based on the operating parameters.

[0050] The logarithmic dynamometer prediction module specifically includes the following steps (numbered as before).

[0051] S600: Establish a three-layer BP neural network with standardized operating parameters as input nodes, t hidden layer nodes, and standardized compression and expansion process indices as output nodes respectively. Use training set data as training samples to train the model and obtain the process index prediction model.

[0052] In this step, there are two three-layer BP neural networks, which respectively predict the compression process index and the expansion process index. The input nodes of the two BP neural networks are 5, which are the standardized intake temperature and pressure, exhaust temperature and pressure, and cylinder wall temperature, respectively. The output node is 1. For the compression process index prediction model, the output node is the standardized compression process index. For the expansion process index prediction model, the output node is the standardized expansion process index.

[0053] S700: On the test set, the standardized operating parameters are input into the process index prediction model to obtain the predicted standardized compression and expansion process indices. After de-standardization using the standardization criteria of the training set, the predicted compression and expansion process indices are obtained. These two process indices are used as the slopes of the compression and expansion process lines, respectively. The compression and expansion process lines are constructed using intersection point 1 in step S300 as the compression start point and intersection point 3 as the expansion start point. The predicted compression and expansion process lines are initially obtained, and the linear equations of the process lines are calculated. Then, the logarithmic volume values ​​on the original compression and expansion segments are substituted into the linear equations to obtain the logarithmic pressure values ​​on the predicted process lines. The compression and expansion process lines terminate at the exhaust process line and the intake process line, respectively, to obtain the final predicted expansion and compression process lines. The intake and exhaust processes are simplified to intake and exhaust pressures, respectively, to finally obtain the predicted logarithmic dynamometer diagram of the test sample, as shown below. Figure 3 As shown, the actual logarithmic work plot and the predicted logarithmic work plot of the example are illustrated.

[0054] Preferably, in step S700, the formula for calculating the linear equations of the expansion section and the compression section is as follows:

[0055]

[0056] For the expansion segment, (x1, y1) and (x2, y2) are the intersection points 3 (p) in step S300, respectively. add3 V add3 ) and intersection point 4 (p add4 V add4 For the compressed segment, (x1, y1) and (x2, y2) are the intersection points 1 (p) in step S300, respectively. add1 V add1 ) and intersection point 2 (p add2 V add2 ).

[0057] Furthermore, the logarithmic pressure values ​​under each logarithmic volume of the intake and exhaust processes are replaced with the logarithmic values ​​of the intake and exhaust pressures in the operating parameters.

[0058] Furthermore, the logarithmic pressure values ​​under each logarithmic volume of the process between the intersection point 1 and the actual compression starting point corresponding to the maximum logarithmic volume are all replaced with the logarithmic intake pressure values ​​in the operating parameters. The logarithmic pressure values ​​under each logarithmic volume of the process between the intersection point 4 and the actual expansion starting point corresponding to the minimum logarithmic volume are all replaced with the logarithmic exhaust pressure values ​​in the operating parameters. Finally, a complete predicted logarithmic indicator diagram is obtained.

[0059] The difference comparison module specifically includes the following steps (numbered as before).

[0060] S800: On the test set, firstly, a simplified original logarithmic dynamometer diagram is obtained using the method in step S300. Then, the natural logarithms of the two pressures under the same volume natural logarithm of the original logarithmic dynamometer diagram and the predicted logarithmic dynamometer diagram are subtracted. All differences are then tiled and standardized in one dimension according to the expansion, intake, compression and exhaust processes to obtain a data sequence reflecting the working state of the valve.

[0061] In step S800, the formula for the difference between the two logarithmic pressures under the same logarithmic volume is as follows:

[0062] Δ=lnp′ V -lnp V

[0063] Among them, lnp′ V To predict the natural logarithm of the pressure corresponding to a logarithmic volume of V in a logarithmic indicator diagram; lnp V It is the natural logarithm of the pressure corresponding to the logarithmic volume V in the actual logarithmic dynamometer diagram.

[0064] Depend on Figure 4 , Figure 5 As can be seen, the diagrams are the dynamometer curves under different operating conditions of the valve under normal conditions and the dynamometer curves of the valve under different fault conditions under the same operating conditions. The changes in the expansion, intake, compression, and exhaust sections are affected by both changes in operating conditions and valve faults. Therefore, for compressors operating under varying conditions, using the absolute characteristics of the dynamometer curve shape for valve fault diagnosis is prone to generating false alarms caused by characteristic changes due to changes in operating conditions. Therefore, eliminating the influencing factors of changes in operating conditions can improve the fault diagnosis effect. The logarithmic dynamometer curve difference preprocessing method proposed in this paper uses a logarithmic dynamometer curve prediction model to generate a standard logarithmic dynamometer curve in a logarithmic coordinate system and then performs difference to eliminate operating condition factors, reflecting only the changes in the shape of the dynamometer curve caused by valve faults, and finally obtaining the relative characteristics of the dynamometer curve. Figure 6The graphs show the logarithmic dynamometer difference waveforms under different operating conditions under normal conditions. It can be seen that even under different operating conditions, the difference curves still maintain similar characteristics: under normal valve conditions, the expansion and compression stages tend to be linear, while the intake and exhaust stages fluctuate. However, under fault conditions, such as... Figure 7 The waveform of the logarithmic dynamometer diagram difference between the normal state and various fault states is shown. The four stages mentioned above are different from the normal state. For example, when a valve leakage fault occurs, the waveform will show a large increase or decrease in the expansion and compression stages. When a valve spring fault occurs, the fluctuation will show an increase or decrease in the intake and exhaust stages. The curves show different characteristics under different valve fault types. Therefore, this processing method can fully reflect the abnormal change characteristics of the dynamometer diagram and distinguish between normal and abnormal data. This processing method also has the characteristic of reducing the two-dimensional attributes of the pressure and volume of the dynamometer diagram to one dimension, which greatly improves the efficiency of subsequent calculation and processing.

Claims

1. A method for handling fault characteristics of reciprocating compressor valves based on logarithmic dynamometer diagrams, characterized in that, Includes the following steps: S100: Select N operating conditions within the operating range of the reciprocating compressor, collect the cylinder dynamic pressure signal, flywheel key phase signal and operating condition parameter signal of the compressor under each operating condition for M cycles of stable operation, generate M indicator diagrams under each operating condition, and record the average value of the corresponding operating condition parameter signal in each operating cycle. S200: Take the natural logarithm of the horizontal and vertical coordinates of the dynamometer charts collected under each working condition to obtain the corresponding logarithmic dynamometer chart; S300: Draw the intake and exhaust pressure lines in logarithmic coordinates on the logarithmic dynamometer diagram. These pressure lines intersect the compression and expansion segments of the logarithmic dynamometer diagram at four points. Use linear interpolation to insert these four intersection points into the logarithmic dynamometer diagram. Then, connect the two intersection points closest to the compression start and end points to form a new compression segment, and connect the two intersection points closest to the expansion start and end points to form a new expansion segment. Connect the new compression and expansion segments to the intake and exhaust segments in the original logarithmic dynamometer diagram to obtain a simplified original logarithmic dynamometer diagram. Finally, calculate the slopes of the new compression and expansion segments, which are respectively the compression exponent k. c With expansion process exponent k e ; S4 00: Take the compression process index and expansion process index of each indicator diagram under each working condition and the collected working condition parameters as a set of data, and then get M sets of data for N working conditions. Randomly divide the M sets of data for each working condition into training set and test set according to the ratio of a:b, where the value of a is greater than b. S5 00: Standardize the operating condition parameters, compression process exponent, and expansion process exponent in the training set, and save the standardization criteria. Standardize the operating condition parameters in the test set according to the same criteria. S6 00: Establish a three-layer BP neural network with standardized operating parameters as input nodes, t hidden layer nodes, and standardized compression and expansion process indices as output nodes respectively. Use training set data as training samples to train the model and obtain the process index prediction model. S700: On the test set, the standardized operating parameters are input into the process index prediction model to obtain the predicted standardized compression and expansion process indices. After destandardization using the standardization criteria of the training set, the predicted compression and expansion process indices are obtained. The predicted process indices are used as the slopes of the compression and expansion process lines, and the compression start point of the new compression segment and the expansion start point of the new expansion segment in step S300 are used as the starting points of the compression and expansion process lines to initially obtain the predicted compression and expansion process lines. The linear equation of the predicted process lines is then obtained. The logarithmic volume values ​​on the original compression and expansion segments are substituted into the linear equation to obtain the logarithmic pressure values ​​on the predicted process lines. The predicted compression and expansion process lines terminate at the exhaust process line and the intake process line, respectively, to obtain the final predicted expansion and compression process lines. The intake and exhaust processes are simplified to intake and exhaust pressures, respectively, to finally obtain the predicted logarithmic dynamometer diagram of the test sample. S800: On the test set, firstly, a simplified original logarithmic dynamometer diagram is obtained using the method in step S300. Then, the natural logarithms of the two pressures under the same volume natural logarithm of the original logarithmic dynamometer diagram and the predicted logarithmic dynamometer diagram are subtracted. All differences are then tiled and standardized in one dimension according to the expansion, intake, compression and exhaust processes to obtain a data sequence reflecting the working state of the valve.

2. The method according to claim 1, characterized in that, The operating condition parameter signals include: intake air temperature and pressure signals, exhaust air temperature and pressure signals, and cylinder wall temperature signals.

3. The method according to claim 1, characterized in that, In step S100: M dynamometer diagrams for N operating conditions are represented as follows: Among them, D i This represents all dynamometer diagram data for the i-th working condition out of N working conditions; [·] m This represents the data of the m-th indicator diagram among M indicator diagrams; (p k V k ) represents the k-th data point in the indicator diagram, p k V represents the dynamic pressure of the cylinder; k This indicates the instantaneous volume of the cylinder.

4. The method according to claim 3, characterized in that, In step S200: The cylinder dynamic pressure p on the indicator diagram k With cylinder instantaneous volume V k The specific formulas for calculating the natural logarithm are as follows: pp k '=ln p k V k '=ln V k Where, p k ' indicates the logarithmic pressure value; V k ' represents the logarithmic volume value.

5. The method according to claim 4, characterized in that, In step S300: The instantaneous volume V of the cylinder in logarithmic coordinate system k The formula for linear interpolation is: Where, p a+1 With p a It is the known logarithmic value of the cylinder dynamic pressure; V a+1 With V a It is the known logarithmic value of the instantaneous cylinder volume; p add It is the logarithmic value of the cylinder dynamic pressure at the intersection point; let the intersection point 1, which is closer to the compression start point, be (p). add1 V add1 The intersection point 2, which is close to the end of the compression, is (p) add2 V add2 The intersection point 3, which is close to the starting point of expansion, is (p) add3 V add3 The intersection point 4, which is close to the end of the expansion, is (p) add4 V add4 ); In step S300, the compression process index k c and the expansion process index k e The calculation formula is:

6. The method according to claim 1 or 5, characterized in that, In step S600: The three-layer BP neural network consists of two layers, which respectively predict the compression process index and the expansion process index. The input nodes of the two BP neural networks are 5, namely the standardized intake temperature and pressure, exhaust temperature and pressure, and cylinder wall temperature, respectively. The output node is 1. For the compression process index prediction model, the output node is the standardized compression process index, and for the expansion process index prediction model, the output node is the standardized expansion process index.

7. The method according to claim 5, characterized in that, In step S700: The formulas for calculating the linear equations of the predicted expansion and compression segments are as follows: For the expansion segment, (x1, y1) and (x2, y2) are the intersection points 3 (p) in step S300, respectively. add3 V add3 ) and intersection point 4 (p add4 V add4 For the compressed segment, (x1, y1) and (x2, y2) are the intersection points 1 (p) in step S300, respectively. add1 V add1 ) and intersection point 2 (p add2 V add2 ).

8. The method according to claim 7, characterized in that, Replace the logarithmic pressure values ​​under each logarithmic volume of the intake and exhaust processes with the logarithmic values ​​of the intake and exhaust pressures in the operating parameters. The intersection point 1(p) add1 V add1 The logarithmic pressure values ​​under each logarithmic volume of the process between the actual compression starting point corresponding to the maximum logarithmic volume are all replaced with the logarithmic intake pressure values ​​in the operating condition parameters, and the intersection point 4 (p) is replaced with the logarithmic intake pressure values. add4 V add4 The logarithmic pressure values ​​under each logarithmic volume of the process between the actual expansion starting point corresponding to the minimum logarithmic volume were replaced with the logarithmic exhaust pressure values ​​in the operating parameters, and finally a complete predicted logarithmic indicator diagram was obtained.

9. The method according to claim 8, characterized in that, In step S800: The formula for subtracting the natural logarithms of two pressures under the same natural logarithm of volume is as follows: Δ=ln p′ V -ln p V Where, ln p′ V To predict the natural logarithm of the pressure corresponding to a logarithmic volume of V in a logarithmic indicator diagram; ln p V It is the natural logarithm of the pressure corresponding to the logarithmic volume V in the actual logarithmic dynamometer diagram.

10. The method according to claim 3, characterized in that, The instantaneous volume V of the cylinder k The calculation formula is as follows Below: V k = x k ·π·D 2 / 4 Where, x k Indicates the instantaneous piston displacement; D represents the piston diameter; Where the instantaneous piston displacement x k The calculation formula is: Where r represents the crank radius; L represents the connecting rod length; α k α represents the instantaneous crank angle. k = k × 360 / K, where k represents the k-th data point out of K data points.

Images