A method for fault diagnosis of reciprocating compressor valves based on cylinder dynamic pressure analysis

Abstract

The present invention provides a method for diagnosing reciprocating compressor valve faults based on cylinder dynamic pressure analysis. By establishing an abnormality detection model that can be adaptively updated based on extreme value theory, it better adapts to the micro-fluctuation characteristics of industrial compressor operating conditions. The abnormality detection threshold is calculated from field data, avoiding the situation where the threshold selected based on experience causes the alarm to be too sensitive or the alarm to be untimely. It can get rid of the existing valve fault diagnosis method's high dependence on the fault data set, and realize low-cost, efficient and accurate diagnosis of reciprocating compressor valve faults.

Description

Technical Field

[0001] The invention belongs to the technical field of reciprocating compressor fault diagnosis, and in particular relates to a reciprocating compressor valve fault diagnosis method based on cylinder dynamic pressure analysis. Background Art

[0002] Reciprocating compressors are indispensable core devices in process industries. Their large pressure application range and high compression efficiency make them widely used in industries such as petroleum, chemical industry, machinery, and metallurgy. However, due to the complex structure and many moving parts of reciprocating compressors, the probability of failure is very high during field applications. Among them, the gas valve is one of the parts with the highest failure rate in reciprocating compressors. It controls the intake and discharge of gas and is an indispensable core component for reciprocating compressors to complete the working cycle. Gas valve failure can lead to pressure ratio imbalance, increased exhaust temperature, reduced exhaust volume, decreased efficiency, etc. The valve plate falling into the cylinder may even roughen the cylinder and cause the unit to be scrapped. Therefore, timely and effective detection of gas valve failure is of great significance to ensure the normal operation of reciprocating compressors.

[0003] Since it is difficult to directly reflect the valve failure by monitoring a certain parameter, the current diagnostic method for valves in industrial production mainly relies on manual inspections, so there are safety hazards such as untimely fault detection and fault deterioration. The diagnosis of reciprocating compressor valve failure mainly needs to solve two problems: one is to distinguish the valve failure state from the normal state through a certain parameter; the other is to identify the valve failure type based on the fault state. Existing research on valve fault diagnosis has also revolved around the above two aspects. The main method focuses on fault identification with the help of artificial intelligence, but it relies on the accumulation of various types of fault data. Since fault data is difficult to collect in industrial production, the implementation and application of this type of method is limited.

[0004] In the Chinese invention patent with application number 201610427413.7, a "compressor valve fault diagnosis method based on Bayesian network" is disclosed. It applies the Bayesian network model to valve fault diagnosis, and establishes a fault diagnosis BN model by preprocessing and feature extraction of the diagnostic signals of the valve under normal and various fault conditions. The method is verified to be effective under information uncertainty conditions through application examples. However, on the one hand, this method depends on the fault data set, and on the other hand, the reciprocating compressor itself has the characteristics of large vibration and non-stationarity, so the vibration signal signal-to-noise ratio is low. Therefore, this method has certain limitations. In addition, there are also literatures that propose a quantitative description method of the indicator diagram characteristics based on the geometric properties of plane figures, and extract the geometric property parameters of the indicator diagram under four common states of the reciprocating compressor valve as feature vectors, and use BP neural network as a classifier for training and testing to verify the effectiveness of the proposed method, but this method also depends on the fault data set. Summary of the invention

[0005] Based on the above-mentioned technical status, the purpose of the present invention is to provide a reciprocating compressor valve fault diagnosis method based on cylinder dynamic pressure analysis. With the help of this method, the existing valve fault diagnosis method can get rid of the high dependence of the fault data set, and realize low-cost, efficient and accurate diagnosis of reciprocating compressor valve faults.

[0006] The technical solution adopted by the present invention is as follows: A method for diagnosing a reciprocating compressor valve fault based on cylinder dynamic pressure analysis comprises the following steps:

[0007] S100: Taking the moment when the piston is at the top dead center as the cylinder dynamic pressure sampling starting point, collecting cylinder dynamic pressure signals of N groups of reciprocating compressors in normal working state during a complete operation cycle by triggering sampling with a data collector, and generating a reciprocating compressor indicator diagram to obtain N indicator diagrams;

[0008] S200: performing corresponding point averaging on the N groups of cylinder dynamic pressure signals under normal working conditions obtained in step S100 to obtain a group of averaged cylinder dynamic pressure signals under normal conditions, and generating an averaged dynamometer diagram of the reciprocating compressor based on the averaged cylinder dynamic pressure signals under normal conditions;

[0009] S300: Calculate the area of ​​the closed curve enclosed by the N dynamometer diagrams obtained in step S100 to obtain indicated power data of N reciprocating compressors;

[0010] S400: Calculate the indicated power of the averaged dynamometer diagram obtained in step S200, and sequentially subtract the indicated power value from the N indicated power data in step S300, and take absolute values ​​of all the differences to obtain N indicated power changes;

[0011] S500: Using the N indicated power changes obtained in step S400 as input, establish an abnormality judgment threshold z q Adaptively updated anomaly detection model;

[0012] S600: Based on the abnormality detection model, an abnormality judgment is performed on the cylinder dynamic pressure signal in an unknown state: first, an indicator diagram is drawn based on the cylinder dynamic pressure signal in an unknown state, and then the indicated power is calculated, and the indicated power is subtracted from the indicated power of the aforementioned averaged indicator diagram and the absolute value is taken to obtain the indicated power change. Finally, the indicated power change is input into the abnormality detection model. When the indicated power change exceeds the abnormality judgment threshold value z defined by the abnormality detection model, q If the value exceeds the specified threshold, it is considered to be abnormal; if the value does not exceed the specified threshold, it is considered to be normal.

[0013] S700: The cylinder dynamic pressure curve judged as abnormal and the averaged cylinder dynamic pressure curve under normal conditions are segmented according to the suction and exhaust pressures, and divided into four stages: expansion stage, intake stage, compression stage and exhaust stage, and compared, and the difference between each stage is calculated, and the averaged expansion deviation coefficient of the expansion stage, the averaged compression deviation coefficient of the compression stage, the intake length deviation coefficient and the intake resistance deviation coefficient of the intake stage, and the exhaust length deviation coefficient and the exhaust resistance deviation coefficient of the exhaust stage are compared with the limit values ​​to determine the fault type. Specifically, the valve fault type can be represented in code form according to certain coding rules, and then a fault code is generated according to the comparison value of the above coefficients with the limit values, and the valve fault is diagnosed according to the correspondence between various valve faults and fault codes.

[0014] The present invention has the following advantages: the present invention realizes the diagnosis of valve fault based on the analysis of the dynamic pressure signal of the cylinder. For a double-acting compressor, only two dynamic pressure sensors and one key phase sensor are required. Compared with the vibration analysis method of installing a vibration sensor on each valve, the hardware investment cost is greatly reduced.

[0015] In addition, the anomaly detection model based on extreme value theory can realize the adaptive update of the anomaly detection threshold, which better adapts to the micro-fluctuation characteristics of industrial compressor working conditions. The anomaly detection threshold is calculated from field data, avoiding the threshold selected based on experience, which may lead to overly sensitive alarms or untimely alarms. The process design combines anomaly detection with fault analysis, and only performs fault analysis calculations when an abnormality is detected, reducing unnecessary computing expenses. More importantly, the fault diagnosis method of the present invention can realize fault classification in the absence of a fault data set. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 is a flow chart of the method of the present invention;

[0017] Figure 2 is a work diagram of a reciprocating compressor in the method of the present invention;

[0018] Figure 3 It is a schematic diagram of a dynamic pressure curve of a reciprocating compressor cylinder divided into sections according to suction and exhaust pressures in the method of the present invention;

[0019] Figure 4 It is a comparison chart of the cylinder dynamic pressure curve under normal state and various fault states, among which a is the comparison between normal state and intake valve leakage fault, b is the comparison between normal state and exhaust valve leakage fault, c is the comparison between normal state and intake valve spring stiffness fault, d is the comparison between normal state and exhaust valve spring stiffness fault. DETAILED DESCRIPTION

[0020] At present, reciprocating compressor valve failures mainly rely on two types of methods: manual inspection or intelligent fault identification that is highly dependent on fault data sets. Manual inspection has safety hazards such as untimely fault discovery and fault deterioration, while fault data sets require a large amount of data and high hardware performance requirements, which increases investment costs. Therefore, the present invention provides a reciprocating compressor valve fault diagnosis method based on cylinder dynamic pressure analysis, which belongs to intelligent fault identification. However, with the help of the method of the present invention, the high dependence on fault data sets can be avoided, and low-cost, efficient and accurate diagnosis of reciprocating compressor valve failures can be achieved.

[0021] See also Figure 1 , is a flow chart of the method of the present invention. The reciprocating compressor valve fault diagnosis method based on cylinder dynamic pressure analysis of the present invention can be summarized into three modules: data preprocessing, anomaly detection, and fault analysis.

[0022] In the data preprocessing module, multiple groups of reciprocating compressor indicator diagrams are generated based on the instantaneous cylinder volume and the cylinder dynamic pressure signal under the normal working state of the gas valve, and the indicated power of each indicator diagram is calculated. By performing the subtraction and taking the absolute value operation with the indicated power of the averaged indicator diagram, the indicated power change of each cylinder dynamic pressure signal is finally obtained, which specifically includes the following steps:

[0023] Step S100: When the piston is at the top dead center of the cylinder, align the photoelectric sensor probe with the reflective strip installed on the flywheel. Each time the piston reaches the top dead center, the photoelectric sensor will output a pulse signal. The dynamic pressure of the cylinder is measured by the dynamic pressure sensor. The signal is collected by triggering sampling with a data collector. The dynamic pressure signals of the cylinder under normal working conditions of N operating cycles are collected. m data points are collected in each operating cycle, and the following is generated: Figure 2 The reciprocating compressor indicator diagram is shown, and N indicator diagrams are obtained;

[0024] In this step, the sampling frequency of the data logger is set to:

[0025] f s =(m·c n ) / 60

[0026] Among them, f s represents the sampling frequency, m represents the number of data points, c n Indicates rotation speed, rpm.

[0027] The dynamic pressure signal of the cylinder under normal working condition of group N is expressed as:

[0028]

[0029] in,[·] nRepresents the nth group of data in N groups of cylinder dynamic pressure signals under normal working conditions; N represents the total number of data groups; Represents the i-th data in the n-th group of data.

[0030] When generating the indicator diagram for a reciprocating compressor, the instantaneous cylinder volume is calculated using the following formula:

[0031] V i =x i ·π·D 2 / 4

[0032] Among them, V i represents the instantaneous cylinder volume, x i represents the instantaneous piston displacement; D represents the piston diameter. i The calculation is done using the following formula:

[0033]

[0034] Where r is the crank radius; L is the connecting rod length; α i represents the instantaneous crank angle, α i =i×360 / m.

[0035] In addition, for those skilled in the art, triggered sampling is a data collection method of the data collector, which will only collect data when it receives a specified trigger signal, otherwise it will keep waiting. Specifically in this embodiment, the trigger source of the triggered sampling is the rising edge of the pulse signal output by the photoelectric sensor. Through triggered sampling, a full-cycle dynamic pressure signal with the start time of the expansion phase during the operation of the reciprocating compressor as the starting point can be obtained.

[0036] Step S200: average the cylinder dynamic pressure signals under normal working conditions of N operating cycles in step S100 at corresponding points to obtain a set of averaged cylinder dynamic pressure signals under normal conditions, and generate an averaged indicator diagram of the corresponding reciprocating compressor. In this step:

[0037] The specific formula for averaging the cylinder dynamic pressure signal under normal conditions is:

[0038]

[0039] in, Represents the i-th data in the averaged cylinder dynamic pressure signal data array under normal conditions.

[0040] Furthermore, the averaged cylinder dynamic pressure signal under normal conditions is expressed as:

[0041]

[0042] Step S300: Based on the N indicator diagrams obtained in step S100, the area of ​​the closed curve enclosed by the N indicator diagrams is calculated. The physical meaning of the area is the indicated power of the reciprocating compressor, that is, N indicated power data are obtained. In this step:

[0043] The calculation formula of the indicated power of the reciprocating compressor is:

[0044] n is [1, N]

[0045] Further, the N indicated powers are expressed as:

[0046] W=[N1,N2,…,N n ,…,N N ]

[0047] Step S400: Calculate the indicated power of the averaged dynamometer diagram in step S200 by using the area method, and sequentially subtract the indicated power value from N indicated power data, take the absolute value of all the differences, and obtain N indicated power changes;

[0048] In this step, the indicated power of the averaged dynamometer diagram is expressed as N mean ; The indicated power change calculation formula is:

[0049] ΔN n =|N mean -N n |

[0050] Furthermore, the N indicated power changes are expressed as:

[0051] ΔW=[ΔN1,ΔN2,…,ΔN n ,…ΔN N ]

[0052] The indicator diagram of a reciprocating compressor is an important analytical diagram that reflects the working status of the gas valve. The change in its shape relative to the normal state means the occurrence of a fault. The indicated power of a reciprocating compressor is calculated based on the area of ​​the closed curve enclosed by the indicator diagram. Therefore, the change in the indicated power can be used to reflect whether there is an abnormality in the gas valve.

[0053] In the abnormality detection module, the present invention uses the method based on extreme value theory to calculate the abnormality judgment threshold value with the indicated power change as input to judge whether the current cylinder dynamic pressure is normal, which specifically includes the following steps (sequence number continues from the previous one):

[0054] Step S500: Using the N indicated power changes obtained in step S400 as input, establish an abnormality judgment threshold z q Adaptively updated anomaly detection model;

[0055] In this step, the abnormality judgment threshold z q The calculation formula is:

[0056]

[0057] Among them, t is the initialization threshold, which is the 98% quantile of the observed data; q is the abnormal probability; n is the number of observed data; N t is the number of observations beyond t; For the part of the observed data that exceeds t, the formula The estimated values ​​of σ and γ calculated using the maximum likelihood estimation method when fitting, where x n >t,x n =ΔN n -t;

[0058] Step S600: Based on the abnormality detection model, the cylinder dynamic pressure signal in the unknown state is judged as abnormal: first, the indicator diagram is drawn based on the cylinder dynamic pressure signal in the unknown state, and then the indicated power is calculated, and the indicated power is subtracted from the indicated power of the averaged indicator diagram and the absolute value is taken to obtain the indicated power change. Finally, the indicated power change is input into the abnormality detection model. When the indicated power change exceeds the abnormality judgment threshold value z defined by the abnormality detection model, the abnormality detection model is used to judge the abnormality. q If it exceeds the specified threshold, it is considered to be normal.

[0059] In this step, the cylinder dynamic pressure signal of the unknown state is expressed as: P′=[p′1, p′2, …, p′ i ,…,p′ m ]; the indicated power calculated from the cylinder dynamic pressure signal of the unknown state is expressed as N′.

[0060] Furthermore, the indicated power change calculation formula is:

[0061] ΔN′=|N mean -N′|

[0062] Furthermore, the indicated power change ΔN′ is related to the abnormality judgment threshold z in step S500. q and the initialization threshold t: when ΔN′<t, the current state is normal; when ΔN′>z q When t≤ΔN′≤z q When , the current ΔN′ is inserted into ΔW, and z in step S500 is recalculated. q , that is, to update the abnormal judgment threshold.

[0063] The fault analysis module is used to further analyze the abnormal data of cylinder dynamic pressure when the above judgment is abnormal, calculate a series of deviation coefficients to generate fault codes according to the changes of cylinder dynamic pressure in suction, compression, exhaust and expansion caused by various valve faults, and determine the fault type with reference to the fault code table, which specifically includes the following steps:

[0064] Step S700: Divide the cylinder dynamic pressure curve judged to be abnormal and the averaged cylinder dynamic pressure curve under normal conditions into segments according to the suction and exhaust pressures, and compare them, calculate the differences between each segment, and perform fault analysis based on changes in the averaged expansion deviation coefficient of the expansion segment, the averaged compression deviation coefficient of the compression segment, the suction length deviation coefficient and the suction resistance deviation coefficient of the suction segment, and the exhaust length deviation coefficient and the exhaust resistance deviation coefficient of the exhaust segment.

[0065] In this step, the pressure values ​​of the expansion section, the intake section, the compression section, and the exhaust section of the abnormal cylinder dynamic pressure signal are respectively expressed as:

[0066] Expansion section: P e ′=[p′1, p′2,…,p′ a ]

[0067] Inspiratory stage: P i ′=[p′ a+1 , p′ a+2 ,…,p′ b ]

[0068] Compression section: P c ′=[p′ b+1 , p′ b+2 ,…,p′ k ]

[0069] Exhaust section: P o ′=[p k ' +1 , p k ' +2 ,…,p d ′]

[0070] Among them, the subscripts a, b, k, and d represent the position index in the cylinder dynamic pressure signal data array.

[0071] The averaged normal cylinder dynamic pressure signal expansion section, intake section, compression section, and exhaust section pressure values ​​are respectively expressed as:

[0072] Expansion section:

[0073] Inspiratory stage:

[0074] Compression section:

[0075] Exhaust section:

[0076] Among them, the subscripts j, q, v, and d represent the position index in the cylinder dynamic pressure signal data array.

[0077] The specific calculation formula of the homogenized expansion deviation coefficient of the expansion section is:

[0078]

[0079] Among them, min() means taking the minimum value; p s Indicates the suction pressure of the cylinder at this level; p d Indicates the exhaust pressure of the cylinder at this level.

[0080] The specific calculation formula of the averaged compression deviation coefficient of the compression section is:

[0081]

[0082] Among them, max() means taking the maximum value.

[0083] The specific calculation formula of the deviation coefficient of the suction length of the suction section is:

[0084]

[0085] The specific calculation formula of the inhalation resistance deviation coefficient of the inhalation section is:

[0086]

[0087] The specific calculation formula of the exhaust length deviation coefficient of the exhaust section is:

[0088]

[0089] The specific calculation formula of the exhaust resistance deviation coefficient of the exhaust section is:

[0090]

[0091] In addition, those skilled in the art can understand that the cylinder dynamic pressure curve can be divided into four stages: expansion stage, suction stage, compression stage and exhaust stage according to the suction pressure and exhaust pressure of the reciprocating compressor. Figure 3 The figure shows a schematic diagram of a dynamic pressure curve of a reciprocating compressor cylinder divided into sections according to the suction and exhaust pressures in the method of the present invention. Figure 3A specific embodiment is used to illustrate the segmentation of the dynamic pressure curve of the cylinder of a reciprocating compressor. From the left to the right on the horizontal axis, the first section of the curve between the suction pressure and the exhaust pressure is the expansion section, the second section of the curve below the suction line is the suction section, the third section of the curve between the suction pressure and the exhaust pressure is the compression section, and the fourth section of the curve above the exhaust pressure is the exhaust section. Further, the averaged expansion deviation coefficient of the expansion section, the averaged compression deviation coefficient of the compression section, the suction length deviation coefficient and the suction resistance deviation coefficient of the suction section, and the exhaust length deviation coefficient and the exhaust resistance deviation coefficient of the exhaust section are calculated, and the limit values ​​of the averaged expansion deviation coefficient and the averaged compression deviation coefficient are specified to be λ1, the limit values ​​of the suction length deviation coefficient and the exhaust length deviation coefficient of the exhaust section are λ2, and the limit values ​​of the suction resistance deviation coefficient and the exhaust resistance deviation coefficient are λ3. Furthermore, a fault code is compiled, and the coding rule is: composed of numbers (1-6) and letters (A, B, C), a single number and a single letter constitute a group, and each code is composed of a combination of 6 groups of numbers and letters. Number 1 represents the averaged expansion deviation coefficient, number 2 represents the averaged compression deviation coefficient, number 3 represents the intake length deviation coefficient, number 4 represents the intake resistance deviation coefficient, number 5 represents the exhaust length deviation coefficient, and number 6 represents the exhaust resistance deviation coefficient. Letter A represents that the deviation coefficient is positive and greater than the limit value, letter B represents that the deviation coefficient is negative and the absolute value is greater than the limit value, and letter C represents that the absolute value of the deviation coefficient does not exceed the limit value. According to the coding rule and combined with the characteristics of various types of valve faults, a fault code table compiled is shown in Table 2. The valve fault type can be determined based on this table.

[0092] Table 2 Fault code table

[0093] Fault type Fault Codes Suction valve leakage 1B2B3A4A5B6C High suction spring stiffness 1C2C3C4A5C6C Exhaust valve leakage 1A2A3B4C5A6A Exhaust spring stiffness is high 1C2C3C4C5C6A Other faults Other codes

[0094] As a specific application example, Figure 4 The comparison diagram of the cylinder dynamic pressure curve under normal state and various fault states is shown, where a is the comparison between the normal state and the intake valve leakage fault, b is the comparison between the normal state and the exhaust valve leakage fault, c is the comparison between the normal state and the intake valve spring stiffness fault, and d is the comparison between the normal state and the exhaust valve spring stiffness fault. The specific values ​​of the indicated power change, averaged expansion deviation coefficient, averaged compression deviation coefficient, intake length deviation coefficient and intake resistance deviation coefficient, exhaust length deviation coefficient and exhaust resistance deviation coefficient of the cylinder dynamic pressure curve under various fault states relative to the normal state are shown in Table 3.

[0095] Table 3 Calculation results data table

[0096]

[0097] It can be seen from the data in Table 3 that the indicated power changes under various fault conditions of the valve are large. Therefore, it is scientific and effective to detect the abnormal state of the valve based on the indicated power change. The limit values ​​of the averaged expansion deviation coefficient and the averaged compression deviation coefficient are set to λ1=10%, the limit values ​​of the intake length deviation coefficient and the exhaust length deviation coefficient of the exhaust section are set to λ2=5%, and the limit values ​​of the intake resistance deviation coefficient and the exhaust resistance deviation coefficient are set to λ3=20%. The intake valve leakage fault code is 1B2B3A4A5B6C, the exhaust valve leakage fault code is 1A2A3B4C5A6A, the intake spring stiffness fault code is: 1C2C3C4A5C6C, and the exhaust spring stiffness fault code is 1C2C3C4C5C6A, which is consistent with the fault type and fault code given in the fault code table in Table 2, verifying the feasibility and effectiveness of the fault analysis method.

Claims

1. A reciprocating compressor valve fault diagnosis method based on cylinder dynamic pressure analysis, characterized in that: The steps include: S100: Taking the moment when the piston is at the top dead center as the cylinder dynamic pressure sampling starting point, collecting cylinder dynamic pressure signals of N groups of reciprocating compressors in normal working state during a complete operation cycle by triggering sampling with a data collector, and generating a reciprocating compressor indicator diagram to obtain N indicator diagrams; S200: performing corresponding point averaging on the N groups of cylinder dynamic pressure signals under normal working conditions obtained in step S100 to obtain a group of averaged cylinder dynamic pressure signals under normal conditions, and generating an averaged dynamometer diagram of the reciprocating compressor based on the averaged cylinder dynamic pressure signals under normal conditions; S300: Calculate the area of ​​the closed curve enclosed by the N dynamometer diagrams obtained in step S100 to obtain indicated power data of N reciprocating compressors; S400: Calculate the indicated power of the averaged dynamometer diagram obtained in step S200, and sequentially subtract the indicated power value from the N indicated power data in step S300, and take absolute values ​​of all the differences to obtain N indicated power changes; S500: Using the N indicated power changes obtained in step S400 as input, establish an abnormality judgment threshold z q Adaptively updated anomaly detection model; S600: Performing an abnormality evaluation on the cylinder dynamic pressure signal in an unknown state based on the abnormality detection model to determine whether there is an abnormality; S700: The cylinder dynamic pressure curve judged to be abnormal and the averaged cylinder dynamic pressure curve under normal conditions are segmented according to the intake and exhaust pressures, and divided into four stages: expansion stage, intake stage, compression stage and exhaust stage, and compared, and the difference between each stage is calculated. The averaged expansion deviation coefficient of the expansion stage, the averaged compression deviation coefficient of the compression stage, the intake length deviation coefficient and the intake resistance deviation coefficient of the intake stage, and the exhaust length deviation coefficient and the exhaust resistance deviation coefficient of the exhaust stage are compared with the limit values ​​to determine the type of fault.

2. The method according to claim 1, further characterized in that: In step S100: The sampling frequency of the data logger is set to: f s =(m·c n ) / 60 Among them, f s represents the sampling frequency, m represents the number of data points, c n Indicates the rotation speed, rpm; The dynamic pressure signal of the cylinder under normal working condition of group N is expressed as: in,[·] n Represents the nth group of data in N groups of cylinder dynamic pressure signals under normal working conditions; N represents the total number of data groups; Represents the i-th data in the n-th group of data.

3. The method according to claim 1, further characterized in that: In step S200: The specific formula for averaging the cylinder dynamic pressure signal under normal conditions is: in, represents the i-th data in the averaged cylinder dynamic pressure signal data array under normal conditions, and m represents the number of data points; The averaged cylinder dynamic pressure signal under normal conditions is expressed as:

4. The method according to claim 1, further characterized in that: In step S300: The indicated power data of N reciprocating compressors is expressed as: W=[N1,N2,…,N n ,…,N N ] The indicated power of the reciprocating compressor is N n The calculation formula is: n is [1, N], c n represents the speed, D represents the piston diameter, represents the i-th data in the n-th group of data, x i Represents the instantaneous piston displacement corresponding to the i-th data.

5. The method according to claim 1, further characterized in that: In step S400: N indicated power changes are expressed as: ΔW=[ΔN1,ΔN2,…,ΔN n ,…ΔN N ] △N n It represents the nth indicated power change, and its calculation formula is: ΔN n =|N mean -N n |, where N mean Indicates the indicated power of the averaged dynamometer diagram, N n Indicates the indicated power of the reciprocating compressor.

6. The method according to claim 1, further characterized in that: In step S500: The abnormality judgment threshold z q The calculation formula is: Among them, t is the initialization threshold, which is the 98% quantile of the observed data; q is the abnormal probability; n is the number of observed data; N t is the number of observations beyond t; For the part of the observed data that exceeds t, the formula The estimated values ​​of σ and γ calculated using the maximum likelihood estimation method when fitting, where x n >t,x n =ΔN n -t; △N n Indicates the nth indicated power change.

7. The method according to claim 6, further characterized in that: In step S600, the method for judging abnormality of the cylinder dynamic pressure signal in an unknown state is as follows: First, the indicator diagram is drawn based on the cylinder dynamic pressure signal in the unknown state, and then the indicated power is calculated. The indicated power is subtracted from the indicated power of the averaged indicator diagram and the absolute value is taken to obtain the corresponding indicated power change. Finally, the indicated power change is input into the anomaly detection model. When the indicated power change exceeds the abnormal judgment threshold z defined by the anomaly detection model, q If it exceeds the specified threshold, it is considered to be normal.

8. The method according to claim 7, further characterized in that: The cylinder dynamic pressure signal of the unknown state is expressed as: P′=[p′1, p′2, …, p′ i ,…,p′ m ]; the indicated power calculated from the cylinder dynamic pressure signal of unknown state is expressed as N′; The calculation formula for the indicated power change corresponding to the unknown cylinder dynamic pressure signal is: ΔN′=|N mean -N′| The indicated power change ΔN′ is compared with the abnormality judgment threshold z in step S500. q and the initialization threshold t: when ΔN′<t, the current state is normal; when ΔN′>z q When t≤ΔN′≤z q When , the current ΔN′ is inserted into ΔW, and z in step S500 is recalculated. q , that is, to update the abnormal judgment threshold.

9. The method according to claim 1, further characterized in that: In step S700: The abnormal cylinder dynamic pressure signal expansion section, suction section, compression section, and exhaust section pressure values ​​are expressed as: Expansion section: P e ′=[p′1, p′2,…,p′ a ] Inspiratory stage: P i ′=[p′ a+1 , p′ a+2 ,…,p′ b ] Compression section: P c ′=[p′ b+1 , p′ b+2 ,…,p′ k ] Exhaust section: P o ′=[p′ k+1 , p′ k+2 ,…,p′ d ] Wherein, the subscripts a, b, k, and d represent the position index in the cylinder dynamic pressure signal data array; The averaged pressure values ​​of the expansion section, intake section, compression section, and exhaust section of the cylinder dynamic pressure signal under normal conditions are expressed as: Expansion section: Inspiratory stage: Compression section: Exhaust section: Wherein, the subscripts j, q, v, and d represent the position index in the cylinder dynamic pressure signal data array; The specific calculation formula of the homogenized expansion deviation coefficient of the expansion section is: Among them, min() means taking the minimum value; p s Indicates the suction pressure of the cylinder at this level; p d Indicates the exhaust pressure of the cylinder at this level; The specific calculation formula of the averaged compression deviation coefficient of the compression section is: Among them, max() means taking the maximum value; The specific calculation formula of the deviation coefficient of the suction length of the suction section is: The specific calculation formula of the inhalation resistance deviation coefficient of the inhalation section is: The specific calculation formula of the exhaust length deviation coefficient of the exhaust section is: The specific calculation formula of the exhaust resistance deviation coefficient of the exhaust section is:

10. The method according to claim 1, further characterized in that: Generate a fault code according to step S700, and diagnose the gas valve fault according to the corresponding relationship between various gas valve faults and the fault codes; The coding rule of the fault code is: composed of numbers 1-6 and letters A, B, and C, a single number and a single letter constitute a group, and each code is composed of a combination of 6 groups of numbers and letters. Number 1 represents the averaged expansion deviation coefficient, number 2 represents the averaged compression deviation coefficient, number 3 represents the intake length deviation coefficient, number 4 represents the intake resistance deviation coefficient, number 5 represents the exhaust length deviation coefficient, and number 6 represents the exhaust resistance deviation coefficient. Letter A represents that the deviation coefficient is positive and greater than the limit value, letter B represents that the deviation coefficient is negative and the absolute value is greater than the limit value, and letter C represents that the absolute value of the deviation coefficient does not exceed the limit value. According to the coding rule and the corresponding relationship between the following fault code and valve fault, the valve fault type can be determined: intake valve leakage: 1B2B3A4A5B6C; intake spring stiffness is large: 1C2C3C4A5C6C; exhaust valve leakage: 1A2A3B4C5A6A; exhaust spring stiffness is large: 1C2C3C4C5C6A; other faults: other codes.

Images