Reciprocating compressor air valve fault identification method

Abstract

The invention belongs to the technical field of compressors, and particularly relates to a reciprocating compressor air valve fault recognition method. The air valve serves as an important component of the reciprocating compressor, the fault occurrence rate of the air valve is relatively high, and the fault problem of the air valve is more prominent especially in the ultrahigh-pressure reciprocating compressor. Existing fault judgment is mostly limited to process parameter and sound analysis, the analysis method is complex, the requirement for professional experience of detection personnel is high, and engineering practice is not widely applied. The invention provides a reciprocating compressor air valve fault detection method, and the method comprises the steps: obtaining the temperature of an air valve cover at the same side of a cylinder according to the temperature data distribution and trend characteristics; according to the method, data are processed, evaluation indexes are constructed according to the similarity of the temperatures of valve covers of the air valves, valve faults are recognized according to the characteristic distribution rule of normal air valve temperature data, and a reference method easy for engineering practice is provided for early-stage rapid recognition of the faults of the air valves of the reciprocating compressor.

Description

Technical Field

[0001] The invention relates to the technical field of gas valve fault diagnosis of reciprocating compressors, in particular to a method for identifying gas valve faults of reciprocating compressors. Background Art

[0002] Reciprocating compressors are widely used in the industrial field. The core component of reciprocating compressors is the valve, which is responsible for periodically controlling the intake and exhaust of gas, directly affecting the efficiency, energy consumption and operational stability of the compressor. At present, valve fault diagnosis is mainly based on signal monitoring and state parameter analysis. Common methods include: (1) vibration signal analysis: using an acceleration sensor to collect valve vibration signals, analyze the impact force change characteristics, and identify valve plate breakage or spring failure; (2) pressure dynamometer method: measuring the pressure change curve in the cylinder. If the intake valve leaks, the expansion process curve will move downward, and the exhaust valve failure will manifest as an abnormal compression process; (3) temperature and pressure monitoring: real-time monitoring of interstage pressure, exhaust temperature and valve cover temperature. Leakage failure is often accompanied by abnormal temperature rise and pressure fluctuation; (4) lubricating oil analysis: detecting the metal particle content in the lubricating oil to indirectly determine the degree of valve wear; (5) acoustic and displacement detection: using noise signals or displacement sensors to capture the valve plate motion trajectory deviation and evaluate the real-time status of the valve action.

[0003] Despite the abundance of existing technologies, the following problems still exist in practical applications: theoretical research is based on ideal situations, while actual field conditions are complex. For example, vibration signals are easily affected by noise from other mechanical components, and the complexity of high-precision sensor installation and signal processing algorithms limits field applications. These factors make it difficult for the above-mentioned commonly used methods to guarantee diagnostic accuracy. Summary of the Invention

[0004] In order to overcome the defect in the above-mentioned prior art that the accuracy of valve failure is affected by complex situations and it is difficult to ensure accuracy, the present invention proposes a method for detecting abnormal temperature data of reciprocating compressor valves, and constructs a comprehensive evaluation index for early fault judgment of valves based on the analysis of the temperature monitoring data of the valve cover of the reciprocating compressor; the present invention is based on the existing valve monitoring equipment, explores its distribution regularity, and realizes an efficient and non-invasive valve fault identification method, which is of great significance to improving the operating reliability of the compressor and reducing the maintenance cost.

[0005] The present invention provides a method for identifying a reciprocating compressor valve fault, comprising the following steps:

[0006] St1. Synchronously collect the valve cover temperature data of the same cylinder and the same side of the reciprocating compressor and align them in time to construct the temperature time series data of each valve;

[0007] St2. Construct temperature time series data for each valve and calculate the similarity index and consistency index of the temperature time series data of each pair of valves;

[0008] St3. For each pair of valves, calculate the comprehensive weight of the similarity index and consistency index as the comprehensive evaluation index s(p,q);

[0009] St4. If all comprehensive evaluation indicators s(p,q) corresponding to the valve p are less than the set threshold, it is determined that the valve p is in a fault state.

[0010] Preferred:

[0011]

[0012] Wherein, w represents the weight coefficient; d(p,q) represents the similarity index of valve p and valve q; ρ(p,q) represents the consistency index of valve p and valve q.

[0013] Preferably, the consistency index adopts the correlation coefficient;

[0014]

[0015] Among them, Tp n Indicates the valve cover temperature of the upper air valve p at time n, Tq n represents the valve cover temperature of the upper gas valve q at time n, and N represents the time series length of the temperature time series data; is the mean value of the temperature time series data T(p), is the mean value of the temperature time series data T(q).

[0016] Preferably, the weight coefficient w takes a value in the interval [0.3, 0.5].

[0017] Preferably, the similarity index d(p,q) adopts Fréchet distance.

[0018] Preferred:

[0019] Let D[i][j]=max(d(Tp i ,Tq j ),min(D[i-1][j],D[i][j-1],D[i-1][j-1]))

[0020] d(Tp i ,Tq j )=|Tp i -Tq j |

[0021] D[0][0]=0

[0022] D[i][0]=∞

[0023] D[0][j]=∞

[0024] i and j represent the time, 1≤i≤N, 1≤j≤N, N represents the length of the temperature time series data; Tp i is the valve cover temperature of valve p at time i, Tq j is the valve cover temperature of valve q at time j; d(Tp i ,Tq j ) represents Tp i and Tq j the distance between them;

[0025] The Fréchet distance between valve p and valve q = D[N][N].

[0026] Preferably, in step St1, the temperature data of the valve covers on the same side of the same cylinder of the reciprocating compressor are first collected and abnormal value processing is performed; then the temperature data of each valve cover are aligned in time points, and the temperature time series data of the valve is constructed.

[0027] The present invention proposes a reciprocating compressor valve fault identification device, which includes a data receiving module, a data preprocessing module, a time alignment module and a comprehensive evaluation module;

[0028] The data receiving module is used to receive the valve cover temperature of the gas valve on the same side of the same cylinder of the reciprocating compressor collected by the monitoring equipment;

[0029] The data preprocessing module is used to preprocess the valve cover temperature time series data of each gas valve, including outlier deletion and missing value interpolation;

[0030] The time alignment module is used to perform time alignment sampling on the temperature time series data of the valve covers after preprocessing of different gas valves;

[0031] The comprehensive evaluation module is used to calculate the correlation index and consistency index of the temperature time series data between different gas valves, and calculate the comprehensive evaluation index.

[0032] The present invention proposes a reciprocating compressor valve fault identification system, which includes a memory and a processor. The memory stores a computer program, and the processor is connected to the memory. The processor is used to execute the computer program to implement the reciprocating compressor valve fault identification method.

[0033] The present invention provides a storage medium storing a computer program, which is used to implement the method for identifying a reciprocating compressor valve fault when the computer program is executed.

[0034] The advantages of the present invention are:

[0035] (1) Compared with the temperature distribution characteristics of the suction valve cover in the normal operating state, the temperature data distribution of the valve cover in the abnormal condition is significantly different. The application analyzes the similarity and consistency of the temperature distribution of the data of the same side valve of a single cylinder, identifies the abnormal valve cover temperature according to the difference in the data distribution trend, and judges the fault of the valve, so that the monitoring and diagnosis of the reciprocating compressor valve fault are more real-time and have stronger robustness, and an easy-to-engineer method is provided for the identification of the valve fault.

[0036] (2) The discrete Fréchet distance mainly measures the similarity of the temperature data distribution trend, considers the shape and order of the temperature change, and reflects the difference in the change trend of the data. The smaller the distance is, the more similar the curve shape of the two columns of data is. The Pearson correlation coefficient mainly evaluates the temperature data change, reflects the consistency of the temperature data in the rising or falling trend, and the closer the correlation coefficient is to 1 or -1, the stronger the monotonic correlation of the two columns of data is. The discrete Fréchet distance and the Pearson correlation coefficient are combined to construct a comprehensive evaluation index, which more comprehensively reflects the similarity of the two columns of temperature data, and fully utilizes the advantages of the two methods, considering both the overall shape change of the data and the monotonic correlation of the data. The comprehensive evaluation index of the application is the sum of the partial weight of the Fréchet distance and the partial weight of the Pearson correlation coefficient. The comprehensive evaluation index is constructed by combining the use of the Fréchet distance and the Pearson correlation coefficient, and the threshold value is set according to the historical data or field knowledge to assist the identification and diagnosis of the valve fault. By combining the use of the discrete Fréchet distance and the Pearson correlation coefficient, the similarity of the two columns of temperature data is comprehensively evaluated from different angles, which can better capture the similarity of the temperature curve shape and trend and better measure the consistency of the temperature data in the monotonic trend.

[0037] (3) The application also provides a simple and fast discrete Fréchet distance calculation method, which greatly improves the calculation efficiency while ensuring the accuracy of the calculation results.

[0038] (4) In the application, the greater the comprehensive evaluation index value is, the more similar the distribution of the valve cover temperature is, and the closer the comprehensive evaluation index is to 1, the stronger the correlation of the two groups of temperature data is. Both the shape and trend similarity of the data curve (through the discrete Fréchet distance) and the monotonic correlation of the data (through the Pearson correlation coefficient) are considered, which can more comprehensively evaluate the correlation of the temperature data. BRIEF DESCRIPTION OF DRAWINGS

[0039] Figure 1 is a flowchart of the method of the application;

[0040] Figure 2This is the temperature distribution diagram of the suction valve cover of a reciprocating unit A100 within 24 hours;

[0041] Figure 3(a) shows the Fréchet distance matrix of valve cover temperature of unit A100;

[0042] Figure 3(b) shows the Pearson coefficient matrix of valve cover temperature of unit A100;

[0043] Figure 3(c) shows the comprehensive evaluation index matrix of valve cover temperature of unit A100;

[0044] Figure 4 This is the temperature distribution diagram of the suction valve cover of the reciprocating unit B100;

[0045] Figure 5(a) shows the Fréchet distance matrix of valve cover temperature of unit A100;

[0046] Figure 5(b) shows the Pearson coefficient matrix of valve cover temperature of unit A100;

[0047] Figure 5(c) shows the comprehensive evaluation index matrix of valve cover temperature of unit A100. DETAILED DESCRIPTION

[0048] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0049] like Figure 1 As shown, the present invention provides a method for identifying a reciprocating compressor valve fault, which includes the following steps.

[0050] St1. Synchronously collect the valve cover temperature data of the same cylinder and the same side of the reciprocating compressor and align them in time to construct the temperature time series data of each valve.

[0051] In the specific implementation, the valve cover temperature data of the same cylinder and the same side of the reciprocating compressor are first collected, and outlier deletion and missing value interpolation are performed.

[0052] The valve cover temperature data is obtained from relevant monitoring equipment or data recording systems. The data should contain temperature measurement values ​​and corresponding timestamp information to facilitate time alignment.

[0053] Then, the valve cover temperature data of the valve on the same side of the same cylinder of the reciprocating compressor are aligned in time points, and the temperature time series data of the valve are constructed.

[0054] Since the temperature measurement time of different valve covers may vary, the temperature data of all valve covers need to be aligned in chronological order. Through interpolation, each valve cover has a corresponding temperature value at the same time point.

[0055] Assume that a reciprocating compressor has two valves on the same side of the same cylinder, denoted as valve p and valve q. The temperature time series data corresponding to valve p is denoted as T(p), and the temperature time series data corresponding to valve q is denoted as T(q).

[0056] T(p)=[Tp1,Tp2,……,Tp n ,…,Tp N ]

[0057] T(q)=[Tq1,Tq2,……,Tq n ,…,Tq N ]

[0058] Among them, Tp n and Tq n They represent the valve cover temperature of valve p and valve q at time n, 1≤n≤N, N represents the time series length; Tp1, Tp2 and Tp N Represent the valve cover temperature of the gas valve p at time 1, 2 and N, Tq1, Tq2 and Tq N Represent the valve cover temperature of the gas valve q at time 1, 2 and N respectively.

[0059] St2. Calculate the Fréchet distance d(T(p), T(q)) and correlation coefficient ρ(p, q) of the temperature time series data of each pair of valves.

[0060] That is, in this embodiment, the Fréchet distance d(T(p), T(q)) of the temperature time series data of valves p and q is used as the similarity index d(p, q), and the correlation coefficient is used as the consistency index ρ(p, q).

[0061] The Fréchet distance of the temperature time series data of valve p and valve q is used to measure the similarity of the valve cover temperature data.

[0062] Construct the linked sequence L of the temperature data series T(p) and T(q):

[0063]

[0064] Among them, a1=b1=1, a m =b m =N; m is the length of the link sequence L, 1≤k≤m, 1≤a k ≤N,1≤b k ≤N;a1≤a2≤a3≤…≤ak ≤…≤a m ; b1≤b2≤b3≤…≤b k ≤…≤b m ; and a k+1 =a k or a k +1, b k+1 =b k or b k +1, k+1≤m;

[0065] Right now

[0066]

[0067] In this way, the series L also follows the order relationship of the endpoints in T(p) and T(q).

[0068] The essence of Fréchet distance is to find the minimum length of the linked sequence among all the regular linked sequences composed of endpoints; the length of the linked sequence is the maximum distance value of the data pairs it contains, that is, the distance values ​​of the data pairs in the linked sequence are calculated separately, and the maximum value is selected as the length of the linked sequence.

[0069] Define length ||L|| as the length of the longest connection in sequence L, that is:

[0070]

[0071] The discrete Fréchet distance between the two columns of temperature data T(p) and T(q) is defined as follows:

[0072] d(T(p), T(q))=min{||L|||L is the link sequence between T(p) and T(q)}.

[0073] Thus, when calculating the Fréchet distance d(T(p), T(q)), we first need to list all the sequences L, then calculate the length of each sequence, and then select the minimum value as the Fréchet distance.

[0074] For example, T(p) = [1, 2, 3] and T(q) = [5, 3, 6]:

[0075] Let the subscripts corresponding to T(p) in the sequence be a1=1, a2=2, a3=3, satisfying a2=a1+1, a3=a2+1, strictly increasing, and satisfying the non-decreasing rule;

[0076] The subscripts corresponding to T(q) in the sequence are b1=1, b2=2, b3=3, satisfying b2=b1+1, b3=b2+1, strictly increasing, and satisfying the non-decreasing rule;

[0077] We get: linked sequence 1 (1,5), (2,3), (3,6), whose sequence length is:

[0078] ||L||=max{|1-5|,|2-3|,|3-6|}=max{4,1,3}=4.

[0079] Let the subscripts corresponding to T(p) in the sequence be a1=1, a2=1, a3=2, a4=2, a5=3; at this time, a2=a1, a3=a2+1, a4=a3, a5=a4+1, satisfying a i+1 =a i or a i+1 =a i +1;

[0080] The subscripts corresponding to T(q) in the sequence are b1=1, b2=2, b3=2, b4=3, b5=3; at this time, b2=b1+1, b3=b2, b4=b3+1, b5=b4, satisfying b i+1 =b i or b i +1=b i+1 ;

[0081] The resulting connection sequence is (1,5), (1,3), (2,3), (2,6), (3,6), and its length is:

[0082] ||L||=max{|1-5|,|1-3|,|2-3|,|2-6|,|3-6|}=max{4,2,1,4,3}=4.

[0083] Let the subscripts corresponding to T(p) in the sequence be a1=1, a2=2, a3=3, a4=3, a5=3, satisfying a i+1 =a i or a i+1 =a i +1;

[0084] The subscripts corresponding to T(q) in the sequence are b1=1, b2=2, b3=3, b4=3, b5=3, satisfying b i+1 =b i or b i +1=b i+1 ;

[0085] The resulting connection sequence is (1,5), (1,3), (1,6), (2,6), (3,6), and its length is:

[0086] ||L||=max{|1-5|,|1-3|,|1-6|,|2-6|,|3-6|}=max{4,2,5,4,3}=5.

[0087] According to the construction rules of the link sequence L, T(p) and T(q) of length N can construct 4 N-1 In this embodiment, a total of 4 2 link sequences, and finally found that the minimum value in ||L|| is 4. Therefore, the discrete Fréchet distance is 4.

[0088] Obviously, the workload of the above-listed algorithms is extremely high. In order to improve the computational efficiency, a simpler computational method is proposed in this embodiment for computing the Fréchet distance.

[0089] First, set the matrix boundaries: D[0][0] = 0, D[i][0] = ∞, D[0][j] = ∞, i and j represent the time, 1≤i≤N, 1≤j≤N;

[0090] Recursively calculate the distance parameter D[i][j]:

[0091] D[i][j]=max{d(Tp i ,Tq j ),min(D[i-1][j],D[i][j-1],D[i-1][j-1])}

[0092] d(Tp i ,Tq j )=|Tp i -Tq j |

[0093] Among them, D[i-1][j-1], D[i-1][j] and D[i][j-1] are distance parameters, Tp i is the valve cover temperature of valve p at time i, Tq j is the valve cover temperature of valve q at time j; d(Tp i ,Tq j ) represents Tp i and Tq j the distance between them;

[0094] The calculated result D[N][N] of D[i][j] is obtained as the Fréchet distance between valve p and valve q.

[0095] In the above embodiment where T(p) = [1, 2, 3] and T(q) = [5, 3, 6], the recursive formula D[i][j] is used for calculation, and the results are as follows:

[0096] D[3][3]=max{d(Tp3,Tq3),min(D[2][3],D[3][2],D[2][2])}

[0097] d(Tp3,Tq3)=|3-6|=3

[0098] D[2][3]=max{d(Tp2,Tq3),min(D[1][3],D[2][1],D[1][2])}

[0099] D[1][3]=max{d(Tp1,Tq3),min(D[0][3],D[1][2],D[0][1])}=max{5,min(∞,D[1][2],∞)}

[0100] D[2][1]=max{d(Tp2,Tq1),min(D[1][1],D[2][0],D[1][0])}=max{3,min(D[1][1],∞,∞)}

[0101] D[1][2]=max{d(Tp1,Tq2),min(D[0][2],D[1][1],D[0][1])}=max{2,min(∞,D[1][1],∞)}

[0102] D[1][1]=max{d(Tp1,Tq1),min(D[0][1],D[1][0],D[0][0])}=max{4,min(∞,∞,0)}=4

[0103] English:D[1][2]=4,D[2][1]=4,D[1][3]=5,D[2][3]=max{4,min(5,4,4)}=4;D[3][2]=max{d(Tp3,Tq2),min(D[2][2],D[3][1],D[2][1])}

[0104] D[2][2]=max{d(Tp2,Tq2),min(D[1][2],D[2][1],D[1][1])}=max{1,min(4,4,4)}=4

[0105] D[3][1]=max{d(Tp3,Tq1),min(D[2][1],D[3][0],D[2][0])}=max{2,min(4,∞,∞)}=4

[0106] D[2][2]=max{d(Tp2,Tq2),min(D[1][2],D[2][1],D[1][1])}=max{1,min(4,4,4)}=4

[0107] 反推:D[3][2]=max{0,min(4,4,4)}=4

[0108] D[3][3]=max{3,min(4,4,4}=4; that is, the Fréchet distance of the temperature data sequences T(p)=[1,2,3] and T(q)=[5,3,6] calculated by the first method is 4.

[0109] It can be seen that the recursive calculation results are consistent with the chain sequence enumeration calculation results, which proves the effectiveness of the simple calculation provided by the present invention.

[0110] The correlation coefficient ρ(p,q) of the temperature time series data of valves p and q is used to evaluate the consistency of temperature data changes. The calculation process is as follows:

[0111]

[0112] in, is the mean value of the temperature time series data T(p), is the mean value of the temperature time series data T(q).

[0113] St3, calculate the comprehensive evaluation index s(p,q) of each pair of valves based on Fréchet distance and correlation coefficient;

[0114]

[0115] Where w represents the weight coefficient; d(p,q) represents the Fréchet distance between valves p and q, and ρ(p,q) represents the correlation coefficient between valves p and q.

[0116] A larger value for the comprehensive evaluation index s(p,q) indicates a greater similarity between the temperature time series data T(p) and T(q). The weight coefficient w, which ranges from [0 to 1], adjusts the weighting of the discrete Fréchet distance and Pearson correlation coefficient in the comprehensive evaluation index. The weight coefficient is typically determined based on the characteristics of the valve cover temperature of a specific compressor model and domain knowledge. The shape of the valve temperature distribution is generally considered more important for identifying valve conditions, so a value within the range of [0.3, 0.5] is recommended.

[0117] St4. If all comprehensive evaluation indicators s(p,q) corresponding to the valve p are less than the set threshold, it is determined that the valve p is in a fault state.

[0118] According to the judgment basis of this embodiment, the comprehensive evaluation index values ​​of each gas valve are compared with the set threshold value. If the comprehensive evaluation index values ​​of a certain gas valve and multiple other gas valves are all smaller than the threshold value, it indicates that the valve cover temperature distribution of the gas valve is significantly different from that of other gas valves. Then it is judged that the gas valve may have a fault abnormality, which is theoretically valid.

[0119] The threshold is set based on historical data or domain knowledge to assist in the identification and diagnosis of valve faults. For example, when the reciprocating compressor is operating normally, the valve cover temperature data of the valves on the same side of the same cylinder is collected, and then the above steps St1-St3 are used to calculate the comprehensive evaluation index between the valves as the health index. In this way, the threshold needs to be set lower than the health index. Alternatively, the historical data of a reciprocating compressor with multiple valves on the same side of the same level is selected, and the valve cover temperature data of each valve when one of the valves is abnormal is obtained from the historical data. Then, the above steps St1-St3 are used to calculate the comprehensive evaluation index between the valves; the threshold is set to separate the comprehensive evaluation index of the abnormal valve from the other comprehensive evaluation indicators.

[0120] The above-mentioned method for identifying a reciprocating compressor valve fault is verified below in conjunction with specific embodiments.

[0121] In the following embodiment, temperature time series data with a step size of 1 hour and a data volume of 24 is constructed.

[0122] Example 1

[0123] Taking a reciprocating unit A100 as an example, it includes 4 suction valves, which are respectively denoted as 1#-4#.

[0124] In this embodiment, the unit monitoring equipment acquires relevant data, such as the temperature data of the #1 to #4 intake valve covers on the same side of the first-stage cylinder over a specified period (e.g., 24 hours). This data includes the temperature measurements of each valve at different times and their corresponding timestamps. Because the temperature measurement times of different valve covers may vary, interpolation is used to pre-process the data for time alignment, ensuring that each valve has a corresponding temperature value at the same time, ensuring the accuracy of subsequent calculations.

[0125] In this embodiment, data from the time period of the failure of the suction valve 2# is selected.

[0126] In this embodiment, a temperature value is extracted from the pre-processed discrete ordered sequence data set at intervals of 1 hour, thereby obtaining 24 temperature values ​​to form the temperature time series data of each intake valve, such as Figure 2 As shown. In this way, the temperature data matrix of the four suction valves of the reciprocating unit A100 is expressed as follows:

[0127]

[0128] That is, p,q∈[1,2,3,4], i,j∈[1,2,……,24]; Tp# i represents the valve cover temperature of the suction valve p at time i;

[0129] The Fréchet distance d(1#,2#) between suction valve 1# and suction valve 2# is calculated as follows:

[0130] Let: D[0][0]=0, D[i][0]=∞, D[0][j]=∞; i,j∈[1, 2,...,24];

[0131] D[1][0]=D[0][1]=∞

[0132] D[1][1]=max{d(T1#1,T2#1),(D[0][0]; D[0][1]; D[1][0])}

[0133] D[1][2]=max{d(T1#1,T2#2),(D[0][1]; D[0][2]; D[1][1])}

[0134] …

[0135] D

[24]

[24] =max{d(T1# 24 ,T2# 24 ),(D

[23]

[23] ; D

[23]

[24] ; D

[24]

[24] )}

[0136] In this embodiment, after actual measurement and sorting, we get T1#1=53.43, T2#1=50.43, and recursive calculation gives D

[24]

[24] =13.82, that is: d(1#,2#)=13.82.

[0137] Using the same method, the discrete Fréchet distances between suction valve 1# and suction valves 3# and 4#, as well as between suction valves 2#, 3#, and 4# are calculated, and finally the obtained Figure 4 The Fréchet distance matrix of the valve cover temperature of the reciprocating unit A (as shown in Figure 3(a)).

[0138] The Pearson correlation coefficient formula for the suction valve 1# temperature series and the suction valve 2#d is calculated as follows:

[0139]

[0140] T1# n ∈{T1#1;T1#2;...;T1# 24}

[0141]

[0142] T2# n ∈{T2#1;T2#2;…;T2# 24}

[0143]

[0144] Among them, T1#n Indicates the valve cover temperature of suction valve 1# at time n; Indicates the average temperature of the suction valve 1# within 24 hours; T2# n Indicates the valve cover temperature of suction valve 2# at time n; Indicates the average temperature of suction valve 2# within 24 hours.

[0145] The Pearson correlation coefficient of the air intake valves 1# and 2# calculated in this embodiment is ρ(1#, 2#)=0.45.

[0146] Using the same method, the Pearson correlation coefficients between suction valve 1# and suction valves 3# and 4#, as well as between suction valves 2#, 3#, and 4#, are calculated to obtain the Pearson correlation coefficient matrix of reciprocating unit A100 in Figure 3(b).

[0147] The weight of the Fréchet distance is set to 0.45, the weight of the Pearson correlation coefficient is set to 0.55, and the weight coefficient w is set to 0.45.

[0148] Therefore, the comprehensive evaluation indexes of the valve covers of suction valve 1# and suction valve 2# are:

[0149]

[0150] Using the same method, the comprehensive evaluation index values ​​between suction valve 1# and suction valves 3# and 4#, as well as between suction valves 2#, 3#, and 4#, are calculated to obtain the comprehensive evaluation index matrix of reciprocating unit A100 in Figure 3(c).

[0151] Observe the comprehensive evaluation index matrix for reciprocating unit A100 in Figure 3(c) and compare the comprehensive evaluation index values ​​between each valve with the set threshold of 0.5. In the comprehensive evaluation index matrix, the comprehensive evaluation index value between intake valve 2# and intake valve 1# is 0.28, which is less than 0.5, indicating that the valve cover temperature distribution similarity between intake valve 2# and intake valve 1# is low. The comprehensive evaluation index values ​​between intake valve 2# and intake valve 3# are 0.24, and the comprehensive evaluation index value between intake valve 2# and intake valve 4# is 0.29, indicating that the temperature distribution similarity between intake valve 2# and intake valves 1#, 3#, and 4# is significantly different. The comprehensive evaluation index values ​​between the other intake valves 1#, 3#, and 4# are all greater than 0.5, indicating that the temperature change trend and distribution shape of intake valve 2# are inconsistent with those of the other intake valves. Therefore, it is judged that intake valve 2# may have a fault anomaly.

[0152] The judgment result is consistent with the actual situation.

[0153] Example 2

[0154] In this embodiment, taking the reciprocating unit B100 as an example, it includes four air valves, which are respectively denoted as air valves A, B, C, and D.

[0155] In this embodiment, when the valve D is in an abnormal state, the valve cover temperature data of valves A, B, C, and D are collected synchronously, and the 24-hour temperature time series data is obtained by referring to the preprocessing method of Example 1. Figure 4 shown.

[0156] In this embodiment, the Fréchet distance matrix of the reciprocating unit B100 is calculated as shown in FIG5(a), the Pearson correlation coefficient matrix is ​​shown in FIG5(b), and the comprehensive evaluation index matrix is ​​shown in FIG5(c).

[0157] In this embodiment, the threshold value is 0.5.

[0158] Based on the comprehensive evaluation index matrix for reciprocating unit B100 in Figure 5(c), the comprehensive evaluation index values ​​of valve D compared to those of the other valves (valve A, valve B, and valve C) are all less than 0.5, indicating that the valve cover temperature distribution of valve D is significantly different from that of the other valves. This means that the temperature change trend and distribution shape of valve D differ from those of the other normally functioning valves, leading to the possibility that valve A is malfunctioning.

[0159] The judgment result is consistent with the actual situation.

[0160] Of course, it will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, but also encompasses the same or similar structures that can be implemented in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims, not the foregoing description, and it is intended that all variations that fall within the meaning and range of equivalents of the claims be encompassed within the present invention. Any reference signs in the claims should not be construed as limiting the claim to which they relate.

[0161] In addition, it should be understood that although this specification is described in terms of implementation methods, not every implementation method contains only one independent technical solution. This narrative method of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment can also be appropriately combined to form other implementation methods that can be understood by those skilled in the art.

[0162] The technology, shape, and structure not described in detail in the present invention are all well-known technologies.

Claims

1. A method for identifying a reciprocating compressor valve fault, characterized in that: The following steps are involved: St1. Synchronously collect the valve cover temperature data of the same cylinder and the same side of the reciprocating compressor and align them in time to construct the temperature time series data of each valve; St2. Construct temperature time series data for each valve and calculate the similarity index and consistency index of the temperature time series data of each pair of valves; St3. For each pair of valves, calculate the comprehensive weight of the similarity index and consistency index as the comprehensive evaluation index s(p,q); St4. If all comprehensive evaluation indicators s(p,q) corresponding to the valve p are less than the set threshold, it is determined that the valve p is in a fault state.

2. The method for identifying a reciprocating compressor valve fault according to claim 1, wherein: Wherein, w represents the weight coefficient; d(p,q) represents the similarity index of valve p and valve q; ρ(p,q) represents the consistency index of valve p and valve q.

3. The method for identifying a reciprocating compressor valve failure according to claim 2, wherein: The consistency index was the correlation coefficient; Among them, Tp n Indicates the valve cover temperature of the upper air valve p at time n, Tq n represents the valve cover temperature of the upper gas valve q at time n, and N represents the time series length of the temperature time series data; is the mean value of the temperature time series data T(p), is the mean value of the temperature time series data T(q).

4. The method for identifying a reciprocating compressor valve failure according to claim 2, wherein: The weight coefficient w takes values ​​in the interval [0.3, 0.5].

5. The method for identifying a reciprocating compressor valve failure according to claim 2, wherein: The similarity index d(p,q) uses the Fréchet distance.

6. The method for identifying a reciprocating compressor valve fault according to claim 5, wherein: Let D[i][j] = max(d(Tp i , Tq j ), min(D[i - 1][j], D[i][j - 1], D[i - 1][j - 1])) d(Tp i ,Tq j )=|Tp i -Tq j | D[0][0]＝0 D[i][0]=∞ D[0][j]=∞ i and j represent the time, 1≤i≤N, 1≤j≤N, N represents the length of the temperature time series data; Tp i is the valve cover temperature of valve p at time i, Tq j is the valve cover temperature of valve q at time j; d(Tp i ,Tq j ) represents Tp i and Tq j the distance between them; The Fréchet distance between valve p and valve q = D[N][N].

7. The method for identifying a reciprocating compressor valve failure according to claim 1, wherein: In step St1, the temperature data of the valve covers on the same side of the same cylinder of the reciprocating compressor are first collected and abnormal value processing is performed; then the temperature data of each valve cover are aligned in time points, and the temperature time series data of the valve is constructed.

8. A reciprocating compressor valve fault identification device, characterized in that: It includes data receiving module, data preprocessing module, time alignment module and comprehensive evaluation module; The data receiving module is used to receive the valve cover temperature of the gas valve on the same side of the same cylinder of the reciprocating compressor collected by the monitoring equipment; The data preprocessing module is used to preprocess the valve cover temperature time series data of each gas valve, including outlier deletion and missing value interpolation; The time alignment module is used to perform time alignment sampling on the temperature time series data of the valve covers after preprocessing of different gas valves; The comprehensive evaluation module is used to calculate the correlation index and consistency index of the temperature time series data between different gas valves, and calculate the comprehensive evaluation index.

9. A reciprocating compressor valve fault identification system, characterized in that: It includes a memory and a processor, the memory stores a computer program, the processor is connected to the memory, and the processor is used to execute the computer program to implement the method for identifying a reciprocating compressor valve fault as described in any one of claims 1 to 7.

10. A storage medium, characterized in that: A computer program is stored, and when the computer program is executed, it is used to implement the method for identifying a reciprocating compressor valve fault according to any one of claims 1 to 7.

Images