A method for detecting and locating valve leakage faults in variable-operating-condition cementing pumps

Abstract

The present invention discloses a method for detecting and locating valve leakage faults of cementing pumps under variable working conditions, and belongs to the technical field of safety monitoring and fault diagnosis of cementing equipment. The present invention effectively eliminates the influence of rotation speed fluctuations on fault feature extraction by converting time domain vibration signals into angle domain signals, thereby improving the accuracy and stability of fault detection. Fault detection and location based on the Bayesian reasoning model comprehensively considers prior probability, conditional probability and posterior probability, can make full use of prior process knowledge, reduce the demand for a large amount of fault mode data, significantly improve the accuracy of fault location, provide an accurate basis for timely repair of cementing pumps, and reduce production losses caused by long troubleshooting time.

Description

Technical Field

[0001] The invention belongs to the technical field of safety monitoring and fault diagnosis of cementing equipment, and in particular relates to a method for detecting and locating valve leakage faults of cementing pumps under variable working conditions. Background Art

[0002] As a key equipment in cementing operation, the reliability of the cementing pump's operation directly affects the cementing quality and the safety and economy of the entire oil production project. The valve is one of the core components of the cementing pump. Due to long-term operation in a harsh working environment with high pressure and high wear, the valve is very prone to leakage. Once the valve leaks, it will not only cause the cementing pump output pressure to be unstable and the displacement to decrease, affecting the smooth progress of the cementing operation, but in severe cases it may also cause underground accidents, resulting in huge economic losses.

[0003] At present, the detection methods for valve leakage failure of cementing pumps mainly include manual inspection and vibration signal time / frequency domain analysis. The manual inspection method subjectively analyzes whether valve leakage occurs through means such as auscultation and pressure detection. It is inefficient and subjective, and it is difficult to accurately detect early faults in real time. In recent years, with the development of digital monitoring technology for cementing equipment, methods based on vibration signal analysis have gradually become a research hotspot, but such methods have two limitations. On the one hand, the change in cementing pump speed affects the regularity of vibration signal feature extraction. Traditional time / frequency domain analysis generally extracts the intrinsic characteristics of vibration signals based on a fixed data window method, but the extracted features are easily affected by changes in the operating conditions of cementing pumps. Operating condition changes generally come from changes in the crankshaft speed of the hydraulic end. At the same sampling frequency, different crankshaft speeds cause significant changes in data samples within a plunger stroke, which makes it difficult to classify features in fixed data windows. On the other hand, most of the existing fault diagnosis methods based on vibration signal analysis rely on machine learning models to build feature classifiers. They are supervised pattern classification technologies that require a large amount of fault condition data. However, in reality, there is more normal condition data but less fault data, which results in limited pattern recognition capabilities of fault diagnosis models and the inability to accurately locate valve leakage faults. Therefore, how to construct a method for detecting and locating valve leakage faults in cementing pumps that does not require prior fault data and is suitable for speed changes based on existing vibration signal analysis is a very worthy research issue. Summary of the invention

[0004] In view of the above-mentioned technical problems existing in the prior art, the present invention proposes a method for detecting and locating valve leakage faults of cementing pumps under variable operating conditions, so as to solve the difficulties of detecting and locating valve leakage faults of cementing pumps under the conditions of high dependence on prior fault data and poor adaptability to operating condition changes faced by the prior art, thereby realizing real-time and accurate monitoring of the operating status of the cementing pump and ensuring the safe and efficient cementing operation.

[0005] To achieve the above-mentioned purpose, the present invention provides a method for detecting and locating valve leakage faults of cementing pumps under variable working conditions based on Bayesian reasoning of angle domain features. The method makes full use of the process mechanism to overcome the limitation of prior fault data, and uses angle domain feature extraction technology to effectively adapt to the changing working conditions, including the following steps: Step 1: Data acquisition: Collect the vibration signal data of the hydraulic end pump body and the real-time speed data of the hydraulic end crankshaft during the operation of the cementing pump; Step 2: Angle domain conversion: Use the equal angle resampling algorithm to convert the time domain vibration signal within a stroke into the angle domain signal, and evenly divide the vibration data into several angle subspaces in the angle domain; Step 3: Statistical feature extraction: Calculate the statistical feature vector of the vibration signal in each angle subspace and construct the T of the feature vector 2 Statistics; Step 4: Bayesian reasoning; Use Bayesian reasoning to calculate the posterior fault probability of each angle subspace and establish a fault probability vector; Step 5: Fault detection; Calculate the overall fault probability. If it exceeds the threshold, it is determined that a fault has occurred, otherwise the cementing pump valve is in a normal state; Step 6: Fault location; Combined with the valve position of the cementing pump, predefine the valve leakage fault mode library, calculate the matching degree of the fault probability vector and the predefined fault mode, and locate the valve fault position through the maximum matching degree.

[0006] Preferably, in step 1, vibration signal data of the hydraulic end pump body of the cementing pump during operation is collected to capture the vibration anomaly caused by changes in the working state of the valve, and real-time speed data of the crankshaft of the hydraulic end of the cementing pump is obtained to describe the periodic time information of the valve working.

[0007] Preferably, in step 2, an equal angle resampling algorithm is used to convert the time domain vibration signal within a stroke into Convert to angle domain signal ,in The sampling point numbers corresponding to the start and end of the stroke, It is defined as the sampling number when any plunger of the cementing pump moves to the right dead point, calculated by formula (1) The value of (1); among which, for The speed value at the moment, the unit is rpm, that is, revolutions per minute. is the data sampling frequency, in Hz; use formula (2) to calculate the angle domain coordinates and time domain sampling number The corresponding relationship; (2).

[0008] Preferably, step 3 specifically includes the following steps: Step 3.1: Evenly divide the vibration data into Angle Subspace ,in is the angle subspace number, , The value of corresponds to the number of cementing pump valves; Step 3.2: In each angle subspace The eight statistical features of the vibration signal are calculated, including the maximum value, minimum value, peak-to-peak value, absolute average value, root mean square value, standard deviation, average frequency, and root mean square frequency, forming a vibration feature vector [ ],in Representative The angle interval features; Step 3.3: Use formula (3) to calculate the maximum value of the vibration signal in each angle subspace ; (3); Step 3.4: Use formula (4) to calculate the minimum value of the vibration signal in each angle subspace ; (4); Step 3.5: Use formula (5) to calculate the peak-to-peak value of the vibration signal in each angle subspace ; (5); Step 3.6: Use formula (6) to calculate the absolute average value of the vibration signal in each angle subspace ; (6); Step 3.7: Use formula (7) to calculate the RMS value of the vibration signal in each angle subspace: ; (7);

[0009] Step 3.8: Use formula (8) to calculate the standard deviation of the vibration signal in each angle subspace ; (8); among them, for The average value of After frequency domain analysis, the frequency Its power spectrum , where M represents the number of frequency values, and the average frequency of the vibration signal in each angle subspace is calculated using formula (9): ; (9); Step 3.10: Calculate the root mean square frequency of the vibration signal in each angle subspace using formula (10): ; (10).

[0010] Preferably, step 4 specifically includes the following steps: Step 4.1: In each angle subspace Calculate the vibration signal feature vector T 2Statistics, and then use Bayesian reasoning to calculate The posterior probability of failure in the interval , and constitute the subspace fault vector =[ ]; Step 4.2: Use formula (11) to calculate the statistics of the vibration signal feature vector ; (11); among them, Represents the vibration signal feature vector The weight matrix of represents the threshold parameter of the feature vector; Step 4.3: Use formula (12) to perform Bayesian inference calculation The posterior probability of failure in the interval ; (12); among them, represents the prior probability of a valve leakage failure, Indicates a fault condition The conditional probability of occurrence, express Total probability of occurrence; Step 4.4: Calculate the total probability using formula (13) ; (13); among them, represents the prior probability of normal operating conditions, Indicates that under normal circumstances Conditional probability of occurrence; Step 4.5: Use formulas (14) and (15) to calculate the probability ; (14); (15); among them, yes The statistical threshold of .

[0011] Preferably, in step 5, the overall failure probability is calculated If the threshold is exceeded , then it is judged that a fault has occurred, otherwise the cementing pump valve is in a normal state; the overall fault probability is calculated using formula (16) ; (16).

[0012] Preferably, in step 6, valve leakage at different positions corresponds to different angular subspace fault vectors, and the valve leakage fault mode library { }, where C represents the number of failure modes, and the subspace failure vector is calculated With predefined failure modes { }'s matching degree MD 1 , locate the fault position of the valve by the maximum matching degree;

[0013] Use formula (17) to calculate the subspace fault vector With predefined failure modes { }'s matching degree MD ;

[0014] MD (17);

[0015] Among them, MD The range is between [0,1], and the closer it is to 1, the more fault feature vector Failure Mode Vector The more similar they are, the higher the probability of locating it as the cth valve leakage fault.

[0016] Beneficial technical effects brought by the present invention: By converting the time domain vibration signal into the angle domain signal, the present invention effectively eliminates the influence of the speed fluctuation on the fault feature extraction, and improves the accuracy and stability of fault detection. Fault detection and location based on the Bayesian reasoning model comprehensively considers the prior probability, likelihood function and posterior probability, can make full use of prior process knowledge, reduce the demand for a large amount of fault mode data, significantly improve the accuracy of fault location, provide an accurate basis for timely repair of cementing pumps, and reduce production losses caused by long troubleshooting time. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is a schematic diagram of a method for detecting and locating a valve leakage fault in a cementing pump; Figure 2 It is a graphical diagram of the first group of fault data; Figure 3 It is a graphical diagram of the second group of fault data; Figure 4 This is the fault detection effect diagram. DETAILED DESCRIPTION

[0018] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments: The present invention provides a method for detecting and locating valve leakage faults in cementing pumps based on Bayesian reasoning of angle domain features, and the process is as follows: Figure 1 As shown in the figure, this method makes full use of the process mechanism to overcome the limitation of prior fault data and uses the angle domain feature extraction technology to effectively adapt to the changing working conditions, including the following steps: Step 1: Data acquisition: Collect the vibration signal data of the hydraulic end pump body during the operation of the cementing pump , used to capture the vibration anomaly caused by the change of the working state of the valve, and obtain the speed data of the crankshaft of the hydraulic end of the cementing pump , used to describe the periodic time information of the vane work.

[0019] Step 2: Angle domain conversion: Considering that one stroke of the cementing pump plunger corresponds to one crankshaft rotation ( ), the time domain vibration signal within a stroke is converted into Convert to angle domain signal ,in The sampling point numbers corresponding to the start and end of the stroke can be obtained according to the speed data; specifically, the time domain vibration signal within a stroke is converted into Convert to angle domain signal ,in The sampling point numbers corresponding to the start and end of the stroke, It is defined as the sampling number when any plunger of the cementing pump moves to the right dead point, calculated by formula (1) The value of (1); among which, for The speed value at the moment, the unit is rpm, that is, revolutions per minute. is the data sampling frequency, in Hz; use formula (2) to calculate the angle domain coordinates and time domain sampling number The corresponding relationship; (2).

[0020] Step 3: Statistical feature extraction: Divide the vibration data evenly into K angular subspaces in the angle domain ,in is the angle subspace number, , the value of K corresponds to the number of cementing pump vanes, and then in each angle subspace The eight statistical features of the vibration signal are calculated, including the maximum value, minimum value, peak-to-peak value, absolute average value, root mean square value, standard deviation, average frequency, and root mean square frequency, forming a vibration feature vector [ ],in represents the i-th feature of the k-th angle interval; specifically includes the following steps: Step 3.1: Divide the vibration data evenly into Angle Subspace ,in is the angle subspace number, , The value of corresponds to the number of cementing pump valves; Step 3.2: In each angle subspace The eight statistical features of the vibration signal are calculated, including the maximum value, minimum value, peak-to-peak value, absolute average value, root mean square value, standard deviation, average frequency, and root mean square frequency, forming a vibration feature vector [ ],in Representative The angle interval features; Step 3.3: Use formula (3) to calculate the maximum value of the vibration signal in each angle subspace ; (3); Step 3.4: Use formula (4) to calculate the minimum value of the vibration signal in each angle subspace ; (4); Step 3.5: Use formula (5) to calculate the peak-to-peak value of the vibration signal in each angle subspace ; (5); Step 3.6: Use formula (6) to calculate the absolute average value of the vibration signal in each angle subspace ; (6); Step 3.7: Use formula (7) to calculate the RMS value of the vibration signal in each angle subspace: ; (7); Step 3.8: Use formula (8) to calculate the standard deviation of the vibration signal in each angle subspace ; (8); among them, for The average value of After frequency domain analysis, the frequency Its power spectrum , where M represents the number of frequency values, and the average frequency of the vibration signal in each angle subspace is calculated using formula (9): ; (9); Step 3.10: Calculate the root mean square frequency of the vibration signal in each angle subspace using formula (10): ; (10).

[0021] Step 4: Bayesian Inference: In each angular subspace Calculate the vibration signal feature vector T 2 Statistics, and then use Bayesian reasoning to calculate The posterior probability of failure in the interval , and constitute the subspace fault vector =[ ]; specifically includes the following steps: Step 4.1: In each angle subspace Calculate the vibration signal feature vector T 2 Statistics, and then use Bayesian reasoning to calculate The posterior probability of failure in the interval , and constitute the subspace fault vector =[ ]; Step 4.2: Use formula (11) to calculate the statistics of the vibration signal feature vector ; (11); among them, Represents the vibration signal feature vector The weight matrix of represents the threshold parameter of the feature vector; Step 4.3: Use formula (12) to perform Bayesian inference calculation The posterior probability of failure in the interval ; (12); among them, represents the prior probability of a valve leakage failure, Indicates a fault condition The conditional probability of occurrence, express Total probability of occurrence; Step 4.4: Calculate the total probability using formula (13) ; (13); among them, represents the prior probability of normal operating conditions, Indicates that under normal circumstances Conditional probability of occurrence; Step 4.5: Use formulas (14) and (15) to calculate the probability ; (14); (15); among them, yes The statistical threshold of .

[0022] Step 5: Fault Detection: Calculate the Overall Failure Probability If the threshold is exceeded , it is judged that a fault has occurred, otherwise the cementing pump valve is in normal condition.

[0023] The overall failure probability is calculated using formula (16): ; (16).

[0024] Step 6: Fault location: Considering that valve leakage at different locations corresponds to different angle subspace fault vectors, a valve leakage fault mode library is predefined based on the valve location of the cementing pump. }, where C represents the number of failure modes, and the subspace failure vector is calculated With predefined failure modes { }'s matching degree MD 1 , locate the fault position of the valve by maximizing the matching degree.

[0025] Use formula (17) to calculate the subspace fault vector With predefined failure modes { }'s matching degree MD ; MD (17); among them, MD The range is between [0,1], and the closer it is to 1, the more fault feature vector Failure Mode Vector The more similar they are, the higher the probability of locating it as the cth valve leakage fault.

[0026] During the operation of a certain company's 3-cylinder cementing pump, vibration data and speed data were collected, including 10 groups of normal operating data samples and 2 groups of fault samples under different speeds. The 2 groups of fault samples were concurrent faults of the 3-cylinder discharge valve and the 2-cylinder suction valve under different speeds. Figure 2 and Figure 3 It is a graph of two sets of fault data. After the data is converted into the angle domain, the statistical features are extracted and the Bayesian reasoning is performed to calculate the fault vector. The final fault detection result is as follows Figure 4 As shown. It can be seen that the failure probability of the first 10 groups of samples is very low, and the failure probability of the last two groups of samples is very high, indicating that the fault can be correctly detected. The two groups of fault condition data are matched with the 6 feature vectors in the fault mode library respectively, and the matching results are shown in Table 1.

[0027] Table 1 Matching degree between two groups of fault samples and six fault modes

[0028] .

[0029] The pattern matching results show that the two sets of fault data have the highest matching degree with the 3-cylinder exhaust valve and the 2-cylinder intake valve, both around 0.866. Therefore, it can be determined that the fault is related to the leakage of these two valves, which is consistent with the actual fault type and illustrates the correctness of the method.

[0030] Of course, the above description is not a limitation of the present invention, and the present invention is not limited to the above examples. Changes, modifications, additions or substitutions made by technicians in this technical field within the essential scope of the present invention should also fall within the protection scope of the present invention.

Claims

1. A method for detecting and locating valve leakage faults of a variable working condition cementing pump, characterized in that: The steps include: Step 1: Data collection; Collect vibration signal data of the hydraulic end pump body and real-time speed data of the hydraulic end crankshaft during the operation of the cementing pump; Step 2: Angle domain conversion; The time domain vibration signal within a stroke is converted into an angle domain signal using an equal angle resampling algorithm, and the vibration data is evenly divided into several angle subspaces in the angle domain; Step 3: Statistical feature extraction; Calculate the statistical eigenvector of the vibration signal in each angular subspace and construct the T of the eigenvector 2 Statistics; Step 4: Bayesian reasoning; Bayesian inference is used to calculate the posterior fault probability of each angle subspace and establish the fault probability vector; Step 5: Fault detection; Calculate the overall failure probability. If it exceeds the threshold, it is determined that a failure has occurred. Otherwise, the cementing pump valve is in a normal state. Step 6: Fault location; Combined with the predefined valve leakage fault mode library of the valve position of the cementing pump, the matching degree between the fault probability vector and the predefined fault mode is calculated, and the valve fault position is located through the maximum matching degree.

2. The variable operating condition cementing pump valve leakage fault detection and positioning method according to claim 1 is characterized in that: In step 1, the vibration signal data of the hydraulic end pump body of the cementing pump during operation is collected to capture the vibration anomaly caused by the change of the working state of the valve, and the real-time speed data of the crankshaft of the hydraulic end of the cementing pump is obtained to describe the periodic time information of the valve working.

3. The method for detecting and locating valve leakage fault of a variable operating condition cementing pump according to claim 1, characterized in that: In step 2, the time domain vibration signal within a stroke is converted into Convert to angle domain signal ,in The sampling point numbers corresponding to the start and end of the stroke, It is defined as the sampling number when any plunger of the cementing pump moves to the right dead point, calculated by formula (1) The value of （1）； in, for The speed value at the moment, the unit is rpm, that is, revolutions per minute. is the data sampling frequency, in Hz; Calculate the angle domain coordinates using formula (2) and time domain sampling number The corresponding relationship; （2）。 4. The method for detecting and locating valve leakage fault of a variable operating condition cementing pump according to claim 1, characterized in that: Step 3 specifically includes the following steps: Step 3.1: Divide the vibration data evenly into Angle Subspace ,in is the angle subspace number, , The value of corresponds to the number of valves of the cementing pump; Step 3.2: In each angular subspace The eight statistical features of the vibration signal are calculated, including the maximum value, minimum value, peak-to-peak value, absolute average value, root mean square value, standard deviation, average frequency, and root mean square frequency, forming a vibration feature vector [ ],in Representative The angle interval Features Step 3.3: Use formula (3) to calculate the maximum value of the vibration signal in each angle subspace ; （3）； Step 3.4: Use formula (4) to calculate the minimum value of the vibration signal in each angle subspace ; （4）； Step 3.5: Use formula (5) to calculate the peak-to-peak value of the vibration signal in each angle subspace ; （5）； Step 3.6: Use formula (6) to calculate the absolute average value of the vibration signal in each angle subspace ; （6）； Step 3.7: Use formula (7) to calculate the RMS value of the vibration signal in each angle subspace: ; （7）； Step 3.8: Use formula (8) to calculate the standard deviation of the vibration signal in each angle subspace ; （8） in, for The average value of Step 3.9: After frequency domain analysis, the frequency Its power spectrum ,in, , M represents the number of frequency values, and the average frequency of the vibration signal in each angle subspace is calculated using formula (9): ; （9）； Step 3.10: Calculate the RMS frequency of the vibration signal in each angular subspace using formula (10): ; （10）。 5. The method for detecting and locating valve leakage fault of a variable operating condition cementing pump according to claim 1, characterized in that: Step 4 specifically includes the following steps: Step 4.1: In each angular subspace Calculate the vibration signal feature vector T 2 Statistics, and then use Bayesian reasoning to calculate The posterior probability of failure in the interval , and constitute the subspace fault vector =[ ]; Step 4.2: Calculate the statistics of the vibration signal feature vector using formula (11) ; （11）； in, Represents the vibration signal feature vector The weight matrix of represents the threshold parameter of the feature vector; Step 4.3: Perform Bayesian inference calculation using formula (12) The posterior probability of failure in the interval ; （12）； in, represents the prior probability of a valve leakage failure, Indicates a fault condition The conditional probability of occurrence, express Total probability of occurrence; Step 4.4: Calculate the total probability using formula (13) ; （13）； in, represents the prior probability of normal operating conditions, Indicates that under normal circumstances Conditional probability of occurrence; Step 4.5: Use formulas (14) and (15) to calculate the probability ; （14）； （15）； in, yes The statistical threshold of .

6. The method for detecting and locating valve leakage fault of a variable operating condition cementing pump according to claim 1, characterized in that: In step 5, calculate the overall failure probability If the threshold is exceeded , it is judged that a fault occurs, otherwise the cementing pump valve is in normal state; The overall failure probability is calculated using formula (16): ; （16）。 7. The method for detecting and locating valve leakage fault of a variable operating condition cementing pump according to claim 1, characterized in that: In step 6, valve leakage at different locations corresponds to different angle subspace fault vectors. Combined with the valve position of the cementing pump, a predefined valve leakage fault mode library { }, where C represents the number of failure modes, and the subspace failure vector is calculated With predefined failure modes { }'s matching degree MD 1 , locate the fault position of the valve by the maximum matching degree; Use formula (17) to calculate the subspace fault vector With predefined failure modes { }'s matching degree MD ; MD (17); Among them, MD The range is between [0,1], and the closer it is to 1, the more fault feature vector Failure Mode Vector The more similar they are, the higher the probability of locating it as the cth valve leakage fault.

Images