A PMSM degradation experiment device of a downhole environment and a motor life prediction method

Abstract

The application belongs to the technical field of oil field drilling, and discloses a PMSM degradation experiment device of downhole environment and a motor life prediction method. The downhole environment motor residual life prediction method comprises the following steps: collecting measurement signals of a permanent magnet synchronous motor by using a sensor, performing spectrum entropy analysis on three-phase current signals of a stator of the permanent magnet synchronous motor, and extracting the frequency spectrum complexity change of the three-phase current signals of the stator in the degradation process; performing wavelet time-frequency band energy transformation on a vibration signal, and extracting the energy change of the vibration signal in the degradation process; dividing the collected temperature, sound wave, torque, rotating speed, extracted three-phase current signals of the stator and extracted vibration signals into a training set, a test set and a verification set, and conveying the above data sets to a network to train the network. The application can realize real-time residual service life prediction of a permanent magnet synchronous motor of a downhole drilling communication tool, and improve the operation reliability and working safety of a downhole system.

Description

Technical Field

[0001] This invention belongs to the field of oilfield drilling technology, and particularly relates to a PMSM degradation experimental device for downhole environments and a method for predicting the remaining life of downhole motors. Background Technology

[0002] Domestic rotary steerable drilling tools and downhole mud pulse communication tools, as the core of high-end automated drilling equipment, are facing the technical challenge of insufficient reliability. Whether their only actuator - the permanent magnet synchronous motor (PMSM) - can operate normally is an important factor affecting their reliability.

[0003] Especially in the oil drilling field, drilling tools operate for extended periods in harsh environments characterized by high temperatures, high pressures, strong vibrations, and significant impacts. This significantly reduces the lifespan of permanent magnet synchronous motors (PMSMs). If PMSMs cannot be pulled out of the well for repair before failure, it can lead to anything from reduced drilling accuracy to substantial economic losses or even accidents. The reliability of downhole PMSMs has become a critical issue restricting drilling efficiency and cost reduction. Therefore, real-time prediction of the remaining lifespan of downhole PMSMs is crucial for increasing drilling reliability and ensuring drilling safety. Furthermore, due to the complexity of the downhole environment and the precision of drilling equipment, establishing a complete degradation model for downhole drilling communication systems is extremely difficult. Even if a model is established, large errors can lead to prediction failures. Therefore, data-driven prediction of remaining lifespan is more realistic and reliable. Downhole data acquisition is limited by storage constraints, making it difficult to obtain large amounts of data. Additionally, limitations in communication technology make it difficult to transmit real-time data collected by downhole sensors for training machine prediction models. Therefore, designing a surface-based permanent magnet synchronous motor degradation experimental system and device based on the downhole environment to acquire relevant data is essential.

[0004] Strong vibrations, large impacts, and noise in downhole environments can obscure useful data in vibration signals. Therefore, analyzing the patterns of downhole vibrations and removing interference such as strong vibrations and noise from the original vibration signals is a crucial step in training the vibration signal network. Thus, optimizing the sensor layout and designing feature extraction methods based on the downhole environment to mitigate the impact of low-frequency strong vibrations is particularly important.

[0005] Chinese Patent Publication No. CN110208642A, published on September 6, 2019, discloses an invention entitled "Simulation Method and System for Degradation Process of Stator Inter-turn Short Circuit Fault in Permanent Magnet Synchronous Motor (PMSM)". This application discloses a simulation method for the degradation process of stator inter-turn short circuit fault in a PMSM based on an interaction model. Its shortcomings are: 1. The model predicts the future state of the PMSM, but the internal complexity of the PMSM under inter-turn short circuit faults can cause the model to fail; 2. The degradation simulation system for the PMSM does not consider the influence of the product's operating environment.

[0006] Chinese Patent Publication No. CN108614940A, published on October 2, 2018, entitled "A Method and System for Evaluating the Performance Degradation of a Permanent Magnet Synchronous Motor." This application discloses a method for detecting the degree of degradation based on PMSM vibration and stator current data. Its shortcomings include: failing to consider noise processing in the collected data; different data collection durations and intervals for normal and faulty motors, resulting in diverse experimental variables and reduced reliability; and failing to consider the influence of the product's operating environment.

[0007] Chinese Patent Publication No. CN113515846A, published on October 19, 2021, discloses an invention entitled "RUL Prediction Method for Electric Turntable Based on Wiener Process Degradation Model at Turning Point." This application discloses a method to address the problems of existing electric turntable RUL prediction methods, such as the occurrence of sudden changes in health factors when predicting RUL before the turning point and the inability to update new data when predicting RUL after the turning point, leading to low prediction accuracy. The shortcomings of this method are: predicting the future state of the permanent magnet synchronous motor (PMSM) based on a model without considering the variability of the model after a failure; and the rotating platform operating based on the PMSM without considering the impact of the product's operating environment on the permanent magnet synchronous motor.

[0008] Reference [1] analyzed the feasibility of using deep learning networks to predict the remaining life of servo motors, and collected speed, torque and vibration signals in an experimental setup. The feature extraction signals were used as the original training data for the network, and a novel recurrent neural network was proposed to improve the accuracy of the prediction. The shortcomings of this method are: it only uses vibration signals and does not process the collected torque and speed signals, and it only processes the frequency domain signals and does not process the vibration signals in the time and frequency domains, resulting in a waste of data; the experimental setup does not consider the influence of the usage environment on the motor, which may lead to the prediction method not being effective in real usage environments.

[0009] References

[0010] [1]Chen D, Qin Y, Luo J, et al. Gated adaptive hierarchical attention unit neural networks for the life prediction of servo motors[J]. IEEE Transactions on Industrial Electronics, 2021.

[0011] Based on the above analysis, the problems and shortcomings of the existing technology are as follows:

[0012] (1) Existing permanent magnet synchronous motor degradation experimental equipment cannot obtain signals with the same characteristics as actual downhole information, and cannot provide a basic sample set for training deep learning networks for predicting remaining service life, so that drilling systems cannot accurately predict the remaining service life of PMSM in relevant downhole real-time prediction.

[0013] (2) Existing technologies cannot reduce the measurement interference caused by low-frequency vibration in the actual operating environment of downhole PMSM and extract useful fault information for training, resulting in a mismatch between the information of the experimental system above ground and the actual system downhole. Furthermore, they cannot fully extract the degradation information of PMSM, resulting in poor accuracy and reliability of PMSM lifetime prediction.

[0014] (3) Existing technologies cannot adaptively reduce the impact of low-frequency strong vibrations on the prediction data, which reduces the accuracy of prediction of PMSM degradation-related data.

[0015] (4) In existing downhole real-time prediction systems, due to the small downhole storage and computing power, it is not possible to predict the remaining service life of the downhole PMSM in real time and transmit information. Summary of the Invention

[0016] To overcome the problems existing in related technologies, the present invention discloses an experimental device, prediction method, equipment, and terminal for downhole motor degradation. Specifically, it relates to a prediction and health management technology for actuators in downhole drilling and communication systems, and more specifically, to a method for predicting the remaining service life of rotary steered drilling tool actuators and mud pulse generator-permanent magnet synchronous motors in drilling communication systems, and an experimental device for downhole permanent magnet synchronous motor degradation.

[0017] The technical solution is as follows: A method for predicting the remaining life of a motor in a downhole environment includes:

[0018] The measurement signals of the permanent magnet synchronous motor are collected using sensors, including vibration, temperature, sound waves, torque, speed, and stator three-phase current. An adaptive time-frequency transformation method is used to reduce the interference of low-frequency strong vibration noise on the original vibration signal. At the same time, spectral entropy analysis is performed on the stator three-phase current signal of the permanent magnet synchronous motor to extract the change in spectral complexity of the stator three-phase current signal during the degradation process.

[0019] Wavelet time-frequency energy transformation is performed on the vibration signal to extract the energy change of the vibration signal during the degradation process. The collected temperature, sound wave, torque, speed, extracted stator three-phase current signal and extracted vibration signal are divided into training set, test set and validation set. The above datasets are fed into the network for training after feature extraction. If the trained network performs well in the test set, the weight values ​​and network structure are written into the downhole real-time prediction system through computer language. The system is then sent into the formation with the drilling equipment and mud pulse communication system to predict the remaining service life of the permanent magnet synchronous motor based on the downhole environment.

[0020] In one embodiment, before acquiring the measurement signal of the permanent magnet synchronous motor using sensors, the spatial layout of the permanent magnet synchronous motor sensors is performed, including: uniformly arranging eight acceleration sensors (4) in the drill string at the location of the permanent magnet synchronous motor. The acceleration sensors (4) are triaxial orthogonal gravity accelerometers, with the x-axis pointing to the center of the drilling platform axis, the y-axis being the drill string forward direction or its opposite direction, and the z-axis being the drill string circular tangent direction at the installation position. The acceleration in the axis normal direction and the rotational tangent direction, which are greatly affected by low-frequency strong vibrations, is extracted.

[0021] In one embodiment, spatial layout of the permanent magnet synchronous motor sensor includes: mitigating the impact of downhole vibration by calculating the relationship between quantities on the x-axis and z-axis, as expressed by:

[0022]

[0023] P is the angular velocity of the PMSM. During steady-state drilling, P changes continuously. However, because the radius R of the downhole motor is very small, but the downhole vibration intensity is very high, the value of a related to downhole vibration is significant. vz a vx The term is much larger than the terms related to R, so the terms related to R are omitted. zx a represents the gravitational acceleration component in the zx plane; vz and a vx The components of vibrational acceleration along the z-axis and x-axis; rotational acceleration includes tangential and centripetal accelerations, respectively. And R(pπ / 180) 2 / 9.8; Analysis reveals:

[0024]

[0025] The above equation is an analysis of triaxial accelerometer sensors distributed in a circle at arbitrary angular positions. After omitting the terms related to R, the equation yields:

[0026]

[0027] When γ = 180°, take the values ​​of the two accelerometers that are symmetrically distributed.

[0028]

[0029] Only the vibration acceleration interference on a certain axis remains, and the strong downhole vibration is at a vz1 and a vz2 The components in these two directions are completely opposite, but the fault information is unevenly distributed in these two directions, through a vz1 and a vz2 The fault information is extracted by adding the components in these two directions.

[0030] In one embodiment, the step of performing wavelet time-frequency energy transformation on the vibration signal to extract the energy change of the vibration signal during the degradation process includes: converting the vibration signal and the acoustic signal into the time-frequency space, then weakening the information of 0-200 Hz by using the prior knowledge of the downhole vibration frequency band distribution law, and selecting the spectrum with a high frequency band energy ratio as the input for network training.

[0031] The vibration signal is subjected to time-frequency transformation in time intervals using wavelet transform, and the expression is as follows:

[0032]

[0033]

[0034] In the formula, f(t) is the original signal, and e -iwt Representing vibration, wavelet transform uses an integral form to describe the original signal at e -iwt The projection on the screen, g(tu), is a window function. By adding a window, the shortcomings of Fourier transform in that it can only display frequency information and cannot display time-frequency transform are solved. The vibration characteristics and window signal are replaced with wavelet function ψ(a,b,t) to realize video feature extraction under variable window. a and b represent the scaling scale and time factor, respectively.

[0035] In one embodiment, during wavelet transform extraction, the common characteristics of the Morlet wavelet and the vibration signal impact attenuation are:

[0036]

[0037] Where t is time, It is an oscillation component, and v(t) represents the signal of oscillation decay.

[0038] In one embodiment, feeding the aforementioned dataset into the network for training includes: constructing a Temporal Convolutional Neural Network (TCN) using TensorFlow; combining the TCN network with an attention mechanism and a CNN to construct a deep learning network; inputting the grouped data into the network for parameter training; the CNN further extracts a large amount of feature information collected, while the TCN network broadens the network's perspective during training and obtains the changing trends of the data.

[0039] In one embodiment, the downhole real-time prediction system is equipped with a main control board, which includes a data acquisition unit and a rotation speed control unit;

[0040] The data acquisition unit obtains stator three-phase current measurement data through A / D sampling, acquires permanent magnet synchronous motor speed measurement data through serial communication, and processes the acquired data to obtain the measured values ​​of motor speed and current.

[0041] The speed control unit adopts a closed-loop control method. It provides control action based on the deviation between the speed set value and the actual measured value. It controls the PWM signal generated by the inverter bridge of the motor drive board to control the rotation of the permanent magnet synchronous motor so that the PMSM speed value follows the set value.

[0042] Another objective of this invention is to provide a PMSM degradation experimental device for a downhole environment, comprising: a permanent magnet synchronous motor connected to a vibration table via a base, the vibration table being used to provide low-frequency strong vibration to simulate a downhole vibration environment; and an acceleration sensor and a temperature sensor fixedly connected to the housing of the permanent magnet synchronous motor, used to measure the acceleration in the lateral and vertical directions of the PMSM and its surrounding temperature, respectively.

[0043] Heating resistance wires are wrapped around the top and left side of the permanent magnet synchronous motor housing to simulate the temperature environment downhole.

[0044] An acoustic sensor is installed at the bottom of the permanent magnet synchronous motor housing to collect sound signals during the degradation process. The PMSM torque output shaft is connected to the torque generator via a coupling, and the magnetic powder brake is connected to the torque generator via a coupling.

[0045] In one embodiment, the permanent magnet synchronous motor housing has multiple mounting slots for mounting accelerometers; the signal lines and power lines of the accelerometers are connected to the downhole data processing circuit board via wiring channels, the power module of the data processing circuit board provides power to the accelerometers, and the accelerometer measurement signals are transmitted to the data processor via IIC; the accelerometers are mounted on the downhole permanent magnet synchronous motor via mounting surfaces.

[0046] Another object of the present invention is to provide a storage medium for receiving user input programs, wherein the stored computer programs enable electronic devices to execute the method for predicting the remaining life of motors in the downhole environment.

[0047] Another object of the present invention is to provide a computer device including a memory and a processor, the memory storing a computer program, which, when executed by the processor, enables the processor to perform a method for predicting the remaining life of a downhole environment motor.

[0048] Combining all the above technical solutions, the advantages and positive effects of this invention are as follows:

[0049] First, addressing the technical problems existing in the prior art and the difficulty in solving them, this paper closely analyzes, in conjunction with the technical solution to be protected by this invention and the results and data obtained during the research and development process, how the technical solution of this invention solves the technical problems, and the inventive technical effects brought about by solving these problems. The specific description is as follows:

[0050] The permanent magnet synchronous motor (PMSM) degradation experimental device used in this invention can simulate the downhole working environment of a PMSM—high temperature and strong vibration—in a laboratory setting. Using this device, experiments on the remaining service life of PMSMs can be conducted in a laboratory environment, obtaining signals with characteristics identical to actual downhole information. This provides a basic sample set for training a deep learning network for remaining service life prediction, and offers a fundamental experimental platform and means for the design and optimization of downhole PMSMs. The trained network can better predict the real-time remaining service life of PMSMs on relevant downhole real-time prediction systems during drilling.

[0051] This invention presents an optimized triaxial accelerometer layout designed based on the actual operating environment of downhole PMSMs. This layout reduces measurement interference caused by low-frequency vibrations downhole and extracts useful fault information for training. Furthermore, this layout can be generalized to both surface experimental systems and actual downhole systems. The diverse sensor types allow for the extraction of PMSM degradation information from multiple perspectives, facilitating the assessment of remaining lifetime based on fused information and improving the accuracy and reliability of lifetime prediction.

[0052] The ASDN method used in the feature extraction stage of the downhole permanent magnet synchronous motor remaining service life prediction method proposed in this invention can adaptively reduce the impact of low-frequency strong vibration on the prediction data, allowing the deep learning network to focus more on the relevant data related to PMSM degradation, which helps to improve the prediction accuracy of the network.

[0053] In its application, this invention solves the problem of low downhole storage and computing power by transplanting the trained network into the downhole real-time prediction system, enabling real-time prediction of the remaining service life and information transmission for the downhole PMSM.

[0054] Second, considering the technical solution as a whole or from the perspective of the product, the technical effects and advantages of the technical solution to be protected by this invention are specifically described as follows:

[0055] This invention provides a degradation test apparatus for actuators in downhole drilling and communication systems, and a method for predicting the remaining service life of permanent magnet synchronous motors (PMSMs) based on the downhole environment. First, the spatial layout of the PMSM sensors is designed to create feasible conditions for extracting raw characteristic signals (vibration, stator three-phase current, output torque, and output speed). The greatest interference downhole is concentrated in vibration, which is a crucial characteristic reflecting whether downhole equipment is damaged. The characteristics of low-frequency, strong downhole vibrations are mainly concentrated along the circular tangential direction of the drilling cylinder. Based on this characteristic, a ring-shaped sensor layout for triaxial accelerometers is designed. This layout can extract the vibration of the drill string's circular tangential direction and axial vibration, thereby separating the axial vibration and circular tangential vibration, which are less related to low-frequency, strong downhole vibrations but can reflect fault information. Simultaneously, a downhole sensor mounting slot is designed based on the characteristics of strong downhole vibrations to ensure the stability of the sensor's position in the strong vibration environment. Regarding sensor selection, the high temperature and high pressure environment downhole must be considered; therefore, proper sensor selection is crucial to ensuring normal data acquisition. Therefore, the accelerometer adopts a three-axis quartz flexible accelerometer, the gyroscope adopts a high-strength heat-resistant magnesium alloy gyroscope, the torque sensor adopts a grating torque sensor, the current sensor adopts a high-temperature and high-pressure resistant Hall current sensor, and the sound wave sensor adopts a heat-resistant electret condenser microphone.Then, in the designed ground-based experimental platform, signal acquisition was conducted on multiple sets of the same type of permanent magnet synchronous motors throughout their entire life cycle (simulating strong downhole vibration and high-temperature environments during data acquisition). This involved acquiring the aforementioned characteristic signals as the motors transitioned from a fault-free state to a fault state, and dividing the acquired signals into training, testing, and validation sets. After signal acquisition, time-frequency analysis was performed on the vibration characteristic signals that were susceptible to strong vibrations (although optimized accelerometer layouts were used to counteract strong vibrations in the main affected directions, some influence remained in other directions). Since strong downhole vibrations are mainly concentrated in the 0-200 Hz frequency domain, useful fault signals are mostly concentrated in the high-frequency domain above 1000 Hz (because bearing fault information is mainly concentrated around 1000 Hz, and bearings are core components of rotating machinery such as PMSMs). After time-frequency analysis using wavelet transform and other methods, the 0- The 200Hz time-frequency signal is weakened by assigning it a smaller weight, thus reducing its role in network training. After network training, the prediction effect is verified on the test set. If the prediction results are good after multiple rounds of testing, the trained network can be transferred to the downhole real-time prediction system for predicting the remaining service life while drilling. This solves the limitation that a large amount of sensor data needs to be transmitted from the downhole to the surface in real time to achieve real-time PMSM remaining service life prediction. Moreover, since the network has been trained in advance, it also solves the problem of insufficient downhole computing power for continuous network optimization and updates. Of course, if the training results are poor, the full life cycle data of the corresponding PMSM model is collected again for training, increasing the amount of data and thus optimizing the network parameters. Finally, through the above steps, real-time remaining service life prediction of the permanent magnet synchronous motor of the downhole drilling communication tool can be realized, improving the operational reliability and safety of the downhole system.

[0056] Third, as supplementary evidence of the inventive step of the claims of this invention, it is also reflected in the following important aspects:

[0057] (1) The expected benefits and commercial value of the technical solution of the present invention after transformation are: increasing the reliability of downhole drilling tools, avoiding catastrophic accidents caused by actuator damage, thereby saving drilling costs, improving drilling efficiency, and enhancing drilling safety.

[0058] (2) This invention fills the technological gap in China regarding PMSM reliability prediction based on downhole environment. Attached Figure Description

[0059] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0060] Figure 1 This is a flowchart of the method for predicting the remaining life of a motor in a downhole environment provided in an embodiment of the present invention;

[0061] Figure 2 This is a flowchart of data acquisition and network training provided in an embodiment of the present invention;

[0062] Figure 3 This is a platform hardware structure diagram provided in an embodiment of the present invention;

[0063] Figure 4 This is a diagram of the experimental apparatus for simulating downhole system degradation provided in an embodiment of the present invention;

[0064] Figure 5 These are images of the original vibration information and the vibration information after incorporating low-frequency strong vibration, provided in embodiments of the present invention; wherein... Figure 5 (a) shows the degradation information without the addition of low-frequency strong vibrations; Figure 5 (b) is a degradation information diagram after adding low-frequency strong vibration;

[0065] Figure 6 These are wavelet transform images from different periods provided in the embodiments of the present invention; wherein Figure 6 (a) is the time-frequency diagram of the initial stage of PMSM; Figure 6 (b) is a time-frequency diagram of the intermediate stage of PMSM degradation; Figure 6 (c) is a time-frequency diagram of the final stage of PMSM degradation;

[0066] Figure 7 This is a flowchart of the adaptive time-frequency transformation for weakening low-frequency strong vibrations provided in an embodiment of the present invention;

[0067] Figure 8 This is an energy percentage diagram for each frequency band provided in the embodiments of the present invention;

[0068] Figure 9 This is a time-spectrum energy segmentation and extraction diagram provided in an embodiment of the present invention; wherein Figure 9 (a) is a diagram of the decomposed low-frequency strong vibration component; Figure 9 (b)~ Figure 9 (f) is a graph showing the extraction of time-frequency domain information based on the magnitude of spectral energy;

[0069] Figure 10 This is a network training structure diagram provided in an embodiment of the present invention;

[0070] Figure 11 This is a prediction effect diagram provided by an embodiment of the present invention; wherein Figure 11 (a) shows the prediction performance of ASDN-TCN. Figure 11 (b) shows the prediction results of ASDN-CCN-LSTM;

[0071] Figure 12 This is an overall flowchart of ASDN-TCN provided in an embodiment of the present invention;

[0072] Figure 13 This is an optimized downhole accelerometer layout diagram provided in an embodiment of the present invention;

[0073] Figure 14 This is a schematic diagram of a multi-force accelerometer provided in an embodiment of the present invention;

[0074] Figure 15 This is a design drawing of a stabilizing mounting slot for a triaxial accelerometer sensor provided in an embodiment of the present invention;

[0075] Figure 16 This is an overall view of the triaxial accelerometer sensor stabilization mounting slot provided in an embodiment of the present invention;

[0076] Figure 17 These are single-phase current waveforms at different degradation stages provided in embodiments of the present invention; wherein, Figure 17 (a) Single-phase current waveform during the degradation period of approximately 3.5 mins; Figure 17 (b) Single-phase current waveform during the degradation period of approximately 1404.5-1408.5 mins; Figure 17 (c) Single-phase current waveform during the degradation period of approximately 7620.5-7624.5 mins.

[0077] Figure 18 This is a current spectrum complexity variation diagram provided in an embodiment of the present invention;

[0078] In the diagram: 1. Stator of permanent magnet synchronous motor; 2. Rotor of permanent magnet synchronous motor; 3. Permanent magnet synchronous motor; 4. Accelerometer; 5. Torque output shaft; 6. Gyroscope; 7. Torque generator; 8. Magnetic powder brake; 9. Vibration table; 10. Audio sensor; 11. Heating resistance wire; 12. Coupling; 13. Temperature sensor.

[0079] 14. Mounting slot; 15. Cable routing slot; 16. Mounting surface. Detailed Implementation

[0080] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Many specific details are set forth in the following description to provide a thorough understanding of the present invention. However, the present invention can be practiced in many other ways different from those described herein, and those skilled in the art can make similar modifications without departing from the spirit of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0081] I. Explanation of the Implementation Example:

[0082] Example 1

[0083] like Figure 1As shown, the method for predicting the remaining life of a downhole motor in an environment provided by this invention, namely, the method for real-time prediction of the remaining life of a downhole drilling and communication tool actuator (PMSM), includes:

[0084] S101 first collects measurement signals from the sensors in the system, including vibration, temperature, sound waves, torque, rotational speed, and stator three-phase current. The rotational speed is measured using a gyroscope, while the stator three-phase current is measured using the current sensor inside the PMSM, without the need for an external dedicated current sensor.

[0085] S102 uses an adaptive time-frequency transformation method to reduce the impact of low-frequency strong vibration noise on the original vibration signal. At the same time, it performs spectral entropy analysis on the current signal of the permanent magnet synchronous motor to extract the changes in its spectral complexity during the degradation process.

[0086] S103 allows for wavelet time-frequency energy transformation of vibration signals to extract energy changes during degradation, or direct extraction of their spectral energy. Other data can be used directly without feature extraction.

[0087] S104, then the dataset is divided into test set, training set and validation set, and the data is fed into the network to train the network. If the trained network performs well in the test set and validation set, the weight values ​​and network structure can be written into the downhole real-time prediction system through computer language, and the actuator (PMSM) life can be predicted in real time as the drilling equipment and its mud pulse communication system go deep into the formation.

[0088] Example 2

[0089] Based on the method for predicting the remaining life of downhole environmental motors provided in Embodiment 1, namely, the method for real-time prediction of the remaining life of downhole drilling and communication tool actuators (PMSMs), this embodiment provides a preferred embodiment, such as... Figure 2As shown, the process includes: First, acquiring measurement signals from sensors in the system, including vibration, temperature, sound waves, torque, rotational speed, and stator three-phase current. Rotational speed is measured using a gyroscope, while the stator three-phase current is measured using the internal current sensor of the PMSM, eliminating the need for an external dedicated current sensor. An adaptive time-frequency transformation method is used to reduce the impact of low-frequency strong vibration noise on the original vibration signal. Simultaneously, spectral entropy analysis is performed on the current signal of the permanent magnet synchronous motor to extract its spectral complexity changes during degradation. Wavelet time-frequency energy transformation can be performed on the vibration signal to extract its energy changes during degradation, or its spectral energy can be directly extracted. Other data can be used directly without feature extraction. The dataset is then divided into test, training, and validation sets, and the data is fed into the network for training. If the trained network performs well on the test and validation sets, the weight values ​​and network structure can be written into the downhole real-time prediction system using computer language. This allows for real-time prediction of the actuator (PMSM) life as the drilling equipment and its mud pulse communication system penetrate deep into the formation.

[0090] The vibration acquisition device (accelerometer), speed information acquisition device (gyroscope), temperature sensor, acoustic sensor, and torque sensor transmit the acquired information to the data acquisition unit via serial communication. The stator three-phase current sensor built into the permanent magnet synchronous motor obtains the motor current sensor measurement data through A / D sampling. At the same time, in order to provide a stable load torque input to the experimental motor, a magnetic powder brake control device - tension controller - is configured. The speed unit of the permanent magnet synchronous motor adopts a closed-loop control method, which provides control action based on the deviation between the speed set value and the actual measured value, controls the PWM signal generated by the inverter bridge of the motor drive board of the stabilizing platform, and thus controls the motor rotation so that the tool face angle measurement value follows the set value.

[0091] Example 3

[0092] Based on the downhole environment motor remaining life prediction method provided in Embodiment 1 or Embodiment 2, namely the implementation method for real-time prediction of the remaining life of the downhole drilling and communication tool actuator (PMSM), this embodiment provides a PMSM degradation experimental device simulating the downhole environment. The hardware structure diagram is as follows: Figure 3 As shown, it includes: accelerometer 4, which is a triaxial orthogonal gravity accelerometer;

[0093] The gyroscope 6 uses a gyroscope as a speed information acquisition device.

[0094] The audio sensor 10 uses a sound wave sensor, the torque sensor transmits the collected information to the data acquisition unit through serial communication, and the stator three-phase current sensor built into the permanent magnet synchronous motor 3 obtains the motor current sensor measurement data through A / D sampling.

[0095] Meanwhile, in order to provide a stable load torque input to the experimental motor, a magnetic powder brake control device, namely a tension controller, is configured.

[0096] The speed unit of the permanent magnet synchronous motor 3 (including the permanent magnet synchronous motor stator 1 and the permanent magnet synchronous motor rotor 2) adopts a closed-loop control method. It provides control action based on the deviation between the speed set value and the actual measured value, controls the PWM signal generated by the inverter bridge of the motor drive board of the stabilizing platform, and thus controls the rotation of the permanent magnet synchronous motor 3 so that the tool face angle measurement value follows the set value.

[0097] Specific structure, such as Figure 4 As shown, it includes: permanent magnet synchronous motor stator 1, permanent magnet synchronous motor rotor 2, permanent magnet synchronous motor 3, acceleration sensor 4, torque output shaft 5, gyroscope 6, torque generator 7, magnetic powder brake 8, vibration table 9, audio sensor 10, heating resistance wire 11, coupling 12, and temperature sensor 13.

[0098] The permanent magnet synchronous motor 3 is connected to the vibration table 9 via a base. The vibration table provides the device on the table with low-frequency strong vibrations to simulate the downhole vibration environment. Two acceleration sensors 4 and four temperature sensors 13 are fixedly connected to the housing of the permanent magnet synchronous motor 3, which are used to measure the acceleration in the lateral and vertical directions of the PMSM and the surrounding temperature, respectively. It is surrounded by large heating resistance wires 11 on its upper and left sides to simulate the high-temperature environment downhole. At the same time, an acoustic sensor is installed at its lower part to collect sound signals during the degradation process. The PMSM torque output shaft 5 is connected to the torque converter 7 via a coupling 12. The magnetic powder brake 8 is also connected to the torque converter 7 via a coupling 12.

[0099] Example 4

[0100] Based on the method for predicting the remaining life of downhole environment motors provided in Embodiment 1 or Embodiment 2, namely, the method for real-time prediction of the remaining life of downhole drilling and communication tool actuators (PMSMs), this invention proposes an adaptive method for attenuating low-frequency large vibrations in vibration signals (ASDN). This method can adaptively identify and remove low-frequency strong vibration signals in vibration signals through time-frequency transformation. Figure 5 The comparison shows that the original vibration image clearly expresses the vibration change trend from the start to the fault time, but after adding low-frequency strong vibrations, it almost completely masks all useful information in the time domain. Figure 5 (a) is a degradation information diagram without the addition of low-frequency strong vibration; Figure 5 (b) is: Degradation information diagram after adding low-frequency strong vibration.

[0101] At this point, wavelet transform is introduced to perform time-frequency transformation on the vibration image in time segments.

[0102]

[0103] f(t) is the original signal, e -iwt Representing vibration, wavelet transform uses an integral form to describe the original signal at e -iwt The projection on the screen, g(tu) is a window function. By adding a window, the shortcomings of Fourier transform in that it can only display frequency information and cannot display time-frequency transform are solved. The vibration characteristics and window signal are replaced with wavelet function ψ(a,b,t) to realize video feature extraction under variable window. a and b represent the scaling scale and time factor.

[0104] After wavelet transform extraction, such as Figure 6 As shown, where, where Figure 6 (a) is: Time-frequency diagram of the initial stage of PMSM; Figure 6 (b) is a time-frequency diagram of the intermediate stage of PMSM degradation; Figure 6 (c) is a time-frequency diagram of the final stage of PMSM degradation.

[0105] It is evident that there are consistently high-brightness values ​​in the low-frequency range, indicating that large low-frequency vibrations persist throughout the entire lifespan. The Morlet wavelet function with a normally distributed envelope is used in the wavelet transform. This wavelet basis offers a better balance in terms of local optimization of time and frequency compared to other wavelet bases. Furthermore, the Morlet wavelet shares the characteristic of impact attenuation with vibration signals.

[0106]

[0107] Where t is time, It is an oscillation component, and v(t) represents the signal of oscillation decay.

[0108] Example 5

[0109] Based on the adaptive attenuation method (ASDN) for large low-frequency vibrations in vibration signals proposed in Example 4, preferably, as follows: Figure 7 The flowchart shown is a method for adaptively reducing low-frequency large vibrations in vibration signals, including:

[0110] First, vibration signals are acquired, and then multi-level wavelet packet decomposition is performed on the vibration signals to decompose the time-domain signals into the time-frequency domain. The wavelet packet coefficients of the decomposed signals are rearranged to be in the correct order. The decomposed wavelet packet coefficients are reconstructed to construct the signal spectrum. The energy proportion of each segment of the spectrum is calculated and sorted. The part with high energy is selected as the main information of degradation. Due to the characteristics of low-frequency strong vibration in the well, the part with the highest energy can be regarded as the strong vibration part and discarded after extraction. This achieves the purpose of adaptively selecting PMSM degradation information.

[0111] in Figure 8 The energy percentage of each frequency band is graphically displayed. It can be seen that the first frequency band has the highest energy percentage, reaching 98.9%. Therefore, the energy of the other frequency bands is amplified to obtain the desired result. Figure 8 The comparison diagram is shown below. Next, the energy of each frequency band is sorted and segmented, resulting in the following... Figure 9 The time-spectrum energy extraction plot shown is from this Figure 9 As can be seen, low-frequency strong vibrations have been segmented, while other frequency bands with higher energy proportions and related to degradation information have also been segmented and used as training data for later stages. Figure 9 (a) shows the decomposed low-frequency strong vibration component; Figure 9 (b)~ Figure 9 (f) is a graph of time-frequency domain information extracted according to the magnitude of spectral energy.

[0112] like Figure 10 The diagram shown is a network training structure diagram provided in an embodiment of the present invention.

[0113] Example 6

[0114] Based on the adaptive attenuation method (ASDN) for large low-frequency vibrations in vibration signals proposed in Example 5, preferably, as follows: Figure 12 As shown, the overall process of ASDN-TCN (ASDN-TCN-based training method) provided in this embodiment of the invention includes:

[0115] Using time-frequency information extracted from ASDN as input data for the TCN network, this method has the advantages of shorter training time and higher accuracy compared to traditional ASDN-CNN or ASDN-CNN-LSTM methods. Temporal convolutional neural networks are a variant of ordinary convolutional neural networks, incorporating extended causal convolution and ResNet networks. Causal convolution is based on regular convolution with zero-padded input variables on one side of the outermost layer. Adding this variable ensures that the predicted value at time t in the time prediction task is only related to the previous (t+1) time steps, i.e.: p(x t / x t-1 ,x t-2The output value of x1) at time t is related to the previous t-1 times but not to t+1, t+2, etc. This extended causal convolution method, which uses unilateral padding based on the CNN window size, keeps the input and output dimensions consistent. This ensures that future information is not leaked to a past moment. Furthermore, in predicting remaining lifetime, the network should focus more on the trend of the input data, thus reducing excessive focus on changes in a particular part. Therefore, it is necessary to appropriately ignore and sequentially skip certain adjacent data in the network, thereby obtaining a larger receptive field. This operation is called network expansion. In an expanded network, the original data information that can skip a specific stride is combined with the previous causal convolution to form an expanded causal convolutional network. The receptive fields of a regular 1D convolutional network and an expanded causal convolutional network are given by the following formula:

[0116]

[0117] In the formula, k is the length of the convolutional neural network filter, d is the dilation rate, L is the size of the dilated window, and j is the number of layers.

[0118] in, Figure 11 This is a prediction effect diagram provided by an embodiment of the present invention; Figure 11 (a) shows the prediction performance of ASDN-TCN. Figure 11 (b) shows the prediction results of ASDN-CCN-LSTM.

[0119] Example 7

[0120] Based on the downhole environment motor remaining life prediction method provided in Embodiment 1 or Embodiment 2, namely the implementation method for real-time prediction of the remaining life of the downhole drilling and communication tool actuator (PMSM), this embodiment of the invention provides a ring layout of a triaxial orthogonal gravity accelerometer sensor optimized based on downhole vibration characteristics, such as... Figure 13 As shown, eight triaxial orthogonal gravity accelerometer sensors are evenly arranged within the drill string at the location of the permanent magnet synchronous motor. Their x-axis points towards the center of the drilling platform axis, their y-axis represents the drill string's forward direction or its opposite, and their z-axis represents the tangent direction of the drill string at the installation position. This annular arrangement of the triaxial accelerometers allows for the extraction of accelerations in the axial normal and rotational tangent directions, which are significantly affected by low-frequency, large vibrations, leaving only the drilling forward acceleration value, which is less affected by downhole low-frequency, large vibrations. This configuration achieves accelerometer redundancy, preventing data loss due to sensor damage, and also allows for the extraction of useful vibration signals from the low-frequency, large-vibration downhole interference signals around the actuator.

[0121] Of course, based on this layout, the influence of downhole vibration can be mitigated by considering the relationship between the calculated quantities on the x and z axes. Firstly...

[0122]

[0123] P is the angular velocity of the PMSM. During steady-state drilling, P changes continuously. However, because the radius R of the downhole motor is very small, but the downhole vibration intensity is very high, the value of a related to downhole vibration is significant. vz a vx The term is much larger than the terms related to R, so the terms related to R are omitted. zx a represents the gravitational acceleration component in the zx plane; vz and a vx The components of vibrational acceleration along the z-axis and x-axis are given; rotational acceleration consists of tangential and centripetal accelerations, respectively. And R(pπ / 180) 2 / 9.8. Based on this, Figure 14 Analysis shows that:

[0124]

[0125] The above equation is an analysis of triaxial accelerometer sensors distributed in a circle at arbitrary angular positions. Omitting terms related to R, we can obtain the following from the equation:

[0126]

[0127] When γ = 180°, that is, when taking the two accelerometer values ​​of the symmetrical distribution...

[0128]

[0129] At this point, it can be observed that this calculation can eliminate other influencing factors, leaving only the effect of vibration acceleration on a certain axis. However, due to the sensor layout, strong downhole vibrations occur at point a... vz1 and a vz2 The components in these two directions are almost completely opposite, but the fault information is unevenly distributed in these two directions, so adding the two values ​​can extract the fault information. The analysis for the x-axis is similar.

[0130] Example 8

[0131] Based on the PMSM degradation experimental apparatus simulating a downhole environment provided in Example 3, this example provides an accelerometer mounting platform design structure, such as... Figure 15 , Figure 16As shown, the mounting slot 14 of the accelerometer sensor 4 (a triaxial orthogonal gravity accelerometer sensor) is installed at the same distance from the mounting slots of the other triaxial gravity accelerometers, and the coordinate system of the mounting platform carrier coincides with the center of the circular surface. The signal and power lines of the accelerometer are connected to the downhole data processing circuit board through the wiring groove 15. The power module of the data processing circuit board provides power to the accelerometer, and the accelerometer measurement signal is transmitted to the data processor through IIC. The accelerometer mounting platform is mounted on the downhole permanent magnet synchronous motor through the mounting surface 16. After the sensor and mounting platform are installed and fixed, high-temperature silicone rubber is injected for sealing.

[0132] Example 9

[0133] Based on the PMSM degradation experimental apparatus for simulating a downhole environment provided in Example 3, preferably, eight accelerometers 4 (triaxial orthogonal gravity accelerometer sensors) are installed symmetrically in eight directions on the permanent magnet synchronous motor 3 to measure the axial and circular tangential vibrations of the permanent magnet synchronous motor. Their x-axis directions all point to the center of the drilling platform axis, the y-axis is the drill string forward direction or its opposite direction, and the z-axis is the tangential direction of the drill string at the installation position.

[0134] To simulate the resistance experienced by a downhole permanent magnet synchronous motor during rotation, the shaft of the permanent magnet synchronous motor 3 is connected to the magnetic powder brake 8 via a rotational coupler (including torque output shaft 5, torque converter 7, and coupling 12, such as...). Figure 4 The magnetic powder brake 8 and the magnetic powder brake control device (fully automatic tension controller) are connected together to control the magnitude of its resisting torque force. At the same time, a PMSM speed control sensor and a torque sensor are configured between the two rotary couplers to detect the speed and torque force of the PMSM.

[0135] The aforementioned equipment is then connected to a vibration table 9 capable of generating frequencies from 0 to 200 Hz to simulate the generation of low-frequency large vibrations in the well (the vibration table is arranged in a non-axial direction). This vibration is superimposed on a useful vibration signal and detected by a velocity sensor 4.

[0136] Heating resistance wires 11, used solely for heating, are installed around the PMSM. A continuously increasing current is passed through the heating resistance wires 11 to simulate the temperature increasing with drilling depth. Temperature sensors 13 are evenly distributed on each side of the permanent magnet synchronous motor 3 to detect temperature changes. The average value of all temperature sensor readings 13 is used as the true temperature value of the PMSM.

[0137] The three-phase current of the permanent magnet synchronous motor stator 1 of the permanent magnet synchronous motor 3 can be obtained through the current sensor of the PMSM itself. The principle is that the current of each phase can be determined by measuring the voltage of the constantan wire connected to each phase. At the same time, an audio sensor 10, i.e., a sound wave sensor, can be configured around it to collect sound signals during the experiment.

[0138] All the above sensors adopt a data sampling rule of collecting data every 1 second every 1 minute. Since there are many high-frequency components in vibration signals, current signals, and sound wave signals, the sampling rate is set to 25.6 kHz according to Shannon's sampling theorem. This can accurately collect the original signal from 0-5 kHz. For characteristic signals that change relatively slowly, such as speed, torque, and temperature, a sampling frequency of 10 kHz can better restore the original signal.

[0139] After collecting multiple sets of raw full-lifecycle data of the same type of PMSM, the vibration and the vibration in the circular tangential direction (with strong vibrations eliminated) were first processed. Then, feature extraction was performed on the data to reduce the interference of low-frequency strong noise on the sound waves and vibration signals. The feature-extracted data was then grouped into test, training, and validation sets, and fed into a deep learning network for training. After training, the network's prediction accuracy was verified using the validation and test sets. If the accuracy was high, the trained network could be integrated into a downhole real-time prediction system, accompanying the drilling equipment into the formation to predict the real-time remaining service life of the PMSM. This achieved real-time prediction under low computing power and improved the reliability and operational safety of the downhole drilling and communication systems. Furthermore, the predicted remaining service life could be returned to the surface in real-time via a downhole pulser communication device. The remaining service life is just a numerical value; compared to various dynamic signals collected by sensors at high sampling frequencies, its transmissibility is greatly increased.

[0140] A reasonable sensor layout was configured based on the downhole environment, downhole equipment layout, and the internal structure of the permanent magnet synchronous motor 3. A redundant sensor arrangement was adopted to prevent damage to the sensors from the downhole environment, which could affect data acquisition. Furthermore, all sensors are high-performance sensors resistant to high temperatures, high pressure, and strong vibrations to cope with the complex working conditions downhole.

[0141] Example 10

[0142] Based on the PMSM degradation experimental device simulating the downhole environment provided in Embodiment 9, this embodiment of the invention designs a set of sensor data hardware electrical connection structure. The acceleration sensor 4, temperature sensor 13, and torque sensor (mounted on the torque generator 7) transmit data to the computer for recording via serial communication, while the stator three-phase current and speed sensor transmit data to the PMSM control board via the CAN bus.

[0143] The main control board integrates a data acquisition unit and a speed control unit. The data acquisition unit receives data transmitted from the data processor, obtains measurement data from the motor current sensor through A / D sampling, and acquires measurement data from the motor speed sensor through serial communication. It then processes the acquired data to obtain the measured values ​​of the motor speed and current. The speed control unit employs a closed-loop control method, providing control based on the deviation between the set speed value and the actual measured value. It controls the PWM signal generated by the inverter bridge circuit of the motor drive board, thereby controlling the motor rotation so that the PMSM speed value follows the set value.

[0144] In this invention embodiment, an adaptive separation of downhole noise (ASDN) feature extraction method is proposed based on wavelet transform-based time-frequency spectrum extraction. This method first converts the vibration and acoustic signals into a time-frequency space (the frequency of the acoustic signal is directly affected by the frequency of the vibration signal). Then, based on prior knowledge of the downhole vibration frequency band distribution, the information in the 0-200Hz range is attenuated, and the spectrum with a high frequency band energy ratio is selected as the input for network training, reducing redundant information and the influence of downhole vibration noise interference. This method can adaptively remove low-frequency vibration signals and select frequency band signals that reflect PMSM faults. This scheme utilizes the characteristics of strong low-frequency vibrations in the well and employs the ASDN method for feature extraction, achieving good results in extracting the energy change trend of a specific frequency band.

[0145] In embodiments of this invention, a CNN-based TCN network is constructed using TensorFlow. This deep learning network incorporates an attention mechanism, and grouped data is input into the network for iterative training. CNN can better extract the large amount of feature information collected, reducing the number of network parameters and thus reducing training time. The TCN network broadens the network's observation field during training, allowing it to better observe data trends and focusing training attention on these trends, significantly ignoring irrelevant information and greatly improving training speed. The attention mechanism effectively focuses the network's attention on a variable threshold, enhancing its adaptability to different training data. This network better observes data trends, and the computational mechanisms in CNN and TCN help reduce the number of network parameters that need to be trained, greatly reducing training time.

[0146] In this embodiment of the invention, a pre-trained network is used. The pre-trained network weight function and network structure are loaded into the downhole real-time prediction system using computer language. Because it is a pre-trained model, the extracted features used during training do not need to be stored in the cache, which solves the problem that downhole data cannot be stored in large quantities. Only the prediction results of the algorithm need to be transmitted back to the surface. Moreover, there is no large amount of computation based on backpropagation, which greatly reduces the amount of computation, thereby realizing downhole PMSM real-time prediction and real-time surface transmission with low computing power.

[0147] The information interaction and execution process between the above-mentioned devices / units are based on the same concept as the method embodiments of the present invention. For details on their specific functions and technical effects, please refer to the method embodiments section, and they will not be repeated here.

[0148] Those skilled in the art will understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the functions described above can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this invention. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0149] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include at least: any entity or device capable of carrying the computer program code to a photographing device / terminal device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium. Examples include USB flash drives, portable hard drives, magnetic disks, or optical disks.

[0150] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0151] II. Application Examples:

[0152] An application embodiment of the present invention provides a computer device, which includes: at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, wherein the processor executes the computer program to implement the steps in any of the above-described method embodiments.

[0153] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, can implement the steps in the above-described method embodiments.

[0154] The present invention also provides an information data processing terminal, which is used to provide a user input interface to implement the steps in the above method embodiments when executed on an electronic device. The information data processing terminal is not limited to mobile phones, computers, and switches.

[0155] The present invention also provides a server, which, when executed on an electronic device, provides a user input interface to implement the steps of the above method embodiments.

[0156] The present invention provides a computer program product that, when run on an electronic device, enables the electronic device to implement the steps described in the various method embodiments.

[0157] III. Experiments and Simulations

[0158] Experiments and simulations were conducted on the method for predicting the remaining life of a motor in a downhole environment provided in the embodiments of the present invention, such as... Figure 17 (a) Single-phase current waveform during the degradation period of approximately 3.5 mins;

[0159] Figure 17 (b) Single-phase current waveform during the degradation period of approximately 1404.5-1408.5 mins;

[0160] Figure 17 (c) Single-phase current waveform during the degradation period of approximately 7620.5-7624.5 mins.

[0161] As the motor degrades, the spectral complexity of the single-phase current continuously increases until it reaches a certain point where the motor can no longer operate normally, at which point it is considered damaged. The transformation curve of the spectral complexity can be used to extract frequency changes that are not observable in the time domain, serving as a feature extraction method for current data. This data can then be input into the network for training. The change in the current spectral complexity is shown in the figure below. Figure 18 As shown.

[0162] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications, equivalent substitutions and improvements made by those skilled in the art within the scope of the technology disclosed in the present invention and within the spirit and principles of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A method for predicting the remaining life of a motor in a downhole environment, characterized in that, The method for predicting the remaining life of a motor in the downhole environment includes: using sensors to collect measurement signals of a permanent magnet synchronous motor, including vibration, temperature, sound waves, torque, speed, and stator three-phase current; using an adaptive time-frequency transformation method to reduce the interference of low-frequency strong vibration noise on the original vibration signal; and simultaneously performing spectral entropy analysis on the stator three-phase current signal of the permanent magnet synchronous motor to extract the change in spectral complexity of the stator three-phase current signal during the degradation process. Wavelet time-frequency energy transformation is performed on the vibration signal to extract the energy change of the vibration signal during the degradation process. The collected temperature, sound wave, torque, speed, extracted stator three-phase current signal and extracted vibration signal are divided into training set, test set and validation set. The above datasets are fed into the network for training after feature extraction. If the trained network performs well in the test set, the weight values ​​and network structure are written into the downhole real-time prediction system through computer language. The system is then sent into the formation with the drilling equipment and mud pulse communication system to predict the remaining service life of the permanent magnet synchronous motor based on the downhole environment. Before using sensors to collect measurement signals from the permanent magnet synchronous motor, the spatial layout of the permanent magnet synchronous motor sensors is carried out, including: eight acceleration sensors (4) are evenly arranged in the drill string at the location of the permanent magnet synchronous motor. The acceleration sensors (4) are triaxial orthogonal gravity accelerometers. The x-axis direction points to the center of the drilling platform axis, the y-axis is the drill string forward direction or its opposite direction, and the z-axis is the drill string circular tangent direction at the installation position. The acceleration in the axis normal direction and the rotation tangent direction, which are greatly affected by low-frequency strong vibration, is extracted. The spatial layout of the permanent magnet synchronous motor sensor includes: mitigating the impact of downhole vibration by calculating the relationship between quantities on the x-axis and z-axis, expressed as: ； For PMSM, during steady-state drilling, It will constantly change, but because the radius R of the downhole motor is very small, while the intensity of downhole vibration is very high, the a related to downhole vibration is... vz a vx The term is much larger than the terms related to R, so the terms related to R are omitted. zx a represents the gravitational acceleration component in the zx plane; vz and a vx The components of vibrational acceleration along the z-axis and x-axis are given; rotational acceleration consists of tangential and centripetal accelerations, respectively. and Analysis revealed that: ； The above equation is an analysis of triaxial accelerometer sensors distributed in a circle at arbitrary angular positions. After omitting the terms related to R, we get the following from the above equation: ； when =180°, when taking the two accelerometer values ​​of the balanced distribution. ； Only the vibration acceleration interference on a certain axis remains, and the strong downhole vibration is at a vz1 and a vz2 The components in these two directions are completely opposite, but the fault information is unevenly distributed in these two directions, through a vz1 and a vz2 The fault information is extracted by adding the components in these two directions.

2. The method for predicting the remaining life of a motor in a downhole environment according to claim 1, characterized in that, The process of performing wavelet time-frequency energy transformation on the vibration signal to extract the energy change of the vibration signal during the degradation process includes: converting the vibration signal and the acoustic signal into the time-frequency space, then weakening the information of 0-200 Hz through the prior downhole vibration frequency band distribution law, and selecting the spectrum with a high frequency band energy ratio as the input for network training. The vibration signal is subjected to time-frequency transformation in time intervals using wavelet transform, and the expression is as follows: ； In the formula, f(t) is the original signal, and e -iwt Representing vibration, wavelet transform uses an integral form to describe the original signal at e -iwt The projection onto the surface, g(tu), is a window function. Windowing addresses the limitation of Fourier transform in only displaying frequency information, not time-frequency transform, by replacing vibration characteristics and the windowed signal with wavelet functions. This enables time-frequency feature extraction under a variable window, where a and b represent the scaling scale and time factor, respectively.

3. The method for predicting the remaining life of a motor in a downhole environment according to claim 2, characterized in that, In the wavelet transform extraction, the common characteristics of the Morlet wavelet and the impact attenuation of the vibration signal are: ； Where t is time, It is an oscillation component, and v(t) represents the signal of oscillation decay.

4. The method for predicting the remaining life of a motor in a downhole environment according to claim 1, characterized in that, The process of feeding the aforementioned dataset into the network for training includes: constructing a Temporal Convolutional Neural Network (TCN) using TensorFlow; combining the TCN network with an attention mechanism and a CNN to construct a deep learning network; inputting the grouped data into the network for parameter training; the CNN further extracts a large amount of feature information collected, while the TCN network broadens the network's perspective during training and obtains the changing trends of the data.

5. The method for predicting the remaining life of a motor in a downhole environment according to claim 1, characterized in that, The downhole real-time prediction system is equipped with a main control board, which includes a data acquisition unit and a rotation speed control unit. The data acquisition unit obtains stator three-phase current measurement data through A / D sampling, acquires permanent magnet synchronous motor speed measurement data through serial communication, and processes the acquired data to obtain the measured values ​​of motor speed and current. The speed control unit adopts a closed-loop control method. It provides control action based on the deviation between the speed set value and the actual measured value. It controls the PWM signal generated by the inverter bridge of the motor drive board to control the rotation of the permanent magnet synchronous motor so that the PMSM speed value follows the set value.

6. A PMSM degradation experimental apparatus for a downhole environment, implementing the downhole environment motor remaining life prediction method according to any one of claims 1 to 5, characterized in that, The PMSM degradation experimental setup for the downhole environment includes: The permanent magnet synchronous motor (3) is connected to the vibration table (9) via a base. The vibration table (9) is used to provide a low-frequency strong vibration to simulate the downhole vibration environment. An acceleration sensor (4) and a temperature sensor (13) are fixedly connected to the housing of the permanent magnet synchronous motor (3) to measure the acceleration in the horizontal and vertical directions of the PMSM and the surrounding temperature, respectively. The top and left side of the permanent magnet synchronous motor (3) housing are covered with heating resistance wires (11) to simulate the temperature environment downhole.

7. The PMSM degradation experimental apparatus for downhole environments according to claim 6, characterized in that, The permanent magnet synchronous motor (3) has an acoustic sensor installed on the lower part of its housing to collect sound signals during the degradation process. The PMSM torque output shaft (5) is connected to the torque generator (7) via a coupling (12). The magnetic powder brake (8) is connected to the torque generator (7) via a coupling (12).

8. The PMSM degradation experimental apparatus for downhole environments according to claim 6, characterized in that, The permanent magnet synchronous motor (3) has multiple mounting slots (14) on its housing for mounting accelerometers (4); the signal lines and power lines of the accelerometers (4) are connected to the downhole data processing circuit board via wiring channels (15). The power module of the data processing circuit board provides power to the accelerometers (4), and the accelerometer measurement signals are transmitted through... The data is transmitted to the data processor; the accelerometer (4) is mounted on the downhole permanent magnet synchronous motor via the mounting surface (16).

Images