Axial flux motor stator immersion monitoring and control method, device and equipment

Abstract

The application relates to a kind of axial flux motor stator impregnation monitoring and control method, device and equipment, including obtaining the capacitance test data collected by capacitive probe sensor in impregnation tank, the vibration damping data collected by vibrating viscometer sensor and the microwave attenuation data collected by microwave resonant cavity sensor, construct multidimensional acquisition data set;Each data in multidimensional acquisition data set is regarded as observation input to extended Kalman filter;Extended Kalman filter takes dielectric constant, viscosity and moisture content as state variable, establishes state transition matrix and observation matrix, and recursively estimates the estimated state vector corresponding to viscosity by prediction step and update step;According to the deviation value between the estimated state value corresponding to the estimated state vector and the preset target value, the adjusting amount of viscosity is calculated using PID control;The adjusting amount is output to the temperature control system of impregnation tank to adjust the oil viscosity of impregnation oil in the impregnation tank.

Description

Technical Field

[0001] This application relates to the field of motor manufacturing technology, and in particular to a method, apparatus and equipment for monitoring and controlling stator impregnation of an axial flux motor. Background Technology

[0002] Stator winding impregnation typically employs preset parameters such as vacuum pressure, impregnation oil pressure, and oil viscosity, executing a fixed-duration impregnation cycle in an open-loop manner. However, during actual impregnation, factors such as oil aging, moisture contamination, and temperature fluctuations can cause dynamic changes in key physicochemical properties like dielectric constant, viscosity, and water content, thereby affecting the resin's penetration into the winding gaps and the consistency of curing. In existing technologies, operators often rely solely on experience or offline sampling and testing to adjust process parameters, resulting in issues such as response lag and coarse parameter tuning, making it difficult to guarantee the repeatability and stability of stator impregnation quality across different batches. Summary of the Invention

[0003] The purpose of this invention is to provide a method, device, and equipment for monitoring and controlling stator impregnation of axial flux motors. This invention can solve the technical problem of poor consistency between penetration effect and curing caused by dynamic changes in the physical and chemical properties of oil in the existing stator impregnation process. This can improve the adaptability of the impregnation process to complex working conditions and ensure the stability, reliability, and repeatability of the impregnation quality of different batches of stator windings.

[0004] To achieve the above objectives, a first aspect of this disclosure provides a method for monitoring and controlling the stator impregnation of an axial flux motor, comprising: Acquire capacitance test data collected by a capacitive probe sensor set in an impregnation tank, vibration damping data collected by a vibratory viscometer sensor, and microwave attenuation data collected by a microwave resonant cavity sensor to construct a multi-dimensional data collection set; Each data point in the multi-dimensional data set is used as an observation input to the extended Kalman filter; the extended Kalman filter uses dielectric constant, viscosity and water content as state variables to establish a state transition matrix and an observation matrix, and recursively estimates the estimated state vector corresponding to viscosity through prediction and update steps; Based on the deviation between the estimated state value corresponding to the estimated state vector and the preset target value, PID control is used to calculate the viscosity adjustment amount; The adjustment amount is output to the temperature control system of the impregnation tank to adjust the viscosity of the impregnation oil in the impregnation tank.

[0005] In a preferred embodiment, the recursive estimation of the estimated state vector corresponding to the viscosity through the prediction step and the update step includes: In the prediction step, the posterior state estimate of the previous time step is multiplied by the state transition matrix to obtain the prior state estimate of the current time step, and the posterior error covariance matrix of the previous time step is multiplied by the transpose of the state transition matrix and then added to the process noise covariance matrix to obtain the prior error covariance matrix of the current time step. In the update step, the Kalman gain matrix is ​​calculated based on the prior error covariance matrix, the observation matrix, and the measurement noise covariance matrix; The observation vector is formed from the data in the multi-dimensional collection of data at the current moment. The measurement residual at the current moment is determined based on the observation vector, the observation matrix, and the prior state estimate. Based on the Kalman gain matrix, the measurement residual, and the prior state estimate, the posterior state estimate is determined as the estimated state vector corresponding to the current time.

[0006] In a preferred embodiment, the step of assembling an observation vector from the data in the multi-dimensional collected data set at the current moment, and determining the measurement residual at the current moment based on the observation vector, the observation matrix, and the prior state estimate, includes: The data in the multi-dimensional data collection at the current moment are arranged into a multi-dimensional data array according to a preset fixed order, wherein each element in the multi-dimensional data array corresponds to the original measurement value of a sensor channel; The multidimensional array of data is used as the observation vector at the current moment, and the prior state estimate at the current moment is substituted into the pre-established observation model. The observation model describes the linear mapping relationship from the state variable to the observation variable through the observation matrix, and calculates the predicted observation vector under the prior state estimate at the current moment. Subtract the element at the corresponding position in the predicted observation vector from each element in the observation vector to obtain the difference value corresponding to each sensor channel; The differences corresponding to each sensor channel are arranged in the same order as the observation vector to form a multidimensional difference array, which serves as the measurement residual at the current moment.

[0007] In a preferred embodiment, determining the posterior state estimate as the estimated state vector corresponding to the current time moment based on the Kalman gain matrix, the measurement residual, and the prior state estimate includes: Perform matrix multiplication between the Kalman gain matrix and the vector corresponding to the measurement residual to obtain a correction vector equal to the number of state variables. Each row of the Kalman gain matrix corresponds to a state variable dimension, and each column corresponds to a measurement channel. Each element in the correction vector represents the adjustment amount of the corresponding prior state estimate using the measurement residual at the current time. The prior state estimation vector and the correction vector are added element-wise to obtain an intermediate state estimation vector without constraint correction. The intermediate state estimation vector is used to characterize the state estimation after fusing the measurement values. Constraint correction is performed on the intermediate state estimation vector to obtain the posterior state estimation value, and the posterior state estimation value is determined as the estimated state vector corresponding to the current time.

[0008] In a preferred embodiment, performing constraint correction on the intermediate state estimation vector to obtain a posterior state estimate includes: If the value of the intermediate state estimation vector is greater than the maximum allowable value, then it is assigned the corresponding maximum allowable value to obtain the posterior state estimation value corresponding to the current time. If the value of the intermediate state estimation vector is less than the minimum allowable value, then the value is assigned to the minimum allowable value to obtain the posterior state estimation value corresponding to the current time.

[0009] In a preferred embodiment, the step of calculating the viscosity adjustment amount using PID control based on the deviation between the estimated state value corresponding to the estimated state vector and the preset target value includes: The deviation value at the current moment is input into the fuzzy controller, which is used to fuzzify the input deviation into a signed fuzzy value, and to obtain the corresponding proportional gain coefficient by defuzzifying the fuzzy value using the centroid method. The proportional term corresponding to the current time is determined by multiplying the proportional gain coefficient corresponding to the current time with the corresponding deviation value; the integral term is determined by multiplying the cumulative deviation value corresponding to the previous time with the integral gain coefficient; and the differential term is determined by multiplying the difference between the deviation value corresponding to the current time and the deviation value corresponding to the same dimension at the previous time with the differential gain coefficient. The adjustment amount corresponding to the current moment is determined based on the corresponding proportional term, integral term, and derivative term.

[0010] In a preferred embodiment, outputting the adjustment amount to the temperature control system of the impregnation tank to adjust the viscosity of the impregnation oil in the impregnation tank includes: Obtain the measured temperature data of the impregnating oil in the impregnation tank at the current moment, as well as the viscosity-temperature relationship table of the impregnating oil under normal pressure; The reference viscosity at the current moment is obtained by querying the viscosity-temperature relationship table based on the measured temperature data. The viscosity adjustment amount in the adjustment amount is mathematically superimposed with the reference viscosity to obtain the target viscosity at the current moment; The target temperature corresponding to the target viscosity is determined by back interpolation of the viscosity-temperature relationship table, based on the temperature difference between the target temperature at the current moment and the current oil temperature in the measured temperature data; The temperature control system is controlled according to the temperature difference to perform temperature control of the impregnating oil, so as to adjust the viscosity of the impregnating oil in the impregnation tank.

[0011] A second aspect of this disclosure provides a stator impregnation monitoring and control device for an axial flux motor, the device comprising: The acquisition module is configured to acquire capacitance test data collected by a capacitive probe sensor set in the impregnation tank, vibration damping data collected by a vibratory viscometer sensor, and microwave attenuation data collected by a microwave resonant cavity sensor, and construct a multi-dimensional acquisition data set. The input template is configured to take each data point in the multi-dimensional collected data set as the observation input to the extended Kalman filter; the extended Kalman filter uses dielectric constant, viscosity and water content as state variables to establish a state transition matrix and an observation matrix, and recursively estimates the estimated state vector corresponding to viscosity through prediction steps and update steps. The calculation module is configured to calculate the viscosity adjustment amount using PID control based on the deviation between the estimated state value corresponding to the estimated state vector and the preset target value. The output module is configured to output the adjustment amount to the temperature control system of the impregnation tank to adjust the viscosity of the impregnation oil in the impregnation tank.

[0012] A third aspect of this disclosure provides an electronic device, comprising: A memory on which computer programs are stored; A processor for executing the computer program in the memory to implement the steps of the method of any one of the first aspects.

[0013] A fourth aspect of this disclosure provides a computer-readable storage medium storing program code that is executed by a processor to implement the steps of the method described in any of the first aspects.

[0014] This invention provides a method, apparatus, and equipment for monitoring and controlling stator impregnation in an axial flux motor. Compared with the prior art, it has the following advantages: By integrating a capacitive probe sensor, a vibratory viscometer sensor, and a microwave resonant cavity sensor, real-time acquisition of capacitance test data, vibration damping data, and microwave attenuation data is achieved, constructing a multi-dimensional data set. This enhances the comprehensiveness of sensing microscopic changes in the impregnation environment. The multi-dimensional data set is then fed into an extended Kalman filter as an observation input. Using a state transition matrix and an observation matrix with dielectric constant, viscosity, and water content as state variables, the estimated state vector corresponding to viscosity is recursively estimated through prediction and update steps. This enables accurate online reconstruction of key physicochemical indicators even in the presence of noise and nonlinear interference. Furthermore, based on the deviation between the estimated state value corresponding to the estimated state vector and the preset target value, PID control is used to calculate the viscosity adjustment amount. This adjustment amount is then output to the temperature control system of the impregnation tank to adjust the viscosity of the impregnating oil within the tank. It can solve the problem that traditional open-loop control or experience-based parameter tuning cannot respond in time to the dynamic changes in physicochemical properties caused by oil aging, moisture mixing, and temperature fluctuations. It avoids uneven penetration and inconsistent curing caused by rough parameter tuning, improves the closed-loop control accuracy of stator impregnation process, batch-to-batch quality repeatability and production process reliability, enhances the consistency and stability of motor stator insulation treatment under complex working conditions, and improves the effectiveness and robustness of impregnation process to adapt to complex working conditions.

[0015] Other features and advantages of this disclosure will be described in detail in the following detailed description section. Attached Figure Description

[0016] The accompanying drawings are provided to further illustrate the present disclosure and form part of the specification. They are used together with the following detailed description to explain the present disclosure, but do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart illustrating a method for monitoring and controlling the stator immersion of an axial flux motor, as shown in the embodiments of the specification.

[0017] Figure 2 This is one implementation shown in the embodiment of the specification. Figure 1 A flowchart of step S20.

[0018] Figure 3 This is one implementation shown in the embodiment of the specification. Figure 1 A flowchart of step S40.

[0019] Figure 4 This is a block diagram of an axial flux motor stator immersion monitoring and control device according to an embodiment of the specification.

[0020] Figure 5This is a block diagram of another axial flux motor stator immersion monitoring and control device shown in the embodiment of the specification. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0022] This disclosure provides a method for monitoring and controlling the stator impregnation of an axial flux motor. Figure 1 This is a flowchart illustrating a method for monitoring and controlling the stator impregnation of an axial flux motor according to an embodiment. The method includes: In step S10, capacitance test data collected by the capacitive probe sensor set in the impregnation tank, vibration damping data collected by the vibration viscometer sensor, and microwave attenuation data collected by the microwave resonant cavity sensor are acquired to construct a multi-dimensional data collection set. The capacitive probe sensor is installed on the inner wall or bottom of the impregnation tank. It is used to obtain an electrical signal reflecting the change in the dielectric constant of the oil by detecting changes in the electric field between the electrodes. The capacitance test data is used to characterize the changes in resin concentration and polarization in the impregnated oil. The vibratory viscometer sensor is a measuring device that utilizes the relationship between the damping characteristics of a vibrating element in a fluid and viscosity. Vibration damping data is a physical quantity reflecting the flow resistance of the oil obtained by monitoring the attenuation of vibration frequency or changes in amplitude. Vibration damping data corresponds to the dynamic viscosity index of the oil. The microwave resonant cavity sensor is a device that measures based on the attenuation characteristics of microwave propagation in a medium. Microwave attenuation data is an index reflecting the water content and impurity content of the oil obtained by analyzing the energy loss of the microwave signal after passing through the oil. Microwave attenuation data is used to help determine the aging degree of the oil and the degree of water contamination. The structured data package composed of the raw data output by the three sensors at the same sampling time, aligned with timestamps, constitutes a multi-dimensional data acquisition set.

[0023] By synchronously acquiring heterogeneous data from multiple sources, the system can comprehensively cover the key physicochemical parameters that affect impregnation quality. The detection system eliminates the limitations of single sensors, such as susceptibility to environmental noise interference or limited measurement dimensions, by synchronously reading data from each sensor channel, thus obtaining a dataset containing dielectric, rheological, and moisture content characteristics.

[0024] In step S20, each data in the multi-dimensional collected data set is used as an observation input to the extended Kalman filter; the extended Kalman filter uses dielectric constant, viscosity and water content as state variables to establish a state transition matrix and an observation matrix, and recursively estimates the estimated state vector corresponding to viscosity through prediction steps and update steps; The Extended Kalman Filter (EKF) is a recursive algorithm for state estimation of nonlinear systems. It linearizes the nonlinear system model near the current operating point to achieve an unbiased estimate of the system's internal state. State variables are core parameters describing the evolution of the physicochemical properties of the impregnated oil, specifically including dielectric constant, viscosity, and water content. These variables collectively determine the impregnation performance of the oil.

[0025] The state transition matrix is ​​a coefficient matrix obtained by discretizing the kinetic equations of the oil's physicochemical properties over time. It describes the evolution of the state variables from the previous time step to the current time step. The observation matrix is ​​a linearized matrix describing the mapping relationship between the state variables and sensor measurements. It is calculated using the Jacobian matrix and is used to convert state variables that cannot be directly measured into observable output quantities.

[0026] In this embodiment, the prediction step uses a state transition matrix to deduce the prior state, and the update step uses the observation data at the current moment to correct the prediction result. These steps are executed alternately to recursively estimate the estimated state vector corresponding to the viscosity. The estimated state vector is a set of states containing the optimal viscosity estimate obtained after filtering, which can effectively suppress the interference of sensor noise and process noise.

[0027] For example, when the oil temperature fluctuates slightly, the state transition matrix can correct the predicted viscosity trajectory, while the observation matrix can adjust its weights based on the nonlinear relationship between the current dielectric constant and capacitance value. Through the prediction and update mechanism of the extended Kalman filter, the conversion from noisy observation data to high-precision state estimation is achieved. This improves the robustness of viscosity estimation and overcomes the problems of traditional single-point measurements being susceptible to interference and failing to reflect internal state coupling relationships.

[0028] In step S30, the viscosity adjustment amount is calculated using PID control based on the deviation between the estimated state value corresponding to the estimated state vector and the preset target value. The preset target value is a viscosity value pre-set according to the stator winding impregnation process requirements. The deviation value is the difference between the estimated viscosity value in the current estimated state vector and the preset target value, used to reflect the degree to which the current oil viscosity deviates from the ideal process window. PID control calculates the control output by linearly combining the proportional, integral, and derivative terms of the deviation. The adjustment quantity is the control command output by the PID controller to change the oil viscosity, and its magnitude is proportional to the deviation value and its changing trend.

[0029] For example, when the estimated viscosity is detected to be 45 mPa·s while the preset target value is 50 mPa·s, the deviation is -5 mPa·s. The PID controller will calculate a positive heating adjustment based on this negative deviation to promote viscosity recovery. Through a closed-loop feedback mechanism, viscosity drift caused by oil aging or temperature fluctuations can be dynamically compensated.

[0030] In this way, precise control commands are generated based on the real-time state estimation results. The control system continuously monitors the deviation and uses the PID algorithm to solve it, thereby obtaining the adjustment amount that can correct the viscosity deviation in real time. This effectively avoids the lag and roughness of manual experience adjustment and ensures the real-time and accurate matching of impregnation process parameters.

[0031] In step S40, the adjustment amount is output to the temperature control system of the impregnation tank to adjust the viscosity of the impregnation oil in the impregnation tank.

[0032] The temperature control system is a heating, cooling, and circulation device integrated into the impregnation tank. It receives external control commands and adjusts the oil temperature. There is a definite physical relationship between oil viscosity and temperature, typically exhibiting the characteristic that viscosity decreases as temperature increases and increases as temperature decreases. The adjustment amount is output to the temperature control system, and the corresponding viscosity adjustment command is converted into a temperature setpoint or power control signal, driving the temperature control system to perform the corresponding heating or cooling actions.

[0033] For example, if the adjustment indicates an increase in viscosity, the temperature control system will reduce the oil heating power or start a cooling cycle to lower the oil temperature to the target range, thereby utilizing the oil's temperature-viscosity characteristics to increase viscosity. Through indirect control, the oil viscosity can be quickly and stably maintained within the range required by the process.

[0034] In this embodiment, a closed loop from digital calculation to physical execution is completed by translating control commands into actual physical actions. The temperature control system responds by adjusting the amount of change to alter the thermodynamic state of the oil, thereby adjusting the oil viscosity. This improves the uniformity of resin penetration into the winding gap and the consistency of curing, ensuring the stability and repeatability of impregnation quality for different batches of products.

[0035] The above technical solution integrates a capacitive probe sensor, a vibratory viscometer sensor, and a microwave resonant cavity sensor to collect capacitance test data, vibration damping data, and microwave attenuation data in real time and construct a multi-dimensional data set, which can improve the comprehensiveness of sensing microscopic changes in the impregnation environment. Then, the multi-dimensional data set is used as an observation input to an extended Kalman filter. Using a state transition matrix and an observation matrix with dielectric constant, viscosity, and water content as state variables, the estimated state vector corresponding to viscosity is recursively estimated through prediction and update steps. This achieves accurate online reconstruction of key physicochemical indicators even in the presence of noise and nonlinear interference. Furthermore, based on the deviation between the estimated state value corresponding to the estimated state vector and the preset target value, PID control is used to calculate the viscosity adjustment amount. The corresponding adjustment amount is then output to the temperature control system of the impregnation tank to adjust the viscosity of the impregnating oil in the impregnation tank. It can solve the problem that traditional open-loop control or experience-based parameter tuning cannot respond in time to the dynamic changes in physicochemical properties caused by oil aging, moisture mixing, and temperature fluctuations. It avoids uneven penetration and inconsistent curing caused by rough parameter tuning, improves the closed-loop control accuracy of stator impregnation process, batch-to-batch quality repeatability and production process reliability, enhances the consistency and stability of motor stator insulation treatment under complex working conditions, and improves the effectiveness and robustness of impregnation process to adapt to complex working conditions.

[0036] Optionally, see Figure 2 As shown, in step S20, the recursive estimation of the estimated state vector corresponding to the viscosity through the prediction step and the update step includes: In step S201, in the prediction step, the posterior state estimate of the previous time step is multiplied by the state transition matrix to obtain the prior state estimate of the current time step, and the posterior error covariance matrix of the previous time step is multiplied by the transpose of the state transition matrix and then added to the process noise covariance matrix to obtain the prior error covariance matrix of the current time step. The posterior state estimate of the previous moment refers to the optimal estimate of the system state after measurement correction at the end of the previous sampling period, which includes values ​​for dielectric constant, viscosity, and water content. The state transition matrix is ​​a coefficient matrix obtained by discretizing the kinetic model of the changes in the physicochemical properties of the impregnated oil, used to describe the evolution of state variables over time. The process noise covariance matrix is ​​used to characterize the uncertainty of the system model itself, such as the random fluctuation of the oil aging rate. It can be initialized based on the statistical characteristics of historical data and kept constant or adaptively adjusted during operation.

[0037] In this embodiment of the disclosure, by multiplying the posterior error covariance matrix of the previous time step by the transpose of the state transition matrix and superimposing the process noise covariance matrix, the uncertainty range of the prior estimate at the current time step can be quantified.

[0038] For example, when the sampling time is Δt, the transfer coefficient of the viscosity component can be approximated as 1 - k × Δt, where k is the natural viscosity decay rate. In this way, by predicting the state at future moments based on a physical model, the system's ability to maintain continuous estimation is improved even when sensor data is missing or interfered with.

[0039] In step S202, during the update step, the Kalman gain matrix is ​​calculated based on the prior error covariance matrix, the observation matrix, and the measurement noise covariance matrix. The Kalman gain matrix is ​​the core weighting factor of the extended Kalman filter, used to determine the degree of confidence between the predicted values ​​and actual measurements of the dynamic equilibrium model. The observation matrix describes the mapping relationship from the internal state space (dielectric constant, viscosity, water content) to the external observation space (capacitance test data, vibration damping data, microwave attenuation data), and its elements reflect the sensitivity of each state variable to different sensor signals. The measurement noise covariance matrix characterizes the measurement accuracy and noise level of the capacitive probe, vibratory viscometer, and microwave resonant cavity sensor. Calculating the Kalman gain matrix essentially involves finding an optimal weighting coefficient that minimizes the mean square value of the estimation error.

[0040] For example, if the signal-to-noise ratio of a sensor decreases, its corresponding noise covariance increases, causing the weight of that channel in the Kalman gain to automatically decrease. By introducing a prior error covariance matrix, the filter can perceive the reliability of the current prediction: when the prediction uncertainty is high, the gain matrix tends to assign higher weights to the measurement data; conversely, it relies more on model prediction. This improves the system's adaptability under complex operating conditions and avoids estimation divergence caused by anomalies in a single data source.

[0041] In step S203, the data in the multi-dimensional collected data set at the current moment are combined into an observation vector, and the measurement residual at the current moment is determined based on the observation vector, the observation matrix, and the prior state estimate. The observation vector is a column vector composed of capacitance test data, vibration damping data, and microwave attenuation data collected at the current moment, arranged in a fixed order. It reflects the real-time physical characteristics of the immersion environment. The measurement residual is the difference between the actual observation vector and the predicted observation vector calculated by the observation model based on the prior state estimate. It represents the inconsistency between the model prediction and the actual measurement.

[0042] For example, the prior state estimate is first substituted into the nonlinear observation function (or its linearized form) to calculate the theoretical sensor reading, i.e., the predicted observation vector. Then, the predicted observation vector is subtracted from the actual sensor reading to obtain the residual vector. A small residual indicates accurate model prediction and a stable system state; a large residual indicates an unexpected change in the system state or significant measurement noise. Quantifying the prediction bias can improve the timeliness of the filter in detecting sudden changes in key parameters such as oil viscosity.

[0043] In step S204, the posterior state estimate is determined as the estimated state vector corresponding to the current time step based on the Kalman gain matrix, the measurement residual, and the prior state estimate.

[0044] The posterior state estimate is the final state output after correction based on measurement information, serving as the optimal estimate of dielectric constant, viscosity, and water content at the current moment. The specific calculation process involves performing matrix multiplication between the Kalman gain matrix and the measurement residual vector to obtain a correction vector. Each component of this vector represents the magnitude of adjustment required for the corresponding state variable. Subsequently, this correction vector is added element-wise to the prior state estimate to complete the state update.

[0045] For example, when the measurement residuals show that the microwave attenuation data is significantly higher than the predicted value, the Kalman gain guides the estimated water content state to be adjusted upwards to match the new measurement evidence. The resulting posterior state estimate not only incorporates the evolutionary laws of the physical model but also corrects for the deviations in real-time measurements, resulting in higher accuracy and timeliness. This result will be directly used as the input for the prediction step at the next moment, forming a closed-loop recursive mechanism, and simultaneously output to the PID controller as the basis for viscosity control, achieving precise tracking of the state of the impregnated oil.

[0046] Preferably, in step S203, the step of forming an observation vector from the data in the multi-dimensional collected data set at the current moment, and determining the measurement residual at the current moment based on the observation vector, the observation matrix, and the prior state estimate, includes: In step S2031, the data in the multi-dimensional acquisition data set at the current moment are arranged into a multi-dimensional data array according to a preset fixed order, wherein each element in the multi-dimensional data array corresponds to the original measurement value of a sensor channel. The preset fixed order refers to the sensor channel index sequence predefined during the system initialization phase, used to ensure that the physical meaning of the data dimensions remains consistent each time the array is constructed. The multidimensional data array is a standardized data structure that carries the original measurement values; its number of dimensions equals the total number of sensor channels involved in state estimation.

[0047] For example, if the system defines a fixed order as capacitance value, vibration damping value, and microwave attenuation value, then at time t, the acquired capacitance value of 2.5pF, vibration damping value of 0.8N·s / m, and microwave attenuation value of 12dB will be sequentially filled into the first, second, and third element positions of the array to form a vector. This arrangement mechanism eliminates the uncertainty in time alignment and semantic mapping of multi-source heterogeneous data.

[0048] In step S2032, the multidimensional array of data is used as the observation vector at the current time, and the prior state estimate at the current time is substituted into the pre-established observation model. The observation model describes the linear mapping relationship from the state variable to the observation variable through the observation matrix, and calculates the predicted observation vector under the prior state estimate at the current time. The observation vector represents the system's actual perception of external physical quantities at the current moment. The observation model is a mathematical expression based on the physical characteristics of the motor immersion process. Its core is the observation matrix H, which describes how internal state variables (such as dielectric constant, viscosity, and water content) are linearly projected onto externally measurable sensor readings. The predicted observation vector refers to the set of values ​​that the sensor should theoretically read, assuming the current prior state estimate is accurate.

[0049] For example, the system uses the prior state estimation vector x obtained from the previous iteration. k|k-1 Left-multiply the observation matrix H, i.e., z k =H×x k-1 This allows us to obtain the predicted observation vector. If the state variable is viscosity, the mapping coefficient from viscosity to vibration damping in the observation matrix is ​​0.9. When the prior viscosity estimate is 100 mPa·s, the vibration damping value predicted by the model is 90 units. This achieves the reverse inference from an internal state that cannot be directly and comprehensively observed to an observable external signal.

[0050] In step S2033, each element in the observation vector is subtracted from the element at the corresponding position in the predicted observation vector to obtain the difference value corresponding to each sensor channel; The difference reflects the degree of inconsistency between the actual measured values ​​in the physical world and the theoretical predictions based on model derivation, and is the core driving force for the state correction of the filter. Since both the observation vector and the predicted observation vector strictly follow the fixed order defined above, elements at the same index position in the two vectors must correspond to the same physical channel, ensuring the physical validity of the subtraction operation.

[0051] For example, the processor performs subtraction element by element: y k =z k -z k , where z kFor the observation vector elements, z k This is used to predict the elements of the observation vector. If the actual measured microwave attenuation is 12 dB, while the model predicts 11.5 dB, the difference for the corresponding channel is 0.5 dB. This difference quantifies the deviation of the model prediction. If the difference is close to zero, it indicates that the current state estimate closely matches the actual operating conditions; if the difference is large, it indicates that there is an error in the state estimate or that a sudden change has occurred in the environment.

[0052] In step S2034, the differences corresponding to each sensor channel are arranged in the same order as the observation vector to form a multidimensional difference array as the measurement residual at the current time.

[0053] The measurement residual is a vector form that repackages the differences from all channels. Its structure is completely consistent with the observation vector, preserving the independent error information of each sensor channel. The same order means strictly reusing the capacitance-vibration-microwave or other preset index sequence to ensure that the i-th element in the residual vector always corresponds to the error of the i-th sensor channel.

[0054] For example, the final generated measurement residual vector might be represented as 0.1, -0.05, 0.5T, indicating the error distribution of slightly high values ​​in the capacitance channel, slightly low values ​​in the vibration channel, and significantly high values ​​in the microwave channel, respectively. This structured residual representation allows the filter to adjust the correction strength for each state variable differently based on the noise characteristics and sensitivity of different sensors, thereby achieving high-precision dynamic tracking of key parameters such as the viscosity of the impregnated oil.

[0055] Preferably, in step S204, determining the posterior state estimate as the estimated state vector corresponding to the current time based on the Kalman gain matrix, the measurement residual, and the prior state estimate includes: In step S2041, the Kalman gain matrix is ​​multiplied by the vector corresponding to the measurement residual to obtain a correction vector equal to the number of state variables. Each row of the Kalman gain matrix corresponds to a state variable dimension, and each column corresponds to a measurement channel. Each element in the correction vector represents the adjustment amount of the corresponding prior state estimate using the measurement residual at the current time. The Kalman gain matrix is ​​the core weight matrix calculated by the extended Kalman filter in the update step, and its dimension is determined by the number of state variables and the number of measurement channels.

[0056] In this embodiment, the state variables include dielectric constant, viscosity, and water content, therefore the number of state variables is 3; the measurement channels correspond to three sensors: a capacitive probe, a vibratory viscometer, and a microwave resonant cavity, so the number of measurement channels is also 3. Each element K in the Kalman gain matrix...ij It represents the correction weight of the residual of the j-th measurement channel on the estimate of the i-th state variable.

[0057] For example, the processor performs matrix multiplication on the pre-calculated Kalman gain matrix and the real-time generated measurement residual vector. The measurement errors from different sensors are then distributed to each state variable with optimal weights, thereby generating a correction vector.

[0058] If the damping data measured by the vibratory viscometer at the current moment is significantly higher than the predicted value, resulting in a large and positive measurement residual, and the weight coefficient of the viscosity dimension in the Kalman gain matrix is ​​high, then the element representing viscosity in the calculated correction vector will show a large positive adjustment, indicating that the system needs to correct the prior viscosity estimate upward. Through matrix multiplication, the complementarity of information from multiple sensor sources can be fully utilized to achieve synchronous and accurate correction of multiple physical state quantities.

[0059] In step S2042, the prior state estimation vector and the correction vector are added element-wise to obtain an intermediate state estimation vector without constraint correction. The intermediate state estimation vector is used to characterize the state estimation after fusing the measurement values. The prior state estimation vector is the output of the prediction step at the previous time step, representing the theoretical predictions of state variables such as dielectric constant, viscosity, and water content before the introduction of new observation data at the current time. The correction vector is the correction amount calculated based on the latest measurement residuals. Element-by-element addition refers to the arithmetic addition of scalars at corresponding positions in two vectors of the same dimension, i.e., the prior estimate plus the correction amount.

[0060] For example, each component in the prior state estimation vector is added to the corresponding component in the correction vector. This achieves the fusion of theoretical predictions and actual observation information. The intermediate state estimation vector is a preliminary update result, reflecting the state trend after incorporating the latest sensor data. However, due to the suddenness of measurement noise or errors in the linearization approximation of the model, the intermediate state estimate obtained by direct addition may exceed the reasonable range of the physical world. For example, under certain extreme disturbances, the calculated intermediate viscosity value may be negative, or the water content may exceed 100%, which is physically impossible. Therefore, although the intermediate state estimation vector incorporates measurement information, it has not yet been verified by physical boundaries.

[0061] In step S2043, constraint correction is performed on the intermediate state estimation vector to obtain the posterior state estimation value, and the posterior state estimation value is determined as the estimated state vector corresponding to the current time.

[0062] The constraint correction is a protective logic implemented to ensure that the state estimate always remains within the physically feasible region. The dielectric constant, viscosity, and water content of the impregnating oil all have clearly defined physical upper and lower limits; for example, the viscosity must be greater than zero and less than the saturation limit, and the water content is typically between 0% and a certain percentage. The posterior state estimate is the final state value determined after constraint correction. It serves as the optimal estimate output for the current moment and is used in the prediction step for the next moment and subsequent PID control calculations.

[0063] For example, each component in the intermediate state estimation vector is traversed and compared with the preset maximum and minimum allowable values. If the value of a component exceeds the set boundary range, it is forcibly truncated to the corresponding boundary value; if the value is within the allowable range, the original value is kept unchanged.

[0064] If the viscosity value calculated in the intermediate state estimation vector is -0.5 mPa·s (due to noise interference), and the minimum allowable viscosity value is 0.1 mPa·s, then this value is forcibly assigned to 0.1 mPa·s. Through this constraint correction mechanism, estimation distortion caused by abnormal measurement data or algorithm divergence is avoided, ensuring that the state parameters output to the temperature control system always conform to physical laws, thus improving the robustness and reliability of the entire immersion monitoring and control system.

[0065] Preferably, in step S2043, performing constraint correction on the intermediate state estimation vector to obtain the posterior state estimate includes: In step S20431, if the value of the intermediate state estimation vector is greater than the maximum allowable value, then it is assigned the corresponding maximum allowable value to obtain the posterior state estimation value corresponding to the current time. The intermediate state estimation vector is state data that integrates the current measurement values ​​but has not yet undergone physical rationality verification. It includes estimated components for multiple dimensions such as dielectric constant, viscosity, and water content. The maximum allowable value refers to the upper limit threshold pre-set based on the physicochemical properties of the impregnating oil and the safety boundary of the motor stator insulation process. It is used to define the reasonable range of values ​​for state variables in engineering applications. The purpose of this step is to truncate abnormal estimates that exceed the physical upper limit, preventing the state estimates from diverging into non-physical ranges due to sudden sensor noise, model linearization errors, or extreme operating condition interference.

[0066] For example, if the maximum permissible viscosity of the impregnating oil is set to 500 mPa·s based on its type, and the viscosity component in the calculated intermediate state estimation vector is 520 mPa·s, this value is determined to be greater than the maximum permissible value, and the value of this component is forcibly assigned to 500 mPa·s. This upper limit clamping mechanism ensures that the state information output to subsequent control stages is always within a safe and feasible range, avoiding erroneous cooling or heating commands caused by excessively high estimated values, thereby guaranteeing the stability of the impregnation process.

[0067] In step S20432, if the value of the intermediate state estimation vector is less than the minimum allowable value, then the value is assigned to the minimum allowable value to obtain the posterior state estimation value corresponding to the current time.

[0068] The minimum allowable value is a lower threshold set based on the physical limits of the impregnating oil under normal operating temperature and pressure. It represents the minimum acceptable physical quantity of the state variable. Setting a minimum allowable value can solve the problem of state estimates falling below physical limits or process lower limits due to negative fluctuations in measurement noise or overcompensation of the algorithm.

[0069] For example, if the minimum allowable value is set to 300 mPa·s based on the type of impregnating oil, when the viscosity component in the calculated intermediate state estimation vector is 290 mPa·s, it is determined that this value is less than the minimum allowable value, and the value of this component is forcibly assigned to 300 mPa·s. Through lower limit clamping, meaningless negative values ​​or underestimations are eliminated, ensuring the authenticity of the state vector.

[0070] Preferably, in step S30, the step of calculating the viscosity adjustment amount using PID control based on the deviation between the estimated state value corresponding to the estimated state vector and the preset target value includes: In step S301, the deviation value at the current moment is input to the fuzzy controller. The fuzzy controller is used to fuzzify the input deviation into a signed fuzzy value, and to obtain the corresponding proportional gain coefficient by defuzzifying the fuzzy value using the centroid method. The deviation value is the difference between the estimated viscosity value in the calculated estimated state vector and the preset target viscosity value. This difference reflects the degree to which the current viscosity of the impregnating oil deviates from the ideal process window. A fuzzy controller is used to convert precise numerical deviations into fuzzy linguistic variables to adapt to the nonlinear characteristics of oil property changes during the impregnation process.

[0071] Specifically, the fuzzification process first defines the universe of discourse for the deviation quantity and maps continuous deviation values ​​to fuzzy sets with positive and negative signs, such as negative large, negative small, zero, positive small, positive large, etc. Each fuzzy set corresponds to a membership function. The signed fuzzy values ​​can intuitively represent the direction and magnitude of the deviation. After obtaining the fuzzy values, the controller performs inference based on a pre-established fuzzy rule base to determine the fuzzy distribution of the output variables, and then uses the centroid method for defuzzification. The centroid method converts the area under the membership function curve of the fuzzy output set into a precise value, namely the dynamic proportional gain coefficient, by calculating the center of area under the curve.

[0072] For example, when a large positive deviation is detected, the fuzzy rule may determine that a stronger correction force is needed, thus defuzzifying to obtain a larger proportional gain coefficient; conversely, when the deviation is close to zero, a smaller proportional gain coefficient is output to avoid system oscillation. Through the fuzzy adaptive mechanism, the proportional gain coefficient is no longer a fixed constant, but is dynamically adjusted according to real-time operating conditions.

[0073] In step S302, the proportional term corresponding to the current time is determined by multiplying the proportional gain coefficient corresponding to the current time with the corresponding deviation value; the integral term is determined by multiplying the cumulative deviation value corresponding to the previous time with the integral gain coefficient; and the differential term is determined by multiplying the difference between the deviation value corresponding to the current time and the deviation value corresponding to the dimension at the previous time with the differential gain coefficient. The proportional, integral, and derivative terms together constitute the core output components of the PID controller, respectively responding to the system's current error, historical cumulative error, and error trend. The proportional term is a correction quantity that directly reflects the current deviation magnitude. Its value is obtained by multiplying the dynamic proportional gain coefficient obtained above with the current deviation value. Since the proportional gain coefficient has been adaptively adjusted according to the deviation magnitude, the proportional term can provide a fast response when the deviation is large and a fine adjustment when the deviation is small.

[0074] The integral term is used to eliminate the steady-state error of the system. Its calculation depends on the product of the cumulative deviation value at the previous moment and the fixed integral gain coefficient. The cumulative deviation value is the sum of the deviation values ​​at all moments in history, reflecting the overall deviation of viscosity control over a period of time.

[0075] The differential term is used to predict the trend of deviation and suppress overshoot. Its calculation is based on the product of the difference between the deviation value at the current time and the deviation value at the previous time and the differential gain coefficient.

[0076] For example, if the current deviation is decreasing rapidly, the derivative term will have a counteracting effect, preventing excessive viscosity adjustment from causing reverse overshoot. These three calculations are performed in parallel, capturing the characteristics of the control system at different time dimensions, ensuring the comprehensiveness of the adjustment strategy.

[0077] In step S303, the adjustment amount corresponding to the current moment is determined based on the corresponding proportional term, integral term, and derivative term.

[0078] The adjustment amount is the final control signal output to the actuator, used to drive the temperature control system to change the temperature of the impregnating oil, thereby adjusting the viscosity. This step linearly superimposes the proportional, integral, and derivative terms calculated above to generate a comprehensive adjustment command. Specifically, the adjustment amount is equal to the algebraic sum of the proportional, integral, and derivative terms.

[0079] For example, during the adjustment process, if the viscosity of the oil suddenly drops due to temperature fluctuations, the proportional term will immediately output a larger heating command based on the increased proportional gain coefficient, the integral term will gradually compensate for the residual small deviation over time, and the derivative term will provide a braking effect when the viscosity rises too quickly. The three work together to ensure that the final determined adjustment amount has both sufficient action range and remains smooth and stable, avoiding the oscillation or response lag problems that are prone to occur in traditional fixed parameter PID control under complex immersion conditions, thereby realizing precise closed-loop control of the viscosity of the immersion oil.

[0080] Preferably, see Figure 3 As shown, in step S40, outputting the adjustment amount to the temperature control system of the impregnation tank to adjust the viscosity of the impregnation oil in the impregnation tank includes: In step S401, the measured temperature data of the impregnating oil in the impregnation tank at the current moment and the viscosity-temperature relationship table of the impregnating oil under normal pressure are obtained. The measured temperature data refers to the oil temperature values ​​collected in real time by high-precision temperature sensors installed inside the impregnation tank or in the circulation pipeline, reflecting the thermodynamic state of the impregnating oil at the current moment. The viscosity-temperature relationship table is a static data mapping table obtained in advance through experimental calibration or by consulting oil technical manuals. This table records the viscosity values ​​of a specific type of impregnating oil at different temperature points under standard atmospheric pressure, used to characterize the nonlinear physical properties of oil viscosity changing with temperature.

[0081] For example, for a certain type of epoxy resin impregnation varnish, this relationship table can contain a corresponding kinematic viscosity value recorded every 1°C within the range of 20°C to 80°C, forming a set of temperature-viscosity key value pairs. This step aims to provide the necessary real-time state benchmark and physical model basis for subsequent viscosity calculations and temperature back-calculation, ensuring that the control strategy is based on accurate oil physicochemical properties.

[0082] In step S402, the viscosity-temperature relationship table is queried based on the measured temperature data to obtain the reference viscosity at the current moment; The reference viscosity refers to the theoretically required standard viscosity value of the impregnated oil at the current measured temperature. This value is obtained by using the established viscosity-temperature relationship table as an index key. When the measured temperature corresponds exactly to a record point in the relationship table, the viscosity value at that point is read directly; when the measured temperature lies between two record points, a precise reference viscosity is calculated using linear interpolation or spline interpolation algorithms.

[0083] For example, if the measured temperature is 45.5℃, but the relationship table only contains data for 45℃ and 46℃, the reference viscosity at 45.5℃ is determined by calculating the weighted average of these two viscosity values. Through a dynamic query mechanism, the current viscosity of the oil due to environmental fluctuations or uneven heating can be sensed in real time, providing a quantitative reference for determining whether adjustment is needed.

[0084] In step S403, the viscosity adjustment amount in the adjustment amount is mathematically superimposed with the reference viscosity to obtain the target viscosity corresponding to the current moment; The adjustment amount originates from the viscosity correction command calculated by the aforementioned PID control loop, representing the viscosity change required to eliminate state deviations such as dielectric constant and water content. Mathematical superposition refers to algebraically adding the viscosity adjustment amount to the aforementioned reference viscosity. If the PID calculation result is positive, it indicates that the oil viscosity needs to be increased, with the target viscosity higher than the reference viscosity; if it is negative, it indicates that the viscosity needs to be decreased.

[0085] For example, if the current reference viscosity is 300 mPa·s, and the adjustment calculated by the PID is +15 mPa·s, then the target viscosity obtained after superposition is 315 mPa·s. This realizes the transformation from error compensation requirements to specific process targets, and clarifies the ideal physicochemical properties that the impregnating oil should achieve in the next control cycle.

[0086] In step S404, the target temperature corresponding to the target viscosity is determined by back interpolation through the viscosity-temperature relationship table, based on the temperature difference between the target temperature at the current moment and the current oil temperature in the measured temperature data; Inverse interpolation refers to the process of using the viscosity-temperature relationship table mentioned above, with the calculated target viscosity as the input variable, to inversely solve for the corresponding temperature value. Since oil viscosity and temperature typically exhibit a monotonically decreasing non-linear relationship, by traversing the relationship table or using a numerical approximation algorithm, a uniquely matching or closest temperature point to the target viscosity can be found; this is the target temperature. The temperature difference, the algebraic difference between the target temperature and the current measured temperature, determines the direction and intensity of the temperature control system's operation.

[0087] For example, if the target viscosity of 315 mPa·s corresponds to 43.2℃ in the relationship table, and the current measured oil temperature is 45.5℃, then the calculated temperature difference is -2.3℃, indicating that the oil temperature needs to be lowered to increase the viscosity. The control logic of deriving the target temperature from the target viscosity and calculating the temperature difference ensures the accuracy of cross-physical quantity control.

[0088] In step S405, the temperature control system is controlled to perform temperature control of the impregnation oil based on the temperature difference, so as to adjust the viscosity of the impregnation oil in the impregnation tank.

[0089] The temperature control system includes actuators such as heaters, cooling units, and circulating pumps, configured to receive temperature differences as control input. Specifically, when the temperature difference is positive, the system activates heating mode, with output power proportional to the difference; when the temperature difference is negative, it activates cooling mode or stops heating and starts refrigerant circulation. This step translates the abstract viscosity control requirement into a specific temperature setpoint, utilizing a mature industrial temperature control loop for rapid response.

[0090] For example, if a temperature reduction of 2.3°C is calculated, the temperature control system will automatically adjust the opening of the cooling water valve or the compressor frequency until the oil temperature stabilizes near the target temperature. Thus, by indirectly controlling the oil viscosity through temperature as an intermediate variable, the problems of difficulty and high hysteresis in directly measuring and controlling viscosity are solved, ensuring the constant physicochemical properties of the oil during impregnation, thereby improving the uniformity and consistency of the stator winding impregnation quality.

[0091] In this way, a comprehensive, multi-dimensional data acquisition system was constructed through the coordinated deployment of capacitive probe sensors, vibratory viscometer sensors, and microwave resonant cavity sensors. An extended Kalman filter was used to deeply integrate multi-source observation data with the oil's physicochemical model. Dynamic derivation of the state transition matrix and observation matrix enabled high-precision online estimation of key state variables such as dielectric constant, viscosity, and water content, filtering out measurement noise and resolving nonlinear coupling issues. Furthermore, based on the deviation between the high-precision viscosity estimate and the preset target value, a PID control algorithm was used to generate an adaptive adjustment quantity, which was then reversed using the temperature control system leveraging the oil's temperature-viscosity characteristics. This multi-source sensing, state fusion, and closed-loop control architecture not only overcomes the shortcomings of traditional open-loop control in responding to oil aging, water contamination, and temperature fluctuations, but also eliminates the uncertainty of human experience intervention. Ultimately, it achieves intelligent, adaptive, and precise control of the impregnated oil viscosity, enhancing the robustness of the stator impregnation process and the consistency of product quality.

[0092] See Figure 4 As shown in the figure, this disclosure also provides an axial flux motor stator impregnation monitoring and control device 400, wherein the axial flux motor stator impregnation monitoring and control device 400 includes: The acquisition module 410 is configured to acquire capacitance test data collected by a capacitive probe sensor set in an impregnation tank, vibration damping data collected by a vibratory viscometer sensor, and microwave attenuation data collected by a microwave resonant cavity sensor, and to construct a multi-dimensional acquisition data set. Input template 420 is configured to take each data in the multi-dimensional collected data set as an observation input to the extended Kalman filter; the extended Kalman filter uses dielectric constant, viscosity and water content as state variables to establish a state transition matrix and an observation matrix, and recursively estimates the estimated state vector corresponding to viscosity through prediction steps and update steps. The calculation module 430 is configured to calculate the viscosity adjustment amount using PID control based on the deviation between the estimated state value corresponding to the estimated state vector and the preset target value. Output module 440 is configured to output the adjustment amount to the temperature control system of the impregnation tank to adjust the viscosity of the impregnation oil in the impregnation tank.

[0093] In a preferred embodiment, the input template 420 is configured as follows: In the prediction step, the posterior state estimate of the previous time step is multiplied by the state transition matrix to obtain the prior state estimate of the current time step, and the posterior error covariance matrix of the previous time step is multiplied by the transpose of the state transition matrix and then added to the process noise covariance matrix to obtain the prior error covariance matrix of the current time step. In the update step, the Kalman gain matrix is ​​calculated based on the prior error covariance matrix, the observation matrix, and the measurement noise covariance matrix. The observation vector is formed from the data in the multi-dimensional collection of data at the current moment. The measurement residual at the current moment is determined based on the observation vector, the observation matrix, and the prior state estimate. Based on the Kalman gain matrix, the measurement residual, and the prior state estimate, the posterior state estimate is determined as the estimated state vector corresponding to the current time.

[0094] In a preferred embodiment, the input template 420 is configured as follows: The data in the multi-dimensional data collection at the current moment are arranged into a multi-dimensional data array according to a preset fixed order, wherein each element in the multi-dimensional data array corresponds to the original measurement value of a sensor channel; The multidimensional array of data is used as the observation vector at the current moment, and the prior state estimate at the current moment is substituted into the pre-established observation model. The observation model describes the linear mapping relationship from the state variable to the observation variable through the observation matrix, and calculates the predicted observation vector under the prior state estimate at the current moment. Subtract the element at the corresponding position in the predicted observation vector from each element in the observation vector to obtain the difference value corresponding to each sensor channel; The differences corresponding to each sensor channel are arranged in the same order as the observation vector to form a multidimensional difference array, which serves as the measurement residual at the current moment.

[0095] In a preferred embodiment, the input template 420 is configured as follows: Perform matrix multiplication between the Kalman gain matrix and the vector corresponding to the measurement residual to obtain a correction vector equal to the number of state variables. Each row of the Kalman gain matrix corresponds to a state variable dimension, and each column corresponds to a measurement channel. Each element in the correction vector represents the adjustment amount of the corresponding prior state estimate using the measurement residual at the current time. The prior state estimation vector and the correction vector are added element-wise to obtain an intermediate state estimation vector without constraint correction. The intermediate state estimation vector is used to characterize the state estimation after fusing the measurement values. Constraint correction is performed on the intermediate state estimation vector to obtain the posterior state estimation value, and the posterior state estimation value is determined as the estimated state vector corresponding to the current time.

[0096] In a preferred embodiment, the input template 420 is configured as follows: If the value of the intermediate state estimation vector is greater than the maximum allowable value, then it is assigned the corresponding maximum allowable value to obtain the posterior state estimation value corresponding to the current time. If the value of the intermediate state estimation vector is less than the minimum allowable value, then the value is assigned to the minimum allowable value to obtain the posterior state estimation value corresponding to the current time.

[0097] In a preferred embodiment, the computing module 430 is configured as follows: The deviation value at the current moment is input into the fuzzy controller, which is used to fuzzify the input deviation into a signed fuzzy value, and to obtain the corresponding proportional gain coefficient by defuzzifying the fuzzy value using the centroid method. The proportional term corresponding to the current time is determined by multiplying the proportional gain coefficient corresponding to the current time with the corresponding deviation value; the integral term is determined by multiplying the cumulative deviation value corresponding to the previous time with the integral gain coefficient; and the differential term is determined by multiplying the difference between the deviation value corresponding to the current time and the deviation value corresponding to the same dimension at the previous time with the differential gain coefficient. The adjustment amount corresponding to the current moment is determined based on the corresponding proportional term, integral term, and derivative term.

[0098] In a preferred embodiment, the output module 440 is configured as follows: Obtain the measured temperature data of the impregnating oil in the impregnation tank at the current moment, as well as the viscosity-temperature relationship table of the impregnating oil under normal pressure; The reference viscosity at the current moment is obtained by querying the viscosity-temperature relationship table based on the measured temperature data. The viscosity adjustment amount in the adjustment amount is mathematically superimposed with the reference viscosity to obtain the target viscosity at the current moment; The target temperature corresponding to the target viscosity is determined by back interpolation of the viscosity-temperature relationship table, based on the temperature difference between the target temperature at the current moment and the current oil temperature in the measured temperature data; The temperature control system is controlled according to the temperature difference to perform temperature control of the impregnating oil, so as to adjust the viscosity of the impregnating oil in the impregnation tank.

[0099] Specific limitations regarding the axial flux motor stator impregnation monitoring and control device can be found in the above-described limitations of the axial flux motor stator impregnation monitoring and control method, and will not be repeated here. Each module in the aforementioned axial flux motor stator impregnation monitoring and control device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in hardware or independently of the processor in a computer device, or stored in software in the memory of a computer device, so that the processor can call and execute the corresponding operations of each module.

[0100] This disclosure also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the methods described in the foregoing embodiments.

[0101] This disclosure also provides an electronic device, including: A memory on which computer programs are stored; A processor for executing the computer program in the memory to implement the steps of any of the methods described in the foregoing embodiments.

[0102] Figure 5 The axial flux motor stator immersion monitoring and control device 100 shown includes a processor 1001 and a memory 1003. The processor 1001 and the memory 1003 are connected, for example, via a bus 1002. Optionally, the axial flux motor stator immersion monitoring and control device 100 may further include a communication component, which can be used for data interaction between the device 100 and other devices, such as sending or receiving data. It should be noted that in actual scheduling, the communication component is not limited to one, and the structure of this axial flux motor stator immersion monitoring and control device 100 does not constitute a limitation on the embodiments of this application.

[0103] Processor 1001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 1001 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0104] Bus 1002 may include a pathway for transmitting information between the aforementioned components. Bus 1002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. Bus 1002 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 5 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0105] The memory 1003 may be ROM (Read Only Memory) or other types of static storage devices capable of storing static information and instructions, RAM (Random Access Memory) or other types of dynamic storage devices capable of storing information and instructions, or EEPROM (Electrically Erasable Programmable Read Only Memory), CD-ROM (Compact Disc Read Only Memory) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media, other magnetic storage devices, or any other medium capable of carrying or storing program code and capable of being read by a computer, without limitation herein.

[0106] The memory 1003 is used to store program code for executing the embodiments of this disclosure, and its execution is controlled by the processor 1001. The processor 1001 is used to execute the program code stored in the memory 1003 to implement the steps shown in the embodiments of the aforementioned axial flux motor stator impregnation monitoring and control method.

[0107] This disclosure also provides a computer-readable storage medium storing program code. When the program code is executed by a processor, it can implement the steps and corresponding content of the aforementioned embodiment of the axial flux motor stator immersion monitoring and control method.

[0108] The preferred embodiments of the present disclosure have been described in detail above with reference to the accompanying drawings. However, the present disclosure is not limited to the specific details of the above embodiments. Within the scope of the technical concept of the present disclosure, various changes, modifications, substitutions and variations can be made to these embodiments, and all such changes, modifications, substitutions and variations fall within the protection scope of the present disclosure.

[0109] It should also be noted that the various specific technical features described in the above embodiments can be combined in any suitable manner without contradiction, and such combinations should also be considered as part of this disclosure. To avoid unnecessary repetition, this disclosure will not further describe the various possible combinations. The technical scope of this application is not limited to the contents of the specification, but must be determined according to the scope of the claims.

Claims

1. A method for monitoring and controlling stator impregnation in an axial flux motor, characterized in that, The method includes: Acquire capacitance test data collected by a capacitive probe sensor set in an impregnation tank, vibration damping data collected by a vibratory viscometer sensor, and microwave attenuation data collected by a microwave resonant cavity sensor to construct a multi-dimensional data collection set; Each data point in the multi-dimensional data set is used as an observation input to the extended Kalman filter; the extended Kalman filter uses dielectric constant, viscosity and water content as state variables to establish a state transition matrix and an observation matrix, and recursively estimates the estimated state vector corresponding to viscosity through prediction and update steps; Based on the deviation between the estimated state value corresponding to the estimated state vector and the preset target value, PID control is used to calculate the viscosity adjustment amount; The adjustment amount is output to the temperature control system of the impregnation tank to adjust the viscosity of the impregnation oil in the impregnation tank.

2. The method according to claim 1, characterized in that, The recursive estimation of the estimated state vector corresponding to the viscosity through prediction and update steps includes: In the prediction step, the posterior state estimate of the previous time step is multiplied by the state transition matrix to obtain the prior state estimate of the current time step, and the posterior error covariance matrix of the previous time step is multiplied by the transpose of the state transition matrix and then added to the process noise covariance matrix to obtain the prior error covariance matrix of the current time step. In the update step, the Kalman gain matrix is ​​calculated based on the prior error covariance matrix, the observation matrix, and the measurement noise covariance matrix; The observation vector is formed from the data in the multi-dimensional collection of data at the current moment. The measurement residual at the current moment is determined based on the observation vector, the observation matrix, and the prior state estimate. Based on the Kalman gain matrix, the measurement residual, and the prior state estimate, the posterior state estimate is determined as the estimated state vector corresponding to the current time.

3. The method according to claim 2, characterized in that, The step of assembling an observation vector from the multi-dimensional collected data set at the current moment, and determining the measurement residual at the current moment based on the observation vector, the observation matrix, and the prior state estimate, includes: The data in the multi-dimensional data collection at the current moment are arranged into a multi-dimensional data array according to a preset fixed order, wherein each element in the multi-dimensional data array corresponds to the original measurement value of a sensor channel; The multidimensional array of data is used as the observation vector at the current moment, and the prior state estimate at the current moment is substituted into the pre-established observation model. The observation model describes the linear mapping relationship from the state variable to the observation variable through the observation matrix, and calculates the predicted observation vector under the prior state estimate at the current moment. Subtract the element at the corresponding position in the predicted observation vector from each element in the observation vector to obtain the difference value corresponding to each sensor channel; The differences corresponding to each sensor channel are arranged in the same order as the observation vector to form a multidimensional difference array, which serves as the measurement residual at the current moment.

4. The method according to claim 2, characterized in that, The step of determining the posterior state estimate as the estimated state vector corresponding to the current time moment based on the Kalman gain matrix, the measurement residual, and the prior state estimate includes: Perform matrix multiplication between the Kalman gain matrix and the vector corresponding to the measurement residual to obtain a correction vector equal to the number of state variables. Each row of the Kalman gain matrix corresponds to a state variable dimension, and each column corresponds to a measurement channel. Each element in the correction vector represents the adjustment amount of the corresponding prior state estimate using the measurement residual at the current time. The prior state estimation vector and the correction vector are added element-wise to obtain an intermediate state estimation vector without constraint correction. The intermediate state estimation vector is used to characterize the state estimation after fusing the measurement values. Constraint correction is performed on the intermediate state estimation vector to obtain the posterior state estimation value, and the posterior state estimation value is determined as the estimated state vector corresponding to the current time.

5. The method according to claim 4, characterized in that, The step of performing constraint correction on the intermediate state estimation vector to obtain the posterior state estimation value includes: If the value of the intermediate state estimation vector is greater than the maximum allowable value, then it is assigned the corresponding maximum allowable value to obtain the posterior state estimation value corresponding to the current time. If the value of the intermediate state estimation vector is less than the minimum allowable value, then the value is assigned to the minimum allowable value to obtain the posterior state estimation value corresponding to the current time.

6. The method according to claim 1, characterized in that, The step of calculating the viscosity adjustment amount using PID control based on the deviation between the estimated state value corresponding to the estimated state vector and the preset target value includes: The deviation value at the current moment is input into the fuzzy controller, which is used to fuzzify the input deviation into a signed fuzzy value, and to obtain the corresponding proportional gain coefficient by defuzzifying the fuzzy value using the centroid method. The proportional term corresponding to the current time is determined by multiplying the proportional gain coefficient corresponding to the current time with the corresponding deviation value; the integral term is determined by multiplying the cumulative deviation value corresponding to the previous time with the integral gain coefficient; and the differential term is determined by multiplying the difference between the deviation value corresponding to the current time and the deviation value corresponding to the same dimension at the previous time with the differential gain coefficient. The adjustment amount corresponding to the current moment is determined based on the corresponding proportional term, integral term, and derivative term.

7. The method according to any one of claims 1-6, characterized in that, The step of outputting the adjustment amount to the temperature control system of the impregnation tank to adjust the viscosity of the impregnation oil in the impregnation tank includes: Obtain the measured temperature data of the impregnating oil in the impregnation tank at the current moment, as well as the viscosity-temperature relationship table of the impregnating oil under normal pressure; The reference viscosity at the current moment is obtained by querying the viscosity-temperature relationship table based on the measured temperature data. The viscosity adjustment amount in the adjustment amount is mathematically superimposed with the reference viscosity to obtain the target viscosity at the current moment; The target temperature corresponding to the target viscosity is determined by back interpolation of the viscosity-temperature relationship table, based on the temperature difference between the target temperature at the current moment and the current oil temperature in the measured temperature data; The temperature control system is controlled according to the temperature difference to perform temperature control of the impregnating oil, so as to adjust the viscosity of the impregnating oil in the impregnation tank.

8. A stator impregnation monitoring and control device for an axial flux motor, characterized in that, The device includes: The acquisition module is configured to acquire capacitance test data collected by a capacitive probe sensor set in the impregnation tank, vibration damping data collected by a vibratory viscometer sensor, and microwave attenuation data collected by a microwave resonant cavity sensor, and construct a multi-dimensional acquisition data set. The input template is configured to take each data point in the multi-dimensional collected data set as the observation input to the extended Kalman filter; the extended Kalman filter uses dielectric constant, viscosity and water content as state variables to establish a state transition matrix and an observation matrix, and recursively estimates the estimated state vector corresponding to viscosity through prediction steps and update steps. The calculation module is configured to calculate the viscosity adjustment amount using PID control based on the deviation between the estimated state value corresponding to the estimated state vector and the preset target value. The output module is configured to output the adjustment amount to the temperature control system of the impregnation tank to adjust the viscosity of the impregnation oil in the impregnation tank.

9. An electronic device, characterized in that, include: A memory on which computer programs are stored; A processor for executing the computer program in the memory to implement the steps of the method according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores program code that is executed by a processor to implement the steps of the method according to any one of claims 1-7.

Images