An intelligent comprehensive performance test platform for magnetically suspended high-speed permanent magnet synchronous motors

Abstract

The present invention relates to the field of motor testing technology, and proposes an intelligent comprehensive performance testing platform for magnetic levitation high-speed permanent magnet synchronous motors, including: an electrical performance index testing module for detecting the electrical performance of magnetic levitation motors; a dynamic characteristic index testing module for detecting the corresponding change capability during the operation of magnetic levitation motors; a process control module for controlling the detection equipment and automatically switching the test links; a data analysis and diagnosis module for analyzing and diagnosing the data obtained from the electrical performance index test and the dynamic characteristic index test, wherein the dynamic characteristic index is analyzed and diagnosed by a nonlinear autoregressive neural network model with external input. Through the above technical solution, the problem of complicated test items in the prior art, which requires frequent changes and connections to multiple test equipment, is solved. This process is not only time-consuming and labor-intensive, but also greatly increases the workload of testers.

Description

Technical Field

[0001] The present invention relates to the technical field of motor testing, and in particular to an intelligent comprehensive performance testing platform for a magnetically suspended high-speed permanent magnet synchronous motor. Background Art

[0002] As a new type of high-efficiency motor, magnetic levitation high-speed permanent magnet synchronous motor is widely used in high-speed trains, magnetic levitation systems, aerospace and other fields due to its advantages such as no friction, low loss and high speed. In these high-demand application scenarios, the performance of the motor is directly related to the safety, stability and energy efficiency of the system, so the comprehensive performance test of the motor is particularly important.

[0003] At present, the performance test of the magnetic levitation high-speed permanent magnet synchronous motor includes but is not limited to the motor's no-load test, load test, overspeed test, temperature rise test, vibration and noise test and other projects. Usually, each type of test corresponds to different test equipment and needs to be completed through complex wiring and parameter configuration. During the test process, technicians need to frequently change test equipment and connect the motor to different instrument systems to complete the measurement and verification of its electrical, mechanical, and thermodynamic parameters. Each equipment replacement is not only time-consuming and labor-intensive, but also prone to problems such as miswiring and loss of test data, which increases the complexity and uncertainty of the test.

[0004] Therefore, the existing technology urgently needs an intelligent and automated magnetic levitation high-speed permanent magnet synchronous motor comprehensive performance test platform that can integrate various test items, simplify the operating process, reduce manual intervention, improve test efficiency, and provide more reliable and accurate data analysis functions to meet the needs of the rapid development of the industry. Summary of the invention

[0005] The present invention proposes an intelligent comprehensive performance testing platform for a magnetically suspended high-speed permanent magnet synchronous motor to improve testing efficiency, reduce testing costs, and alleviate the workload of testers.

[0006] The technical solution of the present invention is as follows:

[0007] An intelligent comprehensive performance test platform for magnetically suspended high-speed permanent magnet synchronous motors, comprising:

[0008] Electrical performance index test module, used to detect the electrical performance of the magnetic levitation motor;

[0009] Dynamic characteristic index test module, used to detect the corresponding change ability of the magnetic levitation motor during operation;

[0010] Process control module, used to control the testing equipment and automatically switch the testing links;

[0011] The data analysis and diagnosis module is used to analyze and diagnose the data obtained from the electrical performance index test and the dynamic characteristic index test, wherein the dynamic characteristic index is analyzed and diagnosed through a nonlinear autoregressive neural network model with external input.

[0012] Furthermore, the electrical performance index test module includes:

[0013] A withstand voltage test unit, used to measure the leakage current value of the stator winding of the magnetic levitation motor multiple times through the process control module;

[0014] The insulation resistance test unit is used to measure the resistance value of the stator winding of the magnetic levitation motor multiple times through the process control module.

[0015] Furthermore, the dynamic characteristic index test module includes:

[0016] The no-load test unit is used to collect the voltage, current, input power, torque, speed, output power and other data of the magnetic levitation motor under no-load conditions through the process control module;

[0017] The load test unit is used to collect data such as voltage, current, input power, torque, speed, output power, temperature rise, vibration and noise of the magnetic suspension motor when it is running forward and reverse at low speed, medium speed and high speed under load through the process control module;

[0018] The stall test unit is used to collect data such as voltage, current, input power, torque, speed, output power, temperature rise, vibration and noise when the magnetic levitation motor runs at rated speed for a period of time and then suddenly adds load until the magnetic levitation motor stalls through the process control module;

[0019] The overspeed test unit is used to collect data such as voltage, current, input power, torque, speed, output power, temperature rise, vibration and noise when the magnetic levitation motor runs at a speed higher than the rated speed for a period of time through the process control module.

[0020] Furthermore, the process control module includes:

[0021] A withstand voltage tester, which is used to output a specified voltage to the stator winding of the magnetic levitation motor and measure the leakage current;

[0022] Insulation resistance tester, used to test the stator winding resistance to ground R of magnetic levitation motor U , R V , R W The resistance R between the stator winding UV , R VW , R WU The resistance value;

[0023] The frequency converter is used to change the frequency and voltage of the magnetic levitation motor input to adjust the speed of the magnetic levitation motor;

[0024] Power analyzers are used to measure parameters such as electrical power, energy and efficiency of electrical systems or equipment;

[0025] Torque and speed sensor, used to detect the torque and speed on the magnetic suspension motor shaft;

[0026] Dynamometer, used to apply different mechanical loads and measure the output torque and power of the magnetic levitation motor;

[0027] Temperature sensor, vibration sensor, and sound sensor are used to detect the temperature, vibration, and sound of the magnetic levitation motor;

[0028] Four-quadrant frequency converter, used to convert the mechanical power of the dynamometer into electrical power to feed back to the power grid;

[0029] The first contactor, the second contactor and the third contactor are respectively used to connect with the frequency converter, the withstand voltage tester and the insulation resistance tester to realize the switching of the frequency converter, the withstand voltage tester or the insulation resistance tester;

[0030] The process control terminal PLC is used to control the withstand voltage tester, the insulation resistance tester, the frequency converter, the power analyzer, the torque speed sensor, the dynamometer, the four-quadrant frequency converter, the first contactor, the second contactor, and the third contactor, wherein the first contactor, the second contactor, and the third contactor are controlled to realize automatic switching of different test links;

[0031] AC busbar, used to connect the magnetic levitation motor with the power analyzer, frequency converter, voltage withstand tester, and insulation resistance tester, so as to realize the detection or control of the magnetic levitation motor by the power analyzer, frequency converter, voltage withstand tester, and insulation resistance tester;

[0032] Communication bus, used to connect and control the withstand voltage tester, insulation resistance tester, frequency converter, power analyzer, torque and speed sensor, dynamometer, four-quadrant frequency converter, temperature sensor, vibration sensor, and sound sensor to achieve data and signal transmission;

[0033] The switch quantity controller circuit is used to realize the control of the first contactor, the second contactor and the third contactor by the process control terminal PLC;

[0034] Among them, the input end of the frequency converter is connected to the power grid, the output ends of the frequency converter, the voltage withstand tester, and the insulation resistance tester are respectively connected to the input ends of the first contactor, the second contactor, and the third contactor, the output ends of the first contactor, the second contactor, and the third contactor are connected to the input end of the AC busbar, the output end of the AC busbar is connected to the input end of the power analyzer and the stator winding of the magnetic levitation motor, the input end of the torque and speed sensor and the output end of the dynamometer are connected to the rotating shaft of the magnetic levitation motor, the input end of the four-quadrant frequency converter is connected to the dynamometer, the output end of the four-quadrant frequency converter is connected to the power grid, the input end of the temperature sensor, the vibration sensor, and the sound sensor are connected to the magnetic levitation motor, the communication bus is connected to the voltage withstand tester, the insulation resistance tester, the frequency converter, the power analyzer, the torque and speed sensor, the dynamometer, the four-quadrant frequency converter, the temperature sensor, the vibration sensor, the sound sensor, the process control terminal PLC, and the data analysis and diagnosis module, and the output end of the switch control circuit is connected to the coils of the first contactor, the second contactor, and the third contactor.

[0035] Furthermore, the data analysis and diagnosis module includes:

[0036] The electrical performance index test judgment unit performs analysis and judgment based on the electrical performance index test data;

[0037] The dynamic characteristic index test and judgment unit analyzes and judges the dynamic characteristics of the magnetic levitation motor through a nonlinear autoregressive neural network model.

[0038] Furthermore, the electrical performance index test judgment unit includes:

[0039] The withstand voltage test judgment unit calculates the withstand voltage test data average value of the multiple test data according to the multiple test data of the electrical performance index test module, and compares the withstand voltage test data average value with the withstand voltage test data standard value. If the difference between the withstand voltage test data average value and the withstand voltage test data standard value is within a preset range, the withstand voltage test is qualified, otherwise, the withstand voltage test is unqualified;

[0040] The insulation resistance test judgment unit calculates the average value of the insulation resistance test data of multiple test data based on the multiple test data of the electrical performance index test module, and compares the average value of the insulation resistance test data with the standard value of the insulation resistance test data. If the difference between the average value of the insulation resistance test data and the standard value of the insulation resistance test data is within a preset range, the insulation resistance test is qualified, otherwise, the insulation resistance test is unqualified.

[0041] Furthermore, the dynamic characteristic index test judgment unit includes:

[0042] The no-load test judgment unit measures the stator resistance and inductance, motor pole pair number, motor no-load back electromotive force, and motor flux constant of the magnetic levitation motor through the dynamic characteristic index test module. The nonlinear autoregressive neural network model obtains the no-load state of the current magnetic levitation motor based on the measured data combined with historical data. If the no-load state of the current magnetic levitation motor meets the no-load qualification standard, the no-load test is qualified, otherwise, the no-load test is unqualified;

[0043] The load test judgment unit measures the stator voltage and current, motor input electric power and output mechanical power, stray loss and efficiency, torque and speed of the magnetic levitation motor through the dynamic characteristic index test module. The nonlinear autoregressive neural network model obtains the current load state of the magnetic levitation motor based on the measured data combined with historical data. If the current load state of the magnetic levitation motor meets the load qualification standard, the load test is qualified, otherwise, the load test is unqualified;

[0044] The stall test judgment unit measures the stator voltage and current, torque, input electric power, and stall loss of the magnetic levitation motor through the dynamic characteristic index test module. The nonlinear autoregressive neural network model obtains the stall state of the current magnetic levitation motor based on the measured data combined with historical data. If the stall state of the current magnetic levitation motor meets the stall qualification standard, the stall test is qualified, otherwise, the stall test is unqualified;

[0045] The overspeed test judgment unit measures the stator voltage and current, torque and speed, motor input electrical power and output mechanical power, and response time of the magnetic levitation motor through the dynamic characteristic index test module. The nonlinear autoregressive neural network model combines the measured data with historical data to obtain the overspeed test status of the current magnetic levitation motor. If the overspeed test status of the current magnetic levitation motor meets the overspeed qualification standard, the overspeed test is qualified. Otherwise, the overspeed test is unqualified.

[0046] Furthermore, the formula of the nonlinear autoregressive neural network model is:

[0047] y(t)=f(y(t-1),y(t-2),...,y(tn y ),u(t-1),u(t-2),...,u(tn u )),

[0048] Where y(t) is the current output variable of the system, f is the nonlinear regression function, y(t-1), y(t-2), ..., y(tn y ) is the historical output variable of the system, n y is the order of the output delay, u(t-1),u(t-2),...,u(tn u ) is the historical input variable of the system, n uis the order of input delay;

[0049] The nonlinear autoregressive neural network model includes an input layer, a hidden layer and an output layer, wherein the input data are normalized voltage, current, input power, speed, torque, output power, temperature rise, vibration and noise data. When selected, the delay order d is 6 and the number of hidden layer neurons h is 12.

[0050] Furthermore, the activation function f1 of the hidden layer is a hyperbolic tangent S-type transfer function, i.e., a tansig function, and the formula is:

[0051]

[0052] The calculation formula of the hidden layer is:

[0053] h(t)=f1(w1u(t)+w2y(t)+b1),

[0054] Where h(t) is the output of the hidden layer, u(t) is the input of each layer of neurons, y(t) is the input of each layer of neurons, w1 and w2 are the connection weights between the input layer nodes and neurons, and b1 is the threshold of the hidden layer;

[0055] The activation function f2 of the output layer is a linear transfer function, namely, purelin function, and the formula is:

[0056] f2(x)=purelin(x)=x;

[0057] The calculation formula of the output layer is:

[0058] y(t)=f2(w3h(t)+b2),

[0059] In the formula, y(t) is the current output of the system, w3 is the connection weight between the hidden layer nodes and the output layer nodes, and b2 is the threshold of the output layer.

[0060] Furthermore, the training function of the nonlinear autoregressive neural network model is the Levenberg-Marquardt algorithm. A judgment index is set in the update iteration process to determine the quality of the approximation. The connection weight adjustment method formula of the Levenberg-Marquardt algorithm is:

[0061] Δw=(J T J+μI) -1 J T e,

[0062] Where μ is an adaptively adjusted scalar.

[0063] The working principle and beneficial effects of the present invention are:

[0064] The present invention can cover the electrical performance indicators (voltage withstand test and insulation resistance test) and dynamic characteristic indicators (no-load test, load test, stall test and overspeed test) of the magnetic levitation permanent magnet synchronous motor through the electrical performance indicator test module and the dynamic characteristic indicator test module, ensuring that the performance of the magnetic levitation motor in multiple aspects can be comprehensively evaluated, and the performance of the magnetic levitation motor is comprehensively analyzed, which is helpful to discover potential problems and optimize the design of the magnetic levitation motor.

[0065] The present invention uses a process control module to manage and control the test process, realizes automatic switching of test links, reduces manual intervention, and improves test efficiency. The data analysis and diagnosis module uses an advanced nonlinear autoregressive neural network model (NARX) for data processing and judgment, which can automatically process data waveforms and quickly and accurately give test results, greatly reducing the time and cost of manual analysis. The data analysis and diagnosis module can not only process test data in real time, but also make intelligent judgments on test results through a trained nonlinear autoregressive neural network model, and quickly identify whether the motor performance meets the design requirements. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] The present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0067] Figure 1 It is a structural block diagram of the present invention;

[0068] Figure 2 It is a flow chart of the electrical performance index test and the dynamic characteristic index test in the present invention;

[0069] Figure 3 This is a structural diagram of the nonlinear autoregressive neural network model in the present invention. DETAILED DESCRIPTION

[0070] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0071] like Figure 1-2 As shown, this embodiment proposes an intelligent comprehensive performance test platform for magnetic levitation high-speed permanent magnet synchronous motor, including:

[0072] Electrical performance index test module, used to detect the electrical performance of the magnetic levitation motor;

[0073] Dynamic characteristic index test module, used to detect the corresponding change ability of the magnetic levitation motor during operation;

[0074] Process control module, used to control the testing equipment and automatically switch the testing links;

[0075] The data analysis and diagnosis module is used to analyze and diagnose the data obtained from the electrical performance index test and the dynamic characteristic index test, wherein the dynamic characteristic index is analyzed and diagnosed through a nonlinear autoregressive neural network model with external input.

[0076] The nonlinear autoregressive neural network model (NARX) is a neural network model for dynamic time series modeling that can predict future outputs based on current and past inputs and past outputs.

[0077] In this embodiment, the electrical performance index test module includes:

[0078] The withstand voltage test unit is used to measure the leakage current value of the stator winding of the magnetic levitation motor multiple times through the process control module to verify whether the insulation system of the motor can withstand the high voltage impact under its working conditions, ensure that the motor will not have electrical breakdown when operating in a high voltage environment, and prevent electrical faults caused by insulation failure;

[0079] The insulation resistance test unit is used to measure the resistance value of the stator winding of the magnetic levitation motor multiple times through the process control module, check the health status of the insulation material, evaluate the degree of deterioration of the motor insulation system, ensure the long-term stable operation of the motor, and prevent safety hazards such as leakage or short circuit.

[0080] In this embodiment, the dynamic characteristic index test module includes:

[0081] The no-load test unit is used to collect the voltage, current, input power, torque, speed, output power and other data of the magnetic levitation motor under no-load conditions through the process control module, evaluate the basic operating performance of the motor, detect whether there are abnormal friction, winding or rotor problems, and confirm the efficiency and power consumption of the motor under no-load conditions;

[0082] The load test unit is used to collect the voltage, current, input power, torque, speed, output power, temperature rise, vibration and noise data of the magnetic levitation motor under load conditions during its low-speed, medium-speed and high-speed forward and reverse operation through the process control module, verify the working performance of the motor under different load conditions, confirm whether it can stably output the designed power, and evaluate its reliability and performance under rated load and overload conditions;

[0083] The stall test unit is used to collect data such as voltage, current, input power, torque, speed, output power, temperature rise, vibration and noise when the magnetic levitation motor runs at rated speed for a period of time and then suddenly adds load until the magnetic levitation motor stalls through the process control module, so as to evaluate the safety and electrical tolerance of the motor under extreme conditions and ensure that the motor will not be damaged or overheated due to excessive current in the stalled state;

[0084] The overspeed test unit is used to collect data such as voltage, current, input power, torque, speed, output power, temperature rise, vibration and noise when the magnetic levitation motor runs at a speed higher than the rated speed for a period of time through the process control module, to confirm that the motor can run safely at extreme speeds without mechanical failures such as rotor instability and bearing damage, thereby ensuring reliability under high-speed operation.

[0085] In this embodiment, the process control module includes:

[0086] A withstand voltage tester, which is used to output a specified voltage to the stator winding of the magnetic levitation motor and measure the leakage current;

[0087] Insulation resistance tester, used to test the stator winding resistance to ground R of magnetic levitation motor U , R V , R W The resistance R between the stator winding UV , R VW , R WU The resistance value;

[0088] The frequency converter is used to change the frequency and voltage of the magnetic levitation motor input to adjust the speed of the magnetic levitation motor;

[0089] Power analyzers are used to measure parameters such as electrical power, energy and efficiency of electrical systems or equipment;

[0090] Torque and speed sensor, used to detect the torque and speed on the magnetic suspension motor shaft;

[0091] Dynamometer, used to apply different mechanical loads and measure the output torque and power of the magnetic levitation motor;

[0092] Temperature sensor, vibration sensor, and sound sensor are used to detect the temperature, vibration, and sound of the magnetic levitation motor;

[0093] Four-quadrant frequency converter, used to convert the mechanical power of the dynamometer into electrical power to feed back to the power grid;

[0094] The first contactor, the second contactor and the third contactor are respectively used to connect with the frequency converter, the withstand voltage tester and the insulation resistance tester to realize the switching of the frequency converter, the withstand voltage tester or the insulation resistance tester;

[0095] The process control terminal PLC is used to control the withstand voltage tester, the insulation resistance tester, the frequency converter, the power analyzer, the torque speed sensor, the dynamometer, the four-quadrant frequency converter, the first contactor, the second contactor, and the third contactor, wherein the first contactor, the second contactor, and the third contactor are controlled to realize automatic switching of different test links;

[0096] AC busbar, used to connect the magnetic levitation motor with the power analyzer, frequency converter, voltage withstand tester, and insulation resistance tester, so as to realize the detection or control of the magnetic levitation motor by the power analyzer, frequency converter, voltage withstand tester, and insulation resistance tester;

[0097] Communication bus, used to connect and control the withstand voltage tester, insulation resistance tester, frequency converter, power analyzer, torque and speed sensor, dynamometer, four-quadrant frequency converter, temperature sensor, vibration sensor, and sound sensor to achieve data and signal transmission;

[0098] The switch quantity controller circuit is used to realize the control of the first contactor, the second contactor and the third contactor by the process control terminal PLC;

[0099] Among them, the input end of the frequency converter is connected to the power grid, the output ends of the frequency converter, the voltage withstand tester, and the insulation resistance tester are respectively connected to the input ends of the first contactor, the second contactor, and the third contactor, the output ends of the first contactor, the second contactor, and the third contactor are connected to the input end of the AC busbar, the output end of the AC busbar is connected to the input end of the power analyzer and the stator winding of the magnetic levitation motor, the input end of the torque and speed sensor and the output end of the dynamometer are connected to the rotating shaft of the magnetic levitation motor, the input end of the four-quadrant frequency converter is connected to the dynamometer, the output end of the four-quadrant frequency converter is connected to the power grid, the input end of the temperature sensor, the vibration sensor, and the sound sensor are connected to the magnetic levitation motor, the communication bus is connected to the voltage withstand tester, the insulation resistance tester, the frequency converter, the power analyzer, the torque and speed sensor, the dynamometer, the four-quadrant frequency converter, the temperature sensor, the vibration sensor, the sound sensor, the process control terminal PLC, and the data analysis and diagnosis module, and the output end of the switch control circuit is connected to the coils of the first contactor, the second contactor, and the third contactor.

[0100] In this embodiment, the withstand voltage test steps are as follows: the second contactor is turned on, and the withstand voltage tester is connected to the stator winding of the magnetic levitation motor. The process control terminal PLC sends a test start command to the withstand voltage tester through the communication bus, and the withstand voltage tester outputs a specified voltage and lasts for a certain period of time. In order to improve the test accuracy, the withstand voltage tester measures the leakage current value multiple times, and transmits the measurement results to the data analysis and diagnosis module using the communication bus. The data analysis and diagnosis module determines whether the withstand voltage test is qualified by comparing the average value of multiple measurement results with the standard value. After the test is completed, the process control terminal PLC sends a shutdown command to the withstand voltage tester and disconnects the second contactor.

[0101] The insulation resistance test steps are as follows: the third contactor is turned on, and the insulation resistance tester is connected to the stator winding of the magnetic levitation motor. The process control terminal PLC sends a test start command to the insulation resistance tester through the communication bus to test the stator winding's resistance to ground R U , R V , R W The resistance R between the stator winding UV , R VW , R WU To improve the test accuracy, the insulation resistance tester performs multiple measurements and transmits the measurement results back to the data analysis and diagnosis module. The data analysis and diagnosis module determines whether the insulation resistance test is qualified by comparing the average value of multiple measurement results with the standard value. After the test is completed, the process control terminal PLC sends a shutdown command to the insulation resistance tester and disconnects the third contactor.

[0102] The no-load test steps are as follows: the first contactor is turned on, and the inverter is connected to the stator winding of the magnetic suspension motor. The process control terminal PLC sets the dynamometer to no-load torque through the communication bus and controls the inverter to output the rated voltage of the industrial frequency. The power analyzer, torque speed sensor, dynamometer and other equipment transmit the data of the stator resistance and inductance of the suspension motor, the number of motor poles, the no-load back electromotive force of the motor, the magnetic flux constant of the motor, etc. back to the data analysis and diagnosis module through the communication bus. The data analysis and diagnosis module determines whether the no-load test meets the requirements through the trained nonlinear autoregressive neural network model.

[0103] The process of measuring the motor's no-load back electromotive force is as follows: the inverter drives the motor to rotate at a constant speed, the dynamometer outputs a certain negative torque to offset the friction torque, and the power analyzer measures the signal output frequency and the corresponding open-circuit voltage of the stator winding until the signal output frequency meets the preset frequency threshold range (90% of the rated frequency). The signal output frequency and the corresponding open-circuit voltage that meet the frequency threshold range are extracted, and the no-load back electromotive force is calculated using the formula, which is as follows:

[0104]

[0105] Where E is the no-load back EMF of the motor, V is the supply voltage of the motor, i.e., the open circuit voltage. I is the current of the motor. R is the resistance of the motor. L is the inductance of the motor. dI / dt is the rate of change of the current over time, i.e., the instantaneous change of the current.

[0106] The motor flux constant measurement process is as follows: the inverter controls the motor to run at a constant speed, the dynamometer outputs a certain negative torque to offset the friction torque, the power analyzer measures the stator voltage waveform, and then uses the formula to calculate the flux constant. The calculation formula is as follows:

[0107]

[0108] Where V pp is the peak value of the back EMF. el It is the time for one electrical rotation cycle, which can be obtained by the mechanical speed and the number of pole pairs through the following calculation formula:

[0109]

[0110] Where n is the mechanical speed and p is the number of motor pole pairs.

[0111] The load test steps are as follows: After other tests are completed, stop the inverter output and let the motor temperature return to room temperature. The process control terminal PLC sends corresponding instructions to the dynamometer and inverter, and the magnetic levitation motor runs forward and reverse at low speed, medium speed, and high speed. During the process, the stator voltage and current of the magnetic levitation motor, the motor input power and output mechanical power, stray loss and efficiency, torque and speed, temperature and other data are collected and sent back to the data analysis and diagnosis module. The data analysis and diagnosis module determines whether the load test meets the requirements through the trained nonlinear autoregressive neural network model.

[0112] The steps of the stall test are as follows: After other tests are completed, stop the inverter output and let the motor temperature return to room temperature. The process control terminal PLC sends corresponding instructions to the dynamometer and inverter, and the magnetic levitation motor runs at the rated speed for a period of time. Then, the process control terminal PLC controls the dynamometer to suddenly add load until the motor stalls. During the collection process, the stator voltage, current, torque, input power, stall loss, temperature and other data of the magnetic levitation motor are transmitted back to the data analysis and diagnosis module. The data analysis and diagnosis module determines whether the stall test meets the requirements through the trained nonlinear autoregressive neural network model.

[0113] The overspeed test steps are as follows: After other tests are completed, stop the inverter output and let the motor temperature return to room temperature. The process control terminal PLC sends corresponding instructions to the dynamometer and inverter. The motor runs at a speed higher than the rated speed for a period of time. During the process, the stator voltage and current, torque and speed, motor input power and output mechanical power, response time and other data of the magnetic suspension motor are collected and transmitted back to the data analysis and diagnosis module. The data analysis and diagnosis module determines whether the overspeed test meets the requirements through the trained nonlinear autoregressive neural network model.

[0114] During the above-mentioned withstand voltage test, insulation resistance test, no-load test, load test, stall test, and overspeed test, the temperature and sound of the motor are closely monitored through temperature sensors, sound sensors, and vibration sensors to ensure the safety of the test process. If the temperature, sound, or vibration output is abnormal, the test is stopped immediately and the tester is notified. The sound and vibration data are not considered as detection data and are only used to determine whether the test process is safe.

[0115] In addition, in the above-mentioned load test and overspeed test links, the four-quadrant inverter can convert the mechanical power of the dynamometer into electrical power and feed it back to the power grid, reducing the energy loss during the experiment.

[0116] In addition, the output end of the four-quadrant inverter is connected to the grid through the first isolation transformer, and the input end of the inverter is connected to the grid through the second isolation transformer. The first isolation transformer and the second isolation transformer are used to provide electrical isolation to prevent electrical faults or overvoltages in the grid from being transmitted to the inverter or dynamometer, protecting the test equipment; filtering electromagnetic contamination of the grid to ensure the stability of the four-quadrant inverter and the inverter.

[0117] In this embodiment, the data analysis and diagnosis module includes:

[0118] An electrical performance index test judgment unit is used to analyze and judge whether the electrical performance of the magnetic levitation motor is qualified according to the electrical performance index test data;

[0119] The dynamic characteristic index test and judgment unit analyzes and judges the dynamic characteristics of the magnetic levitation motor through a nonlinear autoregressive neural network model.

[0120] In this embodiment, the electrical performance index test judgment unit includes:

[0121] The withstand voltage test judgment unit calculates the withstand voltage test data average value of the multiple test data according to the multiple test data of the electrical performance index test module, and compares the withstand voltage test data average value with the withstand voltage test data standard value. If the difference between the withstand voltage test data average value and the withstand voltage test data standard value is within a preset range, the withstand voltage test is qualified, otherwise, the withstand voltage test is unqualified;

[0122] The insulation resistance test judgment unit calculates the average value of the insulation resistance test data of multiple test data based on the multiple test data of the electrical performance index test module, and compares the average value of the insulation resistance test data with the standard value of the insulation resistance test data. If the difference between the average value of the insulation resistance test data and the standard value of the insulation resistance test data is within a preset range, the insulation resistance test is qualified, otherwise, the insulation resistance test is unqualified.

[0123] In this embodiment, the dynamic characteristic index test judgment unit includes:

[0124] The no-load test judgment unit measures the stator resistance and inductance, motor pole pair number, motor no-load back electromotive force, and motor flux constant of the magnetic levitation motor through the dynamic characteristic index test module. The nonlinear autoregressive neural network model obtains the no-load state of the current magnetic levitation motor based on the measured data combined with historical data. If the no-load state of the current magnetic levitation motor meets the no-load qualification standard, the no-load test is qualified, otherwise, the no-load test is unqualified;

[0125] The load test judgment unit measures the stator voltage and current, motor input electric power and output mechanical power, stray loss and efficiency, torque and speed, and temperature of the magnetic levitation motor through the dynamic characteristic index test module. The nonlinear autoregressive neural network model obtains the current load state of the magnetic levitation motor based on the measured data combined with historical data. If the current load state of the magnetic levitation motor meets the load qualification standard, the load test is qualified, otherwise, the load test is unqualified;

[0126] The stall test judgment unit measures the stator voltage and current, torque, input electric power, stall loss, and temperature of the magnetic levitation motor through the dynamic characteristic index test module. The nonlinear autoregressive neural network model obtains the stall state of the current magnetic levitation motor based on the measured data combined with historical data. If the stall state of the current magnetic levitation motor meets the stall qualification standard, the stall test is qualified, otherwise, the stall test is unqualified;

[0127] The overspeed test judgment unit measures the stator voltage and current, torque and speed, motor input electrical power and output mechanical power, and response time of the magnetic levitation motor through the dynamic characteristic index test module. The nonlinear autoregressive neural network model combines the measured data with historical data to obtain the overspeed test status of the current magnetic levitation motor. If the overspeed test status of the current magnetic levitation motor meets the overspeed qualification standard, the overspeed test is qualified. Otherwise, the overspeed test is unqualified.

[0128] In this embodiment, the basic principle of the nonlinear autoregressive neural network model is to use the historical information in the time series data to predict future behavior through a regression model. The formula of the nonlinear autoregressive neural network model is:

[0129] y(t)=f(y(t-1),y(t-2),...,y(tn y ),u(t-1),u(t-2),...,u(tn u )),

[0130] Where y(t) is the current output variable of the system, f is the nonlinear regression function, y(t-1), y(t-2), ..., y(tn y ) is the historical output variable of the system, n y is the order of the output delay, u(t-1),u(t-2),...,u(tn u ) is the historical input variable of the system, n u is the order of input delay, u(t) is the current input variable of the system;

[0131] like Figure 3 As shown in Figure 1, the nonlinear autoregressive neural network model includes an input layer, a hidden layer, and an output layer. The input data are normalized voltage, current, input power, speed, torque, output power, temperature rise, vibration, and noise data. When selected, the delay order d is 6, and the number of hidden layer neurons h is 12. The order of the output delay is the same as the order of the input delay, n y =n u =d.

[0132] In actual testing, the nonlinear autoregressive neural network model predicts the current state of the motor based on the input data (current, voltage, speed, etc.) collected in real time, combined with the previous historical state, and analyzes the output of the nonlinear autoregressive neural network model and compares it with the set qualification standards to determine whether the magnetic levitation motor passes the test. Each test scenario (such as no-load, load, stall, and overspeed tests) has a set of qualification standards. These standards include the allowable range of key parameters such as current, voltage, torque, power, and temperature rise.

[0133] The output results (such as torque, speed, etc.) of the nonlinear autoregressive neural network model are compared with the preset qualified standards. If the output results are within the set qualified range, the test is judged as qualified; otherwise, it is unqualified.

[0134] In this embodiment, the activation function f1 of the hidden layer is a hyperbolic tangent S-type transfer function, that is, a tansig function, and the formula is:

[0135]

[0136] The calculation formula for the hidden layer is:

[0137] h(t)=f1(w1u(t)+w2y(t)+b1),

[0138] Where h(t) is the output of the hidden layer, u(t) is the input of each layer of neurons, y(t) is the input of each layer of neurons, w1 and w2 are the connection weights between the input layer nodes and neurons, and b1 is the threshold of the hidden layer.

[0139] In this embodiment, the activation function f2 of the output layer is a linear transfer function, that is, a purelin function, and the formula is:

[0140] f2(x)=purelin(x)=x;

[0141] The calculation formula of the output layer is:

[0142] y(t)=f2(w3h(t)+b2),

[0143] Where y(t) is the current output of the system, w3 is the connection weight between the hidden layer nodes and the output layer nodes, and b2 is the threshold of the output layer.

[0144] In this embodiment, the training function of the nonlinear autoregressive neural network model is the Levenberg-Marquardt algorithm. A judgment index is set in the update iteration process to determine the quality of the approximation. The connection weight adjustment method formula of the Levenberg-Marquardt algorithm is:

[0145] Δw=(J T J+μI) -1 J T e,

[0146] Where μ is an adaptively adjusted scalar.

[0147] The Levenberg-Marquardt algorithm combines the advantages of gradient descent and Newton's method. Compared with the traditional gradient descent method, it usually has a faster convergence speed. For nonlinear autoregressive neural network models with complex dynamic characteristics, the Levenberg-Marquardt algorithm can adjust weights more efficiently and quickly approach the optimal solution.

[0148] The nonlinear autoregressive neural network model is a type of nonlinear dynamic system modeling, and the Levenberg-Marquardt algorithm can better handle this nonlinear problem. Unlike the simple gradient descent method, the Levenberg-Marquardt algorithm has better robustness when dealing with highly nonlinear functions and can find the global optimal solution in most cases rather than the local optimal solution.

[0149] The Levenberg-Marquardt algorithm automatically adjusts the learning rate during training, showing the fast convergence characteristics of the Newton method when close to the optimal solution, and the conservative nature of gradient descent when far away from the optimal solution. This flexibility enables the Levenberg-Marquardt algorithm to handle diverse and complex systems and reduce the risk of falling into local minima.

[0150] Because the Levenberg-Marquardt algorithm combines second-order derivative information (approximate Hessian matrix), it has stronger numerical stability and accuracy than pure first-order methods (such as gradient descent). This stability can reduce oscillation or divergence for complex nonlinear autoregressive neural network models, especially when a large amount of time series data is involved in the training process.

[0151] The Levenberg-Marquardt algorithm can effectively adjust the model error by considering the minimization of the sum of squared errors. This is very helpful for time series prediction and complex system modeling of nonlinear autoregressive neural network models, especially under conditions of large data noise or complex dynamic conditions, the error correction ability of the Levenberg-Marquardt algorithm is more obvious.

[0152] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. An intelligent comprehensive performance test platform for magnetically suspended high-speed permanent magnet synchronous motors, characterized in that: include: Electrical performance index test module, used to detect the electrical performance of the magnetic levitation motor; Dynamic characteristic index test module, used to detect the corresponding change ability of the magnetic levitation motor during operation; Process control module, used to control the testing equipment and automatically switch the testing links; The data analysis and diagnosis module is used to analyze and diagnose the data obtained from the electrical performance index test and the dynamic characteristic index test, wherein the dynamic characteristic index is analyzed and diagnosed through a nonlinear autoregressive neural network model with external input.

2. According to claim 1, a magnetic suspension high-speed permanent magnet synchronous motor intelligent comprehensive performance test platform is characterized in that: The electrical performance index test module includes: A withstand voltage test unit, used to measure the leakage current value of the stator winding of the magnetic levitation motor multiple times through the process control module; The insulation resistance test unit is used to measure the resistance value of the stator winding of the magnetic levitation motor multiple times through the process control module.

3. According to claim 1, a magnetic suspension high-speed permanent magnet synchronous motor intelligent comprehensive performance test platform is characterized in that: The dynamic characteristic index test module includes: The no-load test unit is used to collect the voltage, current, input power, torque, speed, output power and other data of the magnetic levitation motor under no-load conditions through the process control module; The load test unit is used to collect data such as voltage, current, input power, torque, speed, output power, temperature rise, vibration and noise of the magnetic suspension motor when it is running forward and reverse at low speed, medium speed and high speed under load through the process control module; The stall test unit is used to collect data such as voltage, current, input power, torque, speed, output power, temperature rise, vibration and noise when the magnetic levitation motor runs at rated speed for a period of time and then suddenly adds load until the magnetic levitation motor stalls through the process control module; The overspeed test unit is used to collect data such as voltage, current, input power, torque, speed, output power, temperature rise, vibration and noise when the magnetic levitation motor runs at a speed higher than the rated speed for a period of time through the process control module.

4. The intelligent comprehensive performance test platform for magnetic suspension high-speed permanent magnet synchronous motor according to claim 1 is characterized in that: The process control module includes: A withstand voltage tester, which is used to output a specified voltage to the stator winding of the magnetic levitation motor and measure the leakage current; Insulation resistance tester, used to test the stator winding resistance to ground R of magnetic levitation motor U , R V , R W The resistance R between the stator winding UV , R VW , R WU The resistance value; The frequency converter is used to change the frequency and voltage of the magnetic levitation motor input to adjust the speed of the magnetic levitation motor; Power analyzers are used to measure parameters such as electrical power, energy and efficiency of electrical systems or equipment; Torque and speed sensor, used to detect the torque and speed on the magnetic suspension motor shaft; Dynamometer, used to apply different mechanical loads and measure the output torque and power of the magnetic levitation motor; Temperature sensor, vibration sensor, and sound sensor are used to detect the temperature, vibration, and sound of the magnetic levitation motor; Four-quadrant frequency converter, used to convert the mechanical power of the dynamometer into electrical power to feed back to the power grid; The first contactor, the second contactor and the third contactor are respectively used to connect with the frequency converter, the withstand voltage tester and the insulation resistance tester to realize the switching of the frequency converter, the withstand voltage tester or the insulation resistance tester; The process control terminal PLC is used to control the withstand voltage tester, the insulation resistance tester, the frequency converter, the power analyzer, the torque speed sensor, the dynamometer, the four-quadrant frequency converter, the first contactor, the second contactor, and the third contactor, wherein the first contactor, the second contactor, and the third contactor are controlled to realize automatic switching of different test links; AC busbar, used to connect the magnetic levitation motor with the power analyzer, frequency converter, voltage withstand tester, and insulation resistance tester, so as to realize the detection or control of the magnetic levitation motor by the power analyzer, frequency converter, voltage withstand tester, and insulation resistance tester; Communication bus, used to connect and control the withstand voltage tester, insulation resistance tester, frequency converter, power analyzer, torque and speed sensor, dynamometer, four-quadrant frequency converter, temperature sensor, vibration sensor, and sound sensor to achieve data and signal transmission; The switch quantity controller circuit is used to realize the control of the first contactor, the second contactor and the third contactor by the process control terminal PLC; Among them, the input end of the frequency converter is connected to the power grid, the output ends of the frequency converter, the voltage withstand tester, and the insulation resistance tester are respectively connected to the input ends of the first contactor, the second contactor, and the third contactor, the output ends of the first contactor, the second contactor, and the third contactor are connected to the input end of the AC busbar, the output end of the AC busbar is connected to the input end of the power analyzer and the stator winding of the magnetic levitation motor, the input end of the torque and speed sensor and the output end of the dynamometer are connected to the rotating shaft of the magnetic levitation motor, the input end of the four-quadrant frequency converter is connected to the dynamometer, the output end of the four-quadrant frequency converter is connected to the power grid, the input end of the temperature sensor, the vibration sensor, and the sound sensor are connected to the magnetic levitation motor, the communication bus is connected to the voltage withstand tester, the insulation resistance tester, the frequency converter, the power analyzer, the torque and speed sensor, the dynamometer, the four-quadrant frequency converter, the temperature sensor, the vibration sensor, the sound sensor, the process control terminal PLC, and the data analysis and diagnosis module, and the output end of the switch control circuit is connected to the coils of the first contactor, the second contactor, and the third contactor.

5. The intelligent comprehensive performance test platform for magnetic suspension high-speed permanent magnet synchronous motor according to claim 1 is characterized in that: The data analysis and diagnosis module includes: The electrical performance index test judgment unit performs analysis and judgment based on the electrical performance index test data; The dynamic characteristic index test and judgment unit analyzes and judges the dynamic characteristics of the magnetic levitation motor through a nonlinear autoregressive neural network model.

6. The intelligent comprehensive performance test platform for magnetically suspended high-speed permanent magnet synchronous motor according to claim 5 is characterized in that: The electrical performance index test judgment unit includes: The withstand voltage test judgment unit calculates the withstand voltage test data average value of the multiple test data according to the multiple test data of the electrical performance index test module, and compares the withstand voltage test data average value with the withstand voltage test data standard value. If the difference between the withstand voltage test data average value and the withstand voltage test data standard value is within a preset range, the withstand voltage test is qualified, otherwise, the withstand voltage test is unqualified; The insulation resistance test judgment unit calculates the average value of the insulation resistance test data of multiple test data based on the multiple test data of the electrical performance index test module, and compares the average value of the insulation resistance test data with the standard value of the insulation resistance test data. If the difference between the average value of the insulation resistance test data and the standard value of the insulation resistance test data is within a preset range, the insulation resistance test is qualified, otherwise, the insulation resistance test is unqualified.

7. The intelligent comprehensive performance test platform for magnetic suspension high-speed permanent magnet synchronous motor according to claim 5 is characterized in that: The dynamic characteristic index test judgment unit comprises: The no-load test judgment unit measures the stator resistance and inductance, the number of motor poles, the no-load back electromotive force, and the motor flux constant of the magnetic levitation motor through the dynamic characteristic index test module. The nonlinear autoregressive neural network model obtains the no-load test state of the current magnetic levitation motor based on the measured data combined with historical data. If the no-load test state of the current magnetic levitation motor meets the no-load qualification standard, the no-load test is qualified, otherwise, the no-load test is unqualified; The load test judgment unit measures the stator voltage and current, motor input electric power and output mechanical power, stray loss and efficiency, torque and speed of the magnetic levitation motor through the dynamic characteristic index test module. The nonlinear autoregressive neural network model obtains the current load test state of the magnetic levitation motor based on the measured data combined with historical data. If the current load test state of the magnetic levitation motor meets the load qualification standard, the load test is qualified, otherwise, the load test is unqualified; The stall test judgment unit measures the stator voltage and current, torque, input electric power, and stall loss of the magnetic levitation motor through the dynamic characteristic index test module. The nonlinear autoregressive neural network model obtains the stall test status of the current magnetic levitation motor based on the measured data combined with historical data. If the stall test status of the current magnetic levitation motor meets the stall qualified standard, the stall test is qualified, otherwise, the stall test is unqualified; The overspeed test judgment unit measures the stator voltage and current, torque and speed, motor input electrical power and output mechanical power, and response time of the magnetic levitation motor through the dynamic characteristic index test module. The nonlinear autoregressive neural network model combines the measured data with historical data to obtain the overspeed test status of the current magnetic levitation motor. If the overspeed test status of the current magnetic levitation motor meets the overspeed qualification standard, the overspeed test is qualified. Otherwise, the overspeed test is unqualified.

8. The intelligent comprehensive performance test platform for magnetically suspended high-speed permanent magnet synchronous motor according to claim 1 or 5, characterized in that: The formula of the nonlinear autoregressive neural network model is: y(t)=f(y(t-1),y(t-2),...,y(t-n y ),u(t-1),u(t-2),...,u(t-n u )), Where y(t) is the current output variable of the system, f is the nonlinear regression function, y(t-1), y(t-2), ..., y(tn y ) is the historical output variable of the system, n y is the order of the output delay, u(t-1),u(t-2),...,u(tn u ) is the historical input variable of the system, n u is the order of input delay; The nonlinear autoregressive neural network model includes an input layer, a hidden layer and an output layer, wherein the input data are normalized voltage, current, input power, speed, torque, output power, temperature rise, vibration and noise data. When selected, the delay order d is 6 and the number of hidden layer neurons h is 12.

9. The intelligent comprehensive performance test platform for magnetically suspended high-speed permanent magnet synchronous motor according to claim 8 is characterized in that: The activation function f1 of the hidden layer is a hyperbolic tangent S-type transfer function, i.e., a tansig function, and the formula is: The calculation formula of the hidden layer is: h(t)=f1(w1u(t)+w2y(t)+b1), Where h(t) is the output of the hidden layer, u(t) is the input of each layer of neurons, y(t) is the input of each layer of neurons, w1 and w2 are the connection weights between the input layer nodes and neurons, and b1 is the threshold of the hidden layer; The activation function f2 of the output layer is a linear transfer function, namely, purelin function, and the formula is: f2(x)=purelin(x)=x; The calculation formula of the output layer is: y(t)=f2(w3h(t)+b2), Where y(t) is the current output of the system, w3 is the connection weight between the hidden layer nodes and the output layer nodes, and b2 is the threshold of the output layer.

10. The intelligent comprehensive performance test platform for magnetically suspended high-speed permanent magnet synchronous motor according to claim 1 or 5, characterized in that: The training function of the nonlinear autoregressive neural network model is the Levenberg-Marquardt algorithm. A judgment index is set in the update iteration process to determine the quality of the approximation. The connection weight adjustment method formula of the Levenberg-Marquardt algorithm is: Δw=(J T J+μI) -1 J T e, Where μ is an adaptively adjusted scalar.

Images