Three-phase full-bridge inverter fault diagnosis method, system and equipment

Abstract

The invention relates to the technical field of fault type diagnosis, in particular to a three-phase full-bridge inverter fault diagnosis method, system and equipment. The method comprises the following steps: acquiring a PWM command signal in a three-phase full-bridge inverter and an electric signal and a temperature signal of an IGBT device; based on the bridge arm on-off logic performance, the electric signal and the temperature signal of the IGBT device in the PWM instruction signal period, obtaining an on-off logic analysis coefficient and a conduction efficiency coefficient of the IGBT device, combining a fault determination threshold and a constraint condition, obtaining a fault type label of the IGBT, constructing a fault sample set, training a fault type identification model, and obtaining a fault type identification result; obtaining a trained fault type identification model; according to the method, the on-off logic analysis coefficient and the conduction efficiency coefficient are accurately obtained, so that the accuracy of fault type tag identification is effectively improved, the fault type of the IGBT device can be accurately identified in real time, corresponding processing is carried out in time, and the fault loss of the three-phase full-bridge inverter is effectively reduced.

Description

Technical Field

[0001] This invention relates to the field of fault type diagnosis technology, specifically to a fault diagnosis method, system and equipment for a three-phase full-bridge inverter. Background Technology

[0002] Three-phase full-bridge inverters, as core devices for power conversion, are widely used in key areas such as new energy power generation, motor drives, and uninterruptible power supplies. They contain multiple IGBTs (Insulated Gate Bipolar Transistors) and other core power devices, operating under high-frequency switching, high-current, and high-voltage conditions. They are susceptible to factors such as sudden load changes, component aging, and poor heat dissipation, leading to IGBT device failures. Common fault types include short circuits, open circuits, overloads, and overheating. If these faults are not identified and handled promptly and accurately, they can cause equipment shutdowns and production interruptions, or even permanent damage to IGBT devices and serious safety accidents such as fires. Therefore, real-time and accurate fault diagnosis of three-phase full-bridge inverters is of great significance for ensuring the reliability, safety, and economy of power electronic systems.

[0003] Currently, analyzing the current, voltage, and temperature signals of IGBT devices can determine whether there is a fault in the IGBT device, but it cannot further determine the specific fault type, such as short circuit, open circuit, overload, or overheating. This is because the causes, risk levels, and handling methods of these four types of faults are completely different. This ambiguous diagnostic result forces maintenance personnel to rely on personal experience to blindly and systematically disassemble and troubleshoot, which not only significantly prolongs downtime and increases unplanned maintenance costs, but also carries the risk of fault escalation due to the failure to take timely and correct measures against the specific root cause of the fault, ultimately affecting the reliability and availability of the entire inverter system. Summary of the Invention

[0004] To address the technical problem of the inability to accurately assess IGBT device fault types in real time, the present invention aims to provide a fault diagnosis method, system, and device for three-phase full-bridge inverters. The specific technical solution adopted is as follows: In a first aspect, one embodiment of the present invention provides a fault diagnosis method for a three-phase full-bridge inverter, the method comprising the following steps: The system acquires the PWM command signal, the electrical signal and temperature signal of each IGBT device at each moment in the three-phase full-bridge inverter; the electrical signal includes the collector current and collector-emitter voltage; the temperature signal includes the IGBT junction temperature. Based on the bridge arm switching logic performance and collector current changes of each IGBT device within a preset number of PWM command signal cycles, the switching logic analysis coefficients of each IGBT device are obtained. Based on the collector-emitter voltage fluctuation of each IGBT device within a preset number of PWM command signal cycles, as well as the magnitude of collector current, collector-emitter voltage and IGBT junction temperature, the conduction efficiency coefficient of each IGBT device is obtained. Based on the on / off logic analysis coefficients and conduction efficiency coefficients of each IGBT device, combined with fault determination thresholds and constraints, a fault type label for each IGBT is obtained; the fault type label includes short circuit, open circuit, overload, overheating, and no fault. A fault sample set is constructed based on fault type labels, and a fault type identification model is trained to obtain a trained fault type identification model; the fault type identification model is used to output the fault type of each IGBT device in the three-phase full-bridge inverter.

[0005] Furthermore, the method for obtaining the on / off logic analysis coefficients is as follows: The compliance level of the on / off logic of each IGBT device is obtained based on the number of cycles in which the bridge arm on / off logic of each IGBT device complies with the preset number of PWM command signal cycles and the number of cycles in which the bridge arm on / off logic of each device violates the preset number of PWM command signal cycles. Based on the collector current change of each IGBT device within a preset number of PWM command signal cycles, obtain the current correlation correction coefficient for each IGBT device. The product of the on / off logic compliance level and the current-related correction coefficient for each IGBT device is used as the on / off logic analysis coefficient for each IGBT device.

[0006] Furthermore, the method for obtaining the compliance level of the on / off logic is as follows: For any IGBT device, the ratio of the number of cycles in which the bridge arm switching logic of the IGBT device complies with the predetermined number is used as the first compliance analysis value of the IGBT device. The number of cycles in which the continuous bridge arm switching logic of the IGBT device violates the rule is used as the first quantity. The result of negatively correlating the ratio of the largest first quantity to the preset quantity is used as the compliance correction weight for the IGBT device. The product of the first compliance analysis value and the compliance correction weight of the IGBT device is taken as the compliance level of the IGBT device's on / off logic.

[0007] Furthermore, the method for obtaining the current-related correction coefficient is as follows: For any IGBT device, the ratio of the collector current rise rate to the preset rated current rise rate within each PWM command signal cycle is used as the current surge level. The ratio of each collector current of the IGBT device to the preset rated current within a preset number of PWM command signal cycles is used as the current over-current analysis value. The product of the maximum current surge and the mean current excess analysis value is negatively correlated and normalized, and the result is used as the current correlation correction coefficient for this IGBT device.

[0008] Furthermore, the method for obtaining the conduction efficiency coefficient is as follows: For any IGBT device, the ratio of each collector-emitter voltage of the IGBT device to the preset rated voltage within a preset number of PWM command signal cycles is used as the voltage over-limit analysis value. The standard deviation of the voltage excess analysis value is negatively correlated and normalized to obtain the voltage stability of the IGBT device. For any PWM command signal period, the product of the average collector-emitter voltage and the average collector current of the IGBT device during that PWM command signal period is taken as the actual loss of the IGBT device during that PWM command signal period. The average actual loss of the IGBT device within a preset number of PWM command signal cycles is taken as the overall actual loss of the IGBT device. The result of normalizing the difference between the overall actual loss and the preset rated loss is used as the loss rating analysis value of the IGBT device. The result of normalizing the average IGBT junction temperature of the IGBT device within a preset number of PWM command signal cycles is used as the junction temperature correction coefficient of the IGBT device. The product of the loss rating analysis value of the IGBT device and the junction temperature correction factor is used as the energy transfer analysis value of the IGBT device. The product of the negative correlation result of the energy transfer analysis value of the IGBT device and the voltage stability is used as the conduction efficiency coefficient of the IGBT device.

[0009] Furthermore, the fault determination thresholds include an abnormal on / off logic threshold, a short-circuit-overload distinction threshold, an abnormal conduction performance threshold, an open-circuit-overheat distinction threshold, and a junction temperature critical threshold. For any IGBT device, the abnormal threshold of the on / off logic of the IGBT device is the difference between the mean of the on / off logic analysis coefficients of the IGBT device under normal operating conditions and three times its standard deviation. The short-circuit-overload distinction threshold of this IGBT device is the result of negatively correlated with the ratio of the saturated collector-emitter voltage of this IGBT device to the bus voltage under normal operating conditions. The abnormal threshold for the conduction performance of this IGBT device is the difference between the mean of the conduction performance coefficient of the IGBT device under normal operating conditions and three times its standard deviation. The open-circuit / overheating distinction threshold of this IGBT device is preset based on experimental data; The critical junction temperature threshold of this IGBT device is the maximum IGBT junction temperature of the IGBT device minus three times the standard deviation of the IGBT junction temperature under a specified number of normal operating conditions.

[0010] Furthermore, the method for obtaining the fault type label is as follows: For any IGBT device, when the on / off logic analysis coefficient of the IGBT device is less than the short-circuit-overload distinction threshold and the conduction efficiency coefficient is greater than or equal to the conduction efficiency abnormal boundary threshold, the fault type label of the IGBT device is short circuit. When the on / off logic analysis coefficient of the IGBT device is greater than or equal to the short-circuit-overload distinction threshold and less than the on / off logic abnormal boundary threshold, and the conduction performance coefficient is greater than or equal to the conduction performance abnormal boundary threshold, the fault type label of the IGBT device is overload. When the on / off logic analysis coefficient of the IGBT device is greater than or equal to the on / off logic abnormality threshold, and the conduction efficiency coefficient is less than the open circuit-overheating distinction threshold, the fault type label of the IGBT device is open circuit. When the on / off logic analysis coefficient of the IGBT device is greater than or equal to the on / off logic abnormality threshold, the conduction performance coefficient is greater than or equal to the open circuit-overheating distinction threshold and less than the conduction performance abnormality threshold, and the IGBT junction temperature is greater than the junction temperature critical threshold, the fault type label of the IGBT device is overheating. When the on / off logic analysis coefficient of the IGBT device is greater than or equal to the on / off logic abnormality threshold, and the conduction performance coefficient is greater than or equal to the conduction performance abnormality threshold, the fault type label of the IGBT device is no fault.

[0011] Furthermore, the fault type identification model is a graph neural network model; the fault sample set uses each IGBT device as an independent node, and the physical topological association between devices as edges to construct a graph structure sample. The feature vector of each node includes on / off logic analysis coefficients and conduction efficiency coefficients, and the node label is a fault type label.

[0012] Secondly, another embodiment of the present invention provides a fault diagnosis system for a three-phase full-bridge inverter, the system comprising: The data acquisition module is used to acquire the PWM command signal, the electrical signal and temperature signal of each IGBT device at each moment in the three-phase full-bridge inverter; the electrical signal includes the collector current and collector-emitter voltage; the temperature signal includes the IGBT junction temperature. The on / off logic analysis coefficient acquisition module is used to acquire the on / off logic analysis coefficients of each IGBT device based on the on / off logic performance of the bridge arm and the change of collector current of each IGBT device within a preset number of PWM command signal cycles. The conduction efficiency coefficient acquisition module is used to obtain the conduction efficiency coefficient of each IGBT device based on the collector-emitter voltage fluctuation of each IGBT device within a preset number of PWM command signal cycles, as well as the magnitude of collector current, collector-emitter voltage and IGBT junction temperature. The fault type label acquisition module is used to acquire the fault type label of each IGBT based on the on / off logic analysis coefficient and conduction efficiency coefficient of each IGBT device, combined with the fault judgment threshold and constraints; the fault type label includes short circuit, open circuit, overload, overheat and no fault; The data processing module is used to construct a fault sample set based on fault type labels, train a fault type identification model, and obtain a trained fault type identification model; the fault type identification model is used to output the fault type of each IGBT device in the three-phase full-bridge inverter.

[0013] Thirdly, another embodiment of the present invention provides a fault diagnosis device for a three-phase full-bridge inverter, the device comprising: a memory, a processor, and a computer program stored in the memory and running on the processor, wherein when the processor executes the computer program, it implements the steps of any of the above methods.

[0014] The present invention has the following beneficial effects: This invention first obtains the on / off logic analysis coefficients for each IGBT device based on the bridge arm on / off logic performance and collector current changes of each IGBT device within a preset number of PWM command signal cycles. This accurately reflects the compliance of the IGBT device's bridge arm on / off logic, current surge intensity, and load matching degree, clearly depicting the fault characteristics caused by short circuits and overloads due to abnormal on / off logic or current exceeding limits. This facilitates the subsequent accurate definition of the risk range of on / off related faults and provides a core basis for fault type differentiation. To comprehensively characterize the conduction state and operating characteristics of the IGBT devices, this invention further obtains the conduction efficiency coefficients for each IGBT device based on the collector-emitter voltage fluctuations, collector current, collector-emitter voltage, and IGBT junction temperature within a preset number of PWM command signal cycles. This accurately reflects the voltage stability, energy transfer efficiency, and temperature-loss coupling risk of the IGBT device, precisely capturing the on / off performance. Fault characteristics such as short circuits and overheating caused by device damage or accumulated losses are helpful for effectively distinguishing the types of conduction-related faults, compensating for the deficiency that a single parameter cannot comprehensively evaluate the conduction state. Furthermore, based on the on / off logic analysis coefficients and conduction efficiency coefficients of each IGBT device, combined with fault judgment thresholds and constraints, a fault type label for each IGBT can be obtained. This is beneficial for providing accurate and unique supervision signals for the fault sample set, ensuring the effectiveness and reliability of model training. Subsequently, a fault sample set is constructed based on the fault type label, and a fault type recognition model is trained to obtain a well-trained fault type recognition model. This model can fully integrate the IGBT device's own state characteristics and circuit topology association information, effectively improving the identification accuracy, real-time performance, and robustness of IGBT device fault types in three-phase full-bridge inverters. It can accurately determine five states: short circuit, open circuit, overload, overheating, and no fault, adapting to the application requirements of complex industrial conditions. Attached Figure Description

[0015] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 A schematic flowchart illustrating a fault diagnosis method for a three-phase full-bridge inverter provided in one embodiment of the present invention; Figure 2 A flowchart illustrating a method for obtaining on / off logic analysis coefficients according to an embodiment of the present invention; Figure 3 A flowchart illustrating a method for obtaining the conduction efficiency coefficient according to an embodiment of the present invention; Figure 4 This is a structural diagram of a three-phase full-bridge inverter fault diagnosis system provided in one embodiment of the present invention; Figure 5 This is a schematic diagram of a computer device provided according to an embodiment of the present invention. Detailed Implementation

[0017] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a three-phase full-bridge inverter fault diagnosis method, system, and device proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0019] The following description, in conjunction with the accompanying drawings, details the specific solution of a three-phase full-bridge inverter fault diagnosis method, system, and equipment provided by the present invention.

[0020] Example 1: This invention proposes a fault diagnosis method for a three-phase full-bridge inverter. Please refer to [link / reference]. Figure 1 The diagram illustrates a schematic flowchart of a fault diagnosis method for a three-phase full-bridge inverter according to an embodiment of the present invention. The method includes the following steps: Step S1: Obtain the PWM command signal, the electrical signal and temperature signal of each IGBT device at each moment in the three-phase full-bridge inverter; the electrical signal includes the collector current and collector-emitter voltage; the temperature signal includes the IGBT junction temperature.

[0021] Specifically, fault diagnosis of a three-phase full-bridge inverter relies on accurate sensing of the operating status of IGBT devices. Short circuit, open circuit, overload, and overheating faults in IGBT devices are all manifested through abnormal changes in electrical and temperature signals. Examples include instantaneous current surges during a short circuit, a surge in on-state voltage drop during an open circuit, continuous overcurrent during an overload, and excessive IGBT junction temperature during overheating. Therefore, to accurately detect faults in a three-phase full-bridge inverter, this embodiment employs a multi-sensor collaborative and time-synchronized acquisition architecture. Using the PWM command signal as the timing reference, dedicated current, voltage, and temperature sensors are used to collect the electrical and temperature signals of each IGBT device in the three-phase full-bridge inverter. The sampling frequency of all sensors is synchronized with the PWM command signal, ensuring a one-to-one correspondence between the electrical signal, temperature signal, and control signal (PWM) at the same moment. The electrical signals include collector current and collector-emitter voltage, and the temperature signals include IGBT junction temperature.

[0022] It should be noted that the sampling trigger signals of all sensors are synchronized with the start edge of the PWM command, and the sampling frequency is no less than 10 times the PWM switching frequency. For example, if the PWM switching frequency is 10kHz, the sampling frequency should be greater than or equal to 100kHz to avoid feature calculation errors due to timing misalignment. Furthermore, the measurement range of the sensors must cover the extreme signal values ​​of the IGBT device under normal operation and fault conditions. For example, current sensors must cover the instantaneous inrush current during a short circuit, and voltage sensors must cover the on-state voltage drop close to the bus voltage during an open circuit. The measurement accuracy should be no less than 0.5%, ensuring that even minute fault characteristics can be captured. Each IGBT device is configured with an independent signal acquisition channel to avoid ambiguous fault node location caused by mixed sampling of signals from multiple devices, ensuring accurate correlation between signal anomaly, IGBT device, and fault type.

[0023] The PWM (Pulse Width Modulation) command signal is a drive control signal output by the three-phase full-bridge inverter control system to the IGBT gate. It is used to control the turn-on and turn-off timing of the IGBT devices. Its high level corresponds to the turn-on command of the IGBT device, and its low level corresponds to the turn-off command of the IGBT device. It is a digital control signal and serves as the benchmark for judging the compliance of the bridge arm on / off logic. By comparing the turn-on or turn-off state required by the PWM command signal with the turn-on or turn-off state reflected by the actual electrical signal of the IGBT device, it is determined whether the bridge arm on / off logic is compliant. In this embodiment, the PWM command signal output by the inverter control unit is directly acquired through the digital signal acquisition module, and the start time, high level duration, and low level duration of each command cycle are recorded. Collector current is the current flowing into the collector (C) and out of the emitter (E) of an IGBT device. It is a core parameter reflecting the load state of the IGBT device. During normal operation, the collector current is stable within the rated current range. Under short circuit conditions, the collector current increases explosively; under overload conditions, the collector current continuously exceeds the rated current; and under open circuit conditions, the collector current approaches 0A. To more accurately monitor the collector current, this embodiment uses a closed-loop Hall current sensor (model LT508-S6) to acquire the collector current of each IGBT device at each moment. The LT508-S6 closed-loop Hall current sensor is compatible with the operating current range of IGBT devices and has strong anti-electromagnetic interference capabilities, making it suitable for the strong electromagnetic environment of inverters. The closed-loop Hall current sensor is connected in series in the collector circuit of the IGBT device, installed close to the IGBT device, shortening the lead length to reduce the influence of stray inductance. Simultaneously, electrical isolation between the closed-loop Hall current sensor and the IGBT device is ensured to avoid high-voltage breakdown. Furthermore, the positive and negative directions of the current (reflecting the current flow direction) and the rate of change are recorded simultaneously during collector current acquisition.

[0024] The collector-emitter voltage is the voltage difference between the collector and emitter of an IGBT device. It is a core parameter reflecting the conduction state of the IGBT device. During normal conduction, the collector-emitter voltage is the saturation voltage drop (typically 1V-2V, refer to the IGBT device manual for specific values). During a short circuit, the collector-emitter voltage is close to 0V. During an open circuit, the collector-emitter voltage spikes to near the bus voltage. During overheating, the collector-emitter voltage exceeds the saturation voltage drop. To monitor the collector-emitter voltage more accurately, this embodiment selects a voltage sensor of model LV25-P, as it is compatible with the operating voltage range of the IGBT device and offers high safety and good linearity. In this embodiment, the voltage sensor is connected in parallel across the collector and emitter of the IGBT device, with the voltage polarity matching that of the IGBT device. Shielded cables are used to transmit the signal, reducing voltage fluctuations caused by electromagnetic interference. Furthermore, only the collector-emitter voltage during the PWM command conduction phase of the IGBT device is collected, as the collector-emitter voltage during the turn-off phase is meaningless.

[0025] The IGBT junction temperature is the temperature of the PN junction inside the IGBT device chip. It is a direct parameter reflecting the thermal state of the IGBT device. During normal operation, it is below the maximum allowable IGBT junction temperature. When overheated, it exceeds the critical threshold of the IGBT junction temperature. Overload and heat dissipation failure will cause the IGBT junction temperature to rise continuously. In order to monitor the IGBT junction temperature more accurately, this embodiment selects the WZP-230 temperature sensor because it can completely cover the operating and fault temperatures of the IGBT device. At the same time, it has high accuracy, good stability and small size. In this embodiment, the temperature sensor is mounted on the heat dissipation substrate of the IGBT device chip to ensure good thermal conduction with the chip. In addition, the temperature sensor is electrically isolated from the high voltage area of ​​the IGBT device to avoid the risk of short circuit.

[0026] To ensure signal accuracy, the current, voltage, and temperature sensors need to be pre-calibrated before signal acquisition. Measurement errors are corrected using standard signal sources (such as standard current sources, standard voltage sources, and standard temperature chambers) to ensure that the accuracy meets requirements. When the inverter enters normal operation or preventive maintenance mode, data acquisition is initiated. Valid data is filtered through the PWM command signal timing, retaining only the collector current, collector-emitter voltage, and corresponding IGBT junction temperature during the IGBT's on-phase phase in each PWM command signal cycle. The PWM command state, collector current, collector-emitter voltage, and IGBT junction temperature of the same IGBT device at the same time are aligned based on timestamps.

[0027] The comprehensive, accurate, and synchronous data acquisition scheme described above can obtain complete operating status data of each IGBT device within the PWM command signal cycle, laying a solid foundation for accurate determination of subsequent fault types and ensuring the accuracy and reliability of fault diagnosis.

[0028] Step S2: Based on the bridge arm switching logic performance and collector current changes of each IGBT device within a preset number of PWM command signal cycles, obtain the switching logic analysis coefficients of each IGBT device.

[0029] Specifically, as the core switching device in a three-phase full-bridge inverter, the IGBT device undertakes the critical functions of power conversion and high-frequency on / off control. Its operating state directly determines the stability and reliability of the inverter system. It is known that abnormal on / off logic of the bridge arms is a direct cause of short-circuit faults and a core trigger for overload faults caused by current surges. Among these, abnormal on / off logic of the bridge arms in multiple consecutive command signal cycles often points to an abnormality in the IGBT device itself, while sporadic abnormalities are mostly transient electromagnetic interference. Furthermore, the same on / off violation has a much higher fault risk under heavy load conditions than under light load conditions. Therefore, this embodiment obtains the on / off logic analysis coefficients of each IGBT device based on the bridge arm on / off logic performance and collector current changes of each IGBT device within a preset number of PWM command signal cycles. This accurately reflects the potential risk level of short-circuit or overload faults caused by abnormal on / off logic of each IGBT device, which is beneficial for accurately distinguishing between short-circuit and overload fault types. In this embodiment, the preset number is set to 10, and the 10 PWM command signal cycles are uninterrupted. This ensures the stability of the statistical results (effectively filtering out instantaneous interference and highlighting the true operating state of the device) while also taking into account the real-time nature of the diagnosis (avoiding diagnostic delays due to an excessive number of cycles). Implementers can set the size of the preset number according to the actual situation, and no limitation is imposed here.

[0030] Preferably, in one possible implementation of this embodiment, the method for obtaining the on / off logic analysis coefficients is described in [reference needed]. Figure 2 The flowchart illustrates a method for obtaining on / off logic analysis coefficients provided in this embodiment. The method includes the following steps: Step S201: Based on the number of cycles in which the bridge arm switching logic of each IGBT device complies with the preset number of PWM command signal cycles and the number of cycles in which the bridge arm switching logic violates the preset number of cycles, obtain the switching logic compliance level of each IGBT device.

[0031] Within a preset number of PWM command signal cycles, the more cycles a certain IGBT device has with compliant bridge arm on / off logic, the more normal the on / off logic of that IGBT device is. Conversely, the more cycles a certain IGBT device has with consecutive non-compliant bridge arm on / off logic, the less normal the on / off logic of that IGBT device is. Therefore, this embodiment obtains the compliance level of the on / off logic of each IGBT device based on the number of cycles with compliant bridge arm on / off logic and the number of cycles with consecutive non-compliant bridge arm on / off logic within a preset number of PWM command signal cycles. The higher the compliance level of the on / off logic, the more normal the bridge arm on / off of the corresponding IGBT device is, the higher its overall compliance, and the more stable its on / off state.

[0032] In one possible implementation of this embodiment, the method for obtaining the compliance level of the on / off logic is as follows: For any IGBT device, the ratio of the number of cycles in which the IGBT device's bridge arm on / off logic is compliant to a preset number (10 PWM command signal cycles in this embodiment) is used as the first compliance analysis value of the IGBT device. The larger the first compliance analysis value, the higher the overall compliance level of the IGBT device's bridge arm on / off logic within the statistical period, and the higher the proportion of cycles that comply with the operating rules. Further, the number of cycles in which the IGBT device's continuous bridge arm on / off logic is violated is obtained, and all are used as the first number. When the first number is larger, it indicates that the IGBT device has a persistent on / off logic anomaly, which is more likely a substantial problem caused by device performance degradation or driver failure. Rather than instantaneous electromagnetic interference, the failure risk is significantly higher than that of sporadic violations. To highlight the impact of continuous violation cycles on failure risk, this embodiment negatively correlates the ratio of the largest first quantity to the preset quantity as the compliance correction weight of the IGBT device. The larger the compliance correction weight, the shorter the longest continuous violation cycle of the IGBT device, indirectly indicating a lower failure risk caused by continuous violations. This embodiment negatively correlates the ratio of the largest first quantity to the preset quantity using 1 - the ratio of the largest first quantity to the preset quantity. In order to comprehensively characterize the bridge arm switching logic compliance status of the IGBT device, the product of the first compliance analysis value and the compliance correction weight of the IGBT device is used as the switching logic compliance degree of the IGBT device.

[0033] At this point, the compliance level of the on / off logic of each IGBT device is obtained.

[0034] Step S202: Based on the collector current change of each IGBT device within a preset number of PWM command signal cycles, obtain the current correlation correction coefficient for each IGBT device.

[0035] Within a preset number of PWM command signal cycles, a larger collector current and a faster rise rate for a given IGBT device indicate a stronger current surge, indirectly reflecting a deviation of the IGBT device's bridge arm switching state from normal operating requirements, thus increasing the risk of short circuits or overloads. Therefore, this embodiment obtains a current-related correction coefficient for each IGBT device based on its collector current changes within a preset number of PWM command signal cycles. A smaller current-related correction coefficient indicates a stronger current surge for the corresponding IGBT device, and a higher risk of short circuits or overloads caused by abnormal bridge arm switching logic.

[0036] In one possible implementation of this embodiment, the current-related correction coefficient is obtained as follows: For any IGBT device, the ratio of the collector current rise rate of the IGBT device to the preset rated current rise rate within each PWM command signal cycle is obtained, and both are used as the current surge degree. The greater the current surge degree, the more drastic the current change of the IGBT device within the corresponding command signal cycle, the stronger the instantaneous electrical stress it withstands, and the more likely it is to cause a short circuit fault due to abnormal switching logic. The preset rated current rise rate is the rated current rise rate clearly marked in the IGBT device manual, which is an objective benchmark for judging whether the current surge is abnormal. It should be noted that the method for obtaining the collector current rise rate of the IGBT device within each PWM command signal cycle is as follows: For any PWM command signal cycle, the collector current of the IGBT device within the PWM command signal cycle is fitted into a curve according to the time sequence, and the maximum and minimum points on the curve are obtained; for any maximum point, the difference between the collector current corresponding to the maximum point and its previous adjacent minimum point is taken as the first value; the time interval formed by the time corresponding to the maximum point and its previous adjacent minimum point is taken as the first time interval; the ratio of the first value to the first time interval is taken as the current rise rate corresponding to the maximum point; if the maximum point does not have a previous adjacent minimum point, the collector current at the beginning of the PWM command signal cycle is taken as the collector current corresponding to the previous adjacent minimum point; the current rise rate corresponding to all maximum points is obtained, and the largest current rise rate is taken as the collector current rise rate of the IGBT device within the PWM command signal cycle; the method of fitting the curve is a well-known technique and will not be described in detail here. To more accurately analyze the risk of continuous overload, the ratio of each collector current of the IGBT device to the preset rated current within a preset number of PWM command signal cycles is used as the current overload analysis value. The larger the current overload analysis value, the more the actual load of the IGBT device exceeds the rated range. Long-term operation in this state can easily lead to heat accumulation and overload failure. The preset rated current is the rated operating current marked in the IGBT device manual and is the core parameter for determining whether the load is overloaded. To comprehensively characterize the severity of current surges and the duration of load overload, and to achieve compatibility with on / off logic compliance, the product of the maximum current surge severity and the average current overload analysis value is negatively correlated and normalized, serving as the current-related correction coefficient for this IGBT device. This embodiment uses... The product of the maximum current surge and the mean current excess analysis value is negatively correlated and normalized, where x represents the product of the maximum current surge and the mean current excess analysis value; norm is a linear normalization function.

[0037] At this point, the current-related correction coefficient for each IGBT device is obtained.

[0038] Step S203: The product of the on / off logic compliance level and the current-related correction coefficient of each IGBT device is used as the on / off logic analysis coefficient of each IGBT device.

[0039] A higher compliance level of the on / off logic indicates a higher overall compliance level of the bridge arm on / off logic of the corresponding IGBT device, a shorter continuous violation cycle, a more stable on / off state, and a lower potential risk of short circuits and overload faults caused by abnormal on / off logic. A larger current-related correction coefficient indicates a weaker current surge intensity, a lower degree of continuous overcurrent exceeding the rated current, and a smaller amplification effect of current conditions on fault risk. In order to comprehensively characterize the compliance level of the bridge arm on / off logic and the fault risk of current conditions, and to achieve the organic integration of these two core characteristics, thereby accurately quantifying the comprehensive risk level of short circuits and overload faults caused by abnormal on / off logic for each IGBT device, this embodiment uses the product of the compliance level of the on / off logic of each IGBT device and the current-related correction coefficient as the on / off logic analysis coefficient of each IGBT device. This directly reflects the fault risk level of each IGBT device. A larger on / off logic analysis coefficient indicates a more compliant on / off logic, a lower current surge risk, and a safer operating state for the corresponding IGBT device. A smaller on / off logic analysis coefficient indicates a higher probability of short circuits or overload faults for the corresponding IGBT device.

[0040] At this point, the on / off logic analysis coefficients for each IGBT device are obtained.

[0041] Step S3: Based on the collector-emitter voltage fluctuation of each IGBT device within a preset number of PWM command signal cycles, as well as the magnitude of collector current, collector-emitter voltage and IGBT junction temperature, obtain the conduction efficiency coefficient of each IGBT device.

[0042] Specifically, the conduction performance of an IGBT device directly determines its energy conversion efficiency and thermal stability, serving as the core basis for judging open-circuit and overheating faults. From a fault mechanism perspective, within a preset number (10 in this embodiment) of PWM command signal cycles, under normal conduction conditions, the collector-emitter voltage drop should be stable within the rated range specified in the IGBT device manual (typically 1V-2V). Significant fluctuations in the collector-emitter voltage drop likely indicate two types of anomalies: intermittent open-circuit faults (such as intermittent conduction due to poor pin contact) and IGBT device performance aging caused by overheating (deterioration of conduction characteristics due to prolonged high temperatures). From the perspective of energy transfer, energy transfer efficiency is the core indicator for quantifying the conduction performance of an IGBT device, while conduction losses are the main source of heat generation. Higher losses result in a faster temperature rise rate, directly causing overheating faults. Furthermore, even intermittent open-circuit faults can cause current waveform distortion, indirectly increasing conduction losses.

[0043] The conduction loss of an IGBT device is determined by both the collector-emitter voltage drop and the collector operating current. Energy transfer efficiency is essentially the ratio of effectively transferred energy to total energy consumed. The closer the actual conduction loss is to the rated loss of the IGBT device, the less ineffective energy is consumed during conduction due to abnormal voltage drop or current exceeding limits. This indicates better conduction performance and a lower risk of overheating. Therefore, energy transfer efficiency can be quantified by the difference between rated and actual losses. It should be noted that the on-state voltage drop of an IGBT device changes with the IGBT junction temperature. As the junction temperature increases, the on-state voltage drop decreases linearly. This characteristic directly affects the accuracy of conduction loss calculations. Therefore, a correction term corresponding to the IGBT junction temperature needs to be introduced to compensate for the influence of temperature on conduction characteristics and ensure the accuracy of the quantified energy transfer efficiency results.

[0044] Therefore, this embodiment obtains the conduction efficiency coefficient of each IGBT device based on the collector-emitter voltage fluctuation of each IGBT device within a preset number of PWM command signal cycles, as well as the magnitude of collector current, collector-emitter voltage and IGBT junction temperature, thereby achieving accurate evaluation of the conduction state of the IGBT device.

[0045] Preferably, in one possible implementation of this embodiment, the method for obtaining the conduction efficiency coefficient is described in [reference needed]. Figure 3 The flowchart illustrates a method for obtaining the conduction efficiency coefficient provided in this embodiment. The method includes the following steps: Step S301: Obtain the voltage stability level.

[0046] When the collector-emitter voltage of an IGBT device is more stable, it indicates that the conduction state of the IGBT device is more stable, without intermittent open circuits or poor contact, and the voltage loss fluctuation during energy conversion is small, resulting in better conduction efficiency. Therefore, in this embodiment, for any IGBT device, the ratio of each collector-emitter voltage of the IGBT device to a preset rated voltage within a preset number of PWM command signal cycles is used as the voltage over-limit analysis value. The preset rated voltage is the rated on-state voltage drop (usually 1V-2V) indicated in the IGBT device manual. The closer the voltage over-limit analysis value is to 1, the closer the collector-emitter voltage of the IGBT device is to the rated value, the smaller the deviation of the on-state voltage drop, and the less obvious the voltage abnormality caused by overheating or open circuit. To quantify the overall fluctuation level of the collector-emitter voltage, the standard deviation of the voltage excess analysis value is negatively correlated and normalized to represent the voltage stability of the IGBT device. A higher voltage stability indicates smaller collector-emitter voltage fluctuations within the preset PWM command signal period, stronger consistency in the conduction state, and a lower risk of open-circuit and overheating faults caused by voltage anomalies. In this embodiment, the negative of the aforementioned standard deviation is used as the power of an exponential function with a base of the natural constant. The output of this exponential function is the result of negatively correlated and normalized standard deviations.

[0047] For better description, the following analysis will use this IGBT device as an example.

[0048] Step S302: Obtain the loss amount analysis value.

[0049] Since the conduction loss of the IGBT device is essentially the power loss in its on-state, it can be characterized by the product of voltage and current. Therefore, for any PWM command signal cycle, the product of the average collector-emitter voltage and the average collector current of the IGBT device within that PWM command signal cycle is taken as the actual loss of the IGBT device within that PWM command signal cycle. It should be noted that the average collector-emitter voltage and the average collector current of the IGBT device within that PWM command signal cycle are data from the on-state phase of that PWM command signal cycle. In order to comprehensively evaluate the conduction loss level of the IGBT device within a preset number of PWM command signal cycles (10 cycles in this embodiment) and eliminate the interference of instantaneous fluctuations within a single PWM command signal cycle on the loss evaluation, the average of the actual losses of the IGBT device within the preset number of PWM command signal cycles is taken as the overall actual loss of the IGBT device. To quantify the deviation of the overall actual loss from the rated loss, the difference between the overall actual loss and the preset rated loss is linearly normalized and used as the loss rating analysis value for the IGBT device. The preset rated loss is the rated conduction power loss specified in the IGBT device manual, and is the core benchmark for determining whether the IGBT device's loss exceeds the limit. A larger loss rating analysis value indicates that the overall actual loss of the IGBT device exceeds the preset rated loss, resulting in more ineffective energy loss during energy transfer, poorer conduction performance, and a higher risk of overheating due to accumulated losses over long-term operation, making it more prone to overheating failures. Conversely, a smaller loss rating analysis value indicates that the actual loss is closer to the rated loss, better conduction performance, and a lower risk of overheating.

[0050] Step S303: Obtain the junction temperature correction factor.

[0051] Considering that IGBT junction temperature affects the accuracy of on-state voltage drop and conduction loss calculations, a higher IGBT junction temperature indicates more severe internal heat accumulation and a greater linear decrease in on-state voltage drop. Ignoring the IGBT junction temperature would lead to distorted loss calculation results. Therefore, this embodiment uses the average IGBT junction temperature over a preset number of PWM command signal cycles, linearly normalized, as the junction temperature correction coefficient for the IGBT device. A larger junction temperature correction coefficient indicates that the average IGBT junction temperature is closer to its rated maximum junction temperature (typically 150℃-175℃). Higher thermal stress levels result in a more significant impact of the IGBT junction temperature on on-state voltage drop and losses, requiring a higher correction weight in subsequent conduction efficiency coefficient calculations. The rated maximum junction temperature is the junction temperature parameter specified in the IGBT device datasheet.

[0052] Step S304: Obtain energy transfer analysis values.

[0053] A higher loss rating indicates that the overall actual loss of the IGBT device exceeds the preset rated loss, resulting in more ineffective energy loss during energy transfer, poorer conduction efficiency, and a higher risk of overheating and failure during long-term operation. A higher junction temperature correction factor indicates that the average junction temperature of the IGBT device is closer to its rated maximum junction temperature, resulting in higher thermal stress levels and a more significant impact of junction temperature on conduction voltage drop and loss calculations. To comprehensively characterize the degree of actual loss exceeding the standard and the correction weight of junction temperature on loss, and to eliminate the bias caused by temperature in simple loss assessment, the product of the loss rating analysis value and the junction temperature correction factor is used as the energy transfer analysis value of the IGBT device. This accurately reflects the true energy transfer efficiency and failure risk level of the IGBT device under the dual effects of temperature-coupled losses. Because both the loss rating analysis value and the junction temperature correction factor are normalized data, their values ​​range from 0 to 1, and consequently, the energy transfer analysis value also ranges from 0 to 1. The higher the energy transfer analysis value, the more it indicates that the excessive loss of the IGBT device is further amplified by the high temperature. The lower the energy transfer efficiency, the more serious the deterioration of the conduction performance, and the higher the probability of overheating failure.

[0054] Step S305: Obtain the conduction efficiency coefficient.

[0055] A higher voltage stability indicates smaller collector-emitter voltage fluctuations in the IGBT device, stronger consistency in conduction state, absence of intermittent open circuits or poor contact, and less negative impact of voltage conditions on conduction efficiency. Conversely, a higher energy transfer analysis value indicates that excessive losses in the IGBT device are amplified by high temperatures, resulting in lower energy transfer efficiency, more severe conduction performance degradation, and a higher probability of overheating failures. To comprehensively characterize the combined impact of voltage stability characteristics and temperature-loss coupling risk on the conduction state of the IGBT device, and to organically integrate these two core indicators, the conduction efficiency level of the IGBT device is accurately quantified. The product of the negative correlation between the energy transfer analysis value and the voltage stability is then used as the conduction efficiency coefficient of the IGBT device. A higher conduction efficiency coefficient indicates smaller voltage fluctuations and lower temperature-loss coupling risk, more stable conduction state, higher energy transfer efficiency, and a lower probability of open circuits and overheating failures. In this embodiment, the normalized value range of the energy transfer analysis value is 0 to 1. The result of subtracting the energy transfer analysis value from 1 is taken as the negative correlation processing result of the energy transfer analysis value.

[0056] At this point, the conduction efficiency coefficient of each IGBT device is obtained.

[0057] Step S4: Based on the on / off logic analysis coefficients and conduction efficiency coefficients of each IGBT device, combined with the fault judgment threshold and constraints, obtain the fault type label for each IGBT; the fault type label includes short circuit, open circuit, overload, overheating and no fault.

[0058] Specifically, the on / off logic analysis coefficient ranges from 0 to 1. A smaller on / off logic analysis coefficient indicates poorer compliance of the IGBT's bridge arm on / off logic, greater current surge intensity, and a higher probability of short-circuit or overload faults. Conversely, a larger on / off logic analysis coefficient indicates more stable on / off states of the corresponding IGBT, with lower risks of short circuits and overloads. Similarly, the conduction efficiency coefficient also ranges from 0 to 1. A smaller conduction efficiency coefficient indicates greater voltage fluctuations in the corresponding IGBT, higher temperature-loss coupling risk, and a higher probability of open-circuit or overheating faults. Conversely, a larger conduction efficiency coefficient indicates better conduction states of the corresponding IGBT, with lower risks of open circuits and overheating. Therefore, this embodiment obtains a fault type label for each IGBT based on its on / off logic analysis coefficient and conduction efficiency coefficient, combined with fault determination thresholds and constraints. The fault type labels include short circuit, open circuit, overload, overheating, and no fault.

[0059] The fault determination thresholds include the on / off logic anomaly boundary threshold, the short-circuit-overload differentiation threshold, the conduction performance anomaly boundary threshold, the open-circuit-overheat differentiation threshold, and the junction temperature critical threshold. It should be noted that for any IGBT device, the on / off logic anomaly boundary threshold is the difference between the mean of the on / off logic analysis coefficients under normal operating conditions and three times their standard deviation. This embodiment uses the statistical 3σ principle to define the boundary between normal and abnormal operating conditions. The value of the mean of the on / off logic analysis coefficients minus three times their standard deviation can cover 99.73% of the fluctuations in the coefficients under normal operating conditions. Only when the on / off logic analysis coefficients are lower than the on / off logic anomaly boundary threshold is it determined to be an on / off logic anomaly, thus avoiding transient fluctuations. It can accurately capture sudden changes in the on / off logic analysis coefficients caused by short circuits and overload faults, preventing misjudgments caused by movement. It should be noted that if the difference between the mean value of the on / off logic analysis coefficients of the IGBT device and three times its standard deviation under normal operating conditions is less than 0, then 0 is the default threshold for the abnormal on / off logic of the IGBT device. The 3σ principle is well-known and will not be elaborated here. In this embodiment, the normal operating condition is set as 10 consecutive PWM command signal cycles without any faults. Implementers can set the normal operating condition according to the actual situation, and there is no limitation here. The short-circuit-overload distinction threshold for this IGBT device is the result of negatively correlating the ratio of the saturated collector-emitter voltage to the bus voltage under normal operating conditions. Specifically, 1 minus the difference between the saturated collector-emitter voltage and the bus voltage represents the result of negatively correlating the ratio. The saturated collector-emitter voltage and the bus voltage are explicitly stated in the IGBT device datasheet and can be used directly. It is known that the core characteristic of a short-circuit fault is a sudden drop in the collector-emitter voltage of the IGBT device to the saturation voltage drop, which is far lower than the bus voltage. In contrast, an overload fault only manifests as an excessive current, while the voltage remains within the normal range. By subtracting the difference between the saturated collector-emitter voltage and the bus voltage from 1, the short-circuit condition can be mapped to a high value, and the overload condition to a low value. This allows for accurate differentiation between the two types of faults through the short-circuit-overload distinction threshold. The abnormal conduction performance threshold of this IGBT device is the difference between the mean of the conduction performance coefficient of the IGBT device under normal operating conditions and three times its standard deviation. Consistent with the design logic of the abnormal conduction performance threshold, it filters out small changes in the conduction performance coefficient caused by voltage fluctuations and loss fluctuations under normal operating conditions based on the 3σ principle. Only when the conduction performance coefficient is lower than the abnormal conduction performance threshold is it judged as abnormal conduction performance, corresponding to the risk range of open circuit or overheating fault, ensuring the statistical significance of fault judgment. The open-circuit / overheating distinction threshold for this IGBT device was preset after fitting and calibration based on experimental data under different fault simulation conditions. The core characteristic of an open-circuit fault is a significant fluctuation in the collector-emitter voltage, which is quite obvious. The core characteristics of an overheating fault are excessive losses and an increase in IGBT junction temperature. The two types of faults have different impacts on the conduction efficiency coefficient. By collecting coefficient data from open-circuit and overheating fault simulation experiments and determining the optimal distinction threshold through fitting and calibration, the confusion between the two types of faults can be avoided, improving classification accuracy. Specifically, the larger the open-circuit / overheating distinction threshold, the more likely the IGBT device is to have an overheating fault; the smaller the open-circuit / overheating distinction threshold, the more likely the IGBT device is to have an open-circuit fault. This is because the degradation of the IGBT device's conduction efficiency by an open-circuit fault is far greater than that by an overheating fault. The critical junction temperature threshold of this IGBT device is the maximum IGBT junction temperature minus three times the standard deviation of the IGBT junction temperature under a specified number of normal operating conditions. The maximum IGBT junction temperature specified in the IGBT device datasheet is its limit for safe operation. Subtracting three times the standard deviation of the IGBT junction temperature under normal operating conditions allows for a reasonable temperature safety margin, avoiding false alarms triggered by normal temperature fluctuations. Simultaneously, the critical junction temperature threshold is directly related to the core mechanism of overheating faults. When the actual IGBT junction temperature approaches the critical threshold, it is determined to be a risk of overheating faults, achieving early warning of faults. In this embodiment, the specified number is set to 100, which meets the significance requirements for statistical analysis. Implementers can set the specified number according to actual conditions; no limitation is imposed here.

[0060] It should be noted that the short-circuit-overload differentiation threshold is less than the on / off logic anomaly threshold; the open-circuit-overheat differentiation threshold is less than the conduction performance anomaly threshold. The on / off logic anomaly threshold is the lower limit critical value of the on / off logic analysis coefficient under normal operating conditions; the conduction performance anomaly threshold is the lower limit critical value of the conduction performance coefficient under normal operating conditions.

[0061] Preferably, in one feasible embodiment, the fault type label is obtained as follows: for any IGBT device, when the on / off logic analysis coefficient of the IGBT device is less than the short-circuit-overload distinction threshold and the conduction efficiency coefficient is greater than or equal to the conduction efficiency abnormal boundary threshold, the fault type label of the IGBT device is short circuit; because during a short circuit, the bridge arm on / off logic malfunctions, resulting in a mismatch with the actual voltage and current, but the IGBT device itself is not damaged and can still conduct normally; When the on / off logic analysis coefficient of the IGBT device is greater than or equal to the short-circuit-overload distinction threshold and less than the on / off logic abnormal boundary threshold, and the conduction efficiency coefficient is greater than or equal to the conduction efficiency abnormal boundary threshold, the fault type label of the IGBT device is overload; because overload is a cumulative fault in which the current exceeds the rated range of the IGBT device for a long time, and the IGBT device itself can conduct normally, and the current waveform is distorted only due to the overload. When the on / off logic analysis coefficient of the IGBT device is greater than or equal to the on / off logic abnormality threshold, and the conduction efficiency coefficient is less than the open circuit-overheating distinction threshold, the fault type label of the IGBT device is open circuit. This is because the open circuit fault is caused by damage to the conduction capability of the IGBT device itself, resulting in the inability to form an effective current path, which is a device fault. Although the PWM command (the signal that controls the IGBT to turn on) is output normally, the actual conduction time of the IGBT device is much lower than the command requirement due to its own damage, and the conduction matching degree is reduced.

[0062] When the on / off logic analysis coefficient of the IGBT device is greater than or equal to the on / off logic abnormality threshold, the conduction efficiency coefficient is greater than or equal to the open circuit-overheating distinction threshold but less than the conduction efficiency abnormality threshold, and the IGBT junction temperature is greater than the junction temperature critical threshold, the fault type of the IGBT device is labeled as overheating. Overheating is a cumulative fault caused by temperature. The conduction and switching losses generated during the operation of the IGBT device, in scenarios with insufficient heat dissipation, prevent timely heat dissipation, causing the IGBT junction temperature to continuously rise and eventually exceed the safety threshold. Due to the increased IGBT junction temperature, the conduction capability of the IGBT device decreases slightly, but does not reach the severe failure level of an open circuit. Meanwhile, the PWM command is normal, ruling out faults caused by logic abnormalities or overload. The fault is clearly due to the accumulation of internal losses; heat continues to accumulate and gradually deviates from the normal fluctuation range. When the on / off logic analysis coefficient of the IGBT device is greater than or equal to the on / off logic abnormality threshold, and the conduction performance coefficient is greater than or equal to the conduction performance abnormality threshold, the fault type label of the IGBT device is no fault.

[0063] At this point, the fault type label for each IGBT device is obtained.

[0064] Step S5: Construct a fault sample set based on fault type labels, train a fault type recognition model, and obtain a trained fault type recognition model; the fault type recognition model is used to output the fault type of each IGBT device in the three-phase full-bridge inverter.

[0065] Specifically, in order to achieve efficient and accurate identification of IGBT device fault types in a three-phase full-bridge inverter, this embodiment constructs a fault sample set based on the fault type labels of IGBT devices from historical analysis, as well as the corresponding on / off logic analysis coefficients and conduction efficiency coefficients, to train a fault type identification model. The fault type identification model can directly output the fault type of each IGBT device in the three-phase full-bridge inverter, including short circuit, open circuit, overload, overheating, and no fault.

[0066] The construction of the fault sample set is based on the core principles of data integrity, label accuracy, and feature effectiveness, ensuring that the sample data can comprehensively cover the typical operating conditions and fault modes of IGBT devices, providing reliable supervision data for model training. In this embodiment, the sample data comes from two parts: laboratory fault simulation experiments of three-phase full-bridge inverters and field operation history data. The laboratory fault simulation experiments involve building a three-phase full-bridge inverter test platform consistent with actual applications, artificially simulating different fault types (such as short-circuit faults through bridge arm shoot-through, open-circuit faults through device pin open-circuit, overload faults through applying excessive load, and overheating faults through blocking heat dissipation channels), and collecting full-cycle operation data of each IGBT device before, during, and after the fault occurs. The field operation history data involves collecting the inverter's operation data in actual industrial scenarios, filtering out clearly labeled fault data and normal operation data to supplement the limitations of the laboratory simulation data. Each sample's feature vector consists of the state characteristics and topological correlation characteristics of the IGBT device. The state characteristics are the on / off logic analysis coefficients, conduction efficiency coefficients, and real-time IGBT junction temperature of each IGBT device. The topological correlation characteristics are the average state characteristics of adjacent devices (such as upper and lower tubes of the same bridge arm or parallel devices on the same layer) that are physically connected to the target IGBT device, reflecting the impact of topological correlation between devices on fault propagation. Finally, the feature vector of each IGBT device sample has a 5-dimensional dimension, ensuring that the feature information includes both the device's own state and the correlation characteristics of the circuit topology. Each sample corresponds to only one fault type label to avoid ambiguity caused by multiple labels for the same sample; the labels of laboratory simulation data are directly labeled based on the fault simulation method, while the labels of field historical data are cross-validated with operation and maintenance records and fault judgment rules to ensure that the labels are correct; to address the problem of uneven fault sample quantity (e.g., few short-circuit fault samples and many fault-free samples), a combination of oversampling (data augmentation for scarce fault samples) and undersampling (random screening for fault-free samples) is used to balance the proportion of various fault samples and avoid bias in model training; to eliminate the impact of the difference in the dimensions of different feature parameters on model training, all feature parameters in the sample set are normalized to map the feature values ​​to the [0,1] interval. In this embodiment, the feature parameters are normalized using a linear normalization method, and the normalization method is not limited here; The fault type identification model in this embodiment adopts a graph neural network (GNN) architecture, preferably a graph convolutional network (GCN) as the core model. Its advantage lies in its ability to effectively utilize the topological association information of IGBT devices in a three-phase full-bridge inverter, overcoming the limitation of traditional machine learning models that ignore the physical connections between devices. The overall model architecture consists of an input layer, a graph convolutional layer, and an output layer. The core function of the input layer is to adapt the graph structure data to the model. The input data includes a node feature matrix and an adjacency matrix; the dimension of the node feature matrix is... N represents the total number of IGBT devices in the three-phase full-bridge inverter, and F represents the characteristic dimensions of a single node (5 dimensions, namely, on / off logic analysis coefficients, conduction efficiency coefficients, real-time IGBT junction temperature, and the average on / off logic analysis coefficients and average conduction efficiency coefficients of the corresponding IGBT device and its adjacent IGBT devices); the adjacency matrix has the following dimensions: The adjacency matrix is ​​used to characterize the physical topological relationships between IGBT devices. If two IGBT devices are physically connected in series or parallel, the corresponding element in the adjacency matrix is ​​1; otherwise, it is 0. The adjacency matrix is ​​also normalized to avoid numerical bias during feature aggregation. The graph convolutional layer is the core of the model. Typically, 2-3 graph convolutional layers are used, but this is not a limitation. Their function is to aggregate the features of the target node's neighboring nodes through multiple layers of features, updating the target node's feature representation and achieving deep fusion of local topological information and the node's own state. Through multi-layer graph convolution operations, the feature representation of each IGBT node will gradually fuse with the state information of adjacent devices, effectively capturing fault propagation features caused by topological relationships. The output layer consists of a fully connected layer and a Softmax classifier, mapping node features to fault type probabilities. Furthermore, the training objective of the fault type identification model is to minimize the difference between the predicted fault type and the true label. Therefore, this embodiment uses the cross-entropy loss function as the optimization objective function. The graph convolutional network and the cross-entropy loss function are well-known and will not be described further.

[0067] At this point, a well-trained fault type recognition model has been obtained.

[0068] The trained fault type identification model is deployed to the fault diagnosis system of a three-phase full-bridge inverter. The specific application process is as follows: For the inverter circuit running in real time, the feature parameters of each IGBT device are extracted according to the sample set construction rules to construct a node feature matrix; an adjacency matrix is ​​generated according to the inverter topology, which together with the node feature matrix forms the graph structure input data; the input data is input into the trained fault type identification model, and the model outputs the fault type probability distribution of each IGBT node; the category with the highest probability value is taken as the final fault type of the IGBT device, thereby realizing the fault type identification of all IGBT devices in the inverter.

[0069] For IGBT device fault tags output by the fault type identification model, a graded handling strategy should be adopted based on the different dangers and causes of the faults to achieve rapid response and accurate troubleshooting, avoiding equipment damage or safety risks caused by the escalation of the fault. If the model outputs a short circuit fault tag, since a short circuit fault can cause a sudden large current surge, it can easily lead to serious consequences such as bus voltage collapse and device burnout. The shutdown protection mechanism must be triggered immediately to cut off the inverter main circuit power supply. The key points of fault troubleshooting include three aspects: First, use a multimeter or insulation tester to check whether the IGBT chip is broken down and check the conduction status between the collector and emitter; second, check whether there are short circuit points in the bridge arm circuit, focusing on checking whether the bus capacitor, power cable and terminal have insulation damage, metal bridging and other problems; third, check whether the IGBT drive signal is abnormal. Observe the PWM waveform output by the driver board with an oscilloscope to confirm whether the drive voltage amplitude, pulse width and timing meet the rated standards, and check whether the driver chip and isolation optocoupler are damaged. If the model outputs an open circuit fault label, the machine must be stopped and the power supply disconnected to avoid three-phase output imbalance caused by a single-phase open circuit, which could lead to abnormal vibration or damage to loads such as motors. The troubleshooting steps are as follows: First, check the pin connections of the corresponding IGBT device to confirm whether there are any issues such as poor soldering, detachment, or oxidation / corrosion. Second, check the drive circuit to check whether the drive power supply is normal, whether the drive resistor is burnt out, and whether the drive cable is broken. Finally, perform a single-unit test on the IGBT chip. If an internal open circuit is confirmed, replace it with an IGBT device of the same model. After replacement, perform a continuity test to ensure that the circuit connection is normal. If the model outputs an overload fault label, immediate shutdown is not necessary. A temporary load reduction operation can be performed first, gradually decreasing the inverter's output power to reduce the current load on the IGBT devices and prevent thermal aging caused by prolonged overload. Troubleshooting should include: 1) Checking for abnormalities at the load end, such as mechanical equipment jamming or stalling, or short circuits or poor connections in the load cables; 2) Verifying the transmission system, confirming that components such as the reducer and couplings are functioning correctly to avoid excessive load current due to excessive mechanical resistance; 3) Verifying the inverter's current protection parameter settings to ensure the overload protection threshold matches the actual load requirements, and recalibrating the parameters if necessary. After resolving the load abnormality, gradually restore rated power operation and continuously monitor the IGBT current waveform and temperature changes to ensure stable operation. If the model outputs an overheating fault label, the cooling system should be checked first to prevent the junction temperature from continuously rising, which could lead to increased on-state voltage drop, reduced lifespan, or even thermal breakdown of the IGBT devices. Specific troubleshooting measures include: first, cleaning accumulated dust from the cooling system and checking if the heatsink surface is covered with dust or oil, affecting heat dissipation efficiency; second, checking if the cooling fan is operating normally, checking for fan motor damage and whether the speed is up to standard, and replacing faulty fans if necessary; for liquid cooling systems, checking the coolant level, flow rate, and whether the cooling pipes are blocked to ensure smooth heat dissipation circulation. Simultaneously, an infrared thermometer can be used to monitor the surface temperature of the IGBT module to confirm whether the temperature returns to the rated range after the cooling system is restored to normal. If the model outputs a fault label of no fault, it indicates that the IGBT device's switching logic and conduction performance are in normal working condition, requiring no additional intervention. Only routine monitoring is needed, and the operating parameters of the IGBT (current, voltage, junction temperature) should be collected periodically to record the equipment's operating status and provide data support for subsequent preventive maintenance.

[0070] In summary, this embodiment acquires the PWM command signal, IGBT device electrical signals, and temperature signals in a three-phase full-bridge inverter. Based on the IGBT device's arm switching logic behavior, electrical signals, and temperature signals within the PWM command signal cycle, it acquires the IGBT device's switching logic analysis coefficients and conduction efficiency coefficients. Combining fault judgment thresholds and constraints, it acquires IGBT fault type labels to construct a fault sample set, trains a fault type recognition model, and obtains a well-trained fault type recognition model. This invention effectively improves the accuracy of fault type label recognition by accurately acquiring the switching logic analysis coefficients and conduction efficiency coefficients, which is beneficial for real-time and accurate identification of IGBT device fault types, timely corresponding processing, and effectively reducing fault losses in three-phase full-bridge inverters.

[0071] Example 2: This invention also proposes a fault diagnosis system for a three-phase full-bridge inverter. Please refer to [link / reference]. Figure 4 The diagram illustrates a structural diagram of a three-phase full-bridge inverter fault diagnosis system provided by an embodiment of the present invention. The system includes: a data acquisition module 10, a continuity logic analysis coefficient acquisition module 20, a conduction efficiency coefficient acquisition module 30, a fault type label acquisition module 40, and a data processing module 50.

[0072] The data acquisition module 10 is used to acquire the PWM command signal, the electrical signal and temperature signal of each IGBT device at each moment in the three-phase full-bridge inverter; the electrical signal includes the collector current and the collector-emitter voltage; the temperature signal includes the IGBT junction temperature.

[0073] The on / off logic analysis coefficient acquisition module 20 is used to acquire the on / off logic analysis coefficients of each IGBT device based on the on / off logic performance of the bridge arm and the change of collector current of each IGBT device within a preset number of PWM command signal cycles.

[0074] The conduction efficiency coefficient acquisition module 30 is used to acquire the conduction efficiency coefficient of each IGBT device based on the collector-emitter voltage fluctuation of each IGBT device within a preset number of PWM command signal cycles, as well as the magnitude of collector current, collector-emitter voltage and IGBT junction temperature.

[0075] The fault type label acquisition module 40 is used to acquire the fault type label of each IGBT based on the on / off logic analysis coefficient and conduction efficiency coefficient of each IGBT device, combined with the fault judgment threshold and constraint conditions; the fault type label includes short circuit, open circuit, overload, overheating and no fault.

[0076] The data processing module 50 is used to construct a fault sample set based on fault type labels, train a fault type identification model, and obtain a trained fault type identification model; the fault type identification model is used to output the fault type of each IGBT device in the three-phase full-bridge inverter.

[0077] It should be noted that the system provided in the above embodiments is only an example of the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the computer equipment can be divided into different functional modules to complete all or part of the functions described above. In addition, the three-phase full-bridge inverter fault diagnosis system and the three-phase full-bridge inverter fault diagnosis method embodiment provided in the above embodiments belong to the same concept, and their specific implementation process is detailed in the method embodiment, which will not be repeated here.

[0078] Example 3: This invention also proposes a fault diagnosis device for a three-phase full-bridge inverter. The device includes a memory and a processor. The memory stores executable program code, and the processor calls and executes the executable program code to perform a three-phase full-bridge inverter fault diagnosis method provided in the embodiments of this application. Specifically, the device may be a chip, component, or module. The chip may include a connected processor and memory; the memory stores instructions, and when the processor calls and executes the instructions, the chip can perform the three-phase full-bridge inverter fault diagnosis method provided in the above embodiments.

[0079] Furthermore, this application also protects a computer device; please refer to [link to relevant documentation]. Figure 5The computer device includes a memory 401, a processor 402, and a computer program 403 stored in the memory 401 and running on the processor 402. When the processor 402 executes the computer program 403, the computer device can perform any of the three-phase full-bridge inverter fault diagnosis methods described above.

[0080] Example 4: This embodiment also provides a computer-readable storage medium storing computer program code. When the computer program code is run on a computer, the computer executes the above-described related method steps to implement the three-phase full-bridge inverter fault diagnosis method provided in the above embodiment.

[0081] Example 5: This embodiment also provides a computer program product. When the computer program product is run on a computer, it causes the computer to perform the above-mentioned related steps to realize the three-phase full-bridge inverter fault diagnosis method provided in the above embodiment.

[0082] In this embodiment, the device, computer-readable storage medium, computer program product, or chip are all used to execute the corresponding methods provided above. Therefore, the beneficial effects they can achieve can be referred to the beneficial effects in the corresponding methods provided above, and will not be repeated here.

[0083] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0084] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A fault diagnosis method for a three-phase full-bridge inverter, characterized in that, The method includes the following steps: The system acquires the PWM command signal, the electrical signal and temperature signal of each IGBT device at each moment in the three-phase full-bridge inverter; the electrical signal includes the collector current and collector-emitter voltage; the temperature signal includes the IGBT junction temperature. Based on the bridge arm switching logic performance and collector current changes of each IGBT device within a preset number of PWM command signal cycles, the switching logic analysis coefficients of each IGBT device are obtained. Based on the collector-emitter voltage fluctuation of each IGBT device within a preset number of PWM command signal cycles, as well as the magnitude of collector current, collector-emitter voltage and IGBT junction temperature, the conduction efficiency coefficient of each IGBT device is obtained. Based on the on / off logic analysis coefficients and conduction efficiency coefficients of each IGBT device, combined with fault judgment thresholds and constraints, a fault type label for each IGBT is obtained; the fault type label includes short circuit, open circuit, overload, overheating, and no fault. A fault sample set is constructed based on fault type labels, and a fault type identification model is trained to obtain a trained fault type identification model; the fault type identification model is used to output the fault type of each IGBT device in the three-phase full-bridge inverter.

2. The fault diagnosis method for a three-phase full-bridge inverter as described in claim 1, characterized in that, The method for obtaining the on / off logic analysis coefficients is as follows: The compliance level of the on / off logic of each IGBT device is obtained based on the number of cycles in which the bridge arm on / off logic of each IGBT device complies with the preset number of PWM command signal cycles and the number of cycles in which the bridge arm on / off logic of each device violates the preset number of PWM command signal cycles. Based on the collector current change of each IGBT device within a preset number of PWM command signal cycles, obtain the current correlation correction coefficient for each IGBT device. The product of the on / off logic compliance level and the current-related correction coefficient for each IGBT device is used as the on / off logic analysis coefficient for each IGBT device.

3. The fault diagnosis method for a three-phase full-bridge inverter as described in claim 2, characterized in that, The method for obtaining the compliance level of the on / off logic is as follows: For any IGBT device, the ratio of the number of cycles in which the bridge arm switching logic of the IGBT device complies with the predetermined number is used as the first compliance analysis value of the IGBT device. The number of cycles in which the continuous bridge arm switching logic of the IGBT device violates the rule is used as the first quantity. The result of negatively correlating the ratio of the largest first quantity to the preset quantity is used as the compliance correction weight for the IGBT device. The product of the first compliance analysis value and the compliance correction weight of the IGBT device is taken as the compliance level of the IGBT device's on / off logic.

4. The fault diagnosis method for a three-phase full-bridge inverter as described in claim 2, characterized in that, The method for obtaining the current-related correction coefficient is as follows: For any IGBT device, the ratio of the collector current rise rate to the preset rated current rise rate within each PWM command signal cycle is used as the current surge level. The ratio of each collector current of the IGBT device to the preset rated current within a preset number of PWM command signal cycles is used as the current over-current analysis value. The product of the maximum current surge and the mean current excess analysis value is negatively correlated and normalized, and the result is used as the current correlation correction coefficient for this IGBT device.

5. The fault diagnosis method for a three-phase full-bridge inverter as described in claim 1, characterized in that, The method for obtaining the conduction efficiency coefficient is as follows: For any IGBT device, the ratio of each collector-emitter voltage of the IGBT device to the preset rated voltage within a preset number of PWM command signal cycles is used as the voltage over-limit analysis value. The standard deviation of the voltage excess analysis value is negatively correlated and normalized to obtain the voltage stability of the IGBT device. For any PWM command signal period, the product of the average collector-emitter voltage and the average collector current of the IGBT device during that PWM command signal period is taken as the actual loss of the IGBT device during that PWM command signal period. The average actual loss of the IGBT device within a preset number of PWM command signal cycles is taken as the overall actual loss of the IGBT device. The result of normalizing the difference between the overall actual loss and the preset rated loss is used as the loss rating analysis value of the IGBT device. The result of normalizing the average IGBT junction temperature of the IGBT device within a preset number of PWM command signal cycles is used as the junction temperature correction coefficient of the IGBT device. The product of the loss rating analysis value of the IGBT device and the junction temperature correction factor is used as the energy transfer analysis value of the IGBT device. The product of the negative correlation result of the energy transfer analysis value of the IGBT device and the voltage stability is used as the conduction efficiency coefficient of the IGBT device.

6. The fault diagnosis method for a three-phase full-bridge inverter as described in claim 1, characterized in that, The fault determination thresholds include the on / off logic abnormality boundary threshold, the short circuit-overload distinction threshold, the conduction performance abnormality boundary threshold, the open circuit-overheat distinction threshold, and the junction temperature critical threshold. For any IGBT device, the abnormal threshold of the on / off logic of the IGBT device is the difference between the mean of the on / off logic analysis coefficients of the IGBT device under normal operating conditions and three times its standard deviation. The short-circuit-overload distinction threshold of this IGBT device is the result of negatively correlated with the ratio of the saturated collector-emitter voltage of this IGBT device to the bus voltage under normal operating conditions. The abnormal threshold for the conduction performance of this IGBT device is the difference between the mean of the conduction performance coefficient of the IGBT device under normal operating conditions and three times its standard deviation. The open-circuit / overheating distinction threshold of this IGBT device is preset based on experimental data; The critical junction temperature threshold of this IGBT device is the maximum IGBT junction temperature of the IGBT device minus three times the standard deviation of the IGBT junction temperature under a specified number of normal operating conditions.

7. A fault diagnosis method for a three-phase full-bridge inverter as described in claim 6, characterized in that, The method for obtaining the fault type label is as follows: For any IGBT device, when the on / off logic analysis coefficient of the IGBT device is less than the short-circuit-overload distinction threshold and the conduction efficiency coefficient is greater than or equal to the conduction efficiency abnormal boundary threshold, the fault type label of the IGBT device is short circuit. When the on / off logic analysis coefficient of the IGBT device is greater than or equal to the short-circuit-overload distinction threshold and less than the on / off logic abnormal boundary threshold, and the conduction performance coefficient is greater than or equal to the conduction performance abnormal boundary threshold, the fault type label of the IGBT device is overload. When the on / off logic analysis coefficient of the IGBT device is greater than or equal to the on / off logic abnormality threshold, and the conduction efficiency coefficient is less than the open circuit-overheating distinction threshold, the fault type label of the IGBT device is open circuit. When the on / off logic analysis coefficient of the IGBT device is greater than or equal to the on / off logic abnormality threshold, the conduction performance coefficient is greater than or equal to the open circuit-overheating distinction threshold and less than the conduction performance abnormality threshold, and the IGBT junction temperature is greater than the junction temperature critical threshold, the fault type label of the IGBT device is overheating. When the on / off logic analysis coefficient of the IGBT device is greater than or equal to the on / off logic abnormality threshold, and the conduction performance coefficient is greater than or equal to the conduction performance abnormality threshold, the fault type label of the IGBT device is no fault.

8. The fault diagnosis method for a three-phase full-bridge inverter as described in claim 1, characterized in that, The fault type identification model is a graph neural network model; the fault sample set uses each IGBT device as an independent node and the physical topological relationship between devices as edges to construct a graph structure sample. The feature vector of each node includes the on / off logic analysis coefficient and the conduction efficiency coefficient, and the node label is the fault type label.

9. A fault diagnosis system for a three-phase full-bridge inverter, characterized in that, The system includes: The data acquisition module is used to acquire the PWM command signal, the electrical signal and temperature signal of each IGBT device at each moment in the three-phase full-bridge inverter; the electrical signal includes the collector current and collector-emitter voltage; the temperature signal includes the IGBT junction temperature. The on / off logic analysis coefficient acquisition module is used to acquire the on / off logic analysis coefficients of each IGBT device based on the on / off logic performance of the bridge arm and the change of collector current of each IGBT device within a preset number of PWM command signal cycles. The conduction efficiency coefficient acquisition module is used to obtain the conduction efficiency coefficient of each IGBT device based on the collector-emitter voltage fluctuation of each IGBT device within a preset number of PWM command signal cycles, as well as the magnitude of collector current, collector-emitter voltage and IGBT junction temperature. The fault type label acquisition module is used to acquire the fault type label of each IGBT based on the on / off logic analysis coefficient and conduction efficiency coefficient of each IGBT device, combined with the fault judgment threshold and constraints; the fault type label includes short circuit, open circuit, overload, overheat and no fault; The data processing module is used to construct a fault sample set based on fault type labels, train a fault type identification model, and obtain a trained fault type identification model; the fault type identification model is used to output the fault type of each IGBT device in the three-phase full-bridge inverter.

10. A fault diagnosis device for a three-phase full-bridge inverter, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the fault diagnosis method for a three-phase full-bridge inverter as described in any one of claims 1-8.

Images