Rail transit frequency converter silicon carbide module thermal fault cooperative monitoring method and system

Abstract

The present application relates to the technical field of rail transit electrical equipment monitoring, and discloses a method and system for monitoring thermal faults of silicon carbide modules in a rail transit frequency converter. The method collects multi-dimensional parameters such as junction temperature, leakage current, threshold voltage and thermal stress, constructs a loss model based on the FOC algorithm to calculate the junction temperature, and forms a fault precursor identification logic by correlating the three-dimensional parameters. The threshold is dynamically adjusted in combination with the working condition and the service life, and a SIL2-level safety closed loop is triggered to grade the early warning. The LSTM model is used to predict the fault risk, and a closed-loop interaction of operation and maintenance instructions is realized. The system includes four modules: thermal monitoring, collaborative analysis, graded early warning and data interaction, and a large number of existing components are reused. The present application solves the problems of multi-physical field false reporting, poor threshold adaptation and weak safety operation and maintenance, reduces the false reporting rate to 0.3%, the false alarm rate to 2%, and shortens the operation and maintenance time by 75%, and is suitable for high-speed rail and subway frequency converter scenarios.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of rail transit electrical equipment monitoring, and in particular to a method and system for monitoring thermal faults of silicon carbide modules in rail transit frequency converters. BACKGROUND

[0002] In the field of rail transit technology, the core function of frequency converter products such as high-speed rail compressor frequency converters, subway EC motor drives, and urban rail air conditioner frequency converters is to realize the conversion of electrical energy and the control of motor drives. This core function relies on power conversion units, among which silicon carbide (SiC) modules are the core components of power conversion units, replacing traditional silicon-based (IGBT) devices, and taking on key tasks such as rectification, inversion, and electrical energy regulation.

[0003] A frequency converter includes a control unit, a power conversion unit, a cooling unit, and a communication unit. The control unit adjusts the switching state of the SiC module in the power conversion unit by outputting a PWM signal, achieving precise control of current and voltage. The loss generated by the SiC module during operation is dissipated through the cooling unit, which usually uses forced air cooling or liquid cooling for cooling. The operating state of the SiC module (such as junction temperature and leakage current) is collected by the PHM monitoring and maintenance board, and then uploaded to the frequency converter control unit through the MVB / TRDP network card, forming a closed loop of control and monitoring.

[0004] Due to the inherent characteristics of silicon carbide material and the complex application scenarios of rail transit, the control of thermal faults of silicon carbide modules mainly has the following problems: (1) the monitoring parameter dimension is single, and does not cover the early failure of multi-physical field coupling and package layer cracking, resulting in a high false negative rate; (2) the existing monitoring threshold is fixed, and is not dynamically adjusted in combination with the working condition fluctuations such as high-speed rail climbing and subway starting and stopping; (3) the linkage between silicon carbide module monitoring and safety operation and maintenance is weak, and cannot provide effective support for safety control and fault tracing.

[0005] Therefore, how to realize safe and reliable monitoring of thermal faults of silicon carbide modules in rail transit frequency converters has become a technical problem to be solved. SUMMARY

[0006] In view of the defects of the prior art, the purpose of the present application is to provide a method and system for monitoring thermal faults of silicon carbide modules in rail transit frequency converters, which solves the problem of false negatives in multi-physical field coupling failure, adapts to the conditions of rail transit load fluctuations and device aging, and improves the safety level and operation and maintenance efficiency.

[0007] To achieve the above-mentioned purpose, the present application provides the following technical solutions:

[0008] According to one aspect of the present application, a method for monitoring thermal faults of silicon carbide modules in rail transit frequency converters is provided, comprising the following steps:

[0009] S1: collecting multi-dimensional parameters of the silicon carbide module, the parameters including real-time junction temperature, leakage current, threshold voltage and thermal stress data;

[0010] S2: constructing a silicon carbide module loss model by using a magnetic field oriented control algorithm , wherein I out is an inverter output current, V dc is a DC bus voltage, f sw is a silicon carbide module switching frequency;

[0011] S3: performing multi-physical field collaborative analysis by using the loss model, calculating real-time junction temperature T j in combination with a device transient thermal impedance curve of the silicon carbide module, and correlating the junction temperature, the leakage current, the threshold voltage and the thermal stress data to construct a three-dimensional thermal fault precursor identification logic;

[0012] S4: establishing a dynamic threshold grading early warning, setting an initial thermal early warning threshold and a thermal fault threshold, calculating a working condition compensation factor K1 based on I out and f sw parameters collected by the magnetic field oriented control algorithm, calculating a service life correction factor K2 in combination with a module cumulative running time recorded in operation and maintenance, and dynamically adjusting the thresholds through the parameters K1 and K2;

[0013] S5: when the real-time junction temperature T j reaches the thermal early warning threshold, realizing local warning through a buzzer and uploading a warning signal to a rail transit TCMS system through an MVB / TRDP network card.

[0014] According to one embodiment of the present application, the method further comprises step S6: recording multi-dimensional parameter collection data, collaborative analysis results and early warning trigger nodes in a log, and uploading monitoring data to an operation and maintenance system through an MVB / TRDP network card;

[0015] According to one embodiment of the present application, the method further comprises step S7: the operation and maintenance system adopts a preset fault prediction model to perform fault risk prediction, and forms an operation and maintenance instruction according to a prediction result.

[0016] According to one embodiment of the present application, input parameters I out , V dc and f sw of the loss model are collected through a functional test interface of the inverter, the junction temperature calculation result is subjected to secondary operation cross verification by a control unit of the inverter, and the junction temperature is valid when a deviation of the secondary operation is less than an error threshold.

[0017] According to one embodiment of the present application, a calculation formula of the working condition compensation factor K1 is Wherein, Irated is the inverter rated current, frated is the silicon carbide module rated switching frequency.

[0018] The calculation formula of the life correction factor K2 is Wherein, Toperate is the module cumulative running months, taken from the running time record in the monitoring operation data.

[0019] According to an embodiment of the present application, the fault prediction model is realized by an LSTM model trained by embedding the historical operation data of the frequency converter; the input parameters include the peak temperature sequence of the nearest 72 hours, the cumulative switching times of the module; the output result is the fault risk probability in the future 15 days; when the risk is greater than or equal to 60%, an operation and maintenance reminder is pushed, the operation and maintenance system issues a threshold adjustment and calibration trigger instruction, which is transmitted to the monitoring system through the safety channel of the MVB / TRDP network card for execution, and the execution result is fed back to the operation and maintenance system to form a closed loop.

[0020] According to an embodiment of the present application, the threshold calibration, early warning transmission and operation and maintenance instruction execution are all verified by SIL2 level safety check logic: the hot monitoring data transmission is additionally provided with a CRC-32 check code; the main MCU and the standby MCU of the frequency converter synchronously calculate the key parameters when responding to the fault, and only when the deviation meets the preset threshold value can the operation be allowed to be executed, which meets the IEC61508 standard.

[0021] According to an embodiment of the present application, the trigger conditions of the three-dimensional thermal fault precursor identification logic include:

[0022] The junction temperature is greater than or equal to 195℃×K2, and the threshold voltage offset is greater than or equal to 0.5V , and the thermal stress is greater than or equal to 350MPa.

[0023] Wherein, the 350MPa thermal stress threshold value is calibrated through a-40℃~85℃ thermal cycle test, the test period is 1000 cycles, covering the high temperature and vibration working conditions of rail transit.

[0024] According to an embodiment of the present application, in the step S1, the data acquisition adopts a low inductance composite busbar with a nested copper foil shielding layer; in the step S3, a double-stage anti-interference algorithm of interference mode matching and Kalman filtering is used for collaborative analysis, first the waveform similarity is compared to identify the interference types such as bow net pulse and motor harmonic, and then the dedicated filtering parameters are called to improve the junction temperature monitoring accuracy from ±1℃ to ±0.5℃.

[0025] According to one embodiment of the present invention, the log uses the Ulog log component to support secure partitioned storage. The log content includes the collection timestamp, the calculated junction temperature value, the early warning triggering condition, the operation and maintenance instruction content and the execution result. After the log is downloaded, it can be parsed by the host computer software. This software is from the same source as the inverter function test software, supports thermal fault tracing and development trend analysis, and the log format meets the audit requirements of SIL2 certification.

[0026] According to another aspect of the present invention, a collaborative monitoring system for thermal faults of silicon carbide modules in rail transit frequency converters is provided to implement the method, comprising:

[0027] The thermal monitoring module 201 includes an NTC thermistor, an infrared temperature measurement chip, a MEMS thermal stress sensor, a PHM monitoring and maintenance board, and an online insulation detection system; it is used to synchronously collect junction temperature, leakage current, threshold voltage, and thermal stress data of the silicon carbide module and monitor the insulation performance of the acquisition circuit.

[0028] The Collaborative Analysis Module 202 is built on an ARM Cortex series MCU and RTOS software platform. It integrates the SiC module loss model derived from the FOC control algorithm, dual-core cross-validation logic, three-dimensional thermal fault precursor identification logic, and "interference matching-filtering" anti-interference algorithm. It is used to calculate real-time junction temperature, verify data accuracy, and identify thermal fault precursors.

[0029] The graded early warning module 203 includes a local warning unit and a remote transmission unit using an MVB / TRDP network card; it is used to trigger graded early warnings based on dynamic thresholds, realize local buzzer warnings and TCMS system signal uploads, and the early warning signal transmission complies with SIL2 safety requirements.

[0030] The data interaction and maintenance module 204 consists of a Ulog logging component, an RS232 / USB maintenance interface, an MVB / TRDP communication interface, and maintenance-related units; it is used to record monitoring data, support log download, and realize fault risk prediction and command interaction with the maintenance system.

[0031] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0032] 1. Significantly reduced false alarm rate: By combining three-dimensional parameter monitoring of electrical, thermal, and stress parameters to cover the early signs of package failure, the trial operation test results show that the false alarm rate of thermal failure of SiC modules has dropped from 18% to 0.3%, avoiding air conditioner shutdown caused by solder layer cracking;

[0033] 2. Improved threshold adaptability: The dual-factor dynamic threshold reduces the false alarm rate from 12% to 2% and shortens the heavy load warning lag time from 300ms to 50ms, adapting to the complex working conditions of rail transit.

[0034] 3. Safety Level Compliance: The entire-link SIL2 safety closed loop (dual MCU verification, CRC verification) has passed third-party certification, with a fault false response rate of 0, and complies with EN50155 and IEC61508 standards;

[0035] 4. Optimized operation and maintenance efficiency: LSTM model prediction increases the proportion of planned maintenance from 30% to 85%, and reduces operation and maintenance costs by 40%;

[0036] 5. Strong platform adaptability: 80% of existing hardware components (ARM MCU, PHM board, MVB network card, etc.) are reused, and the software is developed based on a unified RTOS, which can be directly ported to urban rail air conditioning frequency converters and subway EC drives, shortening the industrialization cycle to 3 months. Attached Figure Description

[0037] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this invention, illustrate exemplary embodiments of the invention and are used to explain the invention, but do not constitute an undue limitation of the invention. In the drawings:

[0038] Figure 1 This is a flowchart of a collaborative monitoring method for thermal faults in silicon carbide modules of rail transit frequency converters.

[0039] Figure 2 This is a schematic diagram of a collaborative monitoring system for thermal faults of silicon carbide modules in rail transit frequency converters. Detailed Implementation

[0040] To facilitate a clear description of the technical solutions in the embodiments of the present invention, the terms "first" and "second" are used to distinguish identical or similar items with essentially the same function and effect. For example, the first threshold and the second threshold are merely used to distinguish different thresholds and do not limit their order. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and that the terms "first" and "second" are not necessarily different.

[0041] It should be noted that in this invention, the terms "exemplary" or "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design described as "exemplary" or "for example" in this invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of terms such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0042] In this invention, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one" or similar expressions refer to any combination of these items, including any combination of singular or plural items. For example, "at least one of a, b, or c" can represent: a, b, c, a combination of a and b, a combination of a and c, a combination of b and c, or a, b, and c, where a, b, and c can be single or multiple.

[0043] like Figure 1 The diagram shows a flowchart of a collaborative monitoring method for thermal faults in silicon carbide modules of rail transit frequency converters. The method includes the following steps:

[0044] S101: Acquires multi-dimensional parameters of the silicon carbide module;

[0045] An NTC thermistor, an infrared temperature measurement chip, and a MEMS thermal stress sensor are integrated in the JFET region of the SiC module. Real-time junction temperature (T) is synchronously acquired via the PHM monitoring and maintenance board. j ), leakage current (I) dss Threshold voltage (V) th ), thermal stress (σ) th The sampling frequency is ≥10kHz, and the data is updated every 100μs and transmitted to the control unit via the SPI interface.

[0046] S102: Constructing a loss model for SiC modules using a magnetic field orientation control algorithm;

[0047] Based on the FOC control algorithm, a loss model for SiC modules is constructed: in, For output current, This is the DC bus voltage. The switching frequency is used, and the input parameters are acquired through the inverter's function test interface. The real-time junction temperature is calculated using the transient thermal impedance curve of the SiC device. , For module housing temperature, This represents the junction-to-case transient thermal impedance. In practice, to ensure accurate and reliable temperature readings, the junction temperature results are cross-validated between the ARM Cortex-M7 main MCU and the Cortex-M4 backup MCU, and the results are valid if the deviation is ≤2℃.

[0048] S103: Utilize the aforementioned loss model to perform multi-physics collaborative analysis and construct a three-dimensional thermal fault precursor identification logic;

[0049] The electrical, thermal, and stress parameters collected in step S101 are used to establish a linkage judgment logic, and the triggering conditions must be met simultaneously:

[0050] The electrical parameters are as follows: ;

[0051] The thermal parameter conditions are: or The former indicates a precursor to a thermal warning, while the latter indicates a critical precursor to a thermal failure.

[0052] The mechanical stress conditions are: .

[0053] in, At 350MPa, the probability of cracking of the SiC module solder layer is ≥90%.

[0054] S104: Establish dynamic threshold-based hierarchical early warning;

[0055] Initial threshold setting: Set the heat warning threshold. =170℃, thermal failure threshold =195℃.

[0056] Calculation of working condition compensation factor K1: ,in Rated current, The rated switching frequency is used when the high-speed train is climbing a hill. Lowered to 162℃.

[0057] Calculation of lifetime correction factor K2: , The module's cumulative running months are taken from PHM board records, after 5 years (60 months) of operation. Lowered to 183℃.

[0058] S105: Real-time junction temperature When the heat warning threshold is reached, a local alert is issued via a buzzer;

[0059] The warning was triggered, in achieve When the temperature reaches 170℃ × K1, the buzzer of the door controller (using the BSM-DCU003 door controller warning logic) triggers a local alarm, and simultaneously uploads the data to the TCMS system via the MVB / TRDP network card (compliant with IEC61375); upon arrival... When the temperature reaches 195℃ × K2, an additional redundant module activation signal is output.

[0060] S106: Record multi-dimensional parameter collection data, collaborative analysis results and early warning trigger nodes in the log, and upload the monitoring data to the operation and maintenance system through the MVB / TRDP network card;

[0061] The Ulog logging component stores log content including collection timestamps, calculated junction temperature values, early warning trigger conditions, and operation and maintenance instructions. Logs are stored in partitions (compliant with SIL2 audit requirements) and support RS232 / USB interface download.

[0062] S107: The operation and maintenance system uses a preset fault prediction model to predict fault risks and generates operation and maintenance instructions based on the prediction results.

[0063] Specifically, every 24 hours, the peak junction temperature sequence of the past 72 hours and the cumulative number of module switching times are uploaded to the operation and maintenance cloud platform via the TRDP network card. The LSTM model of the cloud platform (after...) (Training with historical data) Outputs the probability of failure risk in the next 15 days;

[0064] When the risk level is ≥60%, a replacement reminder is pushed out. The operation and maintenance personnel issue a "threshold adjustment / calibration" command, which is transmitted to the monitoring system through the MVB secure channel. The execution result is sent back to the cloud platform to achieve closed-loop operation and maintenance.

[0065] The method is implemented based on existing hardware components (PHM board, ARM MCU, MVB / TRDP network card) and software platform (FOC algorithm, RTOS system).

[0066] like Figure 2 The diagram shows a collaborative monitoring system for thermal faults of SiC modules in a rail transit frequency converter. This system, used to implement the aforementioned method and deeply adapted to existing frequency converter hardware, includes the following modules:

[0067] Thermal monitoring module 201: It consists of an NTC thermistor, an infrared temperature measurement chip, a MEMS thermal stress sensor, a PHM monitoring and maintenance board, and an online insulation detection system; it is responsible for synchronously collecting four-dimensional parameters and monitoring the insulation performance of the acquisition circuit.

[0068] Collaborative Analysis Module 202: Built on the ARM Cortex-M7 MCU (STM32H743) and FreeRTOS platform; integrates FOC loss model, dual MCU cross-validation logic, and three-dimensional fault identification logic; the calculated junction temperature monitoring accuracy is ±0.5℃.

[0069] The graded early warning module 203 includes a local warning unit (reusing the existing gate controller buzzer) and a remote transmission unit (including the existing MVB network card BSM-MVB002 and TRDP network card BSM-TRDP001); the early warning signal transmission is accompanied by a CRC-32 checksum.

[0070] Data Interaction and Operation & Maintenance Module 204: Composed of Ulog logging component, RS232 / USB maintenance interface (multiplexed gating interface), MVB / TRDP communication interface and operation and maintenance related unit; supports local log storage and cloud interaction.

[0071] Example 1: Construction of Logic for Identifying Precursor Thermal Faults

[0072] The three-dimensional thermal fault precursor identification logic takes electrical parameters, thermal parameters, and mechanical stress parameters as core dimensions. Through parameter collaborative acquisition, dynamic threshold calibration, and multi-condition linkage judgment, it solves the problem of missed detection in traditional single-parameter monitoring.

[0073] The electrical parameters focus on the precursors of thermal failure in SiC modules, with the monitored parameters being threshold voltage offset (ΔVth) and leakage current (Idss). These are sensitive indicators of gate oxide aging and abnormal junction temperature in SiC modules. Before thermal failure, changes in carrier mobility cause a significant drift of ≤500mV in ΔVth. When the junction temperature of the SiC module rises or the package insulation degrades, the leakage current increases from the normal μA level to ≥10mA, which is an early electrical signal of thermal runaway.

[0074] The thermal parameter dimension focuses on the core thermal state of the SiC module, with the real-time junction temperature (Tj) monitored through a combination of an NTC thermistor and an infrared temperature measurement chip. The inverter's FOC algorithm is used to control the switching state of the SiC module, and the real-time junction temperature is calculated by combining the transient thermal impedance curve of the SiC device. Tc is the module housing temperature. For junction-shell transient thermal impedance, SiC module losses derived from the FOC algorithm, including output current. DC bus voltage The switching frequency fsw correlation term avoids the deviation problem of traditional methods that only measure the case temperature.

[0075] Mechanical stress parameters focus on early signs of SiC module packaging failure, with added thermal stress. The parameters are used to monitor package cracking issues and are collected by a MEMS thermal stress sensor. The MEMS thermal stress sensor is integrated between the SiC module substrate and the heat sink, with a range of 0-500MPa and an accuracy of ±5MPa. It monitors the thermal stress caused by the difference in thermal expansion coefficients between the SiC chip and the package layer (bonding wires, solder). The thermal stress before bonding wire creep and solder layer cracking is usually greater than 350MPa.

[0076] Electrical parameters (ΔVth, Idss) and thermal parameters (Tj) are synchronously acquired through the analog / digital channels of the PHM monitoring and maintenance board. Thermal stress (σth) is accessed through the analog interface of the PHM board to avoid adding a new independent acquisition link. The sampling frequency is uniformly set to ≥10kHz, and the sampling data is updated every 100μs. It is transmitted to the ARM Cortex series MCU of the inverter control unit through the SPI interface to ensure that the three-dimensional parameters have no timing deviation.

[0077] Since the monitoring using fixed thresholds does not take into account the fluctuations in operating conditions such as high-speed rail climbing and subway start-stop, it is necessary to design dynamic thresholds for the three-dimensional parameters that combine operating condition compensation and lifespan correction to replace the traditional fixed thresholds.

[0078] During the dynamic calibration of electrical parameter thresholds, the baseline value of the ΔVth warning threshold is set to 500mV, and corrected by combining the output current Iout acquired by the FOC algorithm. Irated is the rated current. When the high-speed train is climbing a slope, Iout ≥ 1.2Irated, so the threshold is lowered to 440mV to detect abnormal electrical parameters under heavy load in advance. The Idss warning threshold is set to a base value of 10mA, and is corrected based on the module's cumulative runtime recorded by the PHM board and the Toperate life index. The unit of Toperate is months. After running for 1 year (12 months), the threshold rises to 12mA to adapt to the increase in leakage current baseline caused by aging.

[0079] During the dynamic calibration of the thermal parameter (Tj) threshold, a basic threshold is set: thermal early warning. =170℃, thermal failure =195℃.

[0080] pass Perform operating condition compensation and correction. The rated switching frequency is used when the subway starts and stops. A sudden increase It rose to 1.05. Lowered to 162℃.

[0081] pass Perform lifespan correction after 5 years (60 months) of operation. =0.94, The temperature was lowered to 183℃ to accommodate the decreased heat resistance of SiC after aging.

[0082] The basic early warning threshold for thermal stress (σth) is 350 MPa, combined with ambient temperature. Perform dynamic correction of thermal stress:

[0083] The ambient temperature can be collected by the heat dissipation unit. When the tunnel ambient temperature is 45℃, the threshold is lowered to 340MPa to avoid high temperature aggravating stress accumulation and causing false alarms.

[0084] In an ARM Cortex MCU, a linkage judgment logic is embedded. Only when the dynamic threshold conditions of the three-dimensional parameters are simultaneously met can it be determined as a precursor to a thermal fault. The specific logic is as follows:

[0085] To determine the triggering conditions, electrical parameters, thermal parameters, and mechanical stress conditions must all be met simultaneously for it to be considered a precursor to a thermal fault.

[0086] The electrical parameters are as follows: ;

[0087] The thermal parameter conditions are: or The former indicates a precursor to a thermal warning, while the latter indicates a critical precursor to a thermal failure.

[0088] The mechanical stress conditions are: .

[0089] For example, in the case of high-speed rail climbing uphill, =162℃, When ΔVth=420mV, Idss=11mA, Tj=165℃, and σth=348MPa are detected, it is determined to be a precursor to a heat warning.

[0090] For example, regarding the situation of module aging, =183℃. When ΔVth=480mV, Idss=23mA, Tj=185℃, and σth=352MPa are monitored, it is determined to be a critical precursor to a thermal fault.

[0091] Example 2: LSTM Model Construction

[0092] The LSTM model is used to predict the risk of thermal failure of SiC modules within the next 15 days.

[0093] Synchronous acquisition of 72-hour junction temperature peak sequence: The junction temperature peak of the SiC module was recorded once per hour. The junction temperature peak was analyzed using a loss model. The calculations yielded 72-dimensional input features.

[0094] The PWM signal counting unit of the collaborative analysis module, which can be integrated into the ARM Cortex-M7 MCU, is used to count the cumulative number of switching operations. The count increments by 1 for each detected rising edge of the PWM signal, and the accumulated value is stored in EEPROM (not lost when power is off), forming a 1D input feature.

[0095] Label the occurrence of thermal failure (junction temperature ≥195℃×K2 or thermal stress ≥350MPa) within the next 15 days, with "1" marked as the occurrence and "0" marked as the absence, to obtain the label data.

[0096] The training dataset consists of three years of operational data from frequency converters on rail locomotives. The sample includes three typical rail transit operating conditions. Data from high-speed rail climbing operations accounts for 30%, with parameters ranging from [specific parameters would be inserted here]. to Cruise operating condition samples accounted for 50%, parameter range to The sample size for station start-stop operation was 20%, with a parameter range of... From 0 to The fluctuation frequency is ≥5 times / hour.

[0097] The test dataset uses operational data from 10 trains from January to June 2024, totaling... Use a set of samples to ensure that the data distribution is consistent with the training set.

[0098] Data preprocessing operations, including outlier removal and normalization, are performed on the training and test sets. The junction temperature peak and the cumulative number of switching operations are mapped to the [0,1] interval to avoid the impact of dimensional differences on model training.

[0099] The LSTM model is implemented based on the TensorFlow framework, and the model network structure is shown in Table 1.

[0100] Table 1: Network Structure of LSTM Model

[0101]

[0102] During LSTM model training, the Adam optimizer was used with an initial learning rate of 0.001 (decreasing by a factor of 0.9 after 100 iterations); the binary cross-entropy loss function was employed; the batch size was 128; and the number of epochs was 200. The trained model was then validated offline, achieving a test set accuracy ≥85%, precision ≥82%, and recall ≥80%.

[0103] The model is deployed on the cloud server of the operation and maintenance system. The data interaction and operation and maintenance module uploads the "72-hour junction temperature peak sequence + cumulative switching count" to the cloud once every 24 hours via the TRDP network card, meeting the operation and maintenance requirements of "daily update of fault risk". The model parameters can be further compressed to adapt to the limited computing power of the vehicle-mounted ARM MCU and realize vehicle-mounted deployment.

[0104] Most existing technologies use a single parameter threshold for monitoring and judgment, at most employing two-dimensional monitoring of temperature and current. The technical solution of this invention achieves innovative multi-physics coupling, integrating "electricity (ΔV)" into the frequency converter of rail transit. th / I dss )-heat (T) j ) - Stress (σ th The invention utilizes three-dimensional parameters and a thermal stress threshold of 350 MPa calibrated through 1000 thermal cycle tests (unlike the fixed 300 MPa of IEEE TPEL), solving the problem of false alarms due to encapsulation cracking, demonstrating technological originality. This invention combines operating condition (K1) and lifespan (K2) as two factors to achieve dynamic threshold two-factor calibration, reducing the false alarm rate from 12% to 2%, a non-obvious improvement. While existing technologies largely focus on functional safety concepts, this invention introduces end-to-end SIL2 to link parameter monitoring with safety levels. This invention addresses high-speed rail climbing (K1 compensation) and high-temperature tunnels (σ... th In special working conditions such as dynamic correction, dynamic threshold adjustment can be achieved, which can adapt to more scenarios in actual working conditions.

[0105] Although the invention has been described herein in conjunction with various embodiments, those skilled in the art will understand and implement other variations of the disclosed embodiments by reviewing the accompanying drawings, disclosure, and other materials. In this specification, the word "comprising" does not exclude other components or steps, and "a" or "an" does not exclude multiple components. A single processor or other unit can implement several functions listed in the specification. While certain measures are described in different embodiments, this does not mean that these measures cannot be combined to produce good results.

[0106] Although the invention has been described in conjunction with specific features and embodiments, it is obvious that various modifications and combinations can be made therein without departing from the spirit and scope of the invention. Accordingly, this specification and drawings are merely illustrative of the invention and are considered to cover any and all modifications, variations, combinations, or equivalents within the scope of the invention. Clearly, those skilled in the art can make various alterations and modifications to the invention without departing from its spirit and scope. Thus, if such modifications and modifications fall within the scope of the invention and its equivalents, the invention is also intended to include such modifications and modifications.

Claims

1. A method for collaborative monitoring of thermal faults in silicon carbide modules of rail transit frequency converters, characterized in that, Includes the following steps: S1: Collect multi-dimensional parameters of the silicon carbide module, including real-time junction temperature, leakage current, threshold voltage and thermal stress data; S2: Constructing a loss model for silicon carbide modules using a magnetic field orientation control algorithm. Where Iout is the inverter output current, Vdc is the DC bus voltage, and fsw is the silicon carbide module switching frequency; S3: Perform multi-physics collaborative analysis using the loss model, calculate the real-time junction temperature Tj by combining the transient thermal impedance curve of the silicon carbide module, and correlate junction temperature, leakage current, threshold voltage and thermal stress data to construct a three-dimensional thermal fault precursor identification logic. S4: Establish dynamic threshold hierarchical early warning, set initial thermal warning threshold and thermal fault threshold, calculate the operating condition compensation factor K1 based on the Iout and fsw parameters collected by the magnetic field orientation control algorithm, and dynamically adjust the thermal warning threshold through parameter K1; calculate the lifespan correction factor K2 by combining the module's cumulative running time in the operation and maintenance records, and dynamically adjust the thermal fault threshold through parameter K2. S5: When the real-time junction temperature Tj reaches the thermal warning threshold, a local warning is issued via a buzzer, and the warning signal is uploaded to the rail transit TCMS system via the MVB / TRDP network card; In step S3, the multi-physics collaborative analysis performed using the loss model to construct the three-dimensional thermal fault precursor identification logic specifically includes: Using the collected leakage current and threshold voltage as electrical parameters, real-time junction temperature as a thermal parameter, and thermal stress data as a stress parameter, a linkage judgment logic is established. The triggering condition must simultaneously meet the following conditions: The electrical parameters are as follows: ,in, The thermal parameter conditions are: or The former indicates a precursor to a thermal warning, while the latter indicates a critical precursor to a thermal failure. The mechanical stress conditions are: .

2. The method according to claim 1, characterized in that, The method further includes the following steps: S6: Record multi-dimensional parameter collection data, collaborative analysis results and early warning trigger nodes in the log, and upload the monitoring data to the operation and maintenance system through the MVB / TRDP network card; S7: The operation and maintenance system uses a preset fault prediction model to predict fault risks and generates operation and maintenance instructions based on the prediction results.

3. The method according to claim 1, characterized in that, The input parameters Iout, Vdc, and fsw of the loss model are acquired through the functional test interface of the frequency converter. The real-time junction temperature calculation result is cross-validated by the control unit of the frequency converter through secondary calculation. It takes effect when the junction temperature deviation of the secondary calculation is less than the error threshold.

4. The method according to claim 1, characterized in that, The formula for calculating the operating condition compensation factor K1 is as follows: Where Irated is the rated current of the inverter, and frated is the rated switching frequency of the silicon carbide module; The formula for calculating the lifetime correction factor K2 is as follows: , where Toperate is the cumulative number of months the module has been running, taken from the runtime records in the monitoring and maintenance data.

5. The method according to claim 2, characterized in that, The fault prediction model is implemented using an LSTM model trained with embedded historical operating data of the frequency converter; Input parameters include the junction temperature peak sequence of the most recent 72 hours and the cumulative number of module switching times; The output is the probability of failure risk in the next 15 days; When the risk exceeds 60%, an operation and maintenance reminder is pushed. The operation and maintenance system issues threshold adjustment and calibration trigger commands, which are transmitted to the monitoring system for execution through the secure channel of the MVB / TRDP network card. The execution results are sent back to the operation and maintenance system to form a closed loop.

6. The method according to claim 5, characterized in that, The threshold adjustment, calibration trigger, early warning, and operation and maintenance command execution are all verified by SIL2 level security verification logic: thermal monitoring data transmission is supplemented with CRC-32 check code; during fault response, the main MCU and backup MCU of the frequency converter synchronously calculate key parameters, and the operation is allowed only if the deviation meets the preset threshold.

7. The method according to claim 1, characterized in that, The triggering conditions for the three-dimensional thermal fault precursor identification logic include: junction temperature ≥ 195℃ × K2, threshold voltage deviation. Leakage current Idss≥10mA, thermal stress≥350MPa; the 350MPa thermal stress threshold is calibrated by thermal cycling test at -40℃~85℃, with a test cycle of 1000 cycles, covering high temperature and vibration conditions in rail transit.

8. The method according to claim 1, characterized in that, In step S1, the data is collected using a low-inductance composite busbar with a nested copper foil shielding layer. In step S3, the collaborative analysis employs a two-stage anti-interference algorithm combining interference mode matching and Kalman filtering. First, the types of interference between the pantograph-catenary pulse and the motor harmonics are identified through waveform similarity comparison. Then, specific filtering parameters are called to improve the junction temperature monitoring accuracy from ±1℃ to ±0.5℃.

9. The method according to claim 2, characterized in that, The logs use the Ulog log component to support secure partitioned storage. The log content includes the collection timestamp, calculated junction temperature value, early warning triggering conditions, operation and maintenance instructions and execution results. After the logs are downloaded, they are parsed by the host computer software. This software is from the same source as the inverter function test software, supports thermal fault tracing and development trend analysis, and the log format meets the audit requirements of SIL2 certification.

10. A collaborative monitoring system for thermal faults of silicon carbide modules in rail transit frequency converters, used to implement the method according to any one of claims 1 to 9, characterized in that, include: The thermal monitoring module includes an NTC thermistor, an infrared temperature measurement chip, a MEMS thermal stress sensor, a PHM monitoring and maintenance board, and an online insulation detection system. Used to synchronously acquire junction temperature, leakage current, threshold voltage and thermal stress data of silicon carbide modules, and monitor the insulation performance of the acquisition circuit. The collaborative analysis module is built on the ARM Cortex series MCU and RTOS software platform, integrating the SiC module loss model derived from the FOC control algorithm, dual-core cross-validation logic, three-dimensional thermal fault precursor identification logic, and interference matching filter anti-interference algorithm. Used to calculate real-time junction temperature, verify data accuracy, and identify early signs of thermal failures; The graded early warning module includes a local warning unit and a remote transmission unit using an MVB / TRDP network card; it is used to trigger graded early warnings based on dynamic thresholds, realize local buzzer warnings and TCMS system signal uploads, and the early warning signal transmission complies with SIL2 safety requirements. The data interaction and operation and maintenance module consists of a Ulog logging component, an RS232 / USB maintenance interface, an MVB / TRDP communication interface, and an operation and maintenance related unit; it is used to record monitoring data, support log download, and realize fault risk prediction and command interaction with the operation and maintenance system.

Images