A method and system for diagnosing and classifying thermal transients of an automobile ignition

Abstract

The application relates to the technical field of electronic testing, and discloses a method and system for diagnosing and classifying the thermal transient of an automobile ignition device, which comprises the following steps: applying a current excitation and simultaneously acquiring a response signal of the ignition device; preprocessing the response signal and dividing the response signal into a thermal transient signal sequence; performing multi-scale time-frequency analysis on the thermal transient signal sequence, obtaining a characteristic parameter, and obtaining a thermal response strength and a stability index according to the characteristic parameter to form a multi-dimensional feature vector; fitting a thermal response curve to the multi-dimensional feature vector, determining an electrothermal coupling relationship parameter, and obtaining a steady-state characteristic; collecting environmental temperature data, performing temperature compensation and channel calibration on the multi-dimensional feature vector, and obtaining judgment data; identifying and classifying the judgment data, and outputting a diagnosis result. The application realizes analysis and defect type identification of the thermal transient process of the ignition device, and improves the reliability and environmental adaptability of performance evaluation of the ignition device.

Description

Technical Field

[0001] This invention relates to the field of electronic testing technology, and more specifically, to a method and system for thermal transient diagnosis and classification of automotive ignition devices. Background Technology

[0002] With the development of automotive safety systems, ignition devices, as core actuators in critical components such as airbags and fuel detonation systems, directly impact vehicle safety in terms of performance stability and response characteristics. Traditional ignition device testing methods primarily focus on on / off states and steady-state electrical performance indicators, lacking multi-channel synchronous analysis methods for the initial thermal transient process during energization. Because ignition devices undergo complex electrothermal coupling changes at the moment of ignition, the bridge wire temperature rise, radiation intensity, and electrical parameter changes exhibit strong nonlinear coupling, making single current and voltage detection insufficient to reveal potential defects. With the development of high-precision constant current sources, high-speed synchronous acquisition, and infrared thermometry, combining multimodal response signals with thermal transient characteristic modeling has become an important direction for improving ignition device quality consistency and early defect detection.

[0003] However, existing technologies, such as patent CN118244186B—a method and system for testing high current in ignition devices—while establishing evaluation indicators such as current rise fluctuation index, potential anomaly index, and standard index, can be used to verify the continuous stability of the output current of the high current tester for ignition devices. However, their core focus remains on the current output capability and voltage response consistency of the testing equipment, rather than the thermal response characteristics of the ignition device itself. This method evaluates the performance of the tester through changes in current and voltage waveforms, failing to reflect transient nonlinear behavior in key heat transfer processes such as bridge wire temperature and surface thermal radiation, and lacking the ability to characterize and classify the electrothermal coupling characteristics of the ignition device. Its data processing emphasizes time-domain statistical analysis, without establishing a multi-channel time synchronization mechanism and a temperature compensation calibration mechanism, resulting in significant detection errors under different ambient temperatures or channel drift.

[0004] Therefore, it is necessary to design a method and system for thermal transient diagnosis and classification of automotive ignition devices to solve the problems existing in the current technology. Summary of the Invention

[0005] In view of this, the present invention proposes a method and system for thermal transient diagnosis and classification of automotive ignition devices, aiming to solve the problem that the existing technology has insufficient dynamic characterization capability in the thermal transient stage and cannot diagnose the consistency of thermal response and the difference in charge state of the ignition device.

[0006] In one aspect, this invention proposes a method for thermal transient diagnosis and classification of automotive ignition devices, comprising: A current excitation is applied to the ignition device of the vehicle under test, and the response signal of the ignition device is acquired simultaneously based on multi-channel synchronous acquisition. The response signal includes current signal, voltage signal, surface thermal radiation signal and bridge wire temperature signal. The response signal is preprocessed and then divided into a thermal transient signal sequence within a unified time window. The preprocessing includes time alignment, noise suppression, and amplitude normalization. Multi-scale time-frequency analysis is performed on the thermal transient signal sequence to obtain characteristic parameters, and thermal response intensity and stability index are obtained based on the characteristic parameters to form a multi-dimensional feature vector. The characteristic parameters include thermal response energy peak value, rise time, energy decay time and thermal hysteresis time. Based on the synchronous change pattern of the current signal and the temperature-related signal during the energizing process, the multidimensional feature vector is fitted with a thermal response curve to determine the electrothermal coupling parameters and obtain steady-state characteristics. The steady-state characteristics are then incorporated into the multidimensional feature vector. The steady-state characteristics include equivalent heat capacity parameters, thermal conductivity parameters, and temperature-related resistance change parameters. The temperature-related signals include surface thermal radiation signals and bridge wire temperature signals. Collect ambient temperature data, perform temperature compensation and channel calibration on the multidimensional feature vector after incorporating steady-state features, and obtain judgment data. The channel calibration includes gain correction, zero point correction and time base correction. The judgment data is identified and classified, diagnostic results are output, and confidence levels are generated.

[0007] Furthermore, when applying current excitation to the ignition device of the vehicle under test and simultaneously acquiring the response signal of the ignition device based on multi-channel synchronous acquisition, the process includes: The igniter is excited by a constant current source, which is controlled in a closed loop by a four-terminal sampling resistor to ensure that the current step settling time is no more than 0.5 milliseconds and the overshoot is no more than 5%. Simultaneous acquisition of current, voltage, surface thermal radiation, and bridge wire temperature signals is performed on a unified time reference, with a sampling rate of no less than two million times per second and a channel time error of no more than one hundred nanoseconds. The current signal is acquired via a four-terminal sampling resistor and differential amplification, the voltage signal is acquired using a high input impedance differential method, the surface thermal radiation signal is acquired by an infrared temperature sensor, and the bridge wire temperature signal is acquired by a contact temperature sensor or a fiber optic temperature sensor.

[0008] Furthermore, when preprocessing the response signal and dividing the preprocessed response signal into a thermal transient signal sequence within a unified time window, the process includes: The rising edge of the current signal is used as an alignment marker to perform time alignment on the current signal, voltage signal, surface thermal radiation signal, and bridge wire temperature signal. Band-limited filtering is used to suppress broadband noise and impulse interference, and baseline offset is extracted from the static and stable section at the beginning of the acquisition for baseline correction. The amplitude of the current channel and voltage channel is normalized according to the channel gain coefficient, and the amplitude of the surface thermal radiation signal and bridge wire temperature signal is normalized according to the temperature sensor calibration coefficient. The normalization results are then unified to the same amplitude range. The response signal is synchronously divided into the thermal transient signal sequence, with the starting point of the unified time window being the moment when the current rise edge reaches the current set ratio and the ending point being the moment when the energy decays to the energy set ratio after the bridge wire temperature peaks.

[0009] Furthermore, when constructing a multidimensional feature vector, it includes: The thermal transient signal sequence is divided into a rising phase, a peak phase, and an initial decay phase according to the scale. In the decomposition results, the scale with the highest energy concentration is taken as the principal scale, and the thermal response energy peak is determined according to the energy peak of the principal scale. The rise time is determined by the time difference between the first moment when the accumulated energy reaches the energy rise ratio during the rise phase and the start of a unified time window; the energy decay time is determined by the time difference between the moment when the energy decreases to the energy decrease ratio after the peak phase and the peak moment; the thermal hysteresis time is determined by the time difference between the peak moment of the temperature-related signal and the moment corresponding to the rising edge of the current signal; the thermal response energy peak, rise time, energy decay time, thermal hysteresis time, principal scale position, peak symmetry, and energy distribution width are combined to form the multidimensional feature vector; in generating the multidimensional feature vector, abnormal segments are removed according to the noise sensitivity threshold, and the thermal response intensity index and thermal stability index are calculated according to the ambient temperature range.

[0010] Furthermore, when fitting the thermal response curve of the multidimensional feature vector, the process includes: selecting the time synchronization segment of the current signal, voltage signal and bridge wire temperature signal as the fitting interval during the energizing process, and fitting the thermal response curve with the temperature rise segment corresponding to the current change as the core to obtain the correspondence between the temperature rise rate and the power change rate calculated from the current signal and voltage signal; and adjusting the curve smoothness and fitting accuracy according to the energy balance constraint during the fitting process.

[0011] Furthermore, determining the electrothermal coupling parameters includes: calculating the ratio of energy accumulation rate to temperature rise rate based on the thermal response curve fitting results, and determining the equivalent heat capacity parameter; determining the thermal conductivity parameter based on the time constant difference between the temperature rise phase and the decay phase; and determining the temperature-related resistance change parameter based on the hysteresis between the current signal and the temperature-related signal.

[0012] Furthermore, obtaining steady-state features and incorporating these features into the multidimensional feature vector includes: The equivalent heat capacity parameter, thermal conductivity parameter, and temperature-related resistance change parameter are standardized according to a preset weight ratio to eliminate deviations caused by differences in the structure and charge of different batches of igniters; the standardized steady-state features are then incorporated into the multidimensional feature vector.

[0013] Furthermore, when collecting ambient temperature data and performing temperature compensation and channel calibration on the multidimensional feature vector after incorporating steady-state features, the following steps are taken: During the test, ambient temperature data is collected in real time, and the ambient temperature data is smoothed to obtain the temperature change trend. Based on the temperature change trend, the amplitude drift of the current signal, voltage signal, surface thermal radiation signal and bridge wire temperature signal is corrected. The compensation coefficient is calculated by piecewise linear interpolation and the temperature-related feature terms in the multidimensional feature vector are compensated. During channel calibration, the channel zero-point offset is determined by static sampling under zero-load conditions and the zero-point correction is performed. The channel gain is determined by standard signal input and the gain correction is performed. The sampling time difference of each channel is corrected according to a unified trigger pulse to complete the time reference correction.

[0014] Furthermore, when identifying and classifying the judgment data, outputting diagnostic results, and generating confidence levels, the process includes: Based on historical samples, the judgment rules are trained and then the judgment data is jointly classified using multiple features to obtain a diagnostic result. The diagnostic result includes at least one of the following: normal state, contact abnormality, charge abnormality, and bridge wire abnormality. While outputting diagnostic results, a confidence level is generated based on the similarity to the feature centers of each category and the coordination between features. When the confidence level is lower than the set confidence level threshold or when there is inconsistency in the judgment between different repeated measurements, a review and repeated collection are performed, and the review results and new samples are included in the training sample library.

[0015] Compared with existing technologies, the advantages of this invention are as follows: By simultaneously acquiring current signals, voltage signals, surface thermal radiation signals, and bridge wire temperature signals during the energizing excitation phase, a multi-modal acquisition system integrating electrical and thermal signals is constructed, enabling the capture of the entire process from energy input to thermal response of the ignition device under a unified time reference; after time alignment, noise suppression, and amplitude normalization, the multi-channel data is segmented into a thermal transient signal sequence, and key dynamic features such as energy peak value, rise time, decay time, and thermal hysteresis time are extracted using multi-scale time-frequency analysis, thereby achieving a quantitative characterization of the thermal response intensity and stability; by fitting the electrothermal coupling curve through the synchronous change law of current signals and temperature-related signals, steady-state features such as equivalent heat capacity, thermal conductivity, and temperature-related resistance changes are extracted, and combined with temperature compensation and channel calibration mechanisms, the comparability and stability of data under different environmental and equipment conditions are ensured; and high-confidence diagnostic results are output through multi-dimensional feature fusion and classification judgment. It breaks through the traditional detection mode that relies solely on electrical signals for evaluation, and realizes refined dynamic analysis and defect type identification of the thermal transient process of ignition devices, thereby improving the reliability and environmental adaptability of ignition device performance evaluation.

[0016] On the other hand, this application also provides a thermal transient diagnosis and classification system for automotive ignition devices, used to apply the above-mentioned thermal transient diagnosis and classification method for automotive ignition devices, including: The acquisition unit is configured to apply current excitation to the ignition device of the vehicle under test, and simultaneously acquire the response signal of the ignition device based on multi-channel synchronous acquisition. The response signal includes current signal, voltage signal, surface thermal radiation signal and bridge wire temperature signal. A preprocessing unit is configured to preprocess the response signal and divide the preprocessed response signal into a thermal transient signal sequence within a unified time window. The preprocessing includes time alignment, noise suppression, and amplitude normalization. The extraction unit is configured to perform multi-scale time-frequency analysis on the thermal transient signal sequence, obtain feature parameters, and obtain thermal response intensity and stability index based on the feature parameters to form a multi-dimensional feature vector. The feature parameters include thermal response energy peak value, rise time, energy decay time, and thermal hysteresis time. The processing unit is configured to fit the multidimensional feature vector to a thermal response curve based on the synchronous change pattern of the current signal and the temperature-related signal during the energizing process, determine the electrothermal coupling relationship parameters and obtain steady-state features, and incorporate the steady-state features into the multidimensional feature vector. The steady-state features include equivalent heat capacity parameters, thermal conductivity parameters and temperature-related resistance change parameters. The calibration unit is configured to collect ambient temperature data, perform temperature compensation and channel calibration on the multidimensional feature vector after incorporating steady-state features, and obtain judgment data. The channel calibration includes gain correction, zero-point correction and time base correction. The determination unit is configured to identify and classify the determination data, output diagnostic results, and generate confidence levels.

[0017] It is understandable that the above-mentioned methods and systems for thermal transient diagnosis and classification of automotive ignition devices have the same beneficial effects, and will not be elaborated further here. Attached Figure Description

[0018] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 A flowchart of a method for thermal transient diagnosis and classification of automotive ignition devices provided in an embodiment of the present invention; Figure 2 This is a functional block diagram of the automotive ignition device thermal transient diagnosis and classification system provided in an embodiment of the present invention. Detailed Implementation

[0019] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the disclosure to those skilled in the art. It should be noted that, unless otherwise specified, embodiments and features in the embodiments of the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0020] In traditional automotive ignition device testing, the detection system only collects current and voltage signals for steady-state parameter analysis, without establishing a multi-channel synchronous acquisition mechanism. This results in a time reference deviation between transient thermal signals such as bridge wire temperature and surface thermal radiation and electrical signals. Due to the lack of signal alignment and noise suppression methods within a unified time window, it is difficult to compare the thermal response curves of different batches of ignition devices. Existing technologies do not consider the influence of electrothermal coupling parameters on feature vectors, leading to the absence of steady-state features such as equivalent heat capacity and thermal conductivity parameters in multi-dimensional feature vectors, making it impossible to accurately characterize the internal energy transfer process of the ignition device. Channel drift caused by ambient temperature fluctuations is not eliminated through a dynamic compensation mechanism, resulting in accumulated errors in feature parameter extraction.

[0021] For example, in the testing phase of an automotive airbag ignition production line, the testing equipment employs a single-channel current step excitation and dual-channel voltage monitoring scheme. When there are minute contact defects at the interface between the bridge wire and the propellant, the initial temperature rise rate of the bridge wire and the surface thermal radiation intensity exhibit asymmetrical fluctuations. However, the existing system, lacking synchronous acquisition of temperature-related signals, cannot capture such transient anomalies. A millisecond-level time difference exists between the current and temperature signal acquisition, causing the calculated thermal hysteresis time to deviate by more than 30% of the actual value. Increased ambient temperature causes baseline drift in the infrared sensor; the feature vector, without temperature compensation, mistakenly identifies the drift signal as an anomaly in propellant density.

[0022] If the above issues are not addressed, potential defects such as poor internal contact and uneven propellant loading in the ignition device will go undetected before leaving the factory, leading to an increased false airbag triggering rate. Inconsistent time references in multi-channel signals will mask the true dynamic characteristics of the thermal response, causing quality consistency control across different production batches to fail. Feature drift caused by changes in ambient temperature will reduce the generalization ability of the classification model, increasing the misclassification rate of normal products. The lack of electrothermal coupling parameter modeling will result in inaccurate assessment of propellant aging, shortening the accuracy of ignition device service life prediction.

[0023] For this, please refer to Figure 1 As shown, this application proposes a method for thermal transient diagnosis and classification of automotive ignition devices, including: S100: Apply current excitation to the ignition device of the vehicle under test, and simultaneously acquire the response signal of the ignition device based on multi-channel synchronous acquisition. The response signal includes current signal, voltage signal, surface thermal radiation signal and bridge wire temperature signal. S200: Preprocesses the response signal and divides the preprocessed response signal into a thermal transient signal sequence within a unified time window. The preprocessing includes time alignment, noise suppression, and amplitude normalization. S100: Perform multi-scale time-frequency analysis on thermal transient signal sequences to obtain characteristic parameters and obtain thermal response intensity and stability index based on the characteristic parameters, forming a multi-dimensional feature vector. The characteristic parameters include thermal response energy peak, rise time, energy decay time and thermal hysteresis time. S400: Based on the synchronous change law of current signal and temperature-related signal during the energizing process, the multidimensional feature vector is fitted with thermal response curve to determine the electrothermal coupling relationship parameters and obtain steady-state characteristics. The steady-state characteristics are incorporated into the multidimensional feature vector. The steady-state characteristics include equivalent heat capacity parameters, thermal conductivity parameters and temperature-related resistance change parameters. The temperature-related signals include surface thermal radiation signals and bridge wire temperature signals. S500: Collects ambient temperature data, performs temperature compensation and channel calibration on the multi-dimensional feature vector after incorporating steady-state features, and obtains judgment data. Channel calibration includes gain correction, zero-point correction and time base correction. S600: Identifies and classifies the judgment data, outputs diagnostic results, and generates confidence levels.

[0024] Specifically, multi-channel synchronous acquisition of the ignition device's response signal refers to acquiring current, voltage, surface thermal radiation, and bridge wire temperature signals using multiple sensors under a unified time reference. This can be achieved using a high-speed data acquisition card combined with a four-terminal sampling resistor, an infrared temperature sensor, and a contact temperature sensor, solving the problem that traditional single current and voltage detection cannot capture multi-modal thermal responses. Preprocessing includes time alignment, noise suppression, and amplitude normalization. Specifically, this involves: aligning different channel signals to the same time axis using the rising edge of the current signal as the trigger reference; employing band-limited filtering technology to filter out out-of-band interference noise; and converting the original acquired signals of each channel into standard physical dimensions based on hardware gain coefficients and sensor calibration coefficients, and uniformly scaling them to a preset numerical range, thereby ensuring direct comparability of multi-channel signals within a unified time window. Multi-scale time-frequency analysis to obtain characteristic parameters refers to extracting the energy distribution characteristics of the signal at different time scales using a continuous wavelet transform algorithm. Specifically, the Morlet wavelet basis function is selected, the scale parameter covers a preset low-frequency to high-frequency range, the time window length is adaptively configured according to the sampling rate, and a fixed overlap rate is used for sliding calculation to quantify key parameters such as the energy peak and rise time of the thermal transient process. Thermal response curve fitting to determine electrothermal coupling parameters involves establishing a dynamic relationship model between the current and temperature signals. Specifically, a nonlinear least squares algorithm is used for iterative fitting, with the convergence condition set as the sum of squared residuals falling below a preset small threshold or reaching the maximum number of iterations, addressing the problem that traditional methods cannot characterize electrothermal coupling properties. Temperature compensation and channel calibration correct the impact of ambient temperature and acquisition channel errors on the feature vector. Specifically, piecewise linear interpolation is used to calculate compensation coefficients, zero-load baseline subtraction is used for zero-point correction, and standard signal injection is used for gain correction, eliminating misjudgments caused by temperature drift and channel deviation. The confidence level is generated by evaluating the credibility of the classification results based on feature similarity. Specifically, the feature similarity is obtained by calculating the Mahalanobis distance between the current judgment data and the historical feature centers of each category, and the coordination degree between features is calculated by combining the covariance matrix between the dimensions of the multi-dimensional features, thereby improving the reliability of the diagnostic results.

[0025] This application uses multi-channel synchronous acquisition combined with multi-scale time-frequency analysis to model the thermal transient process of current, voltage, and temperature signals as a multi-dimensional feature vector, and introduces electrothermal coupling parameters and temperature compensation mechanisms to achieve high-precision classification and judgment of early defects in ignition devices.

[0026] The working process and principle of this application are as follows: a current excitation is applied to the ignition device of the vehicle under test, and current signal, voltage signal, surface thermal radiation signal, and bridge wire temperature signal are simultaneously acquired through multi-channel synchronous acquisition. These response signals are preprocessed, including time alignment, noise suppression, and amplitude normalization, and then divided into a thermal transient signal sequence within a unified time window.

[0027] Multi-scale time-frequency analysis is performed on thermal transient signal sequences to extract characteristic parameters such as thermal response energy peak value, rise time, energy decay time, and thermal hysteresis time. Based on these characteristic parameters, thermal response intensity and stability index are calculated, forming a multi-dimensional feature vector.

[0028] Based on the synchronous change pattern of current signal and temperature-related signal during energization, thermal response curves are fitted to the multidimensional feature vector. The electrothermal coupling parameters are determined through fitting, and steady-state characteristics such as equivalent heat capacity parameters, thermal conductivity parameters, and temperature-related resistance change parameters are obtained. These steady-state characteristics are then incorporated into the multidimensional feature vector.

[0029] Ambient temperature data is collected, and temperature compensation and channel calibration are performed on the multidimensional feature vector containing steady-state characteristics, including gain correction, zero-point correction and time base correction, to obtain the final judgment data.

[0030] The system identifies and classifies the judgment data, outputs diagnostic results, and generates confidence levels. This method, through multi-channel synchronous acquisition and thermal transient analysis combined with electrothermal coupling modeling, can comprehensively characterize the dynamic response characteristics of ignition devices, improving the accuracy of diagnosis and classification.

[0031] As a preferred embodiment, the solution of this application is specifically implemented as follows: First, a constant current source is used to apply current excitation to the ignition device of the vehicle under test. The constant current source is controlled in a closed loop through a four-terminal sampling resistor to ensure that the current step settling time does not exceed 0.5 milliseconds and the overshoot does not exceed 5%. A multi-channel synchronous acquisition system is used, with a sampling rate of not less than two million times per second and channel time error controlled within one hundred nanoseconds. The current signal is acquired through a four-terminal sampling resistor and differential amplification, while the voltage signal is acquired using a high input impedance differential method. The surface thermal radiation signal is acquired by an infrared temperature sensor, and the bridge wire temperature signal is acquired by a contact or fiber optic temperature sensor. The acquired response signals are preprocessed. The rising edge of the current signal is used as an alignment marker to time-align the signals of each channel. Band-limited filtering is used to suppress broadband noise and impulse interference, and baseline offset is extracted from the static steady-state section at the beginning of the acquisition for baseline correction. In amplitude normalization, the current and voltage channels are proportionally converted based on the sampling resistor value and amplifier gain. The conversion logic is to divide the original voltage signal by the product of the sampling resistor value and the amplification factor to obtain the actual physical quantity. The typical resistance value of the sampling resistor is set to 0.1 ohms, and the typical gain of the differential amplifier is set between ten and fifty times. The temperature channel is converted into a physical quantity based on the sensor sensitivity coefficient. The typical sensitivity coefficient range for infrared temperature sensors is ten to fifty millivolts per degree Celsius, while contact or fiber optic temperature sensors are converted based on the equivalent voltage temperature coefficient corresponding to the factory calibration curve, thus achieving unified normalization of multi-channel amplitudes. The thermal transient signal sequence is segmented within a unified time window. The preferred start point of the window is when the current rise edge reaches 80% to 95% of the rated current, and the preferred end point is when the energy decays to 10% to 20% of the peak energy after the bridge wire temperature peak. This proportional setting is based on the initial characteristics of thermal transient energy accumulation and the noise characteristics of the thermal decay tail, ensuring complete coverage of the effective response stage. Multi-scale time-frequency analysis is performed on the thermal transient signal sequence to extract characteristic parameters. The signal sequence is divided into a rise phase, a peak phase, and an initial decay phase. The scale with the highest energy concentration is used as the primary scale to determine the peak energy of the thermal response. The rise time and energy decay time are determined based on the time it takes for the accumulated energy to reach a set proportion. The thermal hysteresis time is determined by the time difference between the peak moment of the temperature-related signal and the current rise edge. Thermal response curve fitting is performed, selecting the time synchronization segment of the current signal, voltage signal, and bridge wire temperature signal as the fitting interval. Based on the correspondence between the temperature rise rate and the power change rate, steady-state characteristics such as equivalent heat capacity parameters, thermal conductivity parameters, and temperature-related resistance change parameters are determined. Ambient temperature data is collected in real time and smoothed to obtain the temperature change trend. The amplitude drift of each channel signal is corrected based on the temperature trend. Temperature compensation uses a piecewise linear interpolation method, dividing the ambient temperature into five-degree Celsius intervals. When the current temperature is between the boundaries of two intervals, the compensation coefficients preset at the boundary points are linearly proportionally allocated according to their relative positions to calculate the actual compensation coefficient at the current temperature.Zero-point correction is performed through static sampling under zero-load conditions, gain correction is performed through standard signal input, and time base correction is performed based on the sampling time difference of each channel according to a unified trigger pulse. Finally, the processed data is jointly classified using multi-feature classification based on the judgment rules trained on historical samples, and the diagnostic results are output. The judgment rules are preferably trained using support vector machines or random forest algorithms, and the training sample library contains no less than 3,000 sets of historical samples, covering four types of working conditions: normal, poor contact, loose charge, and bridge wire oxidation. During classification, the feature similarity between the current feature and the feature center of each category is calculated using Mahalanobis distance, and the physical consistency between feature dimensions is evaluated by calculating the Pearson correlation coefficient between feature dimensions. When the consistency is higher than a preset threshold, it is considered a valid association, and a confidence level is generated comprehensively. When the confidence level is lower than the threshold or inconsistencies occur, verification and repeated sampling are performed.

[0032] Through the above-described scheme, this application achieves comprehensive characterization and accurate diagnosis of the thermal transient response of automotive ignition devices. Multi-channel synchronous acquisition eliminates signal time reference deviation and improves the calculation accuracy of key parameters such as thermal hysteresis time. Thermal response curve fitting and electrothermal coupling modeling reveal the internal energy transfer characteristics of the ignition device, helping to detect minute contact defects and abnormal propellant loading. Real-time temperature compensation and channel calibration mechanisms suppress feature drift caused by ambient temperature fluctuations, enhancing the generalization ability of the classification model. Multi-dimensional feature vectors combined with steady-state features improve the ability to identify different types of defects and reduce the false positive rate.

[0033] This application further proposes a method for applying current excitation to an igniter based on a constant current source. The constant current source is controlled in a closed loop through a four-terminal sampling resistor, ensuring that the current step settling time is no greater than 0.5 milliseconds and the overshoot is no greater than 5%. Based on multi-channel synchronous acquisition, current signals, voltage signals, surface thermal radiation signals, and bridge wire temperature signals are simultaneously acquired under a unified time reference, with a sampling rate of no less than two million times per second and a channel time error of no more than one hundred nanoseconds. Specifically, the current signal is acquired through a four-terminal sampling resistor and differential amplification, the voltage signal is acquired using a high input impedance differential method, the surface thermal radiation signal is acquired by an infrared temperature sensor, and the bridge wire temperature signal is acquired by a contact temperature sensor or a fiber optic temperature sensor.

[0034] The constant current source employs a four-terminal sampling resistor structure, achieving rapid current step establishment and low overshoot characteristics through closed-loop control, ensuring the stability of the excitation signal. Multi-channel synchronous acquisition ensures signal time synchronization through a unified time base and high sampling rate, eliminating time deviation between channels. Current signals are acquired through four-terminal resistors and differential amplification, reducing the impact of contact resistance and line noise. Voltage signals are acquired using a high input impedance differential method, avoiding the impact of load effects on measurement accuracy. Surface thermal radiation signals and bridge wire temperature signals employ non-contact infrared sensors and contact or fiber optic sensors, respectively, to meet the measurement needs of different temperature signals.

[0035] Specifically, during the current excitation application phase, the closed-loop control of the four-terminal sampling resistor ensures that the current establishes a step within 0.5 milliseconds, while keeping the overshoot within 5% to avoid thermal shock to the ignition bridge wire from overshoot current. Multi-channel synchronous acquisition uses a unified trigger signal as a reference, synchronously capturing current, voltage, and temperature signals at a sampling rate of over two million times per second. The time error between channels is limited to the order of hundreds of nanoseconds, ensuring synchronous recording of transient processes. The current signal eliminates lead resistance errors through a four-terminal resistor structure and achieves a high signal-to-noise ratio after differential amplification. The voltage signal suppresses common-mode interference through a high-input-impedance differential circuit, maintaining measurement accuracy. The surface thermal radiation signal is measured non-contactly by an infrared sensor, avoiding interference with the ignition structure. The bridge wire temperature signal is directly acquired through a contact or fiber optic sensor, improving the spatial resolution of temperature measurement. Through the combination of these techniques, rapid and stable output of the excitation signal and high-precision synchronous acquisition of multi-mode signals are achieved.

[0036] As a preferred embodiment, the solution of this application is specifically implemented as follows: A constant current source applies current excitation to the igniter, and the constant current source is controlled in a closed loop through a four-terminal sampling resistor. The current step settling time is set to 0.3 milliseconds, and the overshoot is controlled within 3%. Multi-channel synchronous acquisition uses a unified time base, simultaneously acquiring current signals, voltage signals, surface thermal radiation signals, and bridge wire temperature signals. The sampling rate is set to 5 million times per second, and the channel time error is controlled within 50 nanoseconds.

[0037] Current signals are acquired via a four-terminal sampling resistor and a differential amplifier circuit. A 0.1-ohm precision sampling resistor is used, and the differential amplifier circuit gain is set to 10. Voltage signals are acquired using a high-input-impedance differential amplifier circuit with an input impedance greater than 10 megohms. Surface thermal radiation signals are acquired using an infrared temperature sensor, employing a mid-wave infrared detector with a wavelength of 3-5 micrometers and a temperature resolution better than 0.1 degrees Celsius. Bridge wire temperature signals are acquired using a fiber optic temperature sensor with a probe diameter less than 0.5 mm and a response time less than 1 millisecond.

[0038] Through the above technical solutions, this application achieves high-precision synchronous acquisition of multi-channel response signals from ignition devices. Closed-loop control of the constant current source ensures rapid establishment and stable output of the current excitation, providing a reliable excitation source for thermal transient process analysis. High-speed synchronous acquisition across multiple channels guarantees the time consistency of current, voltage, thermal radiation, and temperature signals. The high input impedance differential acquisition method suppresses common-mode interference and improves the acquisition accuracy of weak signals. The application of infrared and fiber optic temperature sensors enables non-contact rapid temperature measurement of the ignition device surface and bridge wire, avoiding interference from traditional contact temperature measurement methods on thermal transient processes.

[0039] This application further proposes a preprocessing method, including using the rising edge of the current signal as an alignment marker to perform time alignment on the current signal, voltage signal, surface thermal radiation signal, and bridge wire temperature signal; suppressing broadband noise and pulse interference based on band-limited filtering, and extracting baseline offset for baseline correction using the static stable section at the beginning of acquisition; normalizing the amplitude of the current channel and voltage channel according to the channel gain coefficient, and normalizing the amplitude of the surface thermal radiation signal and bridge wire temperature signal according to the temperature sensor calibration coefficient, and unifying the normalization results to the same amplitude range; using the start of a unified time window as the moment when the current rising edge reaches the current set ratio, and the end point as the moment when the energy decays to the energy set ratio after the bridge wire temperature peak, synchronously dividing the response signal into a thermal transient signal sequence.

[0040] Among these features, time alignment is achieved by triggering multi-channel synchronization via the rising edge of the current, eliminating differences in sensor response delay; band-limited filtering uses a filter whose cutoff frequency matches the signal bandwidth to suppress high-frequency noise and low-frequency drift; baseline correction uses the statically stable section to calculate the initial offset, eliminating sensor zero-point drift; amplitude normalization converts current and voltage to standard units and temperature signals to degrees Celsius based on the gain coefficients of each channel and the sensor calibration coefficients; and a unified time window dynamically adjusts the window length based on the rising edge of the current and temperature decay to ensure complete coverage of the thermal transient process.

[0041] Specifically, in the preprocessing stage, the rising edge of the current triggers multi-channel synchronous acquisition. The signals of each channel are aligned using rising edge markers to eliminate phase deviations caused by differences in sensor response time. A band-limited filter is designed with a passband range based on the signal's spectral characteristics, filtering out high-frequency noise and power frequency interference while retaining effective signal components. The static steady-state section is selected from the stable phase before power-on, calculating the baseline offset of each channel and performing subtraction correction to eliminate sensor zero-point drift. Current and voltage signals are proportionally converted based on the sampling resistor value and amplifier gain, while the temperature signal is converted into a physical quantity based on the sensor's sensitivity coefficient, achieving unified amplitude across multiple channels. The starting point of the time window is set as the trigger point when the current reaches a set proportion, and the ending point is determined based on the temperature decaying to a set proportion, dynamically capturing the complete thermal transient signal. Through these steps, it is ensured that the multi-channel signals achieve a unified standard in terms of time reference, amplitude range, and noise level.

[0042] As a preferred embodiment, the solution of this application is specifically implemented as follows: The preprocessing of the response signal and its segmentation into a thermal transient signal sequence within a unified time window includes the following steps: First, using the rising edge of the current signal as an alignment marker, time alignment is performed on the current signal, voltage signal, surface thermal radiation signal, and bridge wire temperature signal. Specifically, interpolation algorithms can be used to align the sampling points of each signal to the same time axis, ensuring time synchronization between signals. Second, noise suppression is performed. Band-limited filters are used to suppress broadband noise and impulse interference; the filter bandwidth can be designed according to the frequency characteristics of the signal. Simultaneously, baseline offset is extracted from the initial stable acquisition segment, and baseline correction is performed to eliminate DC bias. Third, signal amplitude normalization is performed. The amplitudes of the current and voltage channels are normalized according to their respective channel gain coefficients to ensure consistent amplitude ranges. The surface thermal radiation signal and bridge wire temperature signal are normalized according to the calibration coefficients of the temperature sensor, converting the temperature signals into a unified temperature unit. Finally, all normalization results are unified to the same amplitude range for easier subsequent analysis and comparison. Finally, signal segmentation is performed. The unified time window begins when the current rises to 90% of the rated current and ends when the energy decays to 10% of the peak energy after the bridge wire temperature peaks. Within this time window, all response signals are synchronously segmented into a thermal transient signal sequence. Through the above technical solutions, this application achieves precise preprocessing of multi-channel response signals from ignition devices. Time alignment ensures the synchronization of each signal, noise suppression and baseline correction improve signal quality, and amplitude normalization makes signals of different physical quantities comparable. The division of a unified time window focuses on the key stages of the thermal transient process. This preprocessing method enhances the quality and analyzability of thermal transient signals, helping to accurately capture the electrothermal coupling characteristics and potential anomalies of ignition devices.

[0043] In some of the above-mentioned schemes of this application, when constructing multi-dimensional feature vectors, since the thermal transient signal of the ignition device has nonlinear and multi-stage characteristics, it is difficult to accurately characterize the thermal response differences of different anomaly types by using only a single scale or simple time-domain parameters, resulting in insufficient feature discrimination and affecting the accuracy of subsequent classification.

[0044] This application further proposes to divide the thermal transient signal sequence into a rise phase, a peak phase, and an initial decay phase according to scale; in the decomposition results, the scale with the highest energy concentration is used as the principal scale, and the thermal response energy peak is determined according to the principal scale energy peak; the rise time is determined by the time difference between the first moment when the energy accumulation reaches the energy rise ratio within the rise phase and the start of the unified time window; the energy decay time is determined by the time difference between the moment when the energy decreases to the energy decrease ratio after the peak phase and the peak moment; the thermal hysteresis time is determined by the time difference between the peak moment of the temperature-related signal and the corresponding moment of the rising edge of the current signal; the thermal response energy peak, rise time, energy decay time, thermal hysteresis time, principal scale position, peak symmetry, and energy distribution width are used to construct a multi-dimensional feature vector; in generating the multi-dimensional feature vector, abnormal segments are removed according to the noise sensitivity threshold, and the thermal response intensity index and thermal stability index are calculated according to the ambient temperature range.

[0045] The division into the rising phase, peak phase, and initial decay phase is based on the energy distribution characteristics of multi-scale time-frequency analysis, with phase boundaries identified through signal decomposition at different scales. The principal scale position is determined by comparing the energy concentration at each scale, for example, selecting the scale with the largest energy modulus after wavelet transform. The energy rise ratio is set to 90%, and the energy fall ratio is set to 10%, with the rise time and energy decay time determined by the cumulative energy ratio threshold. Peak symmetry is obtained by calculating the energy ratio of the first and second halves of the peak, and the energy distribution width is calculated using the half-width at half-maximum (WHM) of the signal duration at the principal scale. The noise sensitivity threshold is set based on three times the standard deviation of the signal baseline noise; abnormal segments refer to signal fluctuation areas exceeding this threshold.

[0046] Specifically, during signal decomposition, multi-scale time-frequency analysis decomposes the thermal transient signal into different frequency band components. The principal scale with the highest energy concentration corresponds to the core frequency band of thermal conduction in the bridge wire material. The rise phase is defined as the interval from the start of current excitation to the principal scale energy reaching its peak value. The peak phase covers the interval where energy maintains high-level fluctuations, and the initial decay phase corresponds to the interval where energy decreases from the peak value to a stable value. The thermal hysteresis time is calculated by comparing the time difference between the current rise edge and the temperature signal peak value, reflecting the electrothermal conversion efficiency. In feature vector construction, the principal scale position and energy distribution width jointly characterize the frequency domain characteristics of the thermal response, and peak shape symmetry is used to identify abnormal decay modes caused by uneven charge or bridge wire breakage. The combination of noise sensitivity threshold and temperature range compensation eliminates the influence of environmental interference and batch differences on feature stability. For example, temperature compensation is achieved by adjusting the weighting coefficient of the thermal response intensity index in high-temperature environments. Through the joint analysis of the above multi-dimensional features, it is possible to distinguish between prolonged rise time caused by contact abnormalities, peak shape asymmetry caused by charge abnormalities, and abnormal thermal hysteresis time caused by bridge wire abnormalities, thereby improving the accuracy of classification.

[0047] As a preferred embodiment, the solution of this application is specifically implemented as follows: The thermal transient signal sequence is divided into a rising phase, a peak phase, and an initial decay phase according to scale. In the decomposition results, the scale with the highest energy concentration is used as the principal scale, and the peak energy of the thermal response is determined according to the energy peak of the principal scale.

[0048] The rise time is determined by the time difference between the first moment when the accumulated energy reaches the energy rise ratio during the rise phase and the start of the unified time window. The energy decay time is determined by the time difference between the moment when the energy decreases to the energy decrease ratio after the peak phase and the peak moment. The thermal hysteresis time is determined by the time difference between the peak moment of the temperature-related signal and the corresponding moment of the rising edge of the current signal.

[0049] The peak thermal response energy, rise time, energy decay time, thermal hysteresis time, principal scale location, peak shape symmetry, and energy distribution width are collectively used to construct a multidimensional feature vector. In generating this multidimensional feature vector, outlier segments are removed based on a noise sensitivity threshold, and the thermal response intensity index and thermal stability index are calculated separately for each ambient temperature range.

[0050] Specifically, the thermal transient signal sequence can be divided into a rising phase, a peak phase, and an initial decay phase according to the rate of energy change. Wavelet decomposition is performed on the signal sequence, and the scale with the highest energy concentration is selected as the principal scale. The energy peak value of the principal scale is taken as the thermal response energy peak value.

[0051] Rise time can be defined as the time difference between when the accumulated energy reaches 80% of its peak and the initial moment. Energy decay time can be defined as the time difference between when the energy decreases to 50% of its peak and the peak moment. Thermal hysteresis time can be defined as the time difference between the peak value of the temperature signal and the rise edge of the current.

[0052] In constructing the feature vector, in addition to the aforementioned temporal features, the principal scale location, peak symmetry coefficient, and energy distribution width are also included. A noise sensitivity threshold is set empirically to remove outlier data segments. The thermal response intensity index and thermal stability index are calculated according to the ambient temperature range to eliminate the influence of temperature.

[0053] Through the above technical solution, this application achieves a comprehensive characterization of the thermal transient response characteristics of ignition devices. By extracting key feature parameters through multi-scale time-frequency analysis and constructing a multi-dimensional feature vector, the thermal response characteristics of the ignition device can be accurately reflected. Using principal-scale parameters such as peak energy and rise time, the thermal response intensity and speed of the ignition device can be characterized. Introducing thermal hysteresis time reflects the electrothermal coupling characteristics of the ignition device. By eliminating abnormal segments and temperature compensation, the stability and reliability of feature extraction are improved. The construction of the multi-dimensional feature vector helps improve the accuracy of ignition device performance evaluation.

[0054] In some of the above-mentioned schemes in this application, during the process of fitting the thermal response curve, the difference in sampling time reference of different signal channels and noise interference may lead to inaccurate selection of the fitting interval, which in turn affects the accuracy of the calculation of the correspondence between the temperature rise rate and the power change rate, and fails to reflect the electrothermal coupling characteristics.

[0055] This application further proposes to select the time synchronization segment of the current signal, voltage signal and bridge wire temperature signal as the fitting interval during the power-on process, and to perform thermal response curve fitting with the temperature rise segment corresponding to the current change as the core, so as to obtain the correspondence between the temperature rise rate and the power change rate calculated from the current signal and voltage signal; and to adjust the curve smoothness and fitting accuracy according to the energy balance constraint during the fitting process.

[0056] The time synchronization section aligns the sampled data of each channel by triggering pulses to eliminate transmission delay errors between channels; the temperature rise section is selected based on the range from the rising edge of the current signal to the peak point of the bridge wire temperature signal; the energy balance constraint dynamically adjusts the regularization parameters of the fitting algorithm by calculating the difference between the input electrical energy and the heat dissipation energy, and balances the smoothness of the curve and the residual convergence threshold.

[0057] Specifically, after the current excitation is applied, signals from each channel are synchronously acquired via trigger pulses to ensure time-domain alignment of current, voltage, and bridge wire temperature data. During the temperature rise phase, the interval where the current change rate exceeds a set threshold is used as the core analysis window, extracting the raw data pairs of power change rate and temperature rise rate within this window. A power-temperature change rate relationship model is established using the least squares method. During iteration, the regularization coefficient is automatically adjusted based on the difference between the input electrical energy integral and the heat dissipation energy integral to suppress overfitting. When the energy difference exceeds the allowable range, the curve smoothness is preferentially reduced to improve residual convergence accuracy and ensure the accuracy of the equivalent heat capacity parameters. This process achieves a balance between fitting accuracy and computational efficiency through a dynamic constraint mechanism.

[0058] As a preferred embodiment, the solution of this application is specifically implemented as follows: When fitting the thermal response curve to a multidimensional feature vector, the time synchronization segment of the current signal, voltage signal, and bridge wire temperature signal during the energizing process is first selected as the fitting interval. Specifically, the thermal response curve is fitted with the temperature rise segment corresponding to the current change as the core. The least squares method is used in the fitting process, with the current signal as the independent variable and the bridge wire temperature signal as the dependent variable, to establish a quadratic polynomial fitting model.

[0059] Furthermore, the correlation between the rate of temperature rise and the rate of power change calculated from the current and voltage signals is obtained by fitting a model. The rate of temperature rise is calculated using the first derivative of the fitted curve, while the rate of power change is represented by the derivative of the product of the current and voltage signals with respect to time.

[0060] Therefore, during the fitting process, the smoothness and fitting accuracy of the curve are adjusted according to the energy balance constraint. Specifically, a regularization term is introduced to control the smoothness of the fitted curve, while iterative optimization ensures that the fitted curve satisfies the principle of energy conservation, i.e., the input electrical energy and the heat energy caused by temperature changes remain in balance. For example, the Tikhonov regularization method can be used, adding the square integral term of the second derivative of the curve to the objective function, and adjusting the regularization parameter to achieve a balance between curve smoothness and fitting accuracy.

[0061] As a preferred implementation, a segmented fitting strategy can be adopted during the fitting process, dividing the temperature rise stage according to the temperature change rate: when the temperature rise rate is higher than a preset rapid rise threshold, it is determined to be the initial rapid rise stage; when the temperature rise rate is lower than or equal to the threshold, it is determined to be the subsequent slow rise stage; and the two stages are independently fitted using an adapted function model to more accurately characterize the thermal response characteristics of different stages.

[0062] Through the above technical solutions, this application achieves accurate modeling of the thermal response process of ignition devices. By selecting the synchronous segment of key signals as the fitting interval, focusing on the temperature change process caused by current changes, the specificity and accuracy of the fitting are improved. Introducing energy balance constraints ensures that the fitting results conform to physical laws, avoiding overfitting. Employing a piecewise fitting strategy better characterizes the thermal response characteristics at different stages, improving the model's adaptability. Therefore, it helps improve the accuracy of ignition device performance evaluation and fault diagnosis.

[0063] This application further proposes to calculate the proportional relationship between the energy accumulation rate and the temperature rise rate based on the fitting results of the thermal response curve, and to determine the equivalent heat capacity parameter; to determine the thermal conductivity parameter based on the difference in time constants between the temperature rise stage and the decay stage; and to determine the temperature-related resistance change parameter based on the degree of hysteresis between the current signal and the temperature-related signal.

[0064] The energy accumulation rate is obtained by measuring the change in the product of the integrated current signal and the voltage signal within the time synchronization period; the temperature rise rate is calculated using the first derivative of the bridge wire temperature signal; the time constant difference is quantified by exponentially fitting the slope ratio of the temperature rise and decay phases; the hysteresis is determined by cross-correlation analysis of the time difference between the rising edge of the current signal and the peak value of the temperature-related signal. The equivalent heat capacity parameter is calculated by linearly regressing the energy accumulation rate and the temperature rise rate using the least squares method, with the proportionality coefficient representing the equivalent heat capacity value; the thermal conductivity parameter is obtained by establishing a function of the difference between the time constant of the temperature rise phase and the time constant of the decay phase, combined with the inversion of the heat conduction equation; the temperature-related resistance change parameter is derived by constructing an empirical formula using the product of the hysteresis time and the rate of change of the temperature signal.

[0065] Specifically, during the energizing process, the time synchronization segment of the current signal, voltage signal, and bridge wire temperature signal is selected as the fitting interval, and the thermal response curve is fitted with the temperature rise segment corresponding to the current change as the core. The ratio of energy accumulation rate to temperature rise rate is obtained through linear regression to obtain the slope coefficient, which directly reflects the equivalent heat capacity parameter and eliminates the fitting error caused by nonlinear heat capacity. The time constant of the temperature rise phase is determined by the exponential fitting time constant of the temperature signal rise edge, and the time constant of the decay phase is determined by the exponential fitting time constant of the temperature signal fall edge. The difference between the two is used to obtain the thermal conductivity parameter through inversion of the heat conduction equation, improving the accuracy of thermal conductivity characteristic assessment. The lag time between the current signal rise edge and the temperature-related signal peak is accurately measured through cross-correlation analysis, and the temperature-related resistance change parameter is calculated in combination with the temperature signal change rate, quantifying the dynamic characteristics of resistance with temperature change. Through the joint calculation of the above parameters, accurate modeling of the electrothermal coupling relationship is achieved, improving the reliability of diagnostic classification.

[0066] As a preferred embodiment, the solution of this application is specifically implemented as follows: When determining the electrothermal coupling parameters, the proportional relationship between the energy accumulation rate and the temperature rise rate is calculated based on the thermal response curve fitting results to determine the equivalent heat capacity parameter. Specifically, during the thermal response curve fitting process, the energy accumulation curve and temperature change curve during the temperature rise segment are obtained. The slopes of these two curves are calculated to obtain the energy accumulation rate and the temperature rise rate. Dividing the energy accumulation rate by the temperature rise rate yields the proportional relationship value. This proportional relationship value is the equivalent heat capacity parameter, reflecting the ignition device's ability to absorb heat.

[0067] Furthermore, the thermal conductivity parameter is determined based on the difference in time constants between the temperature rise and decay phases. In practice, the time constants for the temperature rise and decay phases are calculated separately. The time constant for the temperature rise phase is defined as the time required for the temperature to reach 63.2% of its final steady-state value. The time constant for the decay phase is defined as the time required for the temperature to drop to 36.8% of the difference between the initial and final values. The ratio of the decay phase time constant to the rise phase time constant is used as the thermal conductivity parameter to reflect the heat dissipation characteristics of the ignition device.

[0068] Therefore, the temperature-dependent resistance change parameter is determined based on the hysteresis between the current signal and the temperature-related signal. Specifically, the time difference between when the current signal reaches a set threshold and when the temperature-related signal (such as the bridge wire temperature signal) reaches the corresponding threshold is calculated. This time difference is divided by the rise time of the current signal to obtain the normalized hysteresis value. This value is the temperature-dependent resistance change parameter, reflecting the sensitivity of the ignition fixture's resistance to temperature changes.

[0069] Through the above technical solution, this application achieves a precise characterization of the electrothermal coupling characteristics of the ignition device. By using equivalent heat capacity parameters, thermal conductivity parameters, and temperature-dependent resistance variation parameters, the thermal response characteristics of the ignition device are comprehensively characterized. These parameters reflect the energy absorption, heat dissipation, and resistance change behavior of the ignition device during energization. This improves the accuracy and comprehensiveness of the ignition device performance evaluation.

[0070] In some of the above-mentioned schemes in this application, the original calculated values ​​of equivalent heat capacity parameters, thermal conductivity parameters and temperature-related resistance change parameters may be subject to systematic deviations due to structural differences in different batches of igniters (e.g., changes in bridge wire material thickness) or fluctuations in charge density, which reduces the horizontal comparability of steady-state features in multidimensional feature vectors and affects the generalization ability of the classification and judgment model.

[0071] This application further proposes to standardize the equivalent heat capacity parameter, thermal conductivity parameter and temperature-related resistance change parameter according to a preset weight ratio to eliminate the deviation caused by the difference in structure and charge of different batches of igniters; and to incorporate the standardized steady-state characteristics into the multi-dimensional feature vector.

[0072] The standardization process involves normalizing each parameter using preset weight ratios, determined based on the statistical distribution of parameters from historical batches. For example, the weighting coefficient for the equivalent heat capacity parameter is set to 0.4, for the thermal conductivity parameter to 0.3, and for the temperature-dependent resistance variation parameter to 0.3. Each parameter is linearly scaled and mapped to a unified numerical range. During standardization, the normalization threshold is dynamically adjusted using a moving average algorithm by calculating the relative offset between the current batch parameters and the baseline batch parameters. The standardized parameters complement the dynamic features such as the peak thermal response energy and rise time obtained from multi-scale time-frequency analysis, enhancing the robustness of the feature vector.

[0073] Specifically, in the standardization stage, the equivalent heat capacity parameter is first baseline-calibrated. Using the average equivalent heat capacity of the benchmark batch of igniters as the center point, the current parameter value is divided by this average and multiplied by a weighting coefficient to obtain the normalized equivalent heat capacity parameter. The thermal conductivity parameter is smoothed using an exponential smoothing method to eliminate fluctuations caused by batch-to-batch differences in charge density. The original thermal conductivity value is input into a preset exponential function model, and the smoothed standardized thermal conductivity parameter is output. The temperature-related resistance variation parameter is calculated using a piecewise linear interpolation method. The slope of the resistance variation curve is adjusted according to the ratio between the current ambient temperature and the benchmark temperature, and then multiplied by the corresponding weighting coefficient. After standardization, the three parameters are merged into the end of the multidimensional feature vector in a fixed order, forming a composite feature set containing both dynamic and steady-state features. This process eliminates parameter offsets caused by batch differences, ensuring the comparability of feature vectors from different production batches of igniters under the same judgment rules, thus improving the diagnostic accuracy and stability of the classification model.

[0074] As a preferred embodiment, the solution of this application is specifically implemented as follows: Obtaining steady-state features and incorporating them into a multi-dimensional feature vector includes the following steps: First, the equivalent heat capacity parameter, thermal conductivity parameter, and temperature-dependent resistance change parameter are standardized according to preset weight ratios. Specifically, the equivalent heat capacity parameter is weighted at 0.4, the thermal conductivity parameter at 0.3, and the temperature-dependent resistance change parameter at 0.3. This standardization process eliminates deviations caused by differences in the structure and propellant charge of igniters from different batches.

[0075] Secondly, the standardized steady-state features are incorporated into the multidimensional feature vector. Specifically, the standardized equivalent heat capacity parameter, thermal conductivity parameter, and temperature-dependent resistance change parameter are added as new feature terms to the original multidimensional feature vector. For example, the original multidimensional feature vector includes features such as thermal response energy peak value, rise time, energy decay time, and thermal hysteresis time. Now, the three standardized steady-state features are incorporated to form an extended multidimensional feature vector.

[0076] Through the above technical solution, this application achieves accurate characterization and quantification of the steady-state characteristics of ignition devices. Therefore, ignition devices of different batches and models can be compared and analyzed in a unified feature space, improving the accuracy and reliability of thermal transient diagnosis. Furthermore, by incorporating steady-state characteristics into a multi-dimensional feature vector, a comprehensive analysis of dynamic and steady-state characteristics is achieved, which helps to improve the confidence level of the diagnostic results.

[0077] In some of the solutions described above in this application, the multidimensional feature vector after incorporating steady-state features may still be affected by ambient temperature fluctuations and channel drift, leading to biases in the judgment data. For example, temperature changes cause sensor signal amplitude drift, and the zero-point offset, gain error, and time base differences of different channels are not corrected, resulting in inaccurate feature parameter extraction.

[0078] This application further proposes to collect ambient temperature data, perform temperature compensation and channel calibration on the multidimensional feature vector after incorporating steady-state features, and obtain judgment data. The channel calibration includes gain correction, zero-point correction and time base correction.

[0079] The ambient temperature data is collected in real time by temperature sensors and smoothed using a moving average algorithm to eliminate random noise interference. Temperature change trends are determined by calculating the slope changes between adjacent sampling points, dividing the data into multiple temperature intervals. Amplitude drift correction uses a linear regression model to establish a mapping relationship between temperature and signal drift, with compensation coefficients calculated by piecewise interpolation based on the temperature interval boundary values. Zero-point correction involves collecting static signals from each channel under zero-load conditions, calculating the mean zero-point offset, and storing it as a correction parameter. Gain correction involves inputting a sinusoidal signal of standard amplitude to each channel and comparing the actual sampled values ​​with the theoretical values ​​to calculate the gain error coefficient. Time base correction uses a synchronous trigger pulse signal, measuring the time delay of each channel relative to the trigger pulse, and adjusting the sampling time series using an interpolation algorithm.

[0080] Specifically, during the test, the temperature sensor collects ambient temperature data ten times per second, and generates a temperature change curve after filtering with a ten-point moving average. Based on the curve slope, the test process is divided into a heating phase, a steady-state phase, and a cooling phase. An independent linear interpolation model is used to calculate the compensation coefficient within each phase. The amplitude drift of the current and voltage signals is corrected in real time according to the temperature compensation coefficient, while the surface thermal radiation signal and bridge wire temperature signal are compensated according to the sensor's temperature-voltage calibration curve. For zero-point correction, the connection to the tested firearm is disconnected, and the baseline signal of each channel is collected under zero-input conditions. The average value of one hundred samples is calculated as the zero-point offset. For gain correction, a standard ampere current is input to the current channel, a standard volt voltage is connected to the voltage channel, and an equivalent temperature analog signal is input to the temperature channel. The gain adjustment coefficient is calculated based on the ratio of the actual sampled value to the theoretical value. For time reference calibration, rising edge trigger pulses are sent synchronously to all channels, and the time deviation of each channel signal relative to the trigger pulse is measured. If the deviation exceeds one hundred nanoseconds, the sampling timestamp is adjusted using an interpolation algorithm to ensure that the time synchronization accuracy of the multi-channel data reaches fifty nanoseconds. After temperature compensation and channel calibration, the peak error of thermal response energy in the multidimensional feature vector is reduced to ±2%, the rise time measurement accuracy is improved to ±0.1 milliseconds, and the time synchronization error between channels is controlled within ±50 nanoseconds, eliminating the influence of environmental interference and hardware drift on the judgment data.

[0081] As a preferred embodiment, the specific implementation of this application is as follows: During the ignition device test, ambient temperature data is collected in real time using a high-precision temperature sensor. The original temperature data is smoothed using a moving average algorithm, and a temperature change trend curve is extracted within each 10-second time window. Based on this trend curve, the test process is divided into five temperature fluctuation intervals. Linear regression analysis is performed on the current signal amplitude drift in each interval to generate a temperature-drift mapping table. In the compensation stage, the compensation coefficient of each channel is calculated using cubic spline interpolation, and a dynamic compensation factor is applied to the feature term corresponding to the bridge wire temperature signal in the multidimensional feature vector. During channel calibration, zero-point offset data of each channel is continuously collected under zero-load conditions. A linear change model of the zero-point offset over time is fitted using the least squares method to establish a zero-point correction matrix. The gain deviation of each channel is measured by inputting a standard square wave signal, and a gain correction coefficient table is constructed. Time base correction adopts trigger pulse synchronization technology. By measuring the time delay of each channel relative to the trigger pulse, a time offset compensation sequence between channels is generated.

[0082] Through the above technical solution, this application eliminates the interference of environmental temperature gradient changes on thermal response characteristic parameters, suppresses zero-point drift and gain deviation caused by long-term operation of multi-channel acquisition, and solves the misjudgment problem caused by temperature sensitivity and channel asymmetry in traditional methods. This solution, through dynamic compensation and real-time calibration mechanisms, ensures the consistency of multi-dimensional feature vectors under different batch test conditions, enabling diagnostic results to have cross-environmental stability and repeatability, and reducing the false alarm rate caused by environmental factors.

[0083] In some of the solutions described above in this application, when collecting ambient temperature data and performing temperature compensation and channel calibration on the multidimensional feature vector after incorporating steady-state features, there are problems such as signal amplitude drift, channel zero-point offset, gain error, and time reference deviation caused by changes in ambient temperature. These factors will reduce the accuracy of the judgment data and thus affect the reliability of the diagnostic results.

[0084] This application further proposes to acquire ambient temperature data in real time during the test, and to smooth the ambient temperature data to obtain the temperature change trend; to correct the amplitude drift of the current signal, voltage signal, surface thermal radiation signal and bridge wire temperature signal according to the temperature change trend, to calculate the compensation coefficient using piecewise linear interpolation and to compensate the temperature-related feature terms in the multidimensional feature vector; during channel calibration, to determine the channel zero-point offset and perform zero-point correction by static sampling under zero-load conditions, to determine the channel gain and perform gain correction by standard signal input, and to correct the sampling time difference of each channel according to a unified trigger pulse to complete the time reference correction.

[0085] The system employs a temperature sensor to acquire ambient temperature data in real time at a sampling frequency of ten times per second. Smoothing is achieved through a moving average algorithm, with the average temperature change trend calculated in five-second windows. Piecewise linear interpolation divides the test process into multiple temperature intervals based on the temperature change trend, calculating a compensation coefficient within each interval. This compensation coefficient is determined based on the linear relationship between temperature and signal amplitude. Static sampling under zero-load conditions is performed before testing, collecting the output values ​​of each channel under no-excitation conditions as zero-point offsets. The standard signal input uses a DC signal or square wave signal with known amplitude, and the gain correction coefficient is determined by the ratio of the actual output to the standard signal. A unified trigger pulse is generated by a synchronous controller, and each channel adjusts the sampling clock phase according to the arrival time difference of the trigger pulse, achieving a time reference correction accuracy within ten nanoseconds.

[0086] Specifically, during the test, the temperature sensor collects ambient temperature data in real time. A moving average algorithm filters out random fluctuations in the raw temperature data, generating a smooth temperature change curve. Based on this curve, the test process is divided into multiple temperature intervals, for example, one interval per degree Celsius. Within each interval, temperature compensation coefficients for current, voltage, thermal radiation, and bridge wire temperature signals are calculated. The compensation coefficients are predetermined through calibration experiments. During real-time compensation, linear interpolation is performed from the compensation coefficients of two adjacent temperature points based on the current temperature interval to obtain the correction value at the current temperature, dynamically adjusting the temperature-related feature terms in the multi-dimensional feature vector. During channel calibration, zero-point correction uses the output values ​​of each channel under zero load collected before the test as the zero-point offset, which is subtracted from subsequent data. Gain correction adjusts the gain coefficient based on the ratio of the actual sampled value to the theoretical value by inputting a standard signal to each channel, ensuring consistent sensitivity across channels. Time base correction uses a synchronous controller to send a unified trigger pulse, and each channel records the deviation between the pulse arrival time and the sampling time, adjusting the sampling clock phase to control the time alignment error of multi-channel data to the tens of nanoseconds level.

[0087] As a preferred embodiment, the solution of this application is implemented as follows: A multi-classification model based on support vector machines is deployed in the judgment unit. This model is trained using judgment data from over three thousand sets of ignition devices under different operating conditions in a historical sample database. The historical sample database includes four categories: normal ignition devices, ignition devices with poor contact, ignition devices with loose charge, and ignition devices with oxidized bridge wire. Each sample contains a multi-dimensional feature vector after temperature compensation and channel calibration, along with its manually verified label. The classification model maps the multi-dimensional feature vector to a high-dimensional space using a kernel function, and calculates the distance between each sample and the classification hyperplane to achieve category division. The confidence level is obtained by calculating the Mahalanobis distance between the current judgment data and the center point of each category, combined with the weighted Pearson correlation coefficient between each feature parameter. When the confidence level is below 85% or two different conclusions are obtained in three consecutive measurements, a review process is automatically triggered: the sampling rate is increased to five million times per second for repeated collection, and the newly added data is input into the online incremental learning module to update the classification model. At the same time, the logical consistency between the review conclusion and the original conclusion is checked.

[0088] Through the above technical solution, this application solves the problem of misclassification caused by environmental temperature drift and channel delay differences in traditional methods. It improves the accuracy of anomaly type identification through multi-feature joint classification and confidence assessment mechanisms. Simultaneously, the combination of a review process and an online learning mechanism ensures reliable determination of abnormal states and enables adaptive optimization of the classification model, ensuring stable diagnostic performance during long-term operation.

[0089] In the above embodiments, by simultaneously acquiring current signals, voltage signals, surface thermal radiation signals, and bridge wire temperature signals during the power-on excitation phase, a multi-modal acquisition system integrating electrical and thermal signals is constructed. This system can capture the entire process of the ignition device from energy input to thermal response under a unified time reference. After time alignment, noise suppression, and amplitude normalization, the multi-channel data is segmented into a thermal transient signal sequence. Multi-scale time-frequency analysis is used to extract key dynamic features such as energy peak value, rise time, decay time, and thermal hysteresis time, thereby achieving a quantitative characterization of the thermal response intensity and stability. By fitting the electrothermal coupling curve through the synchronous change law of current signals and temperature-related signals, steady-state features such as equivalent heat capacity, thermal conductivity, and temperature-related resistance changes are extracted. Combined with temperature compensation and channel calibration mechanisms, the comparability and stability of data under different environmental and equipment conditions are ensured. High-confidence diagnostic results are output through multi-dimensional feature fusion and classification judgment. This breaks through the traditional detection mode that relies solely on electrical signal evaluation, realizing refined dynamic analysis and defect type identification of the ignition device's thermal transient process, and improving the reliability and environmental adaptability of ignition device performance evaluation.

[0090] In another preferred embodiment based on the above embodiments, see [reference] Figure 2 As shown, this embodiment provides a thermal transient diagnosis and classification system for automotive ignition devices, used to apply the above-described automotive ignition thermal transient diagnosis and classification method, including: The acquisition unit is configured to apply current excitation to the ignition device of the vehicle under test, and simultaneously acquire the response signal of the ignition device based on multi-channel synchronous acquisition. The response signal includes current signal, voltage signal, surface thermal radiation signal and bridge wire temperature signal. The preprocessing unit is configured to preprocess the response signal and divide the preprocessed response signal into a thermal transient signal sequence within a unified time window. The preprocessing includes time alignment, noise suppression, and amplitude normalization. The extraction unit is configured to perform multi-scale time-frequency analysis on the thermal transient signal sequence, obtain feature parameters, and obtain thermal response intensity and stability index based on the feature parameters to form a multi-dimensional feature vector. The feature parameters include thermal response energy peak value, rise time, energy decay time and thermal hysteresis time. The processing unit is configured to fit the thermal response curve of the multidimensional feature vector according to the synchronous change law of the current signal and the temperature-related signal during the energizing process, determine the electrothermal coupling relationship parameters and obtain steady-state features, and incorporate the steady-state features into the multidimensional feature vector. The steady-state features include equivalent heat capacity parameters, thermal conductivity parameters and temperature-related resistance change parameters. The calibration unit is configured to collect ambient temperature data, perform temperature compensation and channel calibration on the multidimensional feature vector after incorporating steady-state features, and obtain judgment data. Channel calibration includes gain correction, zero-point correction and time base correction. The decision unit is configured to identify and classify decision data, output diagnostic results, and generate confidence levels.

[0091] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for thermal transient diagnosis and classification of automotive ignition devices, characterized in that, include: A current excitation is applied to the ignition device of the vehicle under test, and the response signal of the ignition device is acquired simultaneously based on multi-channel synchronous acquisition. The response signal includes current signal, voltage signal, surface thermal radiation signal and bridge wire temperature signal. The response signal is preprocessed and then divided into a thermal transient signal sequence within a unified time window. The preprocessing includes time alignment, noise suppression, and amplitude normalization. Multi-scale time-frequency analysis is performed on the thermal transient signal sequence to obtain characteristic parameters, and thermal response intensity and stability index are obtained based on the characteristic parameters to form a multi-dimensional feature vector. The characteristic parameters include thermal response energy peak value, rise time, energy decay time and thermal hysteresis time. Based on the synchronous change pattern of the current signal and the temperature-related signal during the energizing process, the multidimensional feature vector is fitted with a thermal response curve to determine the electrothermal coupling parameters and obtain steady-state characteristics. The steady-state characteristics are then incorporated into the multidimensional feature vector. The steady-state characteristics include equivalent heat capacity parameters, thermal conductivity parameters, and temperature-related resistance change parameters. The temperature-related signals include surface thermal radiation signals and bridge wire temperature signals. Collect ambient temperature data, perform temperature compensation and channel calibration on the multidimensional feature vector after incorporating steady-state features, and obtain judgment data. The channel calibration includes gain correction, zero point correction and time base correction. The judgment data is identified and classified, diagnostic results are output, and confidence levels are generated.

2. The method for thermal transient diagnosis and classification of automotive ignition devices according to claim 1, characterized in that, When applying current excitation to the ignition device of the vehicle under test, and simultaneously acquiring the response signal of the ignition device based on multi-channel synchronous acquisition, the process includes: The igniter is excited by a constant current source, which is controlled in a closed loop by a four-terminal sampling resistor to ensure that the current step settling time is no more than 0.5 milliseconds and the overshoot is no more than 5%. Simultaneous acquisition of current, voltage, surface thermal radiation, and bridge wire temperature signals is performed on a unified time reference, with a sampling rate of no less than two million times per second and a channel time error of no more than one hundred nanoseconds. The current signal is acquired via a four-terminal sampling resistor and differential amplification, the voltage signal is acquired using a high input impedance differential method, the surface thermal radiation signal is acquired by an infrared temperature sensor, and the bridge wire temperature signal is acquired by a contact temperature sensor or a fiber optic temperature sensor.

3. The method for thermal transient diagnosis and classification of automotive ignition devices according to claim 1, characterized in that, When preprocessing the response signal and dividing the preprocessed response signal into a thermal transient signal sequence within a unified time window, the process includes: The rising edge of the current signal is used as an alignment marker to perform time alignment on the current signal, voltage signal, surface thermal radiation signal, and bridge wire temperature signal. Band-limited filtering is used to suppress broadband noise and impulse interference, and baseline offset is extracted from the static and stable section at the beginning of the acquisition for baseline correction. The amplitude of the current channel and voltage channel is normalized according to the channel gain coefficient, and the amplitude of the surface thermal radiation signal and bridge wire temperature signal is normalized according to the temperature sensor calibration coefficient. The normalization results are then unified to the same amplitude range. The response signal is synchronously divided into the thermal transient signal sequence, with the starting point of the unified time window being the moment when the current rise edge reaches the current set ratio and the ending point being the moment when the energy decays to the energy set ratio after the bridge wire temperature peaks.

4. The method for thermal transient diagnosis and classification of automotive ignition devices according to claim 1, characterized in that, When constructing a multidimensional feature vector, the following are included: The thermal transient signal sequence is divided into a rising phase, a peak phase, and an initial decay phase according to the scale. In the decomposition results, the scale with the highest energy concentration is taken as the principal scale, and the thermal response energy peak is determined according to the energy peak of the principal scale. The rise time is determined by the time difference between the first moment when the accumulated energy reaches the energy rise ratio during the rise phase and the start of a unified time window; the energy decay time is determined by the time difference between the moment when the energy decreases to the energy decrease ratio after the peak phase and the peak moment; the thermal hysteresis time is determined by the time difference between the peak moment of the temperature-related signal and the moment corresponding to the rising edge of the current signal; the thermal response energy peak, rise time, energy decay time, thermal hysteresis time, principal scale position, peak symmetry, and energy distribution width are combined to form the multidimensional feature vector; in generating the multidimensional feature vector, abnormal segments are removed according to the noise sensitivity threshold, and the thermal response intensity index and thermal stability index are calculated according to the ambient temperature range.

5. The method for thermal transient diagnosis and classification of automotive ignition devices according to claim 1, characterized in that, When fitting the thermal response curve of the multidimensional feature vector, the process includes: selecting the time synchronization segment of the current signal, voltage signal and bridge wire temperature signal as the fitting interval during the energizing process; fitting the thermal response curve with the temperature rise segment corresponding to the current change as the core; obtaining the correspondence between the temperature rise rate and the power change rate calculated from the current signal and voltage signal; and adjusting the curve smoothness and fitting accuracy according to the energy balance constraint during the fitting process.

6. The method for thermal transient diagnosis and classification of automotive ignition devices according to claim 5, characterized in that, Determining the electrothermal coupling parameters includes: calculating the ratio of energy accumulation rate to temperature rise rate based on the thermal response curve fitting results, and determining the equivalent heat capacity parameter; determining the thermal conductivity parameter based on the time constant difference between the temperature rise phase and the decay phase; and determining the temperature-related resistance change parameter based on the degree of hysteresis between the current signal and the temperature-related signal.

7. The method for thermal transient diagnosis and classification of automotive ignition devices according to claim 6, characterized in that, Obtaining steady-state features and incorporating them into the multidimensional feature vector includes: The equivalent heat capacity parameter, thermal conductivity parameter, and temperature-related resistance change parameter are standardized according to a preset weight ratio to eliminate deviations caused by differences in the structure and charge of different batches of igniters; the standardized steady-state features are then incorporated into the multidimensional feature vector.

8. The method for thermal transient diagnosis and classification of automotive ignition devices according to claim 7, characterized in that, When collecting ambient temperature data and performing temperature compensation and channel calibration on the multidimensional feature vector after incorporating steady-state features, the following steps are included: During the test, ambient temperature data is collected in real time, and the ambient temperature data is smoothed to obtain the temperature change trend. Based on the temperature change trend, the amplitude drift of the current signal, voltage signal, surface thermal radiation signal and bridge wire temperature signal is corrected. The compensation coefficient is calculated by piecewise linear interpolation and the temperature-related feature terms in the multidimensional feature vector are compensated. During channel calibration, the channel zero-point offset is determined by static sampling under zero-load conditions and the zero-point correction is performed. The channel gain is determined by standard signal input and the gain correction is performed. The sampling time difference of each channel is corrected according to a unified trigger pulse to complete the time reference correction.

9. The method for thermal transient diagnosis and classification of automotive ignition devices according to claim 8, characterized in that, When identifying and classifying the judgment data, outputting diagnostic results, and generating confidence levels, the process includes: Based on historical samples, the judgment rules are trained and then the judgment data is jointly classified using multiple features to obtain a diagnostic result. The diagnostic result includes at least one of the following: normal state, contact abnormality, charge abnormality, and bridge wire abnormality. While outputting diagnostic results, a confidence level is generated based on the similarity to the feature centers of each category and the coordination between features. When the confidence level is lower than the set confidence level threshold or when there is inconsistency in the judgment between different repeated measurements, a review and repeated collection are performed, and the review results and new samples are included in the training sample library.

10. A thermal transient diagnosis and classification system for automotive ignition devices, used to apply the thermal transient diagnosis and classification method for automotive ignition devices as described in any one of claims 1-9, characterized in that, include: The acquisition unit is configured to apply current excitation to the ignition device of the vehicle under test, and simultaneously acquire the response signal of the ignition device based on multi-channel synchronous acquisition. The response signal includes current signal, voltage signal, surface thermal radiation signal and bridge wire temperature signal. A preprocessing unit is configured to preprocess the response signal and divide the preprocessed response signal into a thermal transient signal sequence within a unified time window. The preprocessing includes time alignment, noise suppression, and amplitude normalization. The extraction unit is configured to perform multi-scale time-frequency analysis on the thermal transient signal sequence, obtain feature parameters, and obtain thermal response intensity and stability index based on the feature parameters to form a multi-dimensional feature vector. The feature parameters include thermal response energy peak value, rise time, energy decay time, and thermal hysteresis time. The processing unit is configured to fit the multidimensional feature vector to a thermal response curve based on the synchronous change pattern of the current signal and the temperature-related signal during the energizing process, determine the electrothermal coupling relationship parameters and obtain steady-state features, and incorporate the steady-state features into the multidimensional feature vector. The steady-state features include equivalent heat capacity parameters, thermal conductivity parameters and temperature-related resistance change parameters. The calibration unit is configured to collect ambient temperature data, perform temperature compensation and channel calibration on the multidimensional feature vector after incorporating steady-state features, and obtain judgment data. The channel calibration includes gain correction, zero-point correction and time base correction. The determination unit is configured to identify and classify the determination data, output diagnostic results, and generate confidence levels.

Images