PEM fuel cell metal bipolar plate carbon-based composite coating defect detection method

Abstract

The invention relates to the technical field of defect detection, and discloses a PEM fuel cell metal bipolar plate carbon-based composite coating defect detection method, which comprises: applying stepped thermal excitation to a bipolar plate carbon-based composite coating to obtain a plurality of temperature stepped points corresponding to the bipolar plate carbon-based composite coating; after each temperature step point is stable, generating an optical response data sequence according to the optical response data under each temperature step point; analyzing a thermal teratogenesis coefficient sequence corresponding to the temperature step point based on the reference optical characteristic data and the optical response data sequence; performing differential processing on the thermal teratogenesis coefficient sequence to obtain a gradient change curve, and analyzing inflection point features in the gradient change curve; and analyzing the defect evolution behavior of the bipolar plate carbon-based composite coating according to the inflection point characteristics, and detecting the defect type and defect grade of the bipolar plate carbon-based composite coating through the defect evolution behavior. According to the invention, the precision of PEM fuel cell metal bipolar plate carbon-based composite coating defect detection can be improved.

Description

Technical Field

[0001] This invention relates to the field of defect detection technology, and in particular to a method for detecting defects in the carbon-based composite coating of PEM fuel cell metal bipolar plates. Background Technology

[0002] The carbon-based composite coating of PEM fuel cell metal bipolar plates is prone to defects such as microcracks and interface peeling during preparation and use. These defects can significantly reduce the efficiency and lifespan of the fuel cell.

[0003] In existing technologies, defect detection methods mainly rely on single optical detection or electrochemical testing, such as visible light imaging or electrochemical impedance spectroscopy. These methods have significant technical limitations: single optical detection cannot effectively identify defects in the interface bonding state, while electrochemical testing cannot accurately locate the defect position, and both are greatly affected by environmental interference, resulting in low defect detection rates and high false positive rates. Summary of the Invention

[0004] This invention provides a method for detecting defects in the carbon-based composite coating of PEM fuel cell metal bipolar plates, the main purpose of which is to solve the problem of low accuracy in detecting defects in the carbon-based composite coating of PEM fuel cell metal bipolar plates.

[0005] To achieve the above objectives, the present invention provides a method for detecting defects in the carbon-based composite coating of a PEM fuel cell metal bipolar plate, comprising:

[0006] Optical scanning is performed on the surface area of ​​the preset bipolar plate carbon-based composite coating at a preset reference temperature to obtain reference optical feature data;

[0007] A stepped thermal excitation was applied to the bipolar plate carbon-based composite coating to obtain multiple temperature step points corresponding to the bipolar plate carbon-based composite coating.

[0008] After stabilization at each temperature step point, an optical response data sequence is generated based on the optical response data at each temperature step point;

[0009] Based on the reference optical feature data and the optical response data sequence, the thermal distortion coefficient sequence corresponding to the temperature step point is analyzed.

[0010] The thermal distortion coefficient sequence is differentiated to obtain a gradient change curve, and the inflection point characteristics in the gradient change curve are analyzed.

[0011] The defect evolution behavior of the bipolar plate carbon-based composite coating is analyzed based on the inflection point characteristics, and the defect type and defect level of the bipolar plate carbon-based composite coating are detected by the defect evolution behavior.

[0012] Optionally, the step of optically scanning the surface area of ​​the preset bipolar plate carbon-based composite coating at a preset reference temperature to obtain reference optical feature data includes:

[0013] The bipolar plate carbon-based composite coating is placed on a temperature control platform, and the temperature control platform is adjusted to the preset reference temperature.

[0014] Based on the reference temperature, the optical scanning unit is controlled to perform a full-coverage scan of the surface area according to a preset serpentine scanning path, while recording the spatial coordinates of each scanning point.

[0015] After the full-coverage scan is completed, the reflected light intensity data of the surface area under multiple target wavelengths are collected;

[0016] The spatial coordinates of each scanning point are correlated and integrated with the reflected light intensity data to generate reference optical feature data.

[0017] Optionally, applying a stepped thermal excitation to the bipolar carbon-based composite coating to obtain multiple temperature step points corresponding to the bipolar carbon-based composite coating includes:

[0018] A nonlinear temperature increment sequence is determined based on the material heat capacity and thermal conductivity of the bipolar plate carbon-based composite coating, and the nonlinear temperature increment sequence is used as the target temperature sequence.

[0019] According to the target temperature sequence, the infrared heating source is controlled to heat the bipolar plate carbon-based composite coating according to a power loading strategy of fast first and slow later.

[0020] During the heating process, the actual temperature of the surface area of ​​the bipolar plate carbon-based composite coating is monitored in real time. When the actual temperature reaches and stabilizes at any step point in the target temperature sequence, the step point is recorded as the temperature step point that has reached stability.

[0021] By traversing all the step parameters in the target temperature sequence, multiple temperature step points are obtained.

[0022] Optionally, after stabilization at each temperature step point, generating an optical response data sequence based on the optical response data at each temperature step point includes:

[0023] After stabilization at each temperature step point, optical response data corresponding to each temperature step point is collected according to the serpentine scanning path and the preset scanning band.

[0024] The optical response data is registered point-by-point with the reference optical feature data based on the spatial coordinates.

[0025] Calculate the reflectance change of each spatial point at each band relative to the reference temperature based on the registered data, and generate a dataset of optical response changes corresponding to temperature step points based on the reflectance change.

[0026] The optical response data sets at all temperature step points are arranged in time sequence according to the temperature step from low to high to generate an optical response data sequence.

[0027] Optionally, the step of analyzing the thermally induced distortion coefficient sequence corresponding to the temperature step point based on the reference optical feature data and the optical response data sequence includes:

[0028] Based on the reference optical feature data and the optical response data sequence, the statistical distribution characteristics of the reflectance change corresponding to each temperature step point are determined.

[0029] The stress coupling factor of the bipolar plate carbon-based composite coating is analyzed, and the statistical distribution characteristics are mapped to the distorted physical quantity of the coating thermomechanical stress response corresponding to each temperature step point based on the stress coupling factor.

[0030] The distortion physical quantities are sorted according to the temperature order of the temperature step points to obtain the thermal distortion coefficient sequence.

[0031] Optionally, the analysis of the stress coupling factor of the bipolar plate carbon-based composite coating includes:

[0032] Based on the microstructure model of bipolar carbon-based composite coating materials, a theoretical constitutive equation relating macroscopic optical response and microscopic interface stress is established.

[0033] Analyze the optical response data and stress response data of preset standard sample data under thermal excitation;

[0034] The optical response data and the stress response data are subjected to stress coupling analysis using the theoretical constitutive equation to obtain the target numerical range of the stress coupling factor.

[0035] The stress coupling factor of the bipolar carbon-based composite coating is selected from the target numerical range based on the composition parameters of the bipolar carbon-based composite coating.

[0036] Optionally, the step of differentiating the thermally induced distortion coefficient sequence to obtain the gradient change curve includes:

[0037] Identify local fluctuation characteristics in the thermally induced distortion coefficient sequence, and adaptively determine the order and step size of the numerical differentiation algorithm based on the amplitude-frequency characteristics of the local fluctuation characteristics;

[0038] The thermal distortion coefficient sequence is processed by the convolution smoothing differential algorithm based on the order and step size to obtain the first derivative sequence.

[0039] The first derivative sequence is used as the initial gradient sequence, and the initial gradient sequence is normalized based on the physical interval of the temperature step points.

[0040] The normalized preliminary gradient sequence is fitted in the preset temperature-gradient coordinate system to generate a gradient change curve.

[0041] Optionally, the analysis of the inflection point characteristics in the gradient change curve includes:

[0042] Calculate the multi-scale curvature of the gradient change curve, determine the curvature extrema under different scale parameters through the multi-scale curvature, and screen preliminary candidate inflection points based on the curvature extrema.

[0043] Identify the neighborhood of the preliminary candidate inflection point and analyze the gradient statistical characteristics of the neighborhood;

[0044] The false inflection points in the preliminary candidate inflection points are tested based on the gradient statistical characteristics, and the true inflection points are selected based on the tested preliminary candidate inflection points.

[0045] Extract the temperature data and inflection point intensity corresponding to the true inflection point, and combine the temperature data and the inflection point intensity into an inflection point feature.

[0046] Optionally, the step of analyzing the defect evolution behavior of the bipolar plate carbon-based composite coating based on the inflection point characteristics includes:

[0047] A defect evolution phase diagram is constructed based on the preset inflection point temperature and inflection point intensity attributes;

[0048] The inflection point feature is projected onto the defect evolution phase diagram, and the stable state, slow evolution state or accelerated evolution state of the bipolar plate carbon-based composite coating is determined according to the region to which the inflection point feature belongs in the defect evolution phase diagram.

[0049] For the inflection point of the slow evolution state or the accelerated evolution state, calculate the Euclidean distance between the inflection point and the defect evolution path template in the defect evolution phase diagram, and analyze the evolution trend and evolution level of the bipolar plate carbon-based composite coating based on the Euclidean distance;

[0050] The defect evolution behavior of the bipolar plate carbon-based composite coating is determined based on the evolution trend and the evolution level.

[0051] Optionally, detecting the defect type and defect level of the bipolar plate carbon-based composite coating through the defect evolution behavior includes:

[0052] A mapping library between defect evolution behavior and defect type is constructed, wherein the steady state is mapped to the inherent heterogeneity of the material, the slow evolution state is mapped to microcrack initiation defects, and the accelerated evolution state is mapped to interface bonding failure defects.

[0053] Based on the mapping relationship library, identify the defect type corresponding to the defect evolution behavior;

[0054] The inflection point feature coordinates of the defect type are calculated in the defect evolution phase diagram, and the minimum Euclidean distance between the defect type and the preset standard evolution path is analyzed. The deviation of the inflection point feature in the direction perpendicular to the standard evolution path is also calculated.

[0055] The bipolar plate carbon-based composite coating is classified into different levels based on the minimum Euclidean distance and the deviation magnitude under the defect type to obtain the defect level.

[0056] This invention establishes an initial state benchmark for the coating through reference optical scanning, providing a foundation for subsequent change detection; it simulates the thermal stress environment of the coating during operation using stepped thermal excitation to induce defect responses; it generates optical response sequences through spatiotemporally aligned data processing to ensure data consistency; it quantifies the coating stress response using thermal distortion coefficients, converting optical signals into physical defect characteristics; it identifies critical defect states through gradient change curve inflection point analysis; and it achieves accurate determination of defect type and level through defect evolution behavior mapping. Therefore, the defect detection method for carbon-based composite coatings of PEM fuel cell metal bipolar plates proposed in this invention can solve the problem of low accuracy in detecting defects in carbon-based composite coatings of PEM fuel cell metal bipolar plates. Attached Figure Description

[0057] Figure 1 This is a schematic flowchart of a method for detecting defects in carbon-based composite coatings of PEM fuel cell metal bipolar plates according to an embodiment of the present invention.

[0058] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0059] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0060] This application provides a method for detecting defects in the carbon-based composite coating of a PEM fuel cell metal bipolar plate. The execution entity of this method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the method can be executed by software or hardware installed on a terminal device or a server device, where the software may be a blockchain platform. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cluster of cloud servers. The server can be an independent server or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0061] Reference Figure 1 The diagram shown is a flowchart illustrating a method for detecting defects in the carbon-based composite coating of a PEM fuel cell metal bipolar plate according to an embodiment of the present invention. In this embodiment, the method for detecting defects in the carbon-based composite coating of a PEM fuel cell metal bipolar plate includes:

[0062] S1. Optical scanning is performed on the surface area of ​​the preset bipolar plate carbon-based composite coating at a preset reference temperature to obtain reference optical feature data.

[0063] In this embodiment of the invention, the reference temperature refers to the stable temperature environment of the coating without thermal stress interference, which is usually set to room temperature of 25 degrees Celsius; the bipolar carbon-based composite coating refers to a composite structure with metal as the substrate and carbon-based material as the coating; the reference optical characteristic data refers to the set of multi-band reflected light intensity and spatial coordinates of the coating at the reference temperature.

[0064] In this embodiment of the invention, the step of optically scanning the surface area of ​​a preset bipolar plate carbon-based composite coating at a preset reference temperature to obtain reference optical feature data includes:

[0065] The bipolar plate carbon-based composite coating is placed on a temperature control platform, and the temperature control platform is adjusted to the preset reference temperature.

[0066] Based on the reference temperature, the optical scanning unit is controlled to perform a full-coverage scan of the surface area according to a preset serpentine scanning path, while recording the spatial coordinates of each scanning point.

[0067] After the full-coverage scan is completed, the reflected light intensity data of the surface area under multiple target wavelengths are collected;

[0068] The spatial coordinates of each scanning point are correlated and integrated with the reflected light intensity data to generate reference optical feature data.

[0069] In detail, the bipolar plate carbon-based composite coating to be tested is flatly fixed on the ceramic stage of the temperature control platform, ensuring that the coating surface is parallel to the platform surface (parallelism error ≤ 0.02mm) to avoid spatial coordinate offset of subsequent scanning points due to coating tilt. Then, the PID temperature control module of the temperature control platform is activated, with a preset reference temperature of 25℃ input. The temperature control platform regulates the temperature through the coordinated operation of the heating element (50W power) and the cooling fan: initially, the heating element operates at full power; when the platform temperature reaches 23℃, the PID algorithm automatically reduces the heating power to 10W, entering the constant temperature fine-tuning stage. Simultaneously, the platform's built-in platinum resistance temperature sensor (accuracy ±0.05℃) collects temperature data in real time and feeds it back to the control module to correct the heating power until the platform temperature stabilizes at 25℃ ±0.1℃ and remains stable for ≥5 minutes (ensuring that the internal temperature of the coating is consistent with the surface temperature, avoiding distortion of the reference data due to temperature gradients). Then, through high-precision PID temperature control, the reference data becomes repeatable. After the temperature control platform stabilizes at 25°C, the displacement drive module of the optical scanning unit is activated (e.g., using a stepper motor with a step angle of 1.8° and a reduction ratio of 1:100). The scanning is performed according to a serpentine scanning path: the scanning starting point is set to the upper left corner of the coating surface (spatial coordinates X=0mm, Y=0mm), and it moves along the positive X-axis. One scanning point is collected for every 0.1mm of movement (corresponding to one pixel column of the linear CCD camera). When it moves to X=100mm (upper right corner of the coating), it moves 0.1mm in the Y-axis direction (along the positive Y-axis direction), and then moves along the negative X-axis direction. The X-axis back-and-forth + Y-axis progressive trajectory is repeated until the entire coating surface is covered. During the scanning process, the encoder of the displacement drive module records the X and Y axis spatial coordinates of each scanning point and stores the coordinate data synchronously with the acquisition trigger signal of the scanning unit (trigger interval of 0.005 seconds to ensure that the coordinates correspond one-to-one with the subsequent optical signals). The coordinate data is bound to the optical signal acquisition timestamp of that point, which solves the defect that blind spots or overlaps in the scanning path lead to missing or redundant data in some areas. It can cover the entire surface and avoid data redundancy.

[0070] Specifically, after the serpentine scanning path covers the entire coating surface, the optical scanning unit switches to a multi-band optical signal acquisition mode: it sequentially switches between a 400-760nm visible light filter and an 800-1100nm near-infrared filter via a motorized filter wheel (switching speed ≤0.1 seconds / filter). For each target band, the linear CCD camera performs a secondary scan of the coating surface (the scanning path is consistent with the steps of controlling the optical scanning unit to perform a full-coverage scan of the surface area according to the preset serpentine scanning path, ensuring that each scanning point has data in both bands). During the acquisition process, the spectrometer simultaneously records the reflected light intensity of each scanning point in the corresponding band: for example, in the visible light band, the reflected light intensity of the 500th scanning point is 2500 counts; in the near-infrared band, the reflected light intensity of this scanning point is 1800 counts. Simultaneously, a dark current correction algorithm (collecting 50 counts of dark current from the camera when there is no light source, and subtracting the dark current value from the actual collected value) eliminates ambient light interference, ensuring the accuracy of reflected light intensity data. The visible light band can capture macroscopic defects such as scratches and dents on the coating surface, while the near-infrared band can detect the interface separation between the carbon matrix and composite particles inside the coating. The XY coordinates of each scanning point are bound to the visible light reflection intensity minus the near-infrared reflection intensity to form the feature data of a single scanning point (e.g., X=50mm, Y=0.49mm, visible light intensity 2450 counts, near-infrared intensity 1750 counts). Then, the feature data of all scanning points are imported into a matrix database to obtain the reference optical feature data.

[0071] Furthermore, the temperature fluctuations during fuel cell operation are simulated by thermal excitation. The changes in the optical characteristics of the coating at different temperatures must be compared with the characteristics at the same reference temperature in order to eliminate the interference of the initial state differences on the thermally induced changes.

[0072] S2. Apply a stepped thermal excitation to the bipolar plate carbon-based composite coating to obtain multiple temperature step points corresponding to the bipolar plate carbon-based composite coating.

[0073] In this embodiment of the invention, the stepped thermal excitation is a heat input method that gradually increases the coating temperature according to a preset temperature step sequence and stabilizes at each step point. This avoids thermal shock damage to the coating caused by sudden temperature increases, while simultaneously acquiring coating response data at different temperatures. A temperature step point is a specific temperature value reached and stabilized in the stepped thermal excitation process. Each step point corresponds to a stable thermal environment used to collect the coating's optical response data at that temperature.

[0074] In this embodiment of the invention, applying a stepped thermal excitation to the bipolar plate carbon-based composite coating to obtain multiple temperature step points corresponding to the bipolar plate carbon-based composite coating includes:

[0075] A nonlinear temperature increment sequence is determined based on the material heat capacity and thermal conductivity of the bipolar plate carbon-based composite coating, and the nonlinear temperature increment sequence is used as the target temperature sequence.

[0076] According to the target temperature sequence, the infrared heating source is controlled to heat the bipolar plate carbon-based composite coating according to a power loading strategy of fast first and slow later.

[0077] During the heating process, the actual temperature of the surface area of ​​the bipolar plate carbon-based composite coating is monitored in real time. When the actual temperature reaches and stabilizes at any step point in the target temperature sequence, the step point is recorded as the temperature step point that has reached stability.

[0078] By traversing all the step parameters in the target temperature sequence, multiple temperature step points are obtained.

[0079] In detail, the material heat capacity of the coating was measured using a differential scanning calorimeter, and the thermal conductivity was measured using a laser flash method. Based on these two parameters, it can be seen that the heat capacity and thermal conductivity of the coating change gradually in the low-temperature range (25-80℃), with a fast temperature response; while in the high-temperature range (80-200℃), the heat capacity increases, the thermal conductivity decreases, and the temperature response slows down (requiring a longer time to stabilize). Based on these characteristics, a nonlinear temperature increment sequence was established using the principle of denser temperature at low temperatures and sparser temperature at high temperatures: starting from 25℃, the temperature interval in the low-temperature range (25-80℃) is 25℃ (25℃→50℃→80℃), and the temperature interval in the high-temperature range (80-160℃) is 40℃ (80℃→120℃→160℃). The final target temperature sequence is determined as [25℃ (baseline), 50℃, 80℃, 120℃, 160℃]. This sequence ensures the detection accuracy in the low-temperature range while avoiding coating damage caused by frequent temperature increases in the high-temperature range. An infrared heating source (power range 0-1000W, radiation wavelength 3-5μm) was fixed 10cm directly above the coating to ensure that the heating spot completely covered the coating surface. For each step point in the target temperature sequence (50℃, 80℃, 120℃, 160℃), a PWM power regulation strategy with a fast-then-slow adjustment was used to control the infrared heating source to heat the bipolar carbon-based composite coating.

[0080] For example, taking the first step point of 50°C as an example: In the initial stage (current temperature 25°C → 45°C), the infrared heating source rapidly heats up at a power of 600W (duty cycle 60%). At this power, the heating rate of the coating is approximately 5°C / minute (based on a heat capacity of 1.2J / (g·°C)), where the power is calculated as follows: ,but For power, For coating quality, For heat capacity, The temperature difference between the initial temperature and the target temperature. This represents the time interval during which the temperature of the carbon-based composite coating on the electrode changes from the initial temperature to the target temperature. Because the heating source needs to simultaneously heat the temperature control platform, the power is amplified by 10 times. When the coating temperature reaches 45℃ (close to 50℃, a difference of 5℃), the power is switched to 100W (duty cycle 10%) for slow heating, with the heating rate reduced to 0.8℃ / minute to avoid overheating. For the high-temperature step point of 160℃: in the initial stage (120℃→155℃), 800W power (duty cycle 80%) is used for rapid heating (heating rate 3℃ / minute; higher power is required due to increased heat capacity at high temperatures); when approaching 160℃ (a difference of 5℃), the power is switched to 150W (duty cycle 15%), with the heating rate reduced to 0.5℃ / minute to ensure the temperature steadily reaches the target value.

[0081] Specifically, an infrared thermometer (resolution 0.01℃, measurement distance 10cm) is used to collect the actual temperature of the coating surface in real time at a frequency of 1 time / second, and the temperature data is transmitted to the control module. When the temperature reaches the target step point (e.g., 50℃), the control module determines whether the actual temperature has reached 50℃ ± 0.2℃. If it has, it enters the stabilization monitoring phase: continuously collecting temperature data for 3 minutes (180 times), calculating the temperature standard deviation within 3 minutes. If the standard deviation is ≤ 0.1℃, the coating temperature is determined to be stable, and 50℃ is recorded as the stable temperature step point. If the standard deviation is > 0.1℃, the infrared heating source power is finely adjusted (e.g., from 100W to 105W) until the temperature standard deviation within 3 minutes is ≤ 0.1℃. Then, heating, monitoring, and stabilization judgment are performed sequentially on the four step points in the target temperature sequence: 50℃, 80℃, 120℃, and 160℃. For example, at 50℃: heating for 5 minutes to reach the target value, stabilizing for 3 minutes, and recording; at 80℃: heating for 6 minutes to reach the target value, stabilizing for 3 minutes, and recording. Finally, all step parameters are traversed to obtain the stable temperature step point. By traversing all the step points of the nonlinear sequence, the full range from room temperature to high temperature is covered, providing complete temperature dimension data for subsequent analysis of the defect evolution of the coating at different operating temperatures.

[0082] Furthermore, the temperature step points serve as the temperature carriers for generating the optical response data sequence. If the temperature at each step point is not kept stable, the acquired optical data will exhibit spurious changes due to temperature fluctuations (e.g., when the temperature is unstable, the intensity of reflected light at the same scanning point fluctuates continuously, making it impossible to determine whether the fluctuation is due to temperature or interference). Therefore, temperature stability is a prerequisite for accurate acquisition of optical response data, and the two together constitute the temperature-optical correspondence.

[0083] S3. After stabilization at each temperature step point, generate an optical response data sequence based on the optical response data at each temperature step point.

[0084] In this embodiment of the invention, the optical response data sequence is an ordered set formed by arranging the optical response change datasets corresponding to all temperature step points in order of temperature from low to high, which can intuitively reflect the continuous trend of the coating's optical properties changing with temperature.

[0085] In this embodiment of the invention, the step of generating an optical response data sequence based on the optical response data at each temperature step point after stabilization includes:

[0086] After stabilization at each temperature step point, optical response data corresponding to each temperature step point is collected according to the serpentine scanning path and the preset scanning band.

[0087] The optical response data is registered point-by-point with the reference optical feature data based on the spatial coordinates.

[0088] Calculate the reflectance change of each spatial point at each band relative to the reference temperature based on the registered data, and generate a dataset of optical response changes corresponding to temperature step points based on the reflectance change.

[0089] The optical response data sets at all temperature step points are arranged in time sequence according to the temperature step from low to high to generate an optical response data sequence.

[0090] In detail, once a temperature step point reaches stability, the optical scanning unit is activated, following the serpentine scanning path in S1 (starting point X=0mm, Y=0mm, X-axis back-and-forth movement + Y-axis progression, scanning interval 0.1mm), while simultaneously switching to two preset target bands (400-760nm visible light band, 800-1100nm near-infrared band) to acquire optical response data. During acquisition, the linear CCD camera acquires the reflected light intensity at each scanning point in both bands: for example, the scanning point with spatial coordinates X=50mm, Y=0.49mm has a visible light reflectance of 2600 counts at 80℃; all scanning points at this temperature step point are acquired one by one, obtaining the raw visible and near-infrared optical response data corresponding to 80℃. By reusing the serpentine path, it is ensured that the optical data acquisition position of the same scanning point is completely consistent at different temperatures. The XY spatial coordinates of each scanning point are used as unique identifiers to bind the optical response data of the current temperature step point with the reference optical feature data.

[0091] For example, taking a scanning point with X=50mm and Y=0.49mm as an example, the reference data for this point is 2450 counts of visible light intensity and 1750 counts of near-infrared intensity. At 80℃, the optical response data for this point is 2600 counts of visible light intensity and 1900 counts of near-infrared intensity. Through coordinate matching, the two sets of data are associated as a correspondence between coordinates (50, 0.49) - reference (2450, 1750) - 80℃ (2600, 1900). This operation is performed on all scanning points one by one. If an abnormal coordinate matching occurs (such as missing coordinates of individual scanning points), the average of the reference data of the four adjacent scanning points is used as a substitute to ensure that the registration coverage reaches 100%.

[0092] Specifically, the formula for calculating the change in reflectance is as follows: ,in The change in reflectivity Current temperature The intensity of reflection, Using the baseline temperature reflectance intensity, the reflectance change is calculated for each scan point in both wavelength bands. The coordinates of the scan point and the reflectance change in both wavelength bands are recorded as a single data record. After calculating the reflectance for all scan points at a specific temperature step, the results are imported into a MySQL database to form an optical response change dataset for that temperature step. The dataset fields include X-coordinate, Y-coordinate, visible light reflectance change, and near-infrared reflectance change. The optical response change datasets corresponding to the temperature step points are extracted and arranged sequentially from low to high temperature to obtain different data blocks. For example, the first data block is the 50℃ optical response change dataset (containing the reflectance change in both wavelength bands for all scan points at that temperature), the second is the 80℃ optical response change dataset, and so on. These different data blocks are then sequentially integrated into a three-dimensional data sequence (dimension: number of scan points × number of wavelength bands × number of temperature steps). This sequence is the optical response data sequence.

[0093] Furthermore, without arranging the reflectance changes at multiple temperature steps in sequence, it is impossible to obtain the variation of reflectance with temperature, and thus impossible to deduce the evolution of thermal distortion with temperature. Therefore, multiple temperature steps are a prerequisite for realizing quantitative analysis of thermal distortion.

[0094] S4. Analyze the thermal distortion coefficient sequence corresponding to the temperature step point based on the reference optical feature data and the optical response data sequence.

[0095] In this embodiment of the invention, the thermal distortion coefficient sequence is an ordered set formed by arranging the thermal distortion coefficients calculated at each temperature step point in order from low to high temperature. This sequence can directly reflect the degree of change of the thermal mechanical stress of the coating with temperature.

[0096] In this embodiment of the invention, the step of analyzing the thermally induced distortion coefficient sequence corresponding to the temperature step point based on the reference optical feature data and the optical response data sequence includes:

[0097] Based on the reference optical feature data and the optical response data sequence, the statistical distribution characteristics of the reflectance change corresponding to each temperature step point are determined.

[0098] The stress coupling factor of the bipolar plate carbon-based composite coating is analyzed, and the statistical distribution characteristics are mapped to the distorted physical quantity of the coating thermomechanical stress response corresponding to each temperature step point based on the stress coupling factor.

[0099] The distortion physical quantities are sorted according to the temperature order of the temperature step points to obtain the thermal distortion coefficient sequence.

[0100] In detail, for each temperature step point in the optical response data sequence, the reflectance change of all scanning points in the near-infrared band is extracted (the near-infrared band is more sensitive to the internal stress of the coating). The mean (μ) and standard deviation (σ) of the reflectance change corresponding to each temperature step point are obtained by statistical analysis algorithm, and then μ+σ of each temperature step point is used as the statistical distribution feature of that temperature.

[0101] For example, at 50℃: the near-infrared reflectance change across all scanning points was calculated to have a mean μ1 = 0.045 and a standard deviation σ1 = 0.007 (indicating that the overall reflectance change of the coating is small at this temperature and the change is uniform across all points); at 80℃: the mean μ2 = 0.086 and the standard deviation σ2 = 0.009 were calculated (as the temperature increases, the overall reflectance change increases, and the dispersion increases slightly); at 120℃: the mean μ3 = 0.120 and the standard deviation σ3 = 0.015 were calculated (the reflectance change further increases, and the dispersion increases significantly, indicating that stress differences begin to appear in some areas); at 160℃: the mean μ4 = 0.180 and the standard deviation σ4 = 0.022 were calculated (the overall change is significant at high temperatures, and the dispersion is the greatest, indicating that the unevenness of the coating stress distribution is aggravated).

[0102] In this embodiment of the invention, the stress coupling factor is a coefficient characterizing the correlation between the optical response of the bipolar carbon-based composite coating and the micro-interface stress. Its value is determined by the composition of the coating material (such as the ratio of carbon matrix to composite particles) and the microstructure.

[0103] In this embodiment of the invention, the analysis of the stress coupling factor of the bipolar plate carbon-based composite coating includes:

[0104] Based on the microstructure model of bipolar carbon-based composite coating materials, a theoretical constitutive equation relating macroscopic optical response and microscopic interface stress is established.

[0105] Analyze the optical response data and stress response data of preset standard sample data under thermal excitation;

[0106] The optical response data and the stress response data are subjected to stress coupling analysis using the theoretical constitutive equation to obtain the target numerical range of the stress coupling factor.

[0107] The stress coupling factor of the bipolar carbon-based composite coating is selected from the target numerical range based on the composition parameters of the bipolar carbon-based composite coating.

[0108] In detail, based on the coating microstructure model (establishing a mixed packing model of carbon particles with a diameter of 200 nm and Al2O3 particles with a diameter of 50 nm), and applying the principles of continuum mechanics, the linear coupling relationship between the change in optical reflectivity and the interfacial stress tensor is derived. A basic framework is established using the Fourier heat conduction equation and optical scattering theory, and the stress optical coefficient is introduced as a bridging parameter to establish the theoretical constitutive equation: ,in Thermal stress generated at the micro-interface, The change in reflectivity The initial reflectivity at the reference temperature. This is the stress coupling factor.

[0109] Specifically, the standard sample refers to a coating sample with known defect types and interfacial bonding states. The standard sample is placed on a temperature-controlled platform and subjected to the same stepped thermal excitation as the sample under test. Simultaneously, optical data is acquired using an optical scanning system, and stress data is acquired using a stress sensor. The data acquisition frequency is synchronized with the thermal excitation to ensure time alignment of optical and stress data at each temperature step. The change in reflectivity in the optical response data is used as the dependent variable, and the stress value in the stress response data is used as the independent variable to construct an equation. Then, the stress coupling factor k in the theoretical constitutive equation is solved using the least squares method or Bayesian inversion algorithm. For each standard sample, the fitted value of k is calculated, and the distribution of k values ​​for all samples is statistically analyzed. For example, the target numerical range of k (e.g., k between 0.1 and 0.5) is determined by calculating the mean and standard deviation.

[0110] Furthermore, the compositional parameters include the coating's carbon content, thickness, and dopant element ratio. The actual compositional parameters of the coating under test are obtained through energy dispersive spectroscopy (EDS) analysis or thickness measurement. These parameters are then compared with a standard sample library for similarity, using methods such as Euclidean distance or machine learning classification algorithms to find the best-matching standard sample. Finally, based on the k-value of the matching sample within a target range, interpolation or weighted averaging is performed to select the stress coupling factor of the coating under test. For example, if the coating under test has a high carbon content and a large thickness, a k-value close to the upper limit of the range is selected. This personalized parameter selection adapts to the characteristics of different coatings, improving the universality and accuracy of the detection method.

[0111] In this embodiment of the invention, the stress coupling factor and statistical distribution characteristics are substituted into the calculation formula of the distortion physical quantity characterizing the thermomechanical stress response of the coating, i.e. ,in For the first Temperature step points The thermal distortion coefficient is below. For the first Temperature step points The mean of the change in reflectivity, For the first Temperature step points The standard deviation of the change in reflectivity is used, and k is the stress coupling factor. This yields the distortion physical quantity corresponding to each temperature step point, which is then arranged in ascending order of temperature to form a thermal distortion coefficient sequence. After the sequence is arranged, its validity is verified: if the difference in distortion coefficients between two adjacent temperature step points exceeds 30% of the previous temperature, the sequence is considered valid; if an abnormal difference occurs (such as a decrease), the optical response data for that temperature is re-acquired to ensure the physical rationality of the sequence. Through the ordered distortion coefficient sequence, the increasing trend of thermal stress in the coating with increasing temperature is visually presented, providing continuous mechanical data for subsequent gradient analysis.

[0112] Furthermore, the thermal distortion coefficient sequence is the only data source for obtaining the gradient change curve through differential processing. If the thermal distortion coefficient sequence is not generated, only discrete distortion coefficients can be obtained, and the rate of change cannot be calculated.

[0113] S5. Differentiate the thermal distortion coefficient sequence to obtain the gradient change curve, and analyze the inflection point characteristics in the gradient change curve.

[0114] In this embodiment of the invention, the gradient change curve is a curve formed by fitting the rate of change (gradient value) of the distortion coefficient obtained by differential processing with the midpoint temperature of the corresponding temperature range in the temperature-gradient value coordinate system, which can reflect the speed at which the distortion coefficient changes with temperature.

[0115] In this embodiment of the invention, the step of differentiating the thermally induced distortion coefficient sequence to obtain the gradient change curve includes:

[0116] Identify local fluctuation characteristics in the thermally induced distortion coefficient sequence, and adaptively determine the order and step size of the numerical differentiation algorithm based on the amplitude-frequency characteristics of the local fluctuation characteristics;

[0117] The thermal distortion coefficient sequence is processed by the convolution smoothing differential algorithm based on the order and step size to obtain the first derivative sequence.

[0118] The first derivative sequence is used as the initial gradient sequence, and the initial gradient sequence is normalized based on the physical interval of the temperature step points.

[0119] The normalized preliminary gradient sequence is fitted in the preset temperature-gradient coordinate system to generate a gradient change curve.

[0120] In detail, the adjacent differences of the thermally induced distortion coefficient sequence are calculated. For example, the difference between adjacent temperatures from 50℃ to 80℃ is The temperature difference between adjacent values ​​from 80℃ to 120℃ is The temperature difference between adjacent values ​​from 120℃ to 160℃ is And analyze the fluctuation characteristics, if and A smaller difference indicates less fluctuation. Comparison An increase in amplitude indicates increased volatility, with the amplitude-frequency characteristics showing stability in the low and mid frequencies and fluctuations in the high frequencies. Based on this characteristic, a third-order convolutional smoothing differential algorithm is adaptively selected (higher order results in stronger smoothing, suitable for sequences with local fluctuations), with a step size of 2 (the step size matches the temperature step interval; here, the temperature interval is 40℃ / 20℃, and a step size of 2 can cover two adjacent intervals). By adaptively selecting the order and step size, it is ensured that the differential result both preserves the trend of change and eliminates interference from abnormal fluctuations. The third-order convolutional kernel is used to perform differential calculations on the distortion coefficient sequence to obtain the first derivative for each temperature interval.

[0121] For example, the temperature range of 50℃→80℃ (midpoint temperature 65℃): gradient value Even after smoothing by 3rd-order convolution, it is still... (Small fluctuations, small differences before and after smoothing); 80℃→120℃ range (midpoint temperature 100℃): gradient value After smoothing, it becomes ; 120℃→160℃ range (midpoint temperature 140℃): gradient value After smoothing, it becomes (Eliminating some fluctuations); Correlating the midpoint temperature with the gradient value to form a first derivative sequence, and converting the discrete distortion coefficients into continuous gradient values ​​through differential calculation, providing data points for gradient curve fitting.

[0122] Specifically, the physical intervals of the temperature gradient points are 50-80℃ (30℃), 80-120℃ (40℃), and 120-160℃ (40℃). Normalization is performed using the gradient value / temperature interval method to eliminate the influence of interval differences on the gradient values. This allows the normalized gradient values ​​to be combined into a preliminary gradient sequence. Normalization ensures a unified comparison standard for gradient values ​​across different temperature ranges, guaranteeing the accuracy of subsequent curve fitting. A coordinate system is constructed with temperature (horizontal axis, range 50-170℃) and normalized gradient values ​​(vertical axis, range 0-6×10³ Pa / ℃²). The least squares method is used to perform polynomial fitting on the three data points of the preliminary gradient sequence (fitting order 2; since there are 3 data points, order 2 fitting can pass through all points), resulting in the fitted curve equation, as shown below. ( For temperature, (The normalized gradient value) is used to plot the gradient change curve according to the equation. After the fitting is completed, the goodness of fit is calculated. If the goodness of fit is close to 1, it indicates that the fitted curve has a high degree of agreement with the actual data points, and the curve is deemed to be effective.

[0123] In this embodiment of the invention, the inflection point feature refers to the temperature data and gradient intensity data corresponding to the turning point in the gradient change curve where the gradient value changes from increasing to decreasing (or vice versa). This feature is a key indicator for judging the evolution stage of coating defects.

[0124] In this embodiment of the invention, analyzing the inflection point characteristics in the gradient change curve includes:

[0125] Calculate the multi-scale curvature of the gradient change curve, determine the curvature extrema under different scale parameters through the multi-scale curvature, and screen preliminary candidate inflection points based on the curvature extrema.

[0126] Identify the neighborhood of the preliminary candidate inflection point and analyze the gradient statistical characteristics of the neighborhood;

[0127] The false inflection points in the preliminary candidate inflection points are tested based on the gradient statistical characteristics, and the true inflection points are selected based on the tested preliminary candidate inflection points.

[0128] Extract the temperature data and inflection point intensity corresponding to the true inflection point, and combine the temperature data and the inflection point intensity into an inflection point feature.

[0129] In detail, multi-scale curvature analysis refers to obtaining curvature information at different scales by changing the size of the curvature calculation window. First, the curvature of the gradient change curve is calculated using the quadratic differential method. Then, three calculation windows at different scales are set; for example, a small-scale window contains 3 data points, a medium-scale window contains 5 data points, and a large-scale window contains 7 data points. Within each scale window, the curvature value is calculated by sliding the calculation to identify curvature extrema, i.e., local maxima or minima of curvature. Finally, the extrema points that coexist at all three scales are selected as preliminary candidate inflection points. Multi-scale analysis can capture subtle changes at small scales and smooth noise at larger scales, ensuring the comprehensiveness and reliability of candidate inflection points and solving the problem of single-scale analysis being susceptible to noise interference or missing true inflection points.

[0130] Specifically, the neighboring region refers to the set of data points within a certain range before and after the candidate inflection point. For each preliminary candidate inflection point, the gradient values ​​corresponding to the five temperature step points before and after it are taken to form the neighboring region. The gradient statistical characteristics include the mean, standard deviation, and skewness of the gradient values ​​in this region. The gradient values ​​of all points in the neighboring region are extracted, the average value of the gradient values ​​is used to reflect the overall trend, the standard deviation is calculated to characterize the degree of fluctuation, and the skewness is analyzed to understand the distribution pattern.

[0131] Furthermore, the false inflection point test refers to distinguishing between true inflection points and false inflection points caused by noise using statistical hypothesis testing methods. If the absolute value of the curvature of a candidate inflection point is less than twice the standard deviation of the gradient in the neighboring region, it is determined to be a false inflection point; if the gradient value at the candidate inflection point has the same sign as the mean of the neighboring region, it is also determined to be a false inflection point. These rules are applied to each candidate inflection point for testing, and all points that meet the false inflection point conditions are eliminated. The remaining points are the true inflection points. The statistical significance test effectively removes false signals generated by random fluctuations, ensuring the accuracy of inflection point detection and solving the problem of false alarms caused by data noise. Inflection point intensity is an indicator that quantifies the significance of an inflection point, defined as the ratio of the absolute value of the curvature at the inflection point to the standard deviation of the curvature in the neighboring region. For each true inflection point, its corresponding temperature value is recorded as temperature data, and the inflection point intensity value at that point is calculated. Finally, the temperature data and inflection point intensity of each true inflection point are combined into a two-dimensional feature vector as a complete feature description of that inflection point.

[0132] Furthermore, this multi-level, multi-feature inflection point analysis method can effectively capture the key critical points in the evolution of coating defects, laying a solid foundation for the accurate determination of subsequent defect types and levels.

[0133] S6. Analyze the defect evolution behavior of the bipolar plate carbon-based composite coating based on the inflection point characteristics, and detect the defect type and defect level of the bipolar plate carbon-based composite coating through the defect evolution behavior.

[0134] In this embodiment of the invention, the defect evolution behavior refers to the development process of internal defects (such as microcracks and interface separation) of the bipolar plate carbon-based composite coating from nothing to something, and from small to large, during the process of temperature increase. It is usually divided into three stages: stable state, slow evolution state, and accelerated evolution state.

[0135] In this embodiment of the invention, the step of analyzing the defect evolution behavior of the bipolar plate carbon-based composite coating based on the inflection point characteristics includes:

[0136] A defect evolution phase diagram is constructed based on the preset inflection point temperature and inflection point intensity attributes;

[0137] The inflection point feature is projected onto the defect evolution phase diagram, and the stable state, slow evolution state or accelerated evolution state of the bipolar plate carbon-based composite coating is determined according to the region to which the inflection point feature belongs in the defect evolution phase diagram.

[0138] For the inflection point of the slow evolution state or the accelerated evolution state, calculate the Euclidean distance between the inflection point and the defect evolution path template in the defect evolution phase diagram, and analyze the evolution trend and evolution level of the bipolar plate carbon-based composite coating based on the Euclidean distance;

[0139] The defect evolution behavior of the bipolar plate carbon-based composite coating is determined based on the evolution trend and the evolution level.

[0140] In detail, the defect evolution phase diagram is a two-dimensional feature space with inflection point temperature data as the horizontal axis and inflection point intensity as the vertical axis. During construction, a large amount of sample data of known defect types is collected, and the inflection point features of the sample data are obtained. Clustering algorithms (such as K-means clustering) are used to automatically classify these inflection point features, forming different clustering regions in the temperature-intensity coordinate system. Each region corresponds to a typical defect evolution state: low temperature and low intensity regions correspond to a stable state, medium temperature and medium intensity regions correspond to a slow evolution state, and high temperature and high intensity regions correspond to an accelerated evolution state. Region boundaries are determined through statistical methods, such as calculating the 95% confidence ellipse of each cluster center as the boundary. The shortest distance from the coordinate point to each region boundary is calculated, and the evolution state is determined by judging the region where the point is located. For example, when the inflection point falls within the stable state region, it indicates that the coating structure is stable near that temperature point; when it falls within the slow evolution state region, it indicates that the coating has begun to show minor damage but is expanding slowly; when it falls within the accelerated evolution state region, it indicates that the damage has entered a rapid expansion stage. This spatial location mapping achieves an objective determination of the defect evolution state, solving the problem of inaccurate subjective experience judgment.

[0141] Specifically, the defect evolution path template is a typical defect development trajectory established through analysis of a large amount of historical data. Typical evolution path curves for different types of defects are plotted in the defect evolution phase diagram. For inflection points in slow or accelerated evolution states, the Euclidean distance to the nearest path template is calculated; the smaller the distance, the better it matches the actual defect evolution pattern. Simultaneously, evolution levels are classified according to the distance value: distances less than the threshold T1 are low-risk, distances between T1 and T2 are medium-risk, and distances greater than T2 are high-risk. Decision rules are established based on the evolution trend and evolution level to determine the defect evolution behavior of the bipolar plate carbon-based composite coating: a stable state combined with any level represents benign behavior; a slow evolution state combined with a medium-to-high-risk level requires early warning; and an accelerated evolution state combined with any level requires immediate action. Through multi-dimensional information fusion, a complete understanding of the defect development trend is formed, providing decision support.

[0142] In this embodiment of the invention, the defect type refers to the classification of the incompleteness of the bipolar plate carbon-based composite coating according to its physical mechanism and manifestation, including but not limited to material inherent heterogeneity, microcrack initiation defects, and interface adhesion failure defects; the defect level refers to the quantitative classification of the severity and development risk of the defect under the determined defect type, such as level one (minor), level two (moderate), and level three (severe).

[0143] In this embodiment of the invention, detecting the defect type and defect level of the bipolar plate carbon-based composite coating through the defect evolution behavior includes:

[0144] A mapping library between defect evolution behavior and defect type is constructed, wherein the steady state is mapped to the inherent heterogeneity of the material, the slow evolution state is mapped to microcrack initiation defects, and the accelerated evolution state is mapped to interface bonding failure defects.

[0145] Based on the mapping relationship library, identify the defect type corresponding to the defect evolution behavior;

[0146] The inflection point feature coordinates of the defect type are calculated in the defect evolution phase diagram, and the minimum Euclidean distance between the defect type and the preset standard evolution path is analyzed. The deviation of the inflection point feature in the direction perpendicular to the standard evolution path is also calculated.

[0147] The bipolar plate carbon-based composite coating is classified into different levels based on the minimum Euclidean distance and the deviation magnitude under the defect type to obtain the defect level.

[0148] In detail, the mapping relation library is a classification model trained using machine learning methods. During training, the evolutionary behavior characteristics of historical samples (including state type, evolution level, etc.) are used as input, and the defect types verified by actual dissection (microcracks, interface delamination, etc.) are used as output. Support vector machines or neural network algorithms are used for training. The trained model can accurately map the steady state to the inherent heterogeneity of the material, the slowly evolving state to microcrack initiation defects, and the accelerated evolution state to interface bonding failure defects. Automatic identification of defect types can be achieved through intelligent classification.

[0149] Specifically, the modal process involves inputting the evolutionary behavior features obtained from real-time detection into a trained classification model, which then outputs the corresponding defect type label. The evolutionary state and level features of the current detected sample are extracted, and a classification model from a mapping database is used for prediction to obtain the most probable defect type. The standard path corresponding to the current defect type is found in the defect evolution phase diagram, and these two parameters are obtained through geometric calculations. The minimum Euclidean distance is the shortest geometric distance from the current inflection point feature to the standard evolution path, reflecting the degree of agreement with typical defect patterns. The deviation magnitude is the deviation value of the inflection point feature in the direction perpendicular to the standard path, reflecting the degree of anomaly.

[0150] Furthermore, a two-dimensional decision matrix method is adopted for grading. Grading rules based on minimum Euclidean distance and deviation magnitude are established: when both minimum Euclidean distance and deviation magnitude are less than 50% of their respective thresholds, it is classified as Level 1 (minor); when either parameter exceeds 50% but is less than 80%, it is classified as Level 2 (moderate); when either parameter exceeds 80%, it is classified as Level 3 (severe). Through multi-parameter comprehensive evaluation, a refined distinction of defect levels is achieved.

[0151] Furthermore, by statistically learning from a standard sample library of known defect levels to determine the threshold for classifying defect levels, a large number of samples with known defect levels are collected. The defect levels of the samples have been precisely calibrated using destructive methods such as metallographic dissection, scanning electron microscopy, or performance degradation testing. The sample library needs to cover different defect types (microcracks, interface peeling, etc.) and various degrees of severity to ensure the comprehensiveness and representativeness of the statistical results. For example, a standard library containing 50 Level 1 samples, 50 Level 2 samples, and 50 Level 3 samples is prepared. A complete testing process is performed on each sample in the standard sample library to obtain the minimum Euclidean distance and deviation amplitude data for each sample. Then, statistical methods are used to analyze the distribution characteristics of these parameters: for Level 1 samples, the upper limit of the 95% confidence interval for their distance and amplitude data is calculated as the normal fluctuation range for that level; for Level 2 samples, the distribution offset of their parameter values ​​relative to Level 1 samples is analyzed; for Level 3 samples, the critical point at which their parameter values ​​are significantly different from other levels is determined. Taking the distinction between Level 1 and Level 2 samples as an example, the distance parameter is used as the classification index. The true positive rate and false positive rate corresponding to different threshold points are calculated. The threshold point that maximizes the Youden index is selected, as this point can optimally balance the risks of missed and false positives. This process is repeated to determine all threshold boundaries between each level. The initially determined thresholds are applied to an independent validation sample set (new samples not involved in training) to evaluate their level classification accuracy. If the accuracy is lower than the preset requirement (e.g., 90%), the thresholds are readjusted. After 3-5 iterations, a stable threshold system is finally determined.

[0152] Furthermore, through defect evolution phase diagram construction, multi-parameter quantitative analysis, and intelligent classification decision-making, a technological leap from simple defect detection to complex evolution behavior prediction has been achieved, upgrading static detection to dynamic behavior prediction. The phase diagram analysis method enables visualized monitoring of defect evolution, effectively solving the industry problem that existing methods cannot predict defect development trends or distinguish defect evolution stages, and providing a new technical approach for the reliability assessment of fuel cell bipolar plate coatings.

[0153] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0154] Therefore, the embodiments should be regarded as exemplary and non-limiting in all respects. The scope of the invention is not limited to the foregoing description, and all variations within the meaning and scope of equivalents falling within the protection scope are intended to be included in the invention.

[0155] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0156] Furthermore, it is clear that the word "including" does not exclude other units or steps, and the singular does not exclude the plural. Terms such as "first," "second," etc., are used to indicate names, but do not indicate any specific order.

[0157] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for detecting defects in the carbon-based composite coating of a PEM fuel cell metal bipolar plate, characterized in that, The method includes: Optical scanning is performed on the surface area of ​​the preset bipolar plate carbon-based composite coating at a preset reference temperature to obtain reference optical feature data; A stepped thermal excitation was applied to the bipolar plate carbon-based composite coating to obtain multiple temperature step points corresponding to the bipolar plate carbon-based composite coating. After stabilization at each temperature step point, an optical response data sequence is generated based on the optical response data at each temperature step point; Based on the reference optical feature data and the optical response data sequence, the thermal distortion coefficient sequence corresponding to the temperature step point is analyzed. The thermal distortion coefficient sequence is differentiated to obtain a gradient change curve, and the inflection point characteristics in the gradient change curve are analyzed. The defect evolution behavior of the bipolar plate carbon-based composite coating is analyzed based on the inflection point characteristics, and the defect type and defect level of the bipolar plate carbon-based composite coating are detected by the defect evolution behavior.

2. The method for detecting defects in the carbon-based composite coating of PEM fuel cell metal bipolar plates as described in claim 1, characterized in that, The process of optically scanning the surface area of ​​the preset bipolar plate carbon-based composite coating at a preset reference temperature to obtain reference optical feature data includes: The bipolar plate carbon-based composite coating is placed on a temperature control platform, and the temperature control platform is adjusted to the preset reference temperature. Based on the reference temperature, the optical scanning unit is controlled to perform a full-coverage scan of the surface area according to a preset serpentine scanning path, while recording the spatial coordinates of each scanning point. After the full-coverage scan is completed, the reflected light intensity data of the surface area under multiple target wavelengths are collected; The spatial coordinates of each scanning point are correlated and integrated with the reflected light intensity data to generate reference optical feature data.

3. The method for detecting defects in the carbon-based composite coating of PEM fuel cell metal bipolar plates as described in claim 1, characterized in that, The stepwise thermal excitation applied to the bipolar carbon-based composite coating yields multiple temperature step points corresponding to the bipolar carbon-based composite coating, including: A nonlinear temperature increment sequence is determined based on the material heat capacity and thermal conductivity of the bipolar plate carbon-based composite coating, and the nonlinear temperature increment sequence is used as the target temperature sequence. According to the target temperature sequence, the infrared heating source is controlled to heat the bipolar plate carbon-based composite coating according to a power loading strategy of fast first and slow later. During the heating process, the actual temperature of the surface area of ​​the bipolar plate carbon-based composite coating is monitored in real time. When the actual temperature reaches and stabilizes at any step point in the target temperature sequence, the step point is recorded as the temperature step point that has reached stability. By traversing all the step parameters in the target temperature sequence, multiple temperature step points are obtained.

4. The method for detecting defects in the carbon-based composite coating of PEM fuel cell metal bipolar plates as described in claim 2, characterized in that, After stabilization at each temperature step point, an optical response data sequence is generated based on the optical response data at each temperature step point, including: After stabilization at each temperature step point, optical response data corresponding to each temperature step point is collected according to the serpentine scanning path and the preset scanning band. The optical response data is registered point-by-point with the reference optical feature data based on the spatial coordinates. Calculate the reflectance change of each spatial point at each band relative to the reference temperature based on the registered data, and generate a dataset of optical response changes corresponding to temperature step points based on the reflectance change. The optical response data sets at all temperature step points are arranged in time sequence according to the temperature step from low to high to generate an optical response data sequence.

5. The method for detecting defects in the carbon-based composite coating of PEM fuel cell metal bipolar plates as described in claim 1, characterized in that, The analysis of the thermally induced distortion coefficient sequence corresponding to the temperature step point based on the reference optical feature data and the optical response data sequence includes: Based on the reference optical feature data and the optical response data sequence, the statistical distribution characteristics of the reflectance change corresponding to each temperature step point are determined. The stress coupling factor of the bipolar plate carbon-based composite coating is analyzed, and the statistical distribution characteristics are mapped to the distorted physical quantity of the coating thermomechanical stress response corresponding to each temperature step point based on the stress coupling factor. The distortion physical quantities are sorted according to the temperature order of the temperature step points to obtain the thermal distortion coefficient sequence.

6. The method for detecting defects in the carbon-based composite coating of PEM fuel cell metal bipolar plates as described in claim 5, characterized in that, The analysis of the stress coupling factor of the bipolar plate carbon-based composite coating includes: Based on the microstructure model of bipolar carbon-based composite coating materials, a theoretical constitutive equation relating macroscopic optical response and microscopic interface stress is established. Analyze the optical response data and stress response data of preset standard sample data under thermal excitation; The optical response data and the stress response data are subjected to stress coupling analysis using the theoretical constitutive equation to obtain the target numerical range of the stress coupling factor. The stress coupling factor of the bipolar carbon-based composite coating is selected from the target numerical range based on the composition parameters of the bipolar carbon-based composite coating.

7. The method for detecting defects in the carbon-based composite coating of PEM fuel cell metal bipolar plates as described in claim 1, characterized in that, The step of differentiating the thermally induced distortion coefficient sequence to obtain the gradient change curve includes: Identify local fluctuation characteristics in the thermally induced distortion coefficient sequence, and adaptively determine the order and step size of the numerical differentiation algorithm based on the amplitude-frequency characteristics of the local fluctuation characteristics; The thermal distortion coefficient sequence is processed by the convolution smoothing differential algorithm based on the order and step size to obtain the first derivative sequence. The first derivative sequence is used as the initial gradient sequence, and the initial gradient sequence is normalized based on the physical interval of the temperature step points. The normalized preliminary gradient sequence is fitted in the preset temperature-gradient coordinate system to generate a gradient change curve.

8. The method for detecting defects in the carbon-based composite coating of PEM fuel cell metal bipolar plates as described in claim 1, characterized in that, The analysis of the inflection point characteristics in the gradient change curve includes: Calculate the multi-scale curvature of the gradient change curve, determine the curvature extrema under different scale parameters through the multi-scale curvature, and screen preliminary candidate inflection points based on the curvature extrema. Identify the neighborhood of the preliminary candidate inflection point and analyze the gradient statistical characteristics of the neighborhood; The false inflection points in the preliminary candidate inflection points are tested based on the gradient statistical characteristics, and the true inflection points are selected based on the tested preliminary candidate inflection points. Extract the temperature data and inflection point intensity corresponding to the true inflection point, and combine the temperature data and the inflection point intensity into an inflection point feature.

9. The method for detecting defects in the carbon-based composite coating of PEM fuel cell metal bipolar plates as described in claim 1, characterized in that, The analysis of the defect evolution behavior of the bipolar plate carbon-based composite coating based on the inflection point characteristics includes: A defect evolution phase diagram is constructed based on the preset inflection point temperature and inflection point intensity attributes; The inflection point feature is projected onto the defect evolution phase diagram, and the stable state, slow evolution state or accelerated evolution state of the bipolar plate carbon-based composite coating is determined according to the region to which the inflection point feature belongs in the defect evolution phase diagram. For the inflection point of the slow evolution state or the accelerated evolution state, calculate the Euclidean distance between the inflection point and the defect evolution path template in the defect evolution phase diagram, and analyze the evolution trend and evolution level of the bipolar plate carbon-based composite coating based on the Euclidean distance; The defect evolution behavior of the bipolar plate carbon-based composite coating is determined based on the evolution trend and the evolution level.

10. The method for detecting defects in the carbon-based composite coating of a PEM fuel cell metal bipolar plate as described in claim 9, characterized in that, The method of detecting the defect type and defect level of the bipolar plate carbon-based composite coating through the defect evolution behavior includes: A mapping library between defect evolution behavior and defect type is constructed, wherein the steady state is mapped to the inherent heterogeneity of the material, the slow evolution state is mapped to microcrack initiation defects, and the accelerated evolution state is mapped to interface bonding failure defects. Based on the mapping relationship library, identify the defect type corresponding to the defect evolution behavior; The inflection point feature coordinates of the defect type are calculated in the defect evolution phase diagram, and the minimum Euclidean distance between the defect type and the preset standard evolution path is analyzed. The deviation of the inflection point feature in the direction perpendicular to the standard evolution path is also calculated. The bipolar plate carbon-based composite coating is classified into different levels based on the minimum Euclidean distance and the deviation magnitude under the defect type to obtain the defect level.

Images