Electrode sheet coating quality detection method and system based on reflection and transmission ultrasonic fusion

Abstract

The present application relates to the technical field of pole piece coating quality control, and discloses a pole piece coating quality detection method and system based on reflection and transmission ultrasonic fusion, comprising: collecting the reflection and transmission ultrasonic signals of the pole piece and completing spatial alignment, extracting the first reflection wave amplitude, interface echo time difference and other characteristics from the reflection signal, extracting the energy, center frequency offset, propagation time and other parameters from the transmission signal, constructing a joint correlation model based on the propagation characteristics of ultrasonic waves in multi-layer media, and synchronously solving the coating thickness and area density; a high-dimensional defect sensitive vector is constructed by integrating the physical consistency characteristics of the two types of signals, and precise defect determination is realized by combining spatial position and change trend verification, the system cooperates six functional units to output quality distribution map and process capability index in real time, the information limitation and interference problem of the traditional single detection mode is overcome, and the high-precision monitoring demand of the whole process of the high-speed coating production line is met.

Description

Technical Field

[0001] This invention relates to the field of electrode coating quality control technology, and in particular to a method and system for detecting electrode coating quality based on the fusion of reflected and transmitted ultrasound. Background Technology

[0002] The coating quality of electrode sheets directly affects the energy density, cycle life, and safety performance of energy storage devices. Uneven coating thickness, fluctuations in areal density, and defects such as pinholes, missed coatings, and coating peeling can lead to decreased charge and discharge efficiency, increased internal resistance, and even the risk of thermal runaway. Currently, the coating process in electrode production is easily affected by factors such as changes in slurry viscosity, fluctuations in coating speed, and deviations in doctor blade gap, requiring real-time and accurate detection of key quality parameters and defect status. At present, the areal density and thickness consistency of coated electrodes mainly rely on offline sampling and weighing or optical contact scanning for random inspection. However, the sampling results lag behind the production cycle and cannot provide immediate feedback for closed-loop coating head control. Optical methods can only obtain surface morphology and cannot penetrate the coating to characterize the internal density and areal density distribution. Surface reflections, leveling ripples, or bubbles on wet coatings often lead to false triggering of optical sensors, resulting in a high false alarm rate.

[0003] Ultrasonic testing offers the advantages of being non-destructive and enabling real-time detection. US Patent 12493281B2 discloses an online monitoring framework for battery manufacturing based on acoustic and process signal analysis. It explicitly points out the need to avoid defects such as inhomogeneity, debonding, and inclusions during electrode coating, and emphasizes the importance of quality monitoring during the wet coating and drying stages. However, this patent only proposes a general approach to signal acquisition and analysis at the system architecture level, without disclosing specific ultrasonic implementation methods for detecting electrode coating thickness and areal density. In particular, it fails to address the following key bottleneck: traditional reflective ultrasound uses a single-channel characteristic (acoustic time or amplitude) for evaluation, and its responses to thickness and density fluctuations are coupled, making it impossible to independently output both thickness and areal density parameters, resulting in a severe lack of detection dimensions.

[0004] Therefore, there is an urgent need for an ultrasonic testing method that can be embedded in existing coating production lines and simultaneously decouple thickness and areal density. Summary of the Invention

[0005] In order to overcome the shortcomings and deficiencies of the existing technology, the present invention provides a method and system for detecting electrode coating quality based on the fusion of reflective and transmissive ultrasound.

[0006] The technical solution adopted in this invention is a method for detecting electrode coating quality based on the fusion of reflected and transmitted ultrasound, comprising the following steps:

[0007] S1. Acquire the reflected and transmitted wave signal sequences from the electrode coating area. Based on the probe's mechanical layout parameters and trigger timing parameters, precisely align the reflected and transmitted signals corresponding to the same measurement point on the time axis. S2. Extract the first reflected wave amplitude and interface echo time difference characteristic parameters from the aligned reflected signals. Extract signal energy, center frequency shift, and propagation time characteristic parameters from the transmitted signals. S3. Based on the acoustic impedance, sound velocity, and attenuation characteristics of ultrasound propagation in the multilayer medium composed of the coating and current collector, construct a joint correlation model between the reflected signal characteristic parameters, the transmitted signal characteristic parameters, and the coating thickness and areal density. S4. Analyze the extracted characteristic parameters... The input is a joint correlation model, which performs real-time calculations using model parameters calibrated with known standard samples to obtain coating thickness and areal density data for each measurement point; S5, by integrating the additional echo characteristics of reflected signals, energy attenuation and waveform distortion characteristics of transmitted signals, and physical consistency characteristics of reflected and transmitted signal change patterns, a high-dimensional defect-sensitive feature vector is constructed and input into the trained pattern recognition model to complete defect category determination; S6, the thickness and areal density data of each measurement point are arranged along the electrode width direction and the coating travel direction, and the lateral range, standard deviation, profile and longitudinal fluctuation parameters, overall distribution standard deviation and process capability index are calculated. The defect type and spatial coordinates are recorded, and a quality distribution map is generated.

[0008] Furthermore, the joint association model in S3 is expressed as follows:

[0009] ,

[0010] in, For coating thickness, The speed at which ultrasound propagates in the coating. The time difference between the reflected signal and the echo at the interface. The angle of ultrasound incidence. For the acoustic impedance of the coating, For the current collector acoustic impedance, The surface density of the coating. For transmitted signal energy, The initial energy of the incident ultrasound. The ultrasonic attenuation coefficient of the coating is denoted as .

[0011] Furthermore, the model solution in S4 is expressed as follows:

[0012]

[0013] ,

[0014] in, For the calculated thickness, This is the standard sample thickness calibration value. The number of characteristic parameters of the reflected signal. The reflection feature weighting coefficient, These are the measured reflection characteristic parameters. These are standard reflection characteristic parameters. This represents the number of characteristic parameters of the transmitted signal. For transmission characteristic weighting coefficients, These are the measured transmission characteristic parameters. These are standard transmission characteristic parameters. To solve for the density later, The areal density calibration value for standard samples. These are the reflection and transmission characteristic weighting coefficients corresponding to the surface density calculation.

[0015] Furthermore, the high-dimensional defect-sensitive feature vector construction in S5 is expressed as follows:

[0016] ,

[0017] in, For defect-sensitive feature vectors, Additional echo amplitude for the reflected signal, This is the reference value for the amplitude of the normal reflected signal. This represents the energy deviation value of the transmitted signal. This is the reference value for the energy of a normal transmitted signal. This represents the center frequency offset of the transmitted signal. The initial center frequency, For reflected signals With transmitted signal covariance, These are the standard deviations of the reflected signal and the transmitted signal, respectively.

[0018] Furthermore, the output of the pattern recognition model in S5 is expressed as follows:

[0019] ,

[0020] in, This is the result of the defect category determination. For the number of support vectors, For Lagrange multipliers, For the category labels corresponding to the support vectors, For kernel function, For the first The feature vectors corresponding to each support vector. This is the hyperplane bias term for classification.

[0021] Furthermore, the process capability index calculation in S6 is expressed as follows:

[0022] ,

[0023] in, For process capability index, To meet the quality characteristics and specification limits, For quality characteristics, the lower specification limit, To measure the mean of the data, To measure the standard deviation of the data, The sample size is denoted by Skew, and Skew is the skewness coefficient of the measurement data.

[0024] Further, S3 includes the following steps: S31, clarifying the acoustic impedance matching relationship, sound wave propagation path, and energy attenuation law of ultrasonic waves propagating in the multilayer medium of coating and current collector, and defining the physical generation mechanism of each characteristic parameter of reflected and transmitted signals; S32, based on the theory of vertical incidence and propagation of sound waves, establishing the correlation between the amplitude of the first wave of reflected wave, the time difference of interface echo, and the coating thickness, and the acoustic impedance of the coating and current collector; S33, combining the principle of energy attenuation of transmitted waves, the law of center frequency shift, and the propagation time characteristics, constructing the correspondence between the characteristic parameters of transmitted signals and the coating thickness and surface density; S34, integrating the correlation between reflected and transmitted signals, eliminating the cross-interference of thickness and density changes, and forming a joint correlation model structure including multiple variables.

[0025] Further, step S4 includes the following steps: S41, selecting a standard sample with known thickness and areal density, acquiring the reflection and transmission signals of the standard sample through an ultrasonic probe, and extracting the corresponding feature parameters as calibration reference data; S42, substituting the calibration reference data into the joint correlation model, solving for the undetermined coefficients and calibration parameters in the model, and determining the quantitative calculation relationship of the model; S43, inputting the measured feature parameters extracted in S2 into the calibrated joint correlation model according to the corresponding category, and obtaining preliminary thickness and areal density calculation results by solving the simultaneous equations; S44, comparing the preliminary calculation results with the known parameters of the standard sample, adjusting the iteration step size and convergence conditions in the model solution process, and outputting the final thickness and areal density data.

[0026] Further, S5 includes the following steps: S51, performing waveform analysis on the aligned reflected signal to capture unexpected additional echo signals generated by defects inside the coating, and recording the amplitude and occurrence time parameters of the additional echoes; S52, analyzing the energy change amplitude, waveform distortion degree, and center frequency offset of the transmitted signal to distinguish between normal fluctuations and abnormal changes; S53, verifying the physical consistency between the abnormal reflected signal and the abnormal transmitted signal in terms of spatial location and change trend, and selecting the feature combination corresponding to the real defects; S54, integrating the abnormal parameters of the reflected signal, the abnormal parameters of the transmitted signal, and the consistency verification results into a high-dimensional feature vector, and inputting it into a pre-trained pattern recognition model to complete the defect category classification.

[0027] A electrode coating quality inspection system based on the fusion of reflected and transmitted ultrasound is applied to an electrode coating quality inspection method based on the fusion of reflected and transmitted ultrasound. The system includes: an ultrasonic signal acquisition unit, a multi-dimensional feature parameter extraction unit, a multi-layer medium ultrasonic propagation joint modeling unit, a real-time coating quality parameter calculation unit, a defect feature fusion and category determination unit, and a quality data visualization and transmission unit. The ultrasonic signal acquisition unit acquires the reflected wave signal sequence and the transmitted wave signal sequence of the electrode coating area. Based on the probe's mechanical layout and triggering sequence, it precisely aligns the reflected and transmitted signals at the same measurement point on the time axis. The multi-dimensional feature parameter extraction unit extracts the first reflected wave amplitude and interface echo time difference from the aligned reflected signal, and extracts energy from the transmitted signal. The system includes: a center frequency offset and propagation time characteristic parameters; a multi-layer medium ultrasonic propagation joint modeling unit, which constructs a correlation model between reflection and transmission signal characteristic parameters and coating thickness and areal density based on the propagation characteristics of ultrasonic waves in coatings and current collectors; a coating quality parameter real-time calculation unit, which inputs the extracted characteristic parameters into the joint model and combines them with standard sample calibration parameters to complete the real-time calculation of thickness and areal density; a defect feature fusion and category determination unit, which integrates the extra echo of reflected signals, the anomaly of transmitted signals, and the consistency characteristics of both, and classifies defects through a pattern recognition model; and a quality data visualization and transmission unit, which calculates lateral, longitudinal, and overall consistency indicators, records the spatial coordinates of defects, generates a quality distribution map, and transmits structured data to the manufacturing execution system or programmable logic controller through a standard interface.

[0028] Compared with the prior art, the present invention has at least one beneficial effect:

[0029] This invention proposes a method and system for detecting electrode coating quality based on the fusion of reflected and transmitted ultrasound. Addressing the limitations of traditional single-reflection ultrasonic testing, which is susceptible to interference from coating surface roughness and current collector flatness, making it difficult to simultaneously and accurately obtain thickness and areal density parameters, and the insufficient sensitivity of transmission ultrasonic testing for thin coatings and its inability to effectively capture subtle internal defects, this method acquires reflected and transmitted wave signal sequences. Through precise spatial alignment, multi-dimensional feature parameters are extracted from the two types of signals. A joint correlation model is constructed by combining the propagation characteristics of ultrasound in multilayer media, achieving simultaneous and accurate detection of coating thickness and areal density, thus overcoming the information limitations of single detection methods.

[0030] To address the issues of independent quality parameter calculation and defect identification, unresolved cross-interference, and high rates of false positives and false negatives in traditional methods, this approach integrates the features of additional echo from reflected signals, energy attenuation of transmitted signals, and waveform distortion. By combining the physical consistency characteristics of both, a high-dimensional defect-sensitive feature vector is constructed. The trained pattern recognition model achieves accurate defect classification. Simultaneously, it calculates horizontal, vertical, and overall quality consistency indices to generate a quality distribution map. This fully utilizes the complementary value of the two types of signals, reduces the impact of environmental noise and equipment vibration, and improves the accuracy of defect identification.

[0031] Through the collaborative work of multiple units, this system achieves real-time calculation of quality parameters, accurate defect judgment, and data visualization transmission. It solves the problems of insufficient real-time performance of traditional offline detection and the inability to provide comprehensive data support for coating process optimization. It meets the needs of high-speed coating production lines for high-precision monitoring of the entire quality process and effectively guarantees the energy density, cycle life, and safety performance of energy storage equipment. Attached Figure Description

[0032] Figure 1 This is a flowchart illustrating the overall process of the method of the present invention.

[0033] Figure 2 This is a schematic diagram of an air-coupled linear array ultrasonic probe symmetrically arranged on the upper and lower sides of the electrode under test according to the present invention.

[0034] Figure 3 This is a schematic diagram of an air-coupled linear array ultrasonic probe arranged on the upper side of the electrode under test and a solid-coupled linear array ultrasonic probe arranged on the lower side of the electrode under test according to the present invention.

[0035] Figure 4 This is a flowchart of method step S3 of the present invention;

[0036] Figure 5 This is a flowchart of method step S4 of the present invention;

[0037] Figure 6 This is a flowchart of step S5 of the method of the present invention;

[0038] Figure 7 This is a diagram showing the system unit composition of the present invention. Detailed Implementation

[0039] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. The application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0040] During the research process, the inventors discovered limitations in using single reflection or transmission methods to detect electrode coating quality. Specifically, reflection ultrasonic testing is easily affected by coating surface roughness and current collector flatness, making it difficult to simultaneously and accurately obtain thickness and areal density parameters. Transmission testing, on the other hand, lacks sensitivity for thin coatings and cannot effectively capture subtle internal defects. Therefore, neither method, when used alone, can fully reflect the coating quality status. Furthermore, traditional ultrasonic testing techniques lack sufficient correlation between data calculation and defect identification. In traditional methods, quality parameter calculation and defect identification are often independent modules, failing to fully utilize the physical consistency characteristics of reflection and transmission signals. This results in ineffective elimination of cross-interference during parameter calculation, and defect identification relies on single signal characteristics, making it susceptible to environmental noise and equipment vibration, leading to missed or incorrect defect detection. Moreover, there is a lack of comprehensive analysis of the spatial distribution patterns and process stability of quality parameters, failing to provide comprehensive data support for coating process optimization.

[0041] To address the aforementioned technical problems, this application provides a method for detecting electrode coating quality based on the fusion of reflected and transmitted ultrasound. This method utilizes a dual-sided ultrasonic probe structure to acquire reflected and transmitted ultrasonic signals during scanning. Based on the physical consistency of the reflected and transmitted signals, joint calculations are performed to achieve independent inversion of coating thickness and areal density, as well as accurate identification of internal defects, thereby comprehensively characterizing the coating quality status.

[0042] like Figure 1 As shown, the electrode coating quality detection method based on the fusion of reflected and transmitted ultrasound includes the following steps:

[0043] S1, collect the reflected wave signal sequence and transmitted wave signal sequence of the electrode coating area, and according to the probe mechanical layout parameters and trigger timing parameters, make the reflected signal and transmitted signal corresponding to the same measurement point form a precise spatial alignment on the time axis;

[0044] Specifically, step S1 achieves the acquisition and precise spatial alignment of the reflected wave signal and the transmitted wave signal, laying the foundation for subsequent feature extraction and parameter calculation.

[0045] In one specific implementation, the reflected wave signal and the transmitted wave signal are acquired by symmetrically arranging air-coupled linear array ultrasonic probes on the upper and lower sides of the electrode under test. Specifically, as shown below... Figure 2As shown, during testing, two air-coupled linear array ultrasonic probes are positioned above and below the electrode under test, respectively. Their working distance is determined based on the probe's nominal focal length, the center frequency of the array elements, and the acoustic characteristics of the coating to ensure stable ultrasonic signal transmission. The air-coupled linear array ultrasonic probe contains several uniformly arranged array elements, with each element arranged along the width of the electrode under test. The length of the probe covers the width of the electrode under test. Either the upper or lower air-coupled linear array ultrasonic probe can be selected as the transmitting probe, emitting ultrasonic waves to the electrode and receiving the reflected wave signal. The other air-coupled linear array ultrasonic probe serves as the transmission receiving probe, receiving the transmitted wave signal that passes through the electrode. It should be noted that the frequency of the air-coupled linear array ultrasonic probe is generally selected between 400kHz and 1MHz.

[0046] In another specific implementation, the reflected wave signal and the transmitted wave signal are acquired by arranging an air-coupled linear array ultrasonic probe on the upper side of the electrode under test and a solid-coupled linear array ultrasonic probe on the lower side of the electrode under test, as detailed below. Figure 3 As shown, the air-coupled linear array ultrasonic probe uses air as the coupling medium, making it particularly suitable for wet electrodes that have not yet dried after coating. The vertical distance between the air-coupled linear array ultrasonic probe and the electrode surface is determined comprehensively based on the probe's nominal focal length, the center frequency of the array elements, and the acoustic characteristics of the coating to ensure stable ultrasonic signal transmission. A flexible solid coupling layer is placed between the solid-coupled linear array ultrasonic probe and the electrode under test. This flexible solid coupling layer has a specific elastic modulus and can undergo adaptive deformation during detection, tightly fitting the contact interface between the underside of the electrode under test and the solid-coupled linear array ultrasonic probe. This effectively fills microscopic gaps, reduces acoustic impedance mismatch and energy loss during ultrasonic wave propagation, improves the coupling efficiency and signal quality of ultrasonic detection, and thus enhances detection sensitivity and accuracy. Both the solid-coupled linear array ultrasonic probe and the air-coupled linear array ultrasonic probe contain several uniformly arranged array elements, with each element arranged along the width direction of the electrode under test. The length direction of both probes covers the width direction of the electrode under test. A solid-coupled linear array ultrasonic probe serves as the transmitting probe, emitting ultrasonic waves to the electrode and receiving reflected wave signals. An upper air-coupled linear array ultrasonic probe serves as the transmission receiving probe, receiving transmitted wave signals that pass through the electrode.

[0047] As can be seen from the above explanation, through Figure 2 or Figure 3The ultrasonic device described herein can acquire both reflected and transmitted wave signals. Simultaneously, specific trigger timing parameters and signal sampling frequencies are set, with the sampling frequency generally not lower than 10MHz. The trigger timing parameters are dynamically adjusted according to the coating travel speed to ensure that reflected and transmitted wave signals from the same measurement point can be acquired during the high-speed movement of the electrode. Subsequently, based on the preset probe mechanical layout parameters and trigger timing parameters, a signal processing algorithm performs time calibration on the two types of signals acquired, eliminating signal offsets caused by factors such as fluctuations in the electrode's movement speed. This ensures that the reflected and transmitted signals corresponding to the same measurement point are precisely aligned with the time axis of the electrode's movement speed, guaranteeing that the subsequently extracted feature parameters accurately correspond to the same measurement position and avoiding the influence of spatial position deviations on the detection results.

[0048] S2, extract the first reflection wave amplitude and interface echo time difference characteristic parameters from the aligned reflection signal, and extract the signal energy, center frequency shift and propagation time characteristic parameters from the transmitted signal;

[0049] Specifically, in step S2, multi-dimensional feature parameters are extracted from the aligned reflected and transmitted signals to provide data support for building the correlation model and defect judgment. For the aligned reflected signal, the signal waveform is analyzed using professional signal processing software to accurately extract two core feature parameters: the amplitude of the first reflected wave and the time difference of the interface echo. The extraction of the amplitude of the first reflected wave requires the elimination of background noise interference. A filtering algorithm is used to preprocess the signal, with the filtering frequency range set to 10-80MHz to ensure accurate capture of the peak intensity of the first reflected wave. Its value range is usually between 0.1-5V, depending on factors such as coating thickness and acoustic impedance. The extraction of the time difference of the interface echo is achieved by identifying the arrival time of the echoes from different interfaces (coating surface, coating and current collector interface) in the reflected signal and calculating the time difference between them. The time difference accuracy is controlled within 0.1ns. This parameter directly reflects the propagation path length of the ultrasonic wave in the coating. For the transmitted signal, after preprocessing such as filtering, three characteristic parameters are extracted: signal energy, center frequency offset, and propagation time. The signal energy is obtained by calculating the integral area under the transmitted signal waveform, with the unit being mJ. Under normal circumstances, it should be maintained at 30%-70% of the initial energy of the incident ultrasound. The center frequency offset is determined by comparing the difference between the transmitted signal and the initial center frequency of the incident ultrasound. The initial center frequency is usually set to 20-50MHz, and the offset should be controlled within ±2MHz. Exceeding this range may indicate an abnormality in the coating. The propagation time refers to the total time it takes for the ultrasound to travel from the transmitting probe through the electrode to the receiving probe, with an accuracy controlled within 0.1ns. It is directly related to the coating thickness and material properties.

[0050] S3. Based on the acoustic impedance, sound velocity and attenuation characteristics of ultrasonic waves propagating in a multilayer medium composed of coating and current collector, a joint correlation model of reflected signal characteristic parameters, transmitted signal characteristic parameters and coating thickness and areal density is constructed.

[0051] Specifically, step S3 constructs a joint correlation model between the characteristic parameters of the reflected signal, the characteristic parameters of the transmitted signal, and the coating thickness and areal density, establishing a quantitative relationship between signal characteristics and physical quality parameters. In implementation, the propagation law of ultrasound in the multilayer medium composed of the coating and the current collector is first analyzed in depth to clarify key physical parameters such as acoustic impedance, sound velocity, and attenuation characteristics. The acoustic impedance of the coating is typically 1.5 × 10⁻⁶. 6 -5.0×10 6 kg / (m 2 Between ・s), the acoustic impedance of the current collector is set to a specific value depending on the material. For example, the acoustic impedance of a copper current collector is approximately 4.5 × 10⁻⁶. 7 kg / (m 2 •s), the aluminum current collector is approximately 1.7 × 10 7 kg / (m 2 The propagation speed of ultrasound in coatings is generally between 2000-4000 m / s. The attenuation coefficient describes the rate of energy loss of the ultrasound signal during propagation in the medium, and is related to the coating material, thickness, and frequency. It is measured in dB / cm / MHz, representing the decibel attenuation of the ultrasound signal per centimeter of propagation at a frequency of one megahertz. Based on these physical characteristics, the generation mechanism of the reflected signal characteristic parameters is analyzed. The amplitude of the first reflected wave is directly related to the acoustic impedance matching between the coating and the current collector, while the interface echo time difference is determined by both the propagation speed of ultrasound in the coating and the coating thickness. Simultaneously, the physical meaning of the transmitted signal characteristic parameters is analyzed. Signal energy attenuation is related to the coating thickness, areal density, and attenuation coefficient; the center frequency shift is affected by the uniformity of the coating's internal structure; and the propagation time comprehensively reflects both the coating thickness and the sound wave propagation speed. Based on this, the correlation between reflection and transmission signals is integrated to eliminate the cross-interference caused by changes in thickness and density. Through physical modeling and mathematical derivation, a joint correlation model including multiple variables is constructed. This model needs to fully consider the influence of parameters such as the ultrasonic incident angle (usually set to a small angle of 0°-15° to reduce reflection loss) to ensure that the coating thickness and areal density can be accurately inferred from the characteristic parameters.

[0052] S4. Input the extracted feature parameters into the joint correlation model, and perform real-time calculation using the model parameters calibrated by the known standard sample to obtain the coating thickness and areal density data for each measurement point.

[0053] Specifically, step S4 utilizes the calibrated joint correlation model to perform real-time calculations on the measured characteristic parameters, obtaining coating thickness and areal density data for each measurement point. Before implementation, standard samples with known thickness and areal density must be selected. The thickness deviation of the standard samples must be controlled within ±1 μm, and the areal density deviation within ±1 g / m². 2 Furthermore, the model needs to cover the thickness and areal density range commonly encountered in actual production, typically selecting 5-10 standard samples of different specifications. Using the same ultrasonic probe and parameter settings as in actual testing, the reflection and transmission signals of the standard samples are acquired, and the corresponding characteristic parameters are extracted as calibration reference data. These include standard reflection characteristic parameters (reference values ​​for first reflection wave amplitude and interface echo time difference) and standard transmission characteristic parameters (reference values ​​for signal energy, center frequency, and propagation time). These calibration reference data are then substituted into a joint correlation model to solve for the undetermined coefficients and calibration parameters, determining key parameters such as reflection characteristic weight coefficients and transmission characteristic weight coefficients. The weight coefficients are solved using optimization algorithms such as the least squares method to ensure that the deviation between the model's predicted values ​​and the known parameters of the standard samples is minimized. Subsequently, the measured characteristic parameters extracted in step S2 are input into the calibrated joint correlation model according to their corresponding categories, and the simultaneous equations are solved to obtain preliminary thickness and areal density calculation results. To improve the accuracy of the solution, it is necessary to compare the preliminary calculation results with the known parameters of the standard sample and adjust the iteration step size in the model solution process (usually set to a thickness iteration step size of 0.01-0.1 μm and a g / m² iteration step size of 0.01-0.1 g / m²). 2 The areal density iteration step size and convergence conditions (convergence error controlled within ±0.1 μm thickness error and ±0.1 g / m) are used to determine the convergence error. 2 The areal density error is calculated, and after multiple iterations of optimization, the final thickness and areal density data are output. The calculation process needs to be completed within 1ms to meet the real-time detection requirements of the high-speed coating production line.

[0054] S5 integrates the additional echo characteristics of reflected signals, the energy attenuation and waveform distortion characteristics of transmitted signals, and the physical consistency characteristics of the change patterns of reflected and transmitted signals to construct a high-dimensional defect-sensitive feature vector, which is then input into the trained pattern recognition model to complete the defect category determination.

[0055] Specifically, step S5 constructs a high-dimensional defect-sensitive feature vector and uses a pattern recognition model to determine the defect category, achieving accurate identification of coating defects. First, a detailed waveform analysis is performed on the aligned reflected signal. A waveform recognition algorithm is used to capture unexpected additional echo signals generated by internal coating defects (such as pinholes, missed coatings, and coating peeling). The amplitude and occurrence time parameters of the additional echo are accurately recorded. The amplitude of the additional echo needs to be compared with the baseline value of the normal reflected signal amplitude, and the occurrence time needs to be calculated by differing from the normal interface echo time to determine the location and size of the defect. Simultaneously, the energy variation amplitude, waveform distortion degree, and center frequency offset of the transmitted signal are analyzed. By calculating the ratio of the transmitted signal energy deviation value to the baseline value of the normal transmitted signal energy, and the ratio of the center frequency offset to the initial center frequency, normal fluctuations and abnormal changes are distinguished. The normal fluctuation range needs to be set based on the test results of standard samples; typically, the energy deviation does not exceed ±10%, and the center frequency offset does not exceed ±5%. Subsequently, the physical consistency between the reflected signal anomalies and the transmitted signal anomalies in terms of spatial location and trend of change is verified. This involves determining whether the anomaly location captured by the reflected signal matches the anomaly location of the transmitted signal, and whether the trend of change conforms to physical laws. This helps to filter out the feature combinations corresponding to genuine defects and eliminate false anomaly signals caused by environmental noise, equipment vibration, and other factors. Finally, the anomaly parameters of the reflected signal (relative value of the additional echo amplitude), the anomaly parameters of the transmitted signal (relative value of energy deviation, relative value of center frequency shift), and the consistency verification results are integrated into a high-dimensional defect-sensitive feature vector. This vector is then input into a pre-trained pattern recognition model. This model needs to be trained with a large amount of sample data including different defect types (pinholes, missing coatings, coating peeling, etc.). During training, parameters such as the kernel function and the number of support vectors need to be optimized to ensure that the model can accurately output defect category judgment results, with a defect recognition accuracy rate of over 95%.

[0056] S6. Arrange the thickness and areal density data of each measurement point along the electrode width direction and the coating travel direction, calculate the lateral range, standard deviation, profile and longitudinal fluctuation parameters, overall distribution standard deviation and process capability index, record the defect type and spatial coordinates, and generate a quality distribution map.

[0057] Specifically, step S6 comprehensively analyzes and visualizes the test data, providing comprehensive data support for coating process optimization. First, the thickness and areal density data at each measurement point are arranged in an orderly manner according to the electrode width direction and the coating travel direction, forming a complete two-dimensional data matrix. The data arrangement must strictly correspond to the actual spatial position of the electrode, with positional errors controlled within ±1mm. Subsequently, a series of quality evaluation parameters are calculated. In the lateral direction, the range, standard deviation, and profile are calculated. The lateral range is the difference between the maximum and minimum values ​​of the thickness or areal density at each measurement point within the same width cross-section of the electrode. The lateral standard deviation reflects the dispersion of data within the same cross-section and typically needs to be controlled within ±2μm (thickness) or ±2g / m². 2 Within the areal density range, the profile accuracy measures the deviation between the actual thickness or areal density profile and the ideal profile, and must meet the design tolerance range. In the longitudinal direction, the fluctuation parameter is calculated to reflect the quality stability along the coating direction. The fluctuation parameter is usually expressed as the sum of the squares of the differences between adjacent measurement points and must be controlled within a set threshold. At the same time, the overall distribution standard deviation and the process capability index are calculated. The overall distribution standard deviation reflects the overall dispersion of the quality parameters of the entire electrode sheet, while the process capability index comprehensively considers factors such as the upper and lower specification limits of quality characteristics, the mean and standard deviation of measurement data, sample size, and skewness coefficient. The process capability index must be greater than 1.33 to indicate that the process is stable and reliable. In addition, the defect types and their corresponding spatial coordinates are recorded in detail, with the spatial coordinates accurate to ±1mm, to form a defect distribution map. Finally, these data and analysis results are integrated to generate intuitive quality distribution maps, including thickness distribution heatmaps, areal density distribution heatmaps, defect location marking maps, etc. The structured data is then transmitted in real time to the manufacturing execution system or programmable logic controller via standard industrial interfaces (such as Ethernet / IP, PROFINET, etc.) to provide timely and accurate data support for process adjustments and quality traceability in the production process.

[0058] Preferably, the joint association model in S3 is expressed as follows:

[0059] ,

[0060] in, For coating thickness, The speed at which ultrasound propagates in the coating. The time difference between the reflected signal and the echo at the interface. The angle of ultrasound incidence. For the acoustic impedance of the coating, For the current collector acoustic impedance, The surface density of the coating. For transmitted signal energy, The initial energy of the incident ultrasound. The ultrasonic attenuation coefficient of the coating is denoted as .

[0061] Specifically, the joint correlation model establishes a quantitative correlation between coating thickness, areal density, and characteristic parameters of reflected and transmitted signals based on the acoustic impedance matching relationship, sound velocity characteristics, and energy attenuation law of ultrasonic waves propagating in a multilayer medium composed of a coating and a current collector. During model construction, key physical parameters such as the propagation speed of ultrasonic waves in the coating, the echo time difference at the interface of the reflected signal, the ultrasonic incident angle, and the acoustic impedance of the coating and the current collector are fully considered. The coating thickness is obtained through the collaborative calculation of these parameters. Simultaneously, the areal density is calculated by combining parameters such as the transmitted signal energy, the initial energy of the incident ultrasonic wave, the ultrasonic attenuation coefficient of the coating, and the calculated coating thickness. The implementation of this model requires first clarifying the propagation path and energy change law of ultrasonic waves in the multilayer medium. By sorting out the intrinsic relationship between various physical parameters, the interface echo time difference characteristics of the reflected signal and the energy characteristics of the transmitted signal are transformed into calculation indicators for thickness and areal density, respectively. This effectively solves the problem of simultaneously and accurately obtaining two core quality parameters in single signal detection, providing a scientific theoretical basis for subsequent real-time calculations and ensuring the correlation and accuracy of thickness and areal density detection.

[0062] Preferably, the model solution in S4 is expressed as:

[0063]

[0064] ,

[0065] in, For the calculated thickness, This is the standard sample thickness calibration value. The number of characteristic parameters of the reflected signal. The reflection feature weighting coefficient, These are the measured reflection characteristic parameters. These are standard reflection characteristic parameters. This represents the number of characteristic parameters of the transmitted signal. For transmission characteristic weighting coefficients, These are the measured transmission characteristic parameters. These are standard transmission characteristic parameters. To solve for the density later, The areal density calibration value for standard samples. These are the reflection and transmission characteristic weighting coefficients corresponding to the surface density calculation.

[0066] Specifically, the model calculation achieves accurate calculation of coating thickness and areal density by weighting the differences between the benchmark data calibrated from standard samples and the measured characteristic parameters. During implementation, the thickness and areal density calibration values ​​of the standard samples, along with the corresponding standard reflection and transmission characteristic parameters, are first obtained. This allows for the determination of key model parameters such as the reflection characteristic weight coefficient and the transmission characteristic weight coefficient. In actual testing, the differences between the measured reflection characteristic parameters and the standard reflection characteristic parameters, and the differences between the measured transmission characteristic parameters and the standard transmission characteristic parameters, are multiplied by their respective weight coefficients, summed, and then added to the thickness or areal density calibration values ​​of the standard samples to obtain the calculated thickness and areal density data. This calculation method fully considers the influence of both reflection and transmission signal characteristic parameters on quality parameters. It achieves reasonable weighting of different characteristics through weight coefficient adjustment and eliminates systematic errors through standard sample calibration. During the calculation process, the contributions of multiple sets of characteristic parameters are integrated through simultaneous equations to ensure the stability and accuracy of the calculation results. This effectively addresses detection deviations caused by signal fluctuations during production and meets the data accuracy requirements of online real-time testing.

[0067] Preferably, the high-dimensional defect-sensitive feature vector construction in S5 is expressed as follows:

[0068] ,

[0069] in, For defect-sensitive feature vectors, Additional echo amplitude for the reflected signal, This is the reference value for the amplitude of the normal reflected signal. This represents the energy deviation value of the transmitted signal. This is the reference value for the energy of a normal transmitted signal. This represents the center frequency offset of the transmitted signal. The initial center frequency, For reflected signals With transmitted signal covariance, These are the standard deviations of the reflected signal and the transmitted signal, respectively.

[0070] Specifically, the construction of a high-dimensional defect-sensitive feature vector integrates multi-dimensional anomaly features of reflected and transmitted signals, as well as their physical consistency features, to form a comprehensive feature vector that accurately reflects the defect state. During construction, firstly, the unexpected additional echo amplitude generated by internal coating defects in the reflected signal is extracted and its ratio to the normal reflected signal amplitude benchmark is calculated to obtain a quantitative index of reflected signal anomalies. Simultaneously, the ratio of the transmitted signal energy deviation to the normal transmitted signal energy benchmark and the ratio of the transmitted signal center frequency offset to the initial center frequency are calculated to capture energy and frequency anomalies in the transmitted signal. Furthermore, by calculating the covariance of the reflected and transmitted signals and dividing by the product of their standard deviations, the physical consistency of the abnormal changes in the two types of signals is quantified. The implementation of this feature vector requires first establishing a normal benchmark system for various signals. Through multi-dimensional analysis and quantification of the measured signals, scattered anomaly features are integrated into a high-dimensional vector, comprehensively covering signal changes caused by defects. This avoids misjudgment or missed detection of defects due to single features, providing comprehensive and effective feature input for subsequent pattern recognition models, ensuring the comprehensiveness and sensitivity of defect detection.

[0071] Preferably, the pattern recognition model output in S5 is expressed as:

[0072] ,

[0073] in, This is the result of the defect category determination. For the number of support vectors, For Lagrange multipliers, For the category labels corresponding to the support vectors, For kernel function, For the first The feature vectors corresponding to each support vector. This is the hyperplane bias term for classification.

[0074] Specifically, the output logic of the pattern recognition model is based on the Support Vector Machine (SVM) algorithm. It performs kernel function operations on a high-dimensional defect-sensitive feature vector and the trained support vectors, combining Lagrange multipliers, support vector class labels, and a classification hyperplane bias term to achieve accurate defect category determination. Before implementation, the model needs to be trained with a large amount of sample data including various defects and normal states to determine the number of support vectors, the corresponding Lagrange multipliers and class labels for each support vector, and to select an appropriate kernel function to achieve non-linear classification in the high-dimensional feature space, and to label the classification hyperplane bias term. During actual determination, the constructed high-dimensional defect-sensitive feature vector is input into the model, the kernel function result of this vector and each support vector is calculated, multiplied by the corresponding Lagrange multiplier and class label, summed, and then the bias term is added. The final defect category determination result is then output through a sign function. This model fully leverages the advantages of support vector machines in classifying small-sample, high-dimensional data. It solves the nonlinear classification problem of high-dimensional features through kernel function processing. The introduction of Lagrange multipliers and bias terms ensures the optimality of the classification boundary, effectively improving the accuracy and robustness of defect category determination. It can cope with signal noise caused by various interference factors in the production environment, providing reliable algorithmic support for accurate defect identification.

[0075] Preferably, the process capability index calculation in step S6 is expressed as follows:

[0076] ,

[0077] in, For process capability index, To meet the quality characteristics and specification limits, For quality characteristics, the lower specification limit, To measure the mean of the data, To measure the standard deviation of the data, The sample size is denoted by Skew, and Skew is the skewness coefficient of the measurement data.

[0078] Specifically, the process capability index is calculated by comprehensively considering the specification limits of quality characteristics, the mean, standard deviation, sample size, and skewness coefficient of the measurement data to fully quantify the coating process's ability to meet quality requirements. The calculation process first clarifies the upper and lower specification limits of the quality characteristics, i.e., the allowable fluctuation range of coating thickness or areal density. Then, a sufficient number of measurement data are collected, and the mean, standard deviation, sample size, and skewness coefficient are calculated. The mean reflects the central tendency of the data, the standard deviation reflects the degree of dispersion, and the skewness coefficient characterizes the asymmetry of the data distribution. Subsequently, the difference between the upper specification limit and the mean is calculated and divided by three times the standard deviation, and the difference between the mean and the lower specification limit is calculated and divided by three times the standard deviation, taking the smaller of the two values. This smaller value is then multiplied by the sample size correction factor (the square root of the ratio of the sample size minus one to the sample size) and the skewness correction factor (the square of the negative skewness coefficient of the natural index divided by two), finally yielding the process capability index. This calculation method not only considers the centralization and dispersion characteristics of the data, but also eliminates the bias caused by small sample data through sample size correction and compensates for the impact of asymmetric data distribution on the results through skewness correction. It can comprehensively and objectively evaluate the stability and reliability of the coating process, provide quantitative basis for process optimization, and help enterprises determine whether the production process can continuously meet quality standards.

[0079] Preferred, such as Figure 4 As shown, S3 includes the following steps: S31, clarifying the acoustic impedance matching relationship, sound wave propagation path, and energy attenuation law of ultrasonic waves propagating in the multilayer medium of coating and current collector, and defining the physical generation mechanism of each characteristic parameter of reflected and transmitted signals; S32, based on the theory of vertical incidence and propagation of sound waves, establishing the correlation between the amplitude of the first wave of reflected wave, the time difference of interface echo, and the coating thickness, and the acoustic impedance of the coating and current collector; S33, combining the principle of energy attenuation of transmitted waves, the law of center frequency shift, and the propagation time characteristics, constructing the correspondence between the characteristic parameters of transmitted signals and the coating thickness and surface density; S34, integrating the correlation between reflected and transmitted signals, eliminating the cross-interference of thickness and density changes, and forming a joint correlation model structure including multiple variables.

[0080] Specifically, the construction process of the joint correlation model includes four sub-steps, forming a logically coherent and progressive modeling flow. The first step involves systematically analyzing the acoustic impedance matching relationship of ultrasonic waves propagating in the multilayer medium of the coating and current collector, clarifying the complete propagation path of the sound wave from incident to reflection and transmission, analyzing the attenuation law of energy during propagation, and further clarifying the physical generation mechanism of each characteristic parameter of the reflected and transmitted signals, laying a theoretical foundation for subsequent modeling. The second step, based on the fundamental theory of vertical incident and propagation of sound waves, focuses on establishing a quantitative correlation between the amplitude of the first wave of the reflected wave, the time difference of the interface echo, and the coating thickness, as well as the acoustic impedance of the coating and current collector, clarifying the characterization effect of the reflected signal characteristics on the thickness parameter. The third step, combining the physical principle of transmitted wave energy attenuation, the variation law of center frequency shift, and the characteristics of propagation time, constructs the correspondence between the characteristic parameters of the transmitted signal and the coating thickness and areal density, exploring the responsiveness of the transmitted signal to these two core quality parameters. The fourth step integrates the correlation relationships between the reflected and transmitted signals established in the first two steps, eliminating cross-interference generated during thickness and density changes through algorithm design, forming a joint correlation model structure that includes multiple physical variables and can simultaneously calculate thickness and areal density. This step-by-step implementation method ensures the scientific rigor and precision of the model construction. From physical mechanism analysis to single signal correlation, and then to multi-signal integration and interference removal, each step provides support for the accuracy of the model. The final model can make full use of the complementary information of the two types of signals, improving the comprehensiveness and accuracy of quality parameter detection.

[0081] Preferred, such as Figure 5 As shown, step S4 includes the following steps: S41, selecting a standard sample with known thickness and areal density, acquiring the reflection and transmission signals of the standard sample through an ultrasonic probe, and extracting the corresponding feature parameters as calibration reference data; S42, substituting the calibration reference data into the joint correlation model, solving for the undetermined coefficients and calibration parameters in the model, and determining the quantitative calculation relationship of the model; S43, inputting the measured feature parameters extracted in S2 into the calibrated joint correlation model according to the corresponding category, and obtaining preliminary thickness and areal density calculation results by solving the simultaneous equations; S44, comparing the preliminary calculation results with the known parameters of the standard sample, adjusting the iteration step size and convergence conditions in the model solution process, and outputting the final thickness and areal density data.

[0082] Specifically, the model calculation process includes four steps, ensuring the accuracy and consistency of thickness and areal density calculations through a standardized process. The first step involves selecting a standard sample with known thickness and areal density, collecting its reflection and transmission signals, and extracting characteristic parameters such as the amplitude of the first reflection wave, the interface echo time difference, and signal energy according to preset feature extraction rules. These parameters serve as the benchmark data for model calibration. The second step substitutes the calibration benchmark data into the previously constructed joint correlation model, solving for the undetermined coefficients and calibration parameters in the model through mathematical operations. This clarifies the quantitative calculation relationship between each characteristic parameter and the thickness and areal density, completing the model calibration process. The third step involves inputting the characteristic parameters extracted from the measured electrode signals into the calibrated joint correlation model according to the categories of reflection and transmission. Solving the simultaneous equations yields preliminary thickness and areal density calculation results. The fourth step compares and analyzes the preliminary calculation results with the known parameters of the standard sample. Adjusting the iteration step size and convergence conditions in the model solution process based on the magnitude of the deviation, and reducing calculation errors through multiple iterations, ultimately outputting accurate thickness and areal density data. This step-by-step calculation method eliminates inherent system errors through standard sample calibration and corrects random errors in the calculation process through iterative optimization, ensuring that the calculation results can truly reflect the actual quality status of the electrode coating and meet the requirements of high-speed production line online real-time detection for data accuracy and stability.

[0083] Preferred, such as Figure 6 As shown, step S5 includes the following steps: S51, performing waveform analysis on the aligned reflected signal to capture unexpected additional echo signals generated by defects inside the coating, and recording the amplitude and occurrence time parameters of the additional echoes; S52, analyzing the energy change amplitude, waveform distortion degree, and center frequency offset of the transmitted signal to distinguish between normal fluctuations and abnormal changes; S53, verifying the physical consistency between the abnormal reflected signal and the abnormal transmitted signal in terms of spatial location and change trend, and selecting the feature combination corresponding to the real defects; S54, integrating the abnormal parameters of the reflected signal, the abnormal parameters of the transmitted signal, and the consistency verification results into a high-dimensional feature vector, and inputting it into a pre-trained pattern recognition model to complete the defect category classification.

[0084] Specifically, the defect feature fusion and category determination process includes four sub-steps, which improve the accuracy of defect determination through multi-dimensional anomaly capture and consistency verification. The first step involves a detailed waveform analysis of the time-aligned reflected signal, focusing on capturing unexpected additional echo signals caused by defects such as pinholes, missing coatings, and peeling within the coating. Key parameters such as the amplitude and occurrence time of these additional echoes are accurately recorded to provide a basis for defect identification at the reflecting end. The second step analyzes the transmitted signal, examining its energy variation amplitude, waveform distortion, and center frequency offset. By setting reasonable thresholds, the system distinguishes between signal changes caused by normal production fluctuations and abnormal changes caused by defects, filtering out suspected defect signals at the transmitting end. The third step verifies whether the anomalies in the reflected and transmitted signals correspond in spatial location and whether their trends are consistent. Based on the physical principles of ultrasonic propagation, false anomalies caused by environmental noise, equipment vibration, and other interference factors are eliminated, filtering out feature combinations that truly reflect the existence of defects. The fourth step integrates the filtered anomaly parameters of the reflected and transmitted signals, along with the results of their physical consistency verification, to construct a high-dimensional defect-sensitive feature vector. This vector is then input into a pattern recognition model pre-trained with a large number of samples, allowing the model to accurately classify the defect category. This step-by-step implementation method, from capturing single signal anomalies to verifying the consistency of multiple signals, and then to the integration and judgment of high-dimensional features, progressively eliminates interference and focuses on real defects, effectively reducing the probability of missed or false defects and providing reliable technical support for timely detection and handling of defects in the production process.

[0085] like Figure 7As shown, a electrode coating quality inspection system based on the fusion of reflected and transmitted ultrasound is used in electrode coating quality inspection methods based on the fusion of reflected and transmitted ultrasound. The system includes: an ultrasonic signal acquisition unit, a multi-dimensional feature parameter extraction unit, a multi-layer medium ultrasonic propagation joint modeling unit, a real-time coating quality parameter calculation unit, a defect feature fusion and category determination unit, and a quality data visualization and transmission unit. The ultrasonic signal acquisition unit acquires the reflected wave signal sequence and the transmitted wave signal sequence of the electrode coating area, and aligns the reflected and transmitted signals at the same measurement point on the time axis with precise spatial positioning based on the probe mechanical layout and triggering sequence. The multi-dimensional feature parameter extraction unit extracts the first reflected wave amplitude and interface echo time difference from the aligned reflected signal, and extracts the energy from the transmitted signal. The system includes several key parameters: quantity, center frequency offset, and propagation time characteristic parameters; a multi-layer medium ultrasonic propagation joint modeling unit, based on the propagation characteristics of ultrasonic waves in coatings and current collectors, constructs a correlation model between the characteristic parameters of reflected and transmitted signals and the coating thickness and areal density; a coating quality parameter real-time calculation unit inputs the extracted characteristic parameters into the joint model and combines them with standard sample calibration parameters to complete the real-time calculation of thickness and areal density; a defect feature fusion and category determination unit integrates the extra echo of reflected signals, the anomaly of transmitted signals, and the consistency characteristics of both, and classifies defects through a pattern recognition model; and a quality data visualization and transmission unit calculates the lateral, longitudinal, and overall consistency indices, records the spatial coordinates of defects and generates a quality distribution map, and transmits the structured data to the manufacturing execution system or programmable logic controller through a standard interface.

[0086] A method and system for detecting electrode coating quality based on the fusion of reflected and transmitted ultrasonic signals achieves comprehensive, high-precision online detection of electrode coating quality through deep fusion of reflected and transmitted ultrasonic signals. Methodologically, by employing... Figure 2 or Figure 3 The ultrasonic probe acquires reflected and transmitted wave signals and performs precise spatial alignment. It extracts features such as the amplitude of the first reflected wave and the time difference of the interface echo from the reflected signal, and extracts parameters such as energy, center frequency shift, and propagation time from the transmitted signal. Based on the propagation characteristics of ultrasound in multilayer media, a joint correlation model is constructed. This ensures both the synchronous and accurate calculation of coating thickness and areal density, and the integration of the physical consistency characteristics of the two types of signals to construct a high-dimensional defect-sensitive vector, significantly improving the accuracy of defect identification. The system, through the coordinated operation of six functional units, achieves fully automated processing of signal acquisition, feature extraction, model calculation, defect identification, and data visualization. It can output quality distribution maps and process capability indices in real time, providing comprehensive data support for production process optimization.

[0087] This invention addresses the information limitations of single detection methods by complementing and fusing reflected and transmitted signals. This avoids the interference of surface roughness and current collector flatness in reflective ultrasonic testing, while also compensating for the shortcomings of transmitted testing in terms of insufficient sensitivity to thin coatings and difficulty in capturing subtle internal defects. This achieves simultaneous and accurate detection of thickness, areal density parameters, and various defects. To address the insufficient correlation between data processing and defect judgment, a joint correlation model is constructed to eliminate cross-interference from thickness and density changes. The consistency of the spatial location and trend of anomalies in reflected and transmitted signals is used to verify and screen for true defects, effectively reducing the risk of misjudgment and missed detection caused by environmental noise and equipment vibration. Furthermore, by calculating indicators such as lateral range, standard deviation, and longitudinal fluctuation parameters, the spatial distribution and process stability of quality parameters are comprehensively analyzed, providing more comprehensive and reliable technical support for real-time adjustment of the coating process and meeting the needs of high-precision monitoring throughout the entire high-speed coating production line.

[0088] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," "link," and "fix" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal communication between two components. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0089] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various equivalent changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for detecting electrode coating quality based on the fusion of reflected and transmitted ultrasound, characterized in that, Includes the following steps: S1, collect the reflected wave signal sequence and transmitted wave signal sequence of the electrode coating area, and according to the probe mechanical layout parameters and trigger timing parameters, make the reflected signal and transmitted signal corresponding to the same measurement point form a precise spatial alignment on the time axis; S2, extract the first reflection wave amplitude and interface echo time difference characteristic parameters from the aligned reflection signal, and extract the signal energy, center frequency shift and propagation time characteristic parameters from the transmitted signal; S3. Based on the acoustic impedance, sound velocity and attenuation characteristics of ultrasonic waves propagating in a multilayer medium composed of coating and current collector, a joint correlation model of reflected signal characteristic parameters, transmitted signal characteristic parameters and coating thickness and areal density is constructed. S4. Input the extracted feature parameters into the joint correlation model, and perform real-time calculation using the model parameters calibrated by the known standard sample to obtain the coating thickness and areal density data for each measurement point. S5 integrates the additional echo characteristics of reflected signals, the energy attenuation and waveform distortion characteristics of transmitted signals, and the physical consistency characteristics of the change patterns of reflected and transmitted signals to construct a high-dimensional defect-sensitive feature vector, which is then input into the trained pattern recognition model to complete the defect category determination. S6. Arrange the thickness and areal density data of each measurement point along the electrode width direction and the coating travel direction, calculate the lateral range, standard deviation, profile and longitudinal fluctuation parameters, overall distribution standard deviation and process capability index, record the defect type and spatial coordinates, and generate a quality distribution map.

2. The method for detecting electrode coating quality based on the fusion of reflected and transmitted ultrasound according to claim 1, characterized in that, The joint association model in S3 is expressed as follows: ， in, For coating thickness, The speed at which ultrasound propagates in the coating. The time difference between the reflected signal and the echo at the interface. The angle of ultrasound incidence. For the acoustic impedance of the coating, For the current collector acoustic impedance, The surface density of the coating. For transmitted signal energy, The initial energy of the incident ultrasound. The ultrasonic attenuation coefficient of the coating is given.

3. The method for detecting electrode coating quality based on the fusion of reflected and transmitted ultrasound according to claim 1, characterized in that, The model solution in S4 is expressed as follows: ， in, For the calculated thickness, This is the standard sample thickness calibration value. The number of characteristic parameters of the reflected signal. The reflection feature weighting coefficient, These are the measured reflection characteristic parameters. These are standard reflection characteristic parameters. This represents the number of characteristic parameters of the transmitted signal. For transmission characteristic weighting coefficients, These are the measured transmission characteristic parameters. These are standard transmission characteristic parameters. To calculate the density later, The areal density calibration value for standard samples. These are the reflection and transmission characteristic weighting coefficients corresponding to the surface density calculation.

4. The method for detecting electrode coating quality based on the fusion of reflected and transmitted ultrasound according to claim 1, characterized in that, The high-dimensional defect-sensitive feature vector construction in S5 is expressed as follows: ， in, For defect-sensitive feature vectors, Additional echo amplitude for the reflected signal, This is the reference value for the amplitude of the normal reflected signal. This represents the energy deviation value of the transmitted signal. This is the reference value for the energy of a normal transmitted signal. This represents the center frequency offset of the transmitted signal. The initial center frequency, For reflected signals With transmitted signal covariance, These are the standard deviations of the reflected signal and the transmitted signal, respectively.

5. The method for detecting electrode coating quality based on the fusion of reflected and transmitted ultrasound according to claim 1, characterized in that, The output of the pattern recognition model in S5 is expressed as follows: ， in, This is the result of the defect category determination. For the number of support vectors, For Lagrange multipliers, For the category labels corresponding to the support vectors, For kernel function, For the first The feature vectors corresponding to each support vector. This is the hyperplane bias term for classification.

6. The method for detecting electrode coating quality based on the fusion of reflected and transmitted ultrasound according to claim 1, characterized in that, The process capability index calculation in S6 is expressed as follows: ， in, For process capability index, To meet the quality characteristics and specification limits, For quality characteristics, the lower specification limit, To measure the mean of the data, To measure the standard deviation of the data, The sample size is denoted by Skew, and Skew is the skewness coefficient of the measurement data.

7. The method for detecting electrode coating quality based on the fusion of reflected and transmitted ultrasound according to claim 1, characterized in that, S3 includes the following steps: S31, sort out the acoustic impedance matching relationship, sound wave propagation path and energy attenuation law of ultrasonic waves propagating in multilayer media of coating and current collector, and clarify the physical generation mechanism of each characteristic parameter of reflected signal and transmitted signal; S32, based on the theory of vertical incidence and propagation of sound waves, establishes the correlation between the amplitude of the first wave of the reflected wave, the time difference of the interface echo, the coating thickness, and the acoustic impedance of the coating and the current collector. S33, combining the principle of transmitted wave energy attenuation, the law of center frequency shift and propagation time characteristics, constructs the correspondence between the characteristic parameters of transmitted signals and the coating thickness and areal density; S34 integrates the correlation between reflection and transmission signals, eliminates the cross-interference of thickness and density changes, and forms a joint correlation model structure that includes multiple variables.

8. The method for detecting electrode coating quality based on the fusion of reflected and transmitted ultrasound according to claim 1, characterized in that, S4 includes the following steps: S41, Select a standard sample with known thickness and areal density, collect the reflection and transmission signals of the standard sample through an ultrasonic probe, and extract the corresponding characteristic parameters as calibration reference data; S42, Substitute the calibration reference data into the joint correlation model, solve for the undetermined coefficients and calibration parameters in the model, and determine the quantitative calculation relationship of the model; S43. Input the measured feature parameters extracted in S2 into the calibrated joint correlation model according to the corresponding category, and solve the simultaneous equations to obtain the preliminary thickness and areal density calculation results. S44. Compare the preliminary calculation results with the known parameters of the standard sample, adjust the iteration step size and convergence conditions in the model solution process, and output the final thickness and areal density data.

9. The method for detecting electrode coating quality based on the fusion of reflected and transmitted ultrasound according to claim 1, characterized in that, S5 includes the following steps: S51, perform waveform analysis on the aligned reflected signal, capture the unexpected additional echo signal generated by internal defects in the coating, and record the amplitude and occurrence time parameters of the additional echo; S52, analyze the energy variation amplitude, waveform distortion degree and center frequency offset of the transmitted signal to distinguish between normal fluctuations and abnormal changes; S53, verify the physical consistency between the reflected signal anomaly and the transmitted signal anomaly in terms of spatial location and trend of change, and screen out the feature combination corresponding to the real defect. S54 integrates the abnormal parameters of the reflected signal, the abnormal parameters of the transmitted signal, and the consistency verification results into a high-dimensional feature vector, which is then input into a pre-trained pattern recognition model to complete the defect category classification.

10. A electrode coating quality inspection system based on the fusion of reflected and transmitted ultrasound, characterized in that, The system is applied to the electrode coating quality detection method based on the fusion of reflected and transmitted ultrasound as described in claim 1, and includes an ultrasonic signal acquisition unit, a multi-dimensional feature parameter extraction unit, a multi-layer medium ultrasonic propagation joint modeling unit, a coating quality parameter real-time calculation unit, a defect feature fusion and category determination unit, and a quality data visualization and transmission unit. The ultrasonic signal acquisition unit acquires the reflected wave signal sequence and the transmitted wave signal sequence of the electrode coating area, and aligns the reflected and transmitted signals of the same measurement point in a precise spatial position on the time axis according to the probe mechanical layout and trigger timing. The multi-dimensional feature parameter extraction unit extracts the first reflection wave amplitude and interface echo time difference from the aligned reflection signal, and extracts energy, center frequency shift, and propagation time feature parameters from the transmission signal. The multilayer medium ultrasonic propagation joint modeling unit constructs a correlation model between the characteristic parameters of reflected and transmitted signals and the coating thickness and areal density based on the propagation characteristics of ultrasonic waves in coatings and current collectors. The real-time calculation unit for coating quality parameters inputs the extracted feature parameters into the joint model and combines them with the calibration parameters of the standard sample to complete the real-time calculation of thickness and areal density. The defect feature fusion and category determination unit integrates the features of extra echo of reflected signal, anomaly of transmitted signal and consistency between the two, and classifies defects through a pattern recognition model. The quality data visualization and transmission unit calculates horizontal, vertical, and overall consistency indicators, records the spatial coordinates of defects and generates a quality distribution map, and transmits structured data to the manufacturing execution system or programmable logic controller through a standard interface.

Images