A method for detecting internal defects of an oil injector based on thermal imaging

Abstract

The application belongs to the technical field of fuel injector detection, and provides a fuel injector internal defect detection method based on thermal imaging, which comprises the following steps: constructing a linear sweep frequency model to generate a driving frequency variation sequence and a driving pulse sequence; collecting a thermal infrared image sequence of the surface of the fuel injector shell by using a thermal imaging device when the driving pulse sequence is applied; obtaining a shell targeted area corresponding to the spatial position of the internal return spring; calculating a transient temperature rise rate sequence; calculating a historical Pearson correlation coefficient by using the historical data of the same type of normal fuel injector by means of the Pearson correlation coefficient method and setting a linear correlation threshold; calculating the current Pearson correlation coefficient of the transient temperature rise rate sequence of the fuel injector to be detected and the driving frequency variation sequence; comparing the current Pearson correlation coefficient with the linear correlation threshold; and determining whether the return spring has an early mechanical fatigue defect. The application can combine the shell surface thermal infrared imaging and the linear sweep frequency driving to determine whether the return spring in the fuel injector has an early mechanical fatigue defect.

Description

Technical Field

[0001] This invention belongs to the field of fuel injector inspection technology, specifically a method for detecting internal defects in fuel injectors based on thermal imaging. Background Technology

[0002] The fuel injector is a precision component of the engine's fuel injection system. The return spring inside the fuel injector is responsible for providing mechanical return force the moment the solenoid coil is de-energized, ensuring that the injector's needle valve quickly sets and cuts off fuel injection.

[0003] In existing fuel injector factory testing, flow meters, pressure sensors, or electrical characteristic analyzers are typically used to test the fuel injection quantity, return quantity, and opening / closing response time. However, under actual automotive operating conditions, the return spring inside the fuel injector is highly susceptible to mechanical fatigue defects due to long-term reciprocating motion. Early mechanical fatigue defects in the return spring are extremely difficult to detect in traditional factory testing. When these early defects occur, the injector's normal flow and pressure parameters are still within acceptable ranges, leading to the injector being missed during traditional factory testing.

[0004] When the return spring experiences initial mechanical fatigue defects, its axial movement trajectory during high-frequency compression and extension will become laterally unstable, leading to abnormal friction between the outer ring of the return spring and the inner guide hole wall of the injector, which exacerbates local heat accumulation.

[0005] In existing technologies, a few methods utilize thermal imaging to inspect fuel injectors. However, these methods are limited to monitoring the overall steady-state temperature of the fuel injector and are mainly used to troubleshoot faults such as coil short circuits or complete nozzle blockage. Existing technologies have not explored the mapping relationship between initial mechanical fatigue defects of the return spring and the localized heating rate of the fuel injector surface. Therefore, this invention provides a method for detecting internal defects in fuel injectors based on thermal imaging. Summary of the Invention

[0006] In order to overcome the shortcomings of the prior art, at least one technical problem raised in the background art is solved.

[0007] The technical solution adopted by this invention to solve its technical problem is: This invention provides a method for detecting internal defects in fuel injectors based on thermal imaging, comprising: The detection duration and drive frequency range of the injector under test are obtained, a linear sweep frequency model is constructed, a drive frequency change sequence is generated, and the corresponding drive pulse sequence is calculated based on the drive frequency change sequence. During the process of applying a driving pulse sequence to the injector under test, a thermal imaging device is used to acquire a sequence of thermal infrared images of the surface of the injector housing. Obtain the target area on the surface of the injector housing corresponding to the spatial position of the internal reset spring, and calculate the transient temperature rise rate sequence of the target area based on the thermal infrared image sequence; Obtain the historical transient temperature rise rate sequence and the corresponding historical drive frequency change sequence of a historical normal injector of the same model as the injector under test and whose internal return spring does not have early mechanical fatigue defects. Calculate the historical Pearson correlation coefficient between the historical transient temperature rise rate sequence and the corresponding historical drive frequency change sequence using the Pearson correlation coefficient method, and set a linear correlation threshold. Calculate the current Pearson correlation coefficient between the transient temperature rise rate sequence and the drive frequency change sequence of the injector under test. If the current Pearson correlation coefficient is greater than or equal to the linear correlation threshold, it is determined that the return spring inside the injector under test does not have early mechanical fatigue defects. If the current Pearson correlation coefficient is less than the linear correlation threshold, it is determined that the return spring inside the injector under test has early mechanical fatigue defects.

[0008] As a further improvement of the present invention, the specific process of constructing the linear frequency sweep model is as follows: Obtain the preset factory test configuration file of the injector under test, and extract the test duration and drive frequency range for fatigue testing from the factory test configuration file; The driving frequency range includes the starting driving frequency and the ending driving frequency; Based on the detection duration, the starting drive frequency, and the ending drive frequency, the frequency change slope is calculated. The calculation process for the frequency change slope is: the difference between the starting drive frequency and the ending drive frequency, divided by the detection duration. Using time as the independent variable and driving frequency as the dependent variable, the initial driving frequency is used as the initial intercept, and the slope of frequency change is used as the coefficient of the first term to construct a linear function relationship in one variable. This linear function relationship is then used as a linear frequency sweep model.

[0009] As a further improvement of the present invention, the specific process for generating the driving pulse sequence is as follows: Set a discretized time step and divide the detection time into multiple consecutive time nodes according to the time step; Substitute each time point into the linear frequency sweep model in sequence to calculate the instantaneous driving frequency corresponding to each time point; Arrange the instantaneous driving frequencies corresponding to all time points in chronological order and combine them to form a discrete sequence of driving frequency changes; Obtain the preset constant pulse width; Based on each instantaneous driving frequency in the driving frequency change sequence, calculate the pulse period corresponding to each instantaneous driving frequency. The pulse period is the reciprocal of the instantaneous driving frequency. Within each pulse cycle, the first segment, equal to a constant pulse width, is set as a high-level drive signal, and the remaining segment within the pulse cycle is set as a low-level turn-off signal. According to the time sequence of the drive frequency change sequence, the high-level drive signal and the low-level turn-off signal in each pulse cycle are sequentially spliced ​​to generate a drive pulse sequence for controlling the on and off of the solenoid coil of the injector under test.

[0010] As a further improvement of the present invention, the specific process of acquiring the thermal infrared image sequence of the surface of the injector housing under test using a thermal imaging device is as follows: The drive pulse sequence is input into the drive test circuit, which converts the drive pulse sequence into a power drive signal to control the electromagnetic coil inside the injector under test to switch on and off according to the drive pulse sequence. Fix the thermal imaging device on the preset observation bracket, and adjust the focal length and field of view of the thermal imaging device so that the entire surface of the injector housing under test is within the effective field of view of the thermal imaging device. When the drive test circuit starts to output the drive pulse sequence, the thermal imaging device is triggered, so that the thermal imaging device continuously takes pictures of the surface of the injector housing under test at a preset image acquisition frame rate. When the detection time ends, the application of the driving pulse sequence is stopped and the thermal imaging device stops capturing images. The multiple frames of thermal infrared images with absolute timestamps are arranged in chronological order to generate a thermal infrared image sequence.

[0011] As a further improvement of the present invention, the specific process of obtaining the target area on the surface of the injector housing corresponding to the spatial position of the internal reset spring is as follows: Obtain a 3D CAD model of the same model as the injector under test, and extract the axial physical distance range of the return spring relative to the reference point of the injector under test from the 3D CAD model; The reference point is the point where the geometric center of the top surface of the injector or the geometric center of the bottom nozzle is located. The axial physical distance range is the actual physical length from the reference point to the upper and lower end faces of the return spring along the central axis of the injector under test. The first frame image in the thermal infrared image sequence is acquired, the pixel coordinate system corresponding to the first frame image is established, the overall outer contour of the injector under test in the first frame image is identified by the image edge detection algorithm, the pixel coordinates of the reference point corresponding to the three-dimensional CAD model are located on the overall outer contour, and the central axis of the injector is extracted. Based on the pre-calibrated scale of image pixels and actual physical size, starting from the pixel coordinates of the reference point, the axial physical distance interval is converted into a pixel distance interval along the central axis. In the first frame image, the upper horizontal boundary line corresponding to the upper end face of the reset spring and the lower horizontal boundary line corresponding to the lower end face of the reset spring are determined. In the first frame image, the upper and lower horizontal boundary lines are used as the upper and lower boundaries, and the left and right edges of the overall outer contour within the upper and lower boundaries are used as the left and right boundaries. The region of interest is extracted and used as the target region.

[0012] As a further improvement of the present invention, the specific process for calculating the transient heating rate sequence of the target region is as follows: Based on each frame of the thermal infrared image sequence, the absolute temperature value of all pixels in the target area is extracted, and the representative temperature value of the target area is calculated. The representative temperature value is the effective average value of the absolute temperature values ​​of all pixels in the target area after removing the extreme values. Based on the timestamp order of each frame image, the extracted representative temperature values ​​are combined to construct the transient temperature time series of the target region. Set a sliding time window; The sliding time window is moved point by point along the transient temperature time series, and the target calculation point of the sliding window corresponds one-to-one with each of the divided time nodes. When performing calculations for each time node, the difference between the representative temperature value at the end of the sliding time window based on the time node and the representative temperature value at the beginning of the time window is obtained, and then divided by the time width of the sliding time window to obtain the transient heating rate at the time node. The transient heating rates calculated at each time point are arranged in chronological order to generate a transient heating rate sequence with the same length as the driving frequency change sequence.

[0013] As a further improvement of the present invention, the specific process of setting the linear correlation threshold is as follows: Data is retrieved from a pre-established database of injector thermal characteristics standards; The data in the injector thermal characteristic standard database was obtained through the following offline database construction process: Several injectors of the same model as the injector to be tested were pre-selected, and it was confirmed that the internal return spring structure was intact and there were no early mechanical fatigue defects, as historical normal injectors; Under the same ambient temperature and heat dissipation conditions as the injector under test, the same linear sweep frequency model, driving parameters and thermal imaging image processing algorithm were used to conduct benchmark tests on each historical normal injector, and the historical transient temperature rise rate sequence and historical driving frequency change sequence corresponding to each historical normal injector were extracted and saved. For each historical normal injector, obtain the historical transient temperature rise rate sequence and the historical drive frequency change sequence corresponding to the time nodes within the same test cycle, and treat the historical transient temperature rise rate sequence and the historical drive frequency change sequence as a set of one-dimensional discrete variables of equal length. Calculate the covariance between the historical transient heating rate sequence and the historical driving frequency change sequence, and calculate the first standard deviation of the historical transient heating rate sequence and the second standard deviation of the historical driving frequency change sequence, respectively. Divide the calculated covariance by the product of the first standard deviation and the second standard deviation to obtain the historical Pearson correlation coefficient of the historical normal injectors; Obtain the historical Pearson correlation coefficients for all historical normal injectors, and calculate the statistical average of all historical Pearson correlation coefficients, setting it as the linear correlation threshold.

[0014] As a further improvement of the present invention, the specific process for calculating the current Pearson correlation coefficient is as follows: Obtain the transient heating rate sequence and the driving frequency change sequence corresponding to the time nodes of the injector under test within the same test cycle, and treat the transient heating rate sequence and the driving frequency change sequence as a set of one-dimensional discrete variables of equal length. Calculate the covariance between the transient heating rate sequence and the driving frequency change sequence, and calculate the third standard deviation of the transient heating rate sequence and the fourth standard deviation of the driving frequency change sequence, respectively. Divide the calculated covariance by the product of the third and fourth standard deviations to obtain the current Pearson correlation coefficient.

[0015] As a further improvement of the present invention, the specific process for determining that the return spring inside the injector under test does not have early mechanical fatigue defects is as follows: The calculated current Pearson correlation coefficient is numerically compared with the linear correlation threshold. If the current Pearson correlation coefficient is greater than or equal to the linear correlation threshold, it is determined that the return spring inside the injector under test does not have early mechanical fatigue defects.

[0016] As a further improvement of the present invention, the specific process for determining that the return spring inside the injector under test has an early mechanical fatigue defect is as follows: If the current Pearson correlation coefficient is less than the linear correlation threshold, it is determined that the return spring inside the injector under test has an early mechanical fatigue defect.

[0017] The beneficial effects of this invention are as follows: 1. By constructing a linear frequency sweep model, a continuously varying drive frequency sequence and a drive pulse sequence, corresponding one-to-one with time nodes, can be generated within a preset detection time. This controls the on / off switching of the injector's electromagnetic coil according to the frequency variation, causing the internal reset spring to be sequentially excited under the continuously changing drive frequency. This provides a multi-frequency, time-controllable excitation process for thermal response-based defect detection. The detection time and drive frequency range are directly taken from the factory test configuration file, eliminating the need for manual setting of excitation parameters. This ensures consistent excitation conditions for different injectors and batches, improving the comparability of detection results. 2. Using thermal imaging equipment to acquire thermal infrared image sequences of the injector housing surface allows for non-contact acquisition of temperature distribution information. The detection process does not require disassembling the injector, preventing damage to the injector structure and facilitating non-destructive and convenient defect detection.

[0018] 2. By extracting the axial physical distance range of the reset spring through a 3D CAD model and combining image edge detection with pixel-to-physical size scale conversion, the target area corresponding to the spatial position of the internal reset spring can be accurately located on the outer shell surface. This ensures that the collected temperature information corresponds to the state of the reset spring, reducing the interference of heat changes in other structural components on the detection results and improving the targeting accuracy. After removing extreme values ​​from the pixel temperature of the target area, the effective average value is taken as the representative temperature. A sliding time window is used to calculate the transient heating rate at each time node, which can suppress the interference of image noise and local abnormal pixels on temperature characterization and obtain a stable and continuous transient heating rate sequence. This sequence corresponds one-to-one with the driving frequency change sequence at each time node and has the same length, providing directly comparable paired data for subsequent correlation analysis.

[0019] 3. By quantifying the linear correlation between the transient heating rate sequence and the driving frequency change sequence using the Pearson correlation coefficient, the current thermal response characteristics of the return spring inside the injector under test can be numerically characterized. Using the statistical average of historical Pearson correlation coefficients calculated from historically normal injectors of the same model under the same test conditions as the linear correlation threshold ensures that the judgment criteria match the thermal characteristics of the specific injector model, improving the specificity and reliability of the judgment. By numerically comparing the current Pearson correlation coefficient with the linear correlation threshold, and determining whether the return spring has early mechanical fatigue defects, early mechanical fatigue defects in the injector's internal return spring can be objectively and quantitatively identified, reducing reliance on manual inspection experience and facilitating timely detection before defects cause significant functional failure, providing a basis for injector maintenance and replacement. Attached Figure Description

[0020] The invention will now be further described with reference to the accompanying drawings.

[0021] Figure 1 This is a schematic diagram illustrating the principle of extracting the axial physical distance range of the reset spring in an embodiment of the present invention; Explanation of reference numerals in the attached diagram: 10 - Injector housing; 20 - Central axis; 30 - Return spring; 40 - Reference point; L1 - Actual physical length to the upper end face of the return spring; L2 - Actual physical length to the lower end face of the return spring; L3 - Axial physical distance range.

[0022] Figure 2 This is a schematic diagram illustrating the principle of edge detection and coordinate localization of the first frame image in an embodiment of the present invention; Figure labeling: 50 - First frame image; 60 - Pixel coordinate system; 70 - Overall outer contour; 80 - Pixel coordinates of the reference point; 90 - Central axis.

[0023] Figure 3 This is a schematic diagram illustrating the principle of target region conversion and interception in an embodiment of the present invention; Figure labeling: 70 - Overall outer contour; 80 - Pixel coordinates of the reference point; 90 - Central axis; 100 - Upper horizontal boundary line; 110 - Lower horizontal boundary line; 120 - Pixel distance range; 130 - Left edge; 140 - Right edge; 150 - Target area.

[0024] Figure 4 This is a flowchart of the steps of a method for detecting internal defects in fuel injectors based on thermal imaging, according to the present invention. Detailed Implementation

[0025] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments. Example 1:

[0026] like Figure 1 As shown in the embodiment of the present invention, a method for detecting internal defects in fuel injectors based on thermal imaging includes: Step 1: Obtain the detection duration and drive frequency range of the injector under test, construct a linear sweep frequency model, generate a drive frequency change sequence, and calculate and generate the corresponding drive pulse sequence based on the drive frequency change sequence; In some specific embodiments, the specific process of obtaining the detection duration and drive frequency range of the injector under test in step one is as follows: Obtain the preset factory test configuration file of the injector to be tested, and extract the detection time and drive frequency range for fatigue testing from the factory test configuration file. The driving frequency range includes the starting driving frequency and the ending driving frequency.

[0027] It should be noted that in practical engineering applications, in order to fully monitor the potential mechanical fatigue defects of the return spring without burning out the electromagnetic coil, the starting drive frequency is usually set to a low-frequency value simulating the engine idling condition. In some embodiments, the starting drive frequency can be 10Hz. The ending drive frequency is usually set to a high-frequency value simulating the engine's limit speed condition. In some embodiments, the ending drive frequency ranges from 100Hz to 150Hz. The detection duration is usually set to the duration that allows the surface of the injector under test to generate a significant temperature rise. In some embodiments, the detection duration ranges from 30 seconds to 120 seconds.

[0028] In some specific embodiments, the specific process of constructing the linear sweep frequency model in step one is as follows: Based on the acquired detection duration, the starting drive frequency, and the ending drive frequency, the frequency change slope is calculated; the formula for calculating the frequency change slope is: the difference between the starting drive frequency and the ending drive frequency, divided by the detection duration. Using time as the independent variable and driving frequency as the dependent variable, the initial driving frequency is used as the initial intercept, and the slope of frequency change is used as the coefficient of the first-order term to construct a linear function relationship in one variable. This linear function relationship is then used as the linear sweep frequency model.

[0029] It should be understood that the linear sweep frequency model can ensure that the frequency of mechanical actuation applied to the return spring increases strictly and uniformly over time, thus providing a benchmark for subsequent evaluation of the linearity of heat accumulation.

[0030] In some specific embodiments, the specific process of generating the driving frequency change sequence in step one is as follows: A discretized time step is set, and the detection duration is divided into multiple consecutive time nodes according to the time step; Substitute each of the aforementioned time points into the linear frequency sweep model in sequence to calculate the instantaneous driving frequency corresponding to each time point; Arrange the instantaneous driving frequencies corresponding to all time points in chronological order to form a discrete sequence of driving frequency changes.

[0031] It should be noted that the preferred range of the time step is 0.01 seconds to 0.1 seconds. The main purpose of setting a discretized time step is to divide the continuous detection time into multiple discrete time nodes to improve the control resolution of the driving frequency change. The smaller the time step, the denser the discretized time nodes, and the higher the corresponding detection accuracy. In practical engineering applications, the lower limit of the time step value is limited by the sampling frequency of the thermal imaging device. If the set time step is smaller than the single sampling period of the thermal imaging device, the high-frequency and dense driving time nodes will not be able to acquire the corresponding thermal infrared image data, which will not only fail to improve the actual detection accuracy, but will also cause data redundancy in the control system. Therefore, the setting of the time step must be matched with the sampling frequency of the thermal imaging device.

[0032] In some specific embodiments, the specific process of calculating and generating the corresponding driving pulse sequence based on the driving frequency change sequence in step one is as follows: Obtain the preset constant pulse width; It should be noted that the constant pulse width is the power-on duration of a single power-on of the injector under test. In some embodiments, the constant pulse width can be set to 2 milliseconds. Keeping the pulse width constant is to eliminate the interference of changes in heat generation during a single power-on on the overall temperature rise, and to ensure that the temperature rise is only related to the drive frequency. Based on each instantaneous driving frequency in the driving frequency change sequence, calculate the pulse period corresponding to each instantaneous driving frequency, where the pulse period is the reciprocal of the instantaneous driving frequency; Within each pulse cycle, the first segment of the pulse is set to a high-level drive signal, and the remaining segment within the pulse cycle is set to a low-level turn-off signal. According to the time sequence of the drive frequency change sequence, the high-level drive signal and the low-level turn-off signal in each pulse period are sequentially spliced ​​to generate the drive pulse sequence used to control the on / off of the electromagnetic coil of the injector under test.

[0033] Step 2: During the process of applying the drive pulse sequence to the injector under test, use a thermal imaging device to acquire a sequence of thermal infrared images of the surface of the injector housing. In some specific embodiments, the specific process of acquiring a thermal infrared image sequence of the surface of the injector housing using a thermal imaging device during step two, while applying a driving pulse sequence to the injector under test, is as follows: The generated drive pulse sequence is input into a dedicated injector drive test circuit, which converts the drive pulse sequence into a power drive signal to control the electromagnetic coil inside the injector under test to switch on and off at high frequency according to the drive pulse sequence. It should be noted that the high-level drive signal and the low-level turn-off signal in the drive pulse sequence correspond to the on and off states of the injector under test, respectively. When the high-level drive signal is applied, the electromagnetic coil is energized to generate electromagnetic force. The electromagnetic force overcomes the elastic force of the return spring, causing the injector under test to open. At this time, the return spring is compressed. When the low-level turn-off signal is applied, the electromagnetic coil is de-energized, the electromagnetic force disappears, and the return spring releases its mechanical return force to close the injector under test. At this time, the return spring extends. Through the high-frequency alternation of high and low levels in the drive pulse sequence, the internal return spring is forced to compress and extend, thereby effectively stimulating its potential early mechanical fatigue defects.

[0034] The thermal imaging device is fixed on a preset observation bracket, and the focal length and field of view of the thermal imaging device are adjusted so that the surface of the injector housing under test is completely within the effective field of view of the thermal imaging device.

[0035] It should also be noted that, in some embodiments, in order to eliminate the interference of the reflection of the metal casing on the accuracy of infrared temperature measurement, a matte coating with a known infrared emissivity, such as a special black paint with high emissivity, can be pre-coated on the surface of the injector casing to be tested before the test, and the corresponding emissivity parameter can be set in the thermal imaging device. When the drive test circuit starts to output the drive pulse sequence, the thermal imaging device is triggered, so that the thermal imaging device continuously takes pictures of the surface of the injector housing under test at a preset image acquisition frame rate. In some embodiments, the image acquisition frame rate can be set to 10 frames / second to 30 frames / second. Maintaining a frame rate of 10 frames / second to 30 frames / second allows the thermal imaging device to fully record the gradual process of heat conduction without generating excessive data redundancy.

[0036] When the detection duration ends, the application of the driving pulse sequence is stopped and the thermal imaging device stops capturing images. The captured multiple frames of thermal infrared images with absolute timestamps are arranged in chronological order to generate the thermal infrared image sequence.

[0037] Step 3: Obtain the target area on the surface of the injector housing corresponding to the spatial position of the internal reset spring, and calculate the transient temperature rise rate sequence of the target area based on the thermal infrared image sequence; In some specific embodiments, the specific process of obtaining the target area on the surface of the injector housing corresponding to the spatial position of the internal reset spring in step three is as follows: Obtain a 3D CAD model of the same model as the injector under test, and extract the axial physical distance range of the return spring relative to the reference point of the injector under test from the 3D CAD model; wherein, the reference point is the point where the geometric center point of the top surface of the injector or the geometric center point of the bottom nozzle is located, and the axial physical distance range is the actual physical length from the reference point to the upper and lower end surfaces of the return spring along the central axis of the injector under test; Combination Figure 1 As shown, Figure 1 This is a schematic diagram illustrating the principle of extracting the axial physical distance range of the return spring. (See diagram below.) Figure 1 As shown, the reference point (40) is determined on the external structure of the injector housing (10) under test (e.g., Figure 1 The intersection of the top surface of the fuel injector and the central axis (20), along Figure 1 In the direction of the central axis (20) shown, the actual physical length (L1) from the reference point (40) downward to the upper end face of the return spring (30) and the actual physical length (L2) from the reference point (40) downward to the lower end face of the return spring (30) are measured; the spatial span (L3) formed between these two actual physical lengths (L1 and L2) is the axial physical distance range; The first frame image in the thermal infrared image sequence is obtained, and a pixel coordinate system corresponding to the first frame image is established. The overall outer contour of the injector to be tested in the first frame image is identified using an image edge detection algorithm. The pixel coordinates of the reference point corresponding to the three-dimensional CAD model are located on the overall outer contour, and the central axis of the injector is extracted. The specific process for establishing the pixel coordinate system corresponding to the first frame image is as follows: The two-dimensional pixel coordinate system is constructed by taking the top left corner of the first frame image as the origin of the coordinate system; taking the horizontal rightward direction of the first frame image as the positive direction of the x-axis; taking the vertical downward direction of the first frame image as the positive direction of the y-axis; and taking the individual pixel of the image as the basic unit of the coordinate axis.

[0038] It should be understood that constructing a two-dimensional pixel coordinate system with the top left corner of the image as the origin is a standard matrix indexing method in the fields of computer vision and digital image processing. The establishment of the pixel coordinate system can directly transform the physical spatial distribution in the thermal infrared image into the row and column index of the digital matrix, thereby providing a basic mathematical calculation framework for subsequent calls to image edge detection algorithms and for performing ratio conversions from physical size to pixel size.

[0039] Combination Figure 2 As shown, Figure 2This is a schematic diagram of the principle of edge detection and coordinate positioning of the first frame image. The first frame image (50) in the thermal infrared image sequence is obtained, and the corresponding pixel coordinate system (60) is established. The overall outer contour (70) of the injector to be tested in the first frame image (50) is identified by the image edge detection algorithm, and the pixel coordinates (80) of the reference point corresponding to the three-dimensional CAD model are located on the overall outer contour (70). At the same time, the central axis (90) of the injector in the image is extracted based on the symmetry of the overall outer contour (70). Based on the pre-calibrated scale of image pixels and actual physical size, starting from the pixel coordinates of the reference point, the axial physical distance interval is converted into a pixel distance interval along the central axis direction, thereby determining the upper horizontal boundary line corresponding to the upper end face of the reset spring and the lower horizontal boundary line corresponding to the lower end face of the reset spring in the first frame image. In the first frame image, a rectangular or polygonal region of interest (ROI) is extracted using the upper and lower horizontal boundary lines as the upper and lower boundaries, and the left and right edges of the overall outer contour within these boundaries as the left and right boundaries. This ROI is then used as the target region. Since the injector position is fixed during the test, the spatial coordinates of the target region will be directly reused in all subsequent frames of the thermal infrared image sequence.

[0040] Combination Figure 3 As shown, Figure 3 This is a schematic diagram illustrating the principle of targeted region conversion and interception.

[0041] For example, the conversion and extraction process of the above-mentioned target region is as follows: Let the pre-calibrated scale be K (e.g., K = 10 pixels / mm). In the pixel coordinate system of the first frame image, the vertical coordinate of the pixel coordinate (80) of the located reference point is Y0 (e.g., Y0 = 100). The actual physical length to the upper end face of the reset spring extracted according to the aforementioned steps is L1 (e.g., L1 = 20 mm), and the actual physical length to the lower end face of the reset spring is L2 (e.g., L2 = 35 mm).

[0042] like Figure 3 As shown, starting from the pixel coordinates (80) of the reference point, along the direction of the central axis (90) (i.e., the direction in which the vertical coordinate of the image increases), the pixel vertical coordinates of the upper horizontal boundary line (100) are calculated as Y1=Y0+L1×K (i.e., 300); the pixel vertical coordinates of the lower horizontal boundary line (110) are calculated as Y2=Y0+L2×K (i.e., 450); the area formed between these two horizontal boundary lines is the converted pixel distance interval (120).

[0043] In the first frame image, the upper horizontal boundary line (100) with ordinate Y1 and the lower horizontal boundary line (110) with ordinate Y2 are used as the upper and lower boundaries, and the left edge (130) and right edge (140) of the overall outer contour (70) within this interval are used as the left and right boundaries, thus accurately extracting the image. Figure 3 The target region (150) is shown in the shaded area.

[0044] In some specific embodiments, the specific process of calculating the transient heating rate sequence of the target region based on the thermal infrared image sequence in step three is as follows: Based on each frame of the thermal infrared image sequence, the absolute temperature value of all pixels in the target area is extracted, and the representative temperature value of the target area is calculated. It should be understood that the representative temperature value is used to eliminate local noise interference, and the effective average value of the absolute temperature values ​​of all pixels in the target area after removing extreme values ​​is usually used as the representative temperature value. Based on the timestamp order of each frame image, the extracted representative temperature values ​​are combined to construct the transient temperature time series of the target region. Set a sliding time window; In some embodiments, the width of the sliding time window is set to 1 to 2 seconds to smooth out the high-frequency thermal noise of the thermal imaging device itself; The sliding time window is slid point by point along the transient temperature time series. It should be noted that, in order to ensure the time sequence alignment of the data for the subsequent Pearson correlation coefficient calculation, the target calculation point here corresponds one-to-one with each of the time nodes divided in step one. When calculating for each time node, the difference between the representative temperature value at the end of the sliding time window based on the time node and the representative temperature value at the beginning of the sliding time window is obtained, and divided by the time width of the sliding time window to obtain the transient heating rate at the time node. The transient heating rates calculated at each of the aforementioned time points are arranged in chronological order to generate a transient heating rate sequence with the same length as the driving frequency change sequence.

[0045] Step 4: Obtain the historical transient temperature rise rate sequence and the corresponding historical drive frequency change sequence of a historical normal injector of the same model as the injector under test and whose internal return spring does not have early mechanical fatigue defects. Calculate the historical Pearson correlation coefficient between the historical transient temperature rise rate sequence and the corresponding historical drive frequency change sequence using the Pearson correlation coefficient method, and set a linear correlation threshold. In some specific embodiments, the specific process of obtaining the historical transient temperature rise rate sequence and the corresponding historical drive frequency change sequence of a historical normal injector of the same model as the injector under test and whose internal return spring does not have early mechanical fatigue defects is as follows: Data is retrieved from a pre-established database of injector thermal characteristics standards; The data in the injector thermal characteristic standard database is obtained through the following offline database construction process: Several injectors of the same model as the injector to be tested are selected in advance, and have undergone rigorous destructive physical analysis (DPA) or long-term durability bench testing to confirm that the internal return spring structure is intact and there are no early mechanical fatigue defects. These injectors are used as the historical normal injectors. Under the same ambient temperature and heat dissipation conditions as the injector under test, a benchmark test is performed on each of the historical normal injectors using the same linear sweep frequency model, driving parameters, and thermal imaging image processing algorithm as in steps one to three above, thereby extracting and saving the historical transient temperature rise rate sequence and the historical driving frequency change sequence corresponding to each historical normal injector.

[0046] In some specific embodiments, the process of calculating the historical Pearson correlation coefficient between the historical transient heating rate sequence and the corresponding historical driving frequency change sequence using the Pearson correlation coefficient method, and setting the linear correlation threshold, is as follows: For each of the historical normal injectors, the historical transient temperature rise rate sequence and the historical drive frequency change sequence corresponding to the time nodes within the same test cycle are obtained, and the historical transient temperature rise rate sequence and the historical drive frequency change sequence are used as a set of one-dimensional discrete variables of equal length. Calculate the covariance between the historical transient heating rate sequence and the historical driving frequency change sequence, and calculate the first standard deviation of the historical transient heating rate sequence and the second standard deviation of the historical driving frequency change sequence, respectively. The calculated covariance is divided by the product of the first standard deviation and the second standard deviation to obtain the historical Pearson correlation coefficient of the historical normal injector.

[0047] It should be understood that, since the normal internal return spring has a stable trajectory during reciprocating motion and does not experience lateral instability, and the normal mechanical friction heat generation and electromagnetic coil heat generation are both strictly linearly positively correlated with the driving frequency, the calculated historical Pearson correlation coefficient is usually a positive number that is very close to 1 (e.g., between 0.95 and 0.99). Obtain the historical Pearson correlation coefficients of all historical normal injectors, calculate the statistical average of all the historical Pearson correlation coefficients, and directly set the statistical average as the linear correlation threshold.

[0048] It should be understood that the linear correlation threshold is the minimum linear following capability that the local heating rate of the same type of injector must maintain under normal operating conditions in response to changes in the drive frequency.

[0049] Step 5: Calculate the current Pearson correlation coefficient between the transient temperature rise rate sequence and the drive frequency change sequence of the injector under test. If the current Pearson correlation coefficient is greater than or equal to the linear correlation threshold, it is determined that the return spring inside the injector under test does not have early mechanical fatigue defects. If the current Pearson correlation coefficient is less than the linear correlation threshold, it is determined that the return spring inside the injector under test has early mechanical fatigue defects.

[0050] In some specific embodiments, the specific process of calculating the current Pearson correlation coefficient between the transient temperature rise rate sequence and the drive frequency change sequence of the injector under test in step five is as follows: Obtain the transient heating rate sequence and the driving frequency change sequence corresponding to the time nodes of the injector under test in the same test cycle, and treat the transient heating rate sequence and the driving frequency change sequence as a set of one-dimensional discrete variables of equal length. Calculate the covariance between the transient heating rate sequence and the driving frequency change sequence, and calculate the third standard deviation of the transient heating rate sequence and the fourth standard deviation of the driving frequency change sequence, respectively. The calculated covariance is divided by the product of the third and fourth standard deviations to obtain the current Pearson correlation coefficient. This current Pearson correlation coefficient is used to quantify the degree to which the local heat accumulation rate of the tested injector linearly follows changes in drive frequency.

[0051] It should be understood that the first and second standard deviations, as well as the third and fourth standard deviations described in the embodiments of the present invention, correspond to completely different data samples and calculation subjects. Specifically, the first and second standard deviations are calculated based on historical test data of historical normal injectors (i.e., historical transient temperature rise rate sequences and historical drive frequency change sequences), and their purpose is to establish a benchmark for comparison (i.e., the linear correlation threshold); while the third and fourth standard deviations are calculated based on the real-time test data of the current injector under test (i.e., the current transient temperature rise rate sequence and drive frequency change sequence), and their purpose is to quantitatively assess the actual health status of the current injector under test.

[0052] In some specific embodiments, the specific process of determining that the return spring inside the injector under test does not have early mechanical fatigue defects in step five, if the current Pearson correlation coefficient is greater than or equal to the linear correlation threshold, is as follows: The calculated current Pearson correlation coefficient is compared numerically with the preset linear correlation threshold. If the current Pearson correlation coefficient is greater than or equal to the linear correlation threshold, it indicates that the heating rate of the target area of ​​the injector under test can strictly follow the linear increase of the driving frequency and maintain a linear increase. It should be understood that the axial movement trajectory of the internal return spring remains stable during the high-frequency compression and extension process, and there is no abnormal friction with the inner guide hole wall of the injector. Its heat accumulation is completely in line with the normal heating law of electromagnetic coil and conventional mechanical actuation. Based on this, it is determined that the return spring inside the injector under test does not have early mechanical fatigue defects.

[0053] In some specific embodiments, the specific process of determining that the return spring inside the injector under test has an early mechanical fatigue defect in step five, if the current Pearson correlation coefficient is less than the linear correlation threshold, is as follows: If the current Pearson correlation coefficient is less than the linear correlation threshold, it indicates that the heating rate of the target area of ​​the fuel injector under test has deviated from the normal linear following relationship with the driving frequency, and a nonlinear heat surge or irregular fluctuation has occurred. It should be understood that the internal return spring has developed initial mechanical fatigue due to long-term reciprocating motion, resulting in uneven stiffness distribution and lateral instability during high-frequency operation. This lateral instability causes abnormal friction between the outer ring of the return spring and the inner guide hole wall of the injector. The additional heat generated by this abnormal friction seriously disrupts the original linear temperature rise pattern. Based on this, it was determined that the return spring inside the fuel injector under test had an early mechanical fatigue defect, thus identifying hidden fatigue risks in advance before the normal flow and pressure indicators showed abnormalities.

[0054] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for detecting internal defects in fuel injectors based on thermal imaging, characterized in that: include: The detection duration and drive frequency range of the injector under test are obtained, a linear sweep frequency model is constructed, a drive frequency change sequence is generated, and the corresponding drive pulse sequence is calculated based on the drive frequency change sequence. During the process of applying a driving pulse sequence to the injector under test, a thermal imaging device is used to acquire a sequence of thermal infrared images of the surface of the injector housing. Obtain the target area on the surface of the injector housing corresponding to the spatial position of the internal reset spring, and calculate the transient temperature rise rate sequence of the target area based on the thermal infrared image sequence; Obtain the historical transient temperature rise rate sequence and the corresponding historical drive frequency change sequence of a historical normal injector of the same model as the injector under test and whose internal return spring does not have early mechanical fatigue defects. Calculate the historical Pearson correlation coefficient between the historical transient temperature rise rate sequence and the corresponding historical drive frequency change sequence using the Pearson correlation coefficient method, and set a linear correlation threshold. Calculate the current Pearson correlation coefficient between the transient temperature rise rate sequence and the drive frequency change sequence of the injector under test. If the current Pearson correlation coefficient is greater than or equal to the linear correlation threshold, it is determined that the return spring inside the injector under test does not have early mechanical fatigue defects. If the current Pearson correlation coefficient is less than the linear correlation threshold, it is determined that the return spring inside the injector under test has early mechanical fatigue defects.

2. The method for detecting internal defects in fuel injectors based on thermal imaging according to claim 1, characterized in that, The specific process of constructing the linear sweep frequency model is as follows: Obtain the preset factory test configuration file of the injector under test, and extract the test duration and drive frequency range for fatigue testing from the factory test configuration file; The driving frequency range includes the starting driving frequency and the ending driving frequency; Based on the detection duration, the starting drive frequency, and the ending drive frequency, the frequency change slope is calculated. The calculation process for the frequency change slope is: the difference between the starting drive frequency and the ending drive frequency, divided by the detection duration. Using time as the independent variable and driving frequency as the dependent variable, the initial driving frequency is used as the initial intercept, and the slope of frequency change is used as the coefficient of the first term to construct a linear function relationship in one variable. This linear function relationship is then used as a linear frequency sweep model.

3. The method for detecting internal defects in fuel injectors based on thermal imaging according to claim 1, characterized in that, The specific process for generating the driving pulse sequence is as follows: Set a discretized time step and divide the detection time into multiple consecutive time nodes according to the time step; Substitute each time point into the linear frequency sweep model in sequence to calculate the instantaneous driving frequency corresponding to each time point; Arrange the instantaneous driving frequencies corresponding to all time points in chronological order and combine them to form a discrete sequence of driving frequency changes; Obtain the preset constant pulse width; Based on each instantaneous driving frequency in the driving frequency change sequence, calculate the pulse period corresponding to each instantaneous driving frequency. The pulse period is the reciprocal of the instantaneous driving frequency. Within each pulse cycle, the first segment, equal to a constant pulse width, is set as a high-level drive signal, and the remaining segment within the pulse cycle is set as a low-level turn-off signal. According to the time sequence of the drive frequency change sequence, the high-level drive signal and the low-level turn-off signal in each pulse cycle are sequentially spliced ​​to generate a drive pulse sequence for controlling the on and off of the solenoid coil of the injector under test.

4. The method for detecting internal defects in fuel injectors based on thermal imaging according to claim 1, characterized in that, The specific process of acquiring the thermal infrared image sequence of the surface of the injector housing under test using thermal imaging equipment is as follows: The drive pulse sequence is input into the drive test circuit, which converts the drive pulse sequence into a power drive signal to control the electromagnetic coil inside the injector under test to switch on and off according to the drive pulse sequence. Fix the thermal imaging device on the preset observation bracket, and adjust the focal length and field of view of the thermal imaging device so that the entire surface of the injector housing under test is within the effective field of view of the thermal imaging device. When the drive test circuit starts to output the drive pulse sequence, the thermal imaging device is triggered, so that the thermal imaging device continuously takes pictures of the surface of the injector housing under test at a preset image acquisition frame rate. When the detection time ends, the application of the driving pulse sequence is stopped and the thermal imaging device stops capturing images. The multiple frames of thermal infrared images with absolute timestamps are arranged in chronological order to generate a thermal infrared image sequence.

5. The method for detecting internal defects in fuel injectors based on thermal imaging according to claim 1, characterized in that, The specific process for obtaining the target area on the surface of the injector housing corresponding to the spatial position of the internal reset spring is as follows: Obtain a 3D CAD model of the same model as the injector under test, and extract the axial physical distance range of the return spring relative to the reference point of the injector under test from the 3D CAD model; The reference point is the point where the geometric center of the top surface of the injector or the geometric center of the bottom nozzle is located. The axial physical distance range is the actual physical length from the reference point to the upper and lower end faces of the return spring along the central axis of the injector under test. The first frame image in the thermal infrared image sequence is acquired, the pixel coordinate system corresponding to the first frame image is established, the overall outer contour of the injector under test in the first frame image is identified by the image edge detection algorithm, the pixel coordinates of the reference point corresponding to the three-dimensional CAD model are located on the overall outer contour, and the central axis of the injector is extracted. Based on the pre-calibrated scale of image pixels and actual physical size, starting from the pixel coordinates of the reference point, the axial physical distance interval is converted into a pixel distance interval along the central axis. In the first frame image, the upper horizontal boundary line corresponding to the upper end face of the reset spring and the lower horizontal boundary line corresponding to the lower end face of the reset spring are determined. In the first frame image, the upper and lower horizontal boundary lines are used as the upper and lower boundaries, and the left and right edges of the overall outer contour within the upper and lower boundaries are used as the left and right boundaries. The region of interest is extracted and used as the target region.

6. The method for detecting internal defects in fuel injectors based on thermal imaging according to claim 1, characterized in that, The specific process for calculating the transient heating rate sequence of the target region is as follows: Based on each frame of the thermal infrared image sequence, the absolute temperature value of all pixels in the target area is extracted, and the representative temperature value of the target area is calculated. The representative temperature value is the effective average value of the absolute temperature values ​​of all pixels in the target area after removing the extreme values. Based on the timestamp order of each frame image, the extracted representative temperature values ​​are combined to construct the transient temperature time series of the target region. Set a sliding time window; The sliding time window is moved point by point along the transient temperature time series, and the target calculation point of the sliding window corresponds one-to-one with each of the divided time nodes. When performing calculations for each time node, the difference between the representative temperature value at the end of the sliding time window based on the time node and the representative temperature value at the beginning of the time window is obtained, and then divided by the time width of the sliding time window to obtain the transient heating rate at the time node. The transient heating rates calculated at each time point are arranged in chronological order to generate a transient heating rate sequence with the same length as the driving frequency change sequence.

7. The method for detecting internal defects in fuel injectors based on thermal imaging according to claim 1, characterized in that, The specific process for setting the linear correlation threshold is as follows: Data is retrieved from a pre-established database of injector thermal characteristics standards; The data in the injector thermal characteristic standard database was obtained through the following offline database construction process: Several injectors of the same model as the injector to be tested were pre-selected, and it was confirmed that the internal return spring structure was intact and there were no early mechanical fatigue defects, as historical normal injectors; Under the same ambient temperature and heat dissipation conditions as the injector under test, the same linear sweep frequency model, driving parameters and thermal imaging image processing algorithm were used to conduct benchmark tests on each historical normal injector, and the historical transient temperature rise rate sequence and historical driving frequency change sequence corresponding to each historical normal injector were extracted and saved. For each historical normal injector, obtain the historical transient temperature rise rate sequence and the historical drive frequency change sequence corresponding to the time nodes within the same test cycle, and treat the historical transient temperature rise rate sequence and the historical drive frequency change sequence as a set of one-dimensional discrete variables of equal length. Calculate the covariance between the historical transient heating rate sequence and the historical driving frequency change sequence, and calculate the first standard deviation of the historical transient heating rate sequence and the second standard deviation of the historical driving frequency change sequence, respectively. Divide the calculated covariance by the product of the first standard deviation and the second standard deviation to obtain the historical Pearson correlation coefficient of the historical normal injectors; Obtain the historical Pearson correlation coefficients for all historical normal injectors, and calculate the statistical average of all historical Pearson correlation coefficients, setting it as the linear correlation threshold.

8. The method for detecting internal defects in fuel injectors based on thermal imaging according to claim 1, characterized in that, The specific process for calculating the current Pearson correlation coefficient is as follows: Obtain the transient heating rate sequence and the driving frequency change sequence corresponding to the time nodes of the injector under test within the same test cycle, and treat the transient heating rate sequence and the driving frequency change sequence as a set of one-dimensional discrete variables of equal length. Calculate the covariance between the transient heating rate sequence and the driving frequency change sequence, and calculate the third standard deviation of the transient heating rate sequence and the fourth standard deviation of the driving frequency change sequence, respectively. Divide the calculated covariance by the product of the third and fourth standard deviations to obtain the current Pearson correlation coefficient.

9. The method for detecting internal defects in fuel injectors based on thermal imaging according to claim 1, characterized in that, The specific process for determining that the return spring inside the injector under test does not have early mechanical fatigue defects is as follows: The calculated current Pearson correlation coefficient is numerically compared with the linear correlation threshold. If the current Pearson correlation coefficient is greater than or equal to the linear correlation threshold, it is determined that the return spring inside the injector under test does not have early mechanical fatigue defects.

10. The method for detecting internal defects in fuel injectors based on thermal imaging according to claim 1, characterized in that, The specific process for determining that the return spring inside the fuel injector under test has an early mechanical fatigue defect is as follows: If the current Pearson correlation coefficient is less than the linear correlation threshold, it is determined that the return spring inside the injector under test has an early mechanical fatigue defect.

Images