Automobile Parts Processing Monitoring Method Based on Machine Vision

Abstract

The present invention relates to the field of image processing technology, and in particular to a method for monitoring the processing of automotive parts based on machine vision, comprising: dividing the surface of a part into multiple regions, continuously acquiring thermal images, and using a cumulative sum algorithm to calculate the cumulative deviation values ​​of the temperature variation coefficient and fractal dimension of each region in each thermal image. Based on the cumulative deviation values ​​of the temperature variation coefficient and fractal dimension of each region in each thermal image, and the correlation between the two, the degree of abnormality of each region is determined. Based on the starting point of the abnormal trend of the temperature variation coefficient and fractal dimension of each region, the overlapping time of the abnormal trend of each region is determined, and an abnormal risk assessment is performed based on the abnormal degree and trend overlapping time of each region, thereby realizing processing monitoring. This method improves the sensitivity to abnormal temperature changes and the accuracy of processing monitoring.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a method for monitoring automobile parts processing based on machine vision. Background Art

[0002] In the automotive manufacturing industry, the machining accuracy of key components is a core factor in determining the safety and operational reliability of the vehicle. Drilling, a key process for forming and finishing metal parts, is widely used in the manufacturing of cylindrical parts such as engine crankshafts, camshafts, transmission input shafts, and brake discs. Because these parts play a vital role in key systems such as vehicle power transmission and braking, their machining quality directly impacts vehicle performance and user safety. Therefore, ensuring the stability and quality of the drilling process is crucial.

[0003] With the rapid development of industrial vision technology, machine vision-based drilling process monitoring technology, with its advantages of non-contact and high real-time performance, has gradually become a mainstream solution in the industry. Infrared thermal imaging technology, by capturing infrared images of the drilling area and using temperature changes as the core monitoring indicator, indirectly assesses the process status. When problems such as increased tool wear, abnormal drilling parameters, or uneven workpiece material occur during the drilling process, the localized temperature in the drilling area is often accompanied by an abnormal increase. Therefore, monitoring temperature field changes can effectively identify potential risks.

[0004] However, most existing monitoring technologies rely on static temperature thresholds or significant temperature rises in hot spots as the basis for abnormality judgment. Alarms are triggered only when the temperature anomaly reaches a high amplitude and its spatial distribution clearly shows concentrated characteristics. However, in actual production, the drilling process is affected by the coupling of multiple factors such as drilling parameter fluctuations, differences in workpiece material properties, and progressive tool wear, and the temperature field exhibits complex dynamic changes. Early cumulative overheating anomalies often manifest as localized small temperature rises, and their signals are easily overwhelmed by normal temperature fluctuations or environmental noise, making it difficult for the system to capture the initial characteristics of the anomaly. This lagging monitoring mechanism makes it impossible to control processing risks in the budding stage, which may lead to serious consequences such as tool breakage, workpiece scrapping, and even equipment failure, affecting the quality stability and production efficiency of automotive parts.

[0005] Therefore, it is urgent to develop an intelligent monitoring method that can integrate the dynamic fluctuation characteristics of the temperature field and the evolution laws of the spatial structure to achieve early identification of abnormal temperatures, thereby providing more forward-looking decision support for process optimization and fault prevention, and improving the quality control level of the automotive parts processing process. Summary of the Invention

[0006] In order to solve the problem of delayed monitoring of progressive temperature anomalies during drilling of automobile parts, the present invention proposes an automobile parts processing monitoring method based on machine vision, comprising:

[0007] Dividing the surface of the automobile part into multiple areas and continuously collecting multiple thermal images of the surface during the drilling process;

[0008] The cumulative deviation value of the temperature variation coefficient and the cumulative deviation value of the fractal dimension in each thermal image of each region are determined by using the cumulative sum algorithm;

[0009] Taking any thermal image as the current thermal image, the abnormality degree of each area in the current thermal image is determined based on the temperature variation coefficient, the cumulative deviation value of the fractal dimension of each area in the current thermal image and the historical thermal image, as well as the correlation between the temperature variation coefficient and the fractal dimension;

[0010] Determine the starting point of the abnormal trend of the temperature variation coefficient and fractal dimension of each region based on the cumulative deviation value of the temperature variation coefficient and fractal dimension in the current thermal image and the historical thermal image, and then determine the overlapping duration of the abnormal trend of each region;

[0011] Based on the abnormality level of each area in the current thermal image and the overlapping duration of the abnormal trend of the area, an abnormal risk assessment is performed on the corresponding moment of the current thermal image, and the processing process of automobile parts is monitored according to the abnormal risk assessment results.

[0012] This technical solution first divides the surface of the part into multiple areas and collects continuous thermal images, ensuring comprehensive monitoring of the dynamic process of temperature changes during processing. Secondly, the cumulative sum algorithm is used to calculate the cumulative deviation value of the temperature variation coefficient and fractal dimension of each area, effectively extracting key information on the numerical fluctuation of temperature and the spatial complexity of temperature. By comparing the current thermal image and the historical thermal image, combined with the correlation between the temperature variation coefficient and the fractal dimension, the degree of abnormality in each area can be accurately judged, which helps to capture progressive temperature anomalies in a timely manner. Furthermore, by analyzing the overlapping duration of the abnormal trends of the temperature variation coefficient and the fractal dimension, an early warning mechanism for abnormal changes is provided, which helps to detect potential processing problems in advance. Compared with traditional instantaneous monitoring, this monitoring method based on cumulative analysis and trend evaluation improves the sensitivity and accuracy of anomaly detection and avoids quality risks and production risks caused by delayed monitoring.

[0013] Preferably, performing an abnormal risk assessment on the corresponding moment of the current thermal image includes:

[0014] The average of the abnormality levels of all areas in the current thermal image / each historical thermal image is used as the abnormality level of the current thermal image / each historical thermal image, respectively; a linear regression analysis is performed based on the abnormality levels of the current thermal image and all historical thermal images to obtain the rate of change and direction of change of the abnormality level of the current thermal image; the average of the overlapping duration of the abnormal trends of all areas is used as the duration of the abnormality; based on the rate of change and direction of change of the abnormality level of the current thermal image, the duration of the abnormality, and the discrete degree of the abnormality level of all historical thermal images, the abnormal risk index at the corresponding moment of the current thermal image is determined to quantify the abnormal risk assessment results.

[0015] This technical solution more accurately identifies potential issues in the machining process by comprehensively evaluating the rate of change, direction, and duration of anomaly levels. Linear regression analysis not only reveals the dynamics of anomalies but also predicts their development trends, avoiding misjudgment based on single data points. Analysis of overlap duration and discreteness further enhances the assessment of anomaly persistence, making anomaly risk assessment more timely and reliable.

[0016] Preferably, the abnormal trend starting point of the temperature variation coefficient and fractal dimension of each region is determined based on the following method: the cumulative deviation value of the temperature variation coefficient of each region in the current thermal imaging image and the historical thermal imaging image is compared with a preset first threshold value in chronological order; if the abnormal cumulative deviation value of the temperature variation coefficient of the region in a certain thermal imaging image is greater than the first threshold value for the first time, the thermal imaging image is used as the abnormal trend starting point of the temperature variation coefficient of the region; the cumulative deviation value of the fractal dimension of each region in the current thermal imaging image and the historical thermal imaging image is compared with a preset second threshold value in chronological order; if the abnormal cumulative deviation value of the fractal dimension of the region in a certain thermal imaging image is greater than the second threshold value for the first time, the thermal imaging image is used as the abnormal trend starting point of the fractal dimension of the region.

[0017] By comparing the cumulative deviation value with a preset threshold, this technical solution accurately identifies the moment when an anomaly first exceeds a limit, marking the beginning of an abnormal trend. By dually monitoring the temperature coefficient of variation and fractal dimension, this method comprehensively identifies the starting point of potential anomalies from both the heat distribution and surface structure perspectives, providing a more accurate and stable basis for subsequent risk assessment and fault warning.

[0018] Preferably, the current thermal imaging image and the historical thermal imaging image are set based on the following method: I thermal imaging images before the current thermal imaging image are used as historical thermal imaging images, where I is a preset positive integer.

[0019] Preferably, the degree of abnormality of each region in the current thermal image is determined based on the following formula:

[0020]

[0021] In the formula, For the The abnormality of the area in the current thermal image, For the The cumulative deviation value of the temperature variation coefficient of each area in the current thermal image, For the The average of the cumulative deviations of the temperature variation coefficient in all historical thermal images of the region, For the The cumulative deviation value of the fractal dimension of the region in the current thermal image, For the The mean of the cumulative deviation values ​​of the fractal dimension of the region in all historical thermal images, For the The Pearson correlation coefficient of temperature variation coefficient and fractal dimension in all historical thermal images of a region.

[0022] This technical solution quantifies the degree of anomaly in each area by integrating the temperature coefficient of variation, fractal dimension, and correlation analysis. This multi-dimensional comprehensive assessment method not only enhances sensitivity to local anomalies in each area, but also finds more accurate basis for anomaly judgment among multiple data sources, thereby improving the monitoring system's ability to identify gradual and subtle anomalies.

[0023] Preferably, the abnormal risk index of the current thermal image at the corresponding moment is determined based on the following formula:

[0024]

[0025] In the formula, is the abnormal risk index of the current thermal image at the corresponding moment, is the rate of change of the abnormality of the current thermal image, is the absolute value symbol, is the duration of the abnormality, is the total duration corresponding to all historical thermal images, is the natural exponential function, is the variance of the anomaly degree of all historical thermal images, is a sign function that reflects the direction of change in the abnormality of the current thermal image.

[0026] This technical solution comprehensively quantifies anomaly risk by comprehensively considering the rate of change, duration, historical volatility, and direction of change. The combination of rate of change and duration reflects the suddenness and persistence of an anomaly, while the combination of direction of change more accurately captures anomaly trends. Further, by integrating historical data to assess the stability and trend of anomalies, risk assessment becomes more robust.

[0027] Preferably, it is characterized in that the overlapping duration of the abnormal trend of each area is determined based on the following method: the starting point of the abnormal trend of the temperature variation coefficient of each area to the current thermal imaging image is used as the first segment, and the starting point of the abnormal trend of the fractal dimension of each area to the current thermal imaging image is used as the second segment; the duration corresponding to the overlapping part of the first segment and the second segment is used as the overlapping duration of the abnormal trend of the area.

[0028] This technical solution accurately measures the persistence and development of abnormal trends in each region by segmenting the abnormal trends of the temperature coefficient of variation and fractal dimension. Overlapping the time period from the start of the abnormal trends of the two key indicators to the current moment effectively identifies the temporal intersection of these abnormal changes, thereby capturing the duration of the abnormal pattern.

[0029] Preferably, the temperature variation coefficient is used to measure the intensity of temperature fluctuations, and the temperature variation coefficient of each region in each thermal image is the ratio of the standard deviation to the mean of the temperature values ​​of all pixels corresponding to the region in the thermal image.

[0030] Preferably, the fractal dimension is used to measure the morphological complexity of the temperature distribution, and the fractal dimension of each region in each thermal image is determined by a differential box dimension method.

[0031] Preferably, the processing process of automobile parts is monitored according to the abnormal risk assessment results, including: presetting a first abnormal risk index threshold and a second abnormal risk index threshold, and the second abnormal risk index threshold is greater than the first abnormal risk index threshold; if the abnormal risk index at the moment corresponding to the current thermal imaging image is less than the first abnormal risk index threshold, it is determined that the processing is normal; if the abnormal risk index at the moment corresponding to the current thermal imaging image is greater than or equal to the first abnormal risk index threshold and less than the second abnormal risk index threshold, it is determined that the processing has a slight abnormality; if the abnormal risk index at the moment corresponding to the current thermal imaging image is greater than or equal to the second abnormal risk index threshold, it is determined that the processing has a serious abnormality.

[0032] The present invention has the following effects:

[0033] The present invention identifies progressive anomalies during the machining process by comprehensively analyzing the changes in the temperature coefficient of variation and fractal dimension of different surface areas of the part during the machining process. By dynamically monitoring multiple key indicators, the sensitivity to subtle changes in the machining process is improved, and the lag of traditional monitoring methods in dealing with progressive anomalies is avoided, thereby providing accurate risk assessment results and improving the accuracy of machining monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 It is a schematic flow chart of the method of the present invention. DETAILED DESCRIPTION

[0035] The technical solutions in the embodiments of the present invention will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present invention.

[0036] The present invention provides a method for monitoring automobile parts processing based on machine vision, such as Figure 1 As shown in , including:

[0037] S1: Continuously collect multiple thermal images of the surface of automotive parts during the drilling process.

[0038] During the drilling process of automotive parts, drilling temperatures typically fluctuate locally and change dynamically over time. The distribution of drilling temperature is often influenced by factors such as cutting force, cutting speed, tool-workpiece contact area, and the thermal conductivity of the material. Due to these factors, temperatures can vary across the machined surface, typically with higher temperatures near the cutting point and lower temperatures further away. Heat accumulation and diffusion are typically localized, and temperature variations at different locations can reflect different machining conditions and potential problems.

[0039] Therefore, first obtain the position to be drilled on the surface of the part, divide the maximum inscribed rectangle with the position to be drilled as the center, and use the area within the maximum inscribed rectangle as the region of interest. The region of interest is evenly divided into grids of the same size, with each grid as a region.

[0040] During drilling, the temperature changes most dramatically at the drilling point, and these changes gradually spread to surrounding areas, particularly those closer to the cutting point. Therefore, this ROI ensures that critical areas of temperature change during drilling are included in the monitoring scope.

[0041] By dividing the area, the temperature changes in each area can be analyzed independently, which enables more detailed analysis and monitoring of local temperature changes on the surface of the part, thereby more accurately capturing local anomalies and fluctuations.

[0042] Next, infrared thermal imaging equipment was installed around the drilling equipment, with a capture frequency set to 10Hz. High-frequency acquisition can better capture transient abnormal changes. The camera angle was adjusted to ensure that its field of view fully covered the part surface. Thermal images were continuously captured during the drilling process. Each image was of the same size, and the pixel value of each pixel in the image represented the temperature.

[0043] Since the areas of the part surface have been divided before, the corresponding position of each area of ​​the part surface in each thermal image can be obtained. By continuously analyzing the temperature changes of each area of ​​the part surface in multiple thermal images, it is possible to identify whether its temperature changes are abnormal.

[0044] S2: Extract the temperature variation coefficient and fractal dimension of each region in each thermal image.

[0045] In order to achieve precise monitoring of the temperature changes in each area of ​​the part surface, the temperature variation coefficient and fractal dimension are introduced. Through the complementary indicators of these two dimensions, the temperature characteristics of each area in each thermal infrared image are comprehensively analyzed.

[0046] The temperature coefficient of variation (CTV) measures the intensity of temperature fluctuations and is defined as the ratio of the standard deviation of the temperature to the mean. A smaller CTV for a region generally indicates a more uniform and stable temperature distribution, suggesting a higher likelihood of stable drilling. Conversely, a larger CTV indicates greater temperature fluctuations, suggesting a higher likelihood of abnormal drilling.

[0047] For example, the first The region in The temperature variation coefficient in the thermal image is the first The region in The ratio of the standard deviation to the mean of the temperature values ​​of all pixels corresponding to a thermal image.

[0048] The fractal dimension can describe the spatial filling ability of a geometric object. The higher the fractal dimension of the temperature field, the greater the complexity of its spatial distribution (such as temperature gradient changes and irregular patch distribution). In the present invention, the fractal dimension is used to reflect the morphological complexity of the temperature distribution in each area during the drilling process. During the drilling process, temperature anomalies are usually manifested as complex morphology and irregular distribution of local hot zone boundaries. A higher fractal dimension indicates that the edge of the hot zone is rougher and more fragmented, reflecting the existence of local instability or abnormal diffusion in the temperature field. If the fractal dimension of a region is smaller, it means that the temperature distribution in the region is more stable and uniform, indicating that the drilling process is in a relatively normal and stable state. If the fractal dimension of a region is larger, it means that the temperature distribution in the region is complex and irregular, suggesting that there may be processing anomalies or abnormal diffusion of local heat.

[0049] In one embodiment, the fractal dimension of each region in each thermal image is obtained by using a differential box dimension method, which is a common method for calculating fractal dimension and is suitable for complexity analysis of two-dimensional scalar fields (such as temperature, humidity, etc.).

[0050] Specifically, they include:

[0051] S21: Obtain the area corresponding to each area on the surface of the part in each thermal image.

[0052] For each thermal image, a coordinate system is constructed with the lower left corner as the origin, the horizontal rightward as the abscissa, and the vertical upward as the ordinate.

[0053] Get the surface of the part The region in The corresponding area (composed of all pixels) in the thermal image , The length is , width is , the units of length and width are both pixels.

[0054] S22: Set the multi-scale partition box.

[0055] Set initial scale ,Pick , and generate a decreasing scale sequence: ,in, , is the number of scale layers, and the empirical value is 6.

[0056] Scale-by-scale box division and temperature range calculation: For example, first divide the box grid: Divide into sides of If the square box cannot be divided evenly, the edge is filled (the filling value is the average temperature of the region). The number of square boxes obtained after division is:

[0057]

[0058] In this formula, For scale Down, The number of square boxes it is divided into, For

[0059] Ceiling symbol, To take the maximum value.

[0060] S23: Get the valid box at each scale.

[0061] Calculate the temperature range of each box: subtract the minimum temperature value from the maximum temperature value of all pixels contained in each box to obtain the temperature range of the box.

[0062] Effective box statistics: Set the temperature extreme difference threshold ε, the empirical value is 0.5 times the standard deviation of the temperature values ​​of all pixels included, the boxes with a temperature range greater than ε are counted as valid boxes at this scale, and the number is recorded as This operation is to retain only the boxes with significant temperature changes and exclude the uniform low temperature areas that mask the fractal characteristics.

[0063] S24: Determine the fractal dimension through the power-law relationship between the number of valid boxes and the scale.

[0064] For fractal structures, when the scale When the number of effective boxes changes The scale satisfies the following power law relationship:

[0065]

[0066] Where D is the fractal dimension, which represents the spatial complexity of the fractal structure and is usually a non-integer greater than 0. Indicates a proportional relationship. The overall formula shows that the number of effective boxes is proportional to the negative power of the scale, that is, when the scale When the number of effective boxes changes Changes according to the law of power function.

[0067] Therefore, based on the power law relationship, logarithmic transformation and linear fitting are performed to determine the fractal dimension: taking the natural logarithm of both sides of the power law relationship, the linear equation is obtained:

[0068]

[0069] in, is the fractal dimension, and its physical meaning is: , the temperature distribution is approximately linear, such as the uniform heat diffusion zone; , The temperature distribution is highly irregular. Reflects the rate at which the number of effective boxes shrinks as the scale decreases. The larger D is, the higher the spatial complexity of the temperature field is. is a natural constant.

[0070] For each scale ,calculate and , and As the horizontal axis, As the vertical coordinate, the linear fit is done by the least squares method, and the slope of the linear fit is , whose absolute value is the fractal dimension D.

[0071] For each region corresponding to a region in each thermal image, steps S22 to S24 are followed to obtain the fractal dimension of the region in each thermal image.

[0072] In summary, the temperature variation coefficient and fractal dimension characterize the temperature field characteristics of each area from the two dimensions of numerical fluctuation and spatial complexity respectively. The former temperature variation coefficient can sensitively reflect the discrete degree and abnormal amplitude of temperature values, and the fractal dimension reveals the spatial complexity of temperature distribution in each area. The combination of the two improves the accuracy of monitoring.

[0073] S3: Determine the respective cumulative deviation values ​​of the temperature coefficient of variation and the fractal dimension in each thermal image for each region.

[0074] In drilling process monitoring, the temperature variation coefficient and fractal dimension can preliminarily characterize the temperature characteristics of each area. To further determine whether there are progressive anomalies in each area, such as slow wear of the tool and gradual deterioration of heat dissipation conditions, which lead to abnormal temperatures in certain areas, the CUSUM (cumulative sum) algorithm is used to obtain the deviation of the temperature variation coefficient and fractal dimension of each area in each thermal image from the baseline value. This can amplify small trend changes into significant cumulative deviation values. The core advantage of this algorithm lies in its sensitivity to slow offsets. Compared with traditional threshold detection, it can identify persistent abnormal fluctuations in indicators earlier, which is particularly suitable for scenarios such as drilling that require real-time tracking of parameter gradients.

[0075] Therefore, for each thermal image, the previous I thermal images are used as the historical thermal images of the thermal image. The CUSUM (cumulative sum) algorithm is used to calculate the cumulative deviation value of the temperature variation coefficient and the cumulative deviation value of the fractal dimension of each region in each thermal image, where the empirical value of I is 10.

[0076] First Regions and Take the thermal image as an example, The region in The process of obtaining the cumulative deviation value of the temperature variation coefficient in a thermal image is as follows:

[0077] Get the first All historical thermal images of the thermal image, the first The region in The temperature variation coefficients of all historical thermal images of a thermal image constitute a sequence, and the The region in The cumulative deviation of the temperature coefficient of variation in the thermal image is 0, that is, the cumulative deviation of the initialization temperature variation coefficient is 0, then The region in The cumulative deviation of the temperature coefficient of variation in each thermal image is:

[0078]

[0079] In this formula, For the The region in The cumulative deviation value of the temperature variation coefficient in each thermal image, To obtain the maximum value function, For the The region in The temperature variation coefficient in each thermal image is For the The region in The temperature variation coefficient in each thermal image is For the The region in The mean of the temperature variation coefficients in all historical thermal images at the moment, is the tolerance parameter, which indicates the degree of deviation of the tolerated data. Usually, , For the The region in The standard deviation of the temperature coefficient of variation in all historical thermal images at a given moment.

[0080] First Regions and As an example of a thermal image, The region in The process of obtaining the cumulative deviation value of the fractal dimension in a thermal image is as follows:

[0081] First get the All historical thermal images of the thermal image, the first The region in The fractal dimensions of all historical thermal images of a thermal image constitute a sequence, and the The region in The cumulative deviation of the fractal dimension in the thermal image is 0, that is, the cumulative deviation value of the initialized fractal dimension is 0, then The region in The cumulative deviation of the fractal dimension in the thermal image is:

[0082]

[0083] In this formula, For the The region in The cumulative deviation of the fractal dimension in the thermal image, To obtain the maximum value function, For the The region in The fractal dimension of a thermal image, For the The region in The fractal dimension of a thermal image, For the The region in The mean value of the fractal dimension of all historical thermal images at the moment, is the tolerance parameter, which indicates the degree of deviation of the tolerated data. Usually, , For the The region in The standard deviation of the fractal dimension of all historical thermal images at a given moment.

[0084] In summary, both formulas in this step are based on a cumulative sum algorithm. Their core purpose is to capture progressive anomalies in the indicator through recursive cumulative deviations, enabling accurate monitoring of the continuous changes in temperature characteristics during the drilling process. Ultimately, this step yields the cumulative deviation of the temperature coefficient of variation for each region in each thermal image, as well as the cumulative deviation of the fractal dimension for each region in each thermal image.

[0085] S4: Calculate the abnormality degree of each area in the current thermal image.

[0086] During the drilling monitoring process, the abnormal characteristics of the temperature field often show multi-dimensional coupling characteristics, which include significant changes in temperature fluctuations and the complex evolution of spatial distribution patterns.

[0087] This step effectively integrates two types of indicators to obtain the degree of abnormality of each area in the current thermal image, and introduces the Pearson correlation coefficient to deeply explore the synergistic relationship between temperature fluctuations and temperature space complexity. When the changing trends of the two show a significant positive correlation, for example, when the increase in temperature dispersion and the complexity of the hot zone structure occur simultaneously, the comprehensive abnormality degree indicator will amplify the abnormal signal through the coupling mechanism, accurately capturing the multi-dimensional feature collaborative evolution pattern under working conditions such as tool wear and heat dissipation abnormalities.

[0088] In one embodiment, the abnormality level of each region in the current thermal image is determined based on the following formula:

[0089]

[0090] In this formula, For the The abnormality of the area in the current thermal image, For the The cumulative deviation value of the temperature variation coefficient of each area in the current thermal image, For the The cumulative deviation value of the fractal dimension of the region in the current thermal image, For the The average of the cumulative deviations of the temperature variation coefficient in all historical thermal images of the region, For the The mean of the cumulative deviation values ​​of the fractal dimension of the region in all historical thermal images, For the The Pearson correlation coefficient of the temperature variation coefficient and fractal dimension of the region in all historical thermal images is ,use Convert it to the interval [0,1]. The larger the value, the closer it is to 1, the stronger the synergistic change trend is; the smaller the value, the closer it is to 0, the less synergistic change trend exists.

[0091] In this formula, Is to use the The average value of the temperature variation coefficient of the region in the historical thermal image is The temperature variation coefficient of each area in the current thermal image is standardized. The larger the value, the The more the temperature fluctuation of an area in the current thermal image deviates from the historical baseline level, the more the temperature fluctuation of the area in the current thermal image deviates from the historical baseline level. The more severe the temperature fluctuation in a region.

[0092] In this formula, Is to use the The mean value of the fractal dimension of the region in the historical thermal image is The fractal dimension of each area in the current thermal image is standardized. The larger the value, the more complex the local temperature change in the area, and the more complex the morphology is.

[0093] The stronger the trend of synergistic changes in temperature fluctuations and spatial complexity, the more likely the area is to be a localized temperature anomaly; conversely, the more likely it is to be a normal temperature area. In summary, normalizing the cumulative deviations of the temperature coefficient of variation and fractal dimension with their respective means in historical data effectively eliminates the interference of single data fluctuations, making the assessment of anomaly severity more stable. By introducing the Pearson correlation coefficient, the relationship between the temperature coefficient of variation and fractal dimension is further combined, improving the accuracy of anomaly detection.

[0094] This design not only conforms to the multi-physics field coupling nature of the temperature field during the drilling process, but also, through the dual mechanisms of data standardization and correlation analysis, provides an evaluation tool with both physical interpretability and statistical reliability for the quantitative identification of local temperature anomalies, thereby improving the sensitivity and positioning accuracy of anomaly detection under complex working conditions.

[0095] S5: Analyze the abnormal trend starting point of the temperature variation coefficient and the abnormal trend starting point of the fractal dimension in each area, and determine the overlapping duration of the abnormal trend in each area.

[0096] In one embodiment, the abnormal trend starting point of the temperature coefficient of variation and the abnormal trend starting point of the fractal dimension of each region are determined based on the following method:

[0097] Obtain the cumulative deviation value of the temperature variation coefficient of each region in the current thermal image, as well as the cumulative deviation value of the temperature variation coefficient of the region in all historical thermal images (of the current thermal image) (the method for obtaining the cumulative deviation value of the temperature variation coefficient of the region in the current thermal image is consistent).

[0098] According to The cumulative deviation value of the temperature variation coefficient of the area in the current thermal image, and the The cumulative deviation value of the temperature variation coefficient of each area in each historical thermal image of the current thermal image is used to determine the The threshold value of the cumulative deviation value of the temperature coefficient of variation of each region is:

[0099]

[0100] In this formula, For the The threshold of the cumulative deviation value of the temperature coefficient of variation in each area, For the The standard deviation of the cumulative deviation of the temperature coefficient of variation in all historical thermal images of the region, is the total number of all historical thermal images.

[0101] This is an approximate statistical fluctuation boundary setting method. Its core idea is derived from the idea of ​​the central limit theorem and the random walk model: as the number of observation points in the time series increases, even if the cumulative deviation value is a purely random fluctuation, the standard deviation of the cumulative deviation value will increase by increase.

[0102] Therefore, in order to prevent the cumulative sum algorithm from misjudging normal fluctuations as abnormal trends, the threshold value needs to be adjusted accordingly to maintain the consistency and stability of the judgment. The standard deviation of the cumulative deviation value can be used to characterize the natural amplitude of temperature fluctuations in the historical thermal imaging map of the area. This threshold value can be regarded as the reasonable upper limit that the cumulative deviation value may reach in the case of purely random fluctuations. Exceeding this range is considered to be the formation of an abnormal trend.

[0103] Therefore, the The cumulative deviation value of the temperature variation coefficient of the area in the current thermal image, and the The cumulative deviation value of the temperature variation coefficient of the region in all historical thermal images of the current thermal image is calculated in chronological order. In contrast, The cumulative deviation of the temperature variation coefficient in a certain area in a certain thermal image is greater than The moment corresponding to the thermal image is taken as the The starting point of the cumulative abnormal trend of the temperature variation coefficient of the region is The temperature variation coefficients of the regions began to show a persistent cumulative anomaly trend.

[0104] Similarly, according to The cumulative deviation value of the fractal dimension of the region in the current thermal image, and the The cumulative deviation value of the fractal dimension of the region in each historical thermal image of the current thermal image is used to determine the The threshold value of the cumulative deviation value of the fractal dimension of a region is:

[0105]

[0106] In this formula, For the The threshold of the cumulative deviation value of the fractal dimension of the region, For the The standard deviation of the cumulative deviation of the fractal dimension of a region in all historical thermal images, is the total number of all historical thermal images.

[0107] The first The cumulative deviation value of the fractal dimension of the region in the current thermal image, and the The cumulative deviation value of the temperature variation coefficient of the region in all historical thermal images of the current thermal image is calculated in chronological order. In contrast, The cumulative deviation value of the fractal dimension of a region in a certain thermal image is greater than The moment corresponding to the thermal image is taken as the The starting point of the cumulative abnormal trend of the fractal dimension of the region indicates the The fractal dimension of each region begins to produce a continuous cumulative abnormal trend.

[0108] In one embodiment, the overlapping duration of the abnormal trend of each area is determined based on the following method: the starting point of the abnormal trend of the temperature variation coefficient of each area to the current thermal imaging image is used as the first segment, and the starting point of the abnormal trend of the fractal dimension of each area to the current thermal imaging image is used as the second segment, and the duration corresponding to the overlapping part of the first segment and the second segment is used as the overlapping duration of the abnormal trend of the area.

[0109] S6: An abnormality risk assessment is performed based on the abnormality degree of each area in the current thermal image and the overlapping duration of the abnormal trend of the area.

[0110] In one embodiment, the abnormal risk assessment process includes:

[0111] For the current thermal image and all its historical thermal images, the average of the abnormality levels of all regions in the current thermal image is used as the abnormality level of the current thermal image, reflecting the overall abnormality level of the current thermal image. The average of the abnormality levels of all regions in each historical thermal image of the current thermal image is used as the abnormality level of the historical thermal image, reflecting the overall abnormality level of the historical thermal image.

[0112] Next, the average of the overlapping durations of the abnormal trends in all regions is taken as the duration of the abnormality.

[0113] Then, a linear regression analysis is performed based on the abnormality of the current thermal image and all historical thermal images to obtain the change rate and direction of the abnormality of the current thermal image:

[0114]

[0115] In this formula, It is the rate of change of the abnormal degree of the current thermal image. It can not only sensitively reflect the direction and slope of the abnormal trend, but also has good smoothness and anti-interference properties. It is a key parameter reflecting the evolution speed of the abnormality. Indicates the time point corresponding to the current thermal image The offset of the center position of all time points corresponding to all historical thermal images of the current thermal image, also called the centralized time index. is the mean of the serial numbers of all historical thermal images corresponding to the current thermal image, is the serial number of the historical thermal image of the current thermal image, is the total number of historical thermal images, The current thermal image The degree of anomaly of the historical thermal images, It is the mean value of the abnormal degree of all historical thermal images of the current thermal image.

[0116] This formula is a linear regression slope estimate. If the trend of abnormality increases, then , if the trend of abnormality decreases, then If there is no obvious trend in the degree of abnormality, then .

[0117] Next, based on the rate and direction of change of the abnormality of the current thermal image, the duration of the abnormality, and the degree of dispersion of the abnormality of all historical thermal images, the abnormality risk index of the current thermal image at the corresponding moment is determined:

[0118]

[0119] In the formula, The abnormal risk index of the current thermal image at the corresponding moment is used to quantify the abnormal risk assessment results. is the rate of change of the abnormality of the current thermal image, is the absolute value symbol, is the duration of the abnormality, is the total duration corresponding to all historical thermal images, is the natural exponential function, is the variance of the anomaly degree of all historical thermal images, is a sign function that reflects the direction of change in the abnormality of the current thermal image. and All are anti-0 parameters, and the values ​​are . It is to prevent The impact on the overall calculation is, When the abnormality of the current thermal image changes at a rate and in a direction that is almost negligible, it only depends on and Conduct abnormal assessment. It is to prevent The impact on the overall calculation is, When there is no continuous abnormal time, it only depends on and Calculate the abnormal risk index.

[0120] In this formula, It reflects the severity of abnormal temperature changes. The more severe the abnormal changes are, the longer the duration will be. The longer it is, the more likely it is that the temperature of the part surface continues to rise at the corresponding moment of the current thermal imaging image, which means that drilling anomalies are more likely to occur. is the duration of the abnormality, is the total time (the time period corresponding to all historical thermal images of the current thermal image), The larger the value, the more likely it is that there is a long-term anomaly, such as a sustained high temperature. Conversely, the more likely it is an occasional anomaly, such as a momentary fluctuation. is the natural exponential function, is the variance of the abnormality of all historical thermal images of the current thermal image, which is used to measure the discreteness of the abnormality. The smaller it is, the more likely it is that there is a local overheating anomaly. Conversely, the more uniform the temperature change is, the less likely it is that there is an anomaly.

[0121] In this formula, is the trend adjustment item, is a symbolic function, greater than 0, ; is equal to 0, ; Less than 0, Only when the abnormal trend is growing, It will be magnified and increased by 0.5 times (experience value), indicating that the risk is increasing; when the abnormal trend remains unchanged, keep observing; when the abnormal trend decreases, the abnormal trend is alleviated, Reduce by 0.5 times.

[0122] In summary, this formula constructs a comprehensive and dynamic index for identifying abnormality types by factoring in factors such as the rate of change of abnormality, the severity of the change, the proportion of abnormality duration, the degree of concentration or dispersion of the abnormality, and the direction of the abnormality trend. This effectively captures abnormality risks during the drilling process. When severe, sustained, and concentrated temperature anomalies occur in the drilling area, the index will increase significantly, issuing a timely warning. Conversely, when the anomalies are dispersed, short-lived, or showing a mitigating trend, the index will decrease accordingly. This helps users efficiently and accurately assess the safety and stability of the drilling process, providing a scientific basis for preventing drilling failures.

[0123] Abnormality risk is quantified by combining current and historical thermal image data. The average abnormality level for each area reflects temperature fluctuations throughout the entire process. Linear regression is used to analyze the rate and direction of change in abnormality levels, revealing the development trend of abnormalities. By analyzing the overlap of abnormal trends and assessing the duration of abnormalities, the abnormality risk is further quantified. Furthermore, the volatility of historical data is taken into account to ensure the comprehensiveness and accuracy of risk assessments.

[0124] S7: Monitor the machining process of automotive parts based on abnormal risk assessment results.

[0125] The first abnormal risk index threshold is preset to 0.5 (empirical value), and the second abnormal risk index threshold is preset to 0.8 (empirical value). If the abnormal risk index is less than 0.5, it means that the temperature change of the part surface is normal during the drilling process, the drilling process is stable, and the processing is judged to be normal; if the abnormal risk index is greater than or equal to 0.5 and less than 0.8, it means that the temperature change of the part surface is slightly abnormal during the drilling process, which may be the initial wear of the tool or local abnormality. It is necessary to continue to pay attention to the development trend and judge that the processing has a slight abnormality; if the abnormal risk index is greater than 0.8, it means that there is a local temperature abnormality during the drilling process, which may be serious wear of the tool, chipping or system failure, and an immediate warning will be issued to remind the staff to check.

[0126] While various embodiments of the present invention have been shown and described herein, it will be obvious to those skilled in the art that such embodiments are provided by way of example only.

Claims

1. The automobile parts processing monitoring method based on machine vision is characterized by: include: The surface of the automobile part is divided into multiple regions, and multiple thermal images of the surface during the drilling process are continuously collected; a cumulative deviation value of the temperature variation coefficient and the cumulative deviation value of the fractal dimension in each thermal image of each region are determined using a cumulative sum algorithm; Taking any thermal image as the current thermal image, the abnormality degree of each area in the current thermal image is determined based on the temperature variation coefficient, the cumulative deviation value of the fractal dimension of each area in the current thermal image and the historical thermal image, as well as the correlation between the temperature variation coefficient and the fractal dimension; Determine the starting point of the abnormal trend of the temperature variation coefficient and fractal dimension of each region based on the cumulative deviation value of the temperature variation coefficient and fractal dimension in the current thermal image and the historical thermal image, and then determine the overlapping duration of the abnormal trend of each region; Based on the abnormality degree of each area in the current thermal image and the overlapping duration of the abnormal trend of the area, an abnormal risk assessment is performed on the corresponding moment of the current thermal image, including: taking the average of the abnormality degrees of all areas in the current thermal image / each historical thermal image as the abnormality degree of the current thermal image / each historical thermal image; performing linear regression analysis based on the abnormality degrees of the current thermal image and all historical thermal images to obtain the change rate and direction of the abnormality degree of the current thermal image; taking the average of the overlapping duration of the abnormal trends of all areas as the duration of the abnormality; and determining the abnormal risk index at the corresponding moment of the current thermal image to quantify the abnormal risk assessment result. The calculation formula is: ； In the formula, is the abnormal risk index of the current thermal image at the corresponding moment, is the rate of change of the abnormality of the current thermal image, is the absolute value symbol, is the duration of the abnormality, is the total duration corresponding to all historical thermal images, is the natural exponential function, is the variance of the anomaly degree of all historical thermal images, is a sign function that reflects the direction of change in the abnormality of the current thermal image; Monitor the machining process of automotive parts based on abnormal risk assessment results.

2. The automobile parts processing monitoring method according to claim 1, characterized in that: The starting point of the abnormal trend of the temperature coefficient of variation and fractal dimension of each area is determined based on the following method: Compare the cumulative deviation values ​​of the temperature coefficient of variation of each region in the current thermal image and the historical thermal images in chronological order with a preset first threshold value. If the abnormal cumulative deviation value of the temperature coefficient of variation of the region in a thermal image exceeds the first threshold value for the first time, the thermal image is regarded as the starting point of the abnormal trend of the temperature coefficient of variation of the region; The cumulative deviation values ​​of the fractal dimensions of each region in the current thermal image and the historical thermal images are compared with the preset second threshold in chronological order. If the abnormal cumulative deviation value of the fractal dimension of the region in a certain thermal image is greater than the second threshold for the first time, the thermal image is regarded as the starting point of the abnormal trend of the fractal dimension of the region.

3. The automobile parts processing monitoring method according to claim 1, characterized in that: The current thermal image and the historical thermal image are set based on the following method: the I thermal images before the current thermal image are used as the historical thermal image, where I is a preset positive integer.

4. The automobile parts processing monitoring method according to claim 1, characterized in that: The degree of abnormality of each area in the current thermal image is determined based on the following formula: ； In the formula, For the The abnormality of the area in the current thermal image, For the The cumulative deviation value of the temperature variation coefficient of each area in the current thermal image, For the The average of the cumulative deviations of the temperature variation coefficient in all historical thermal images of the region, For the The cumulative deviation value of the fractal dimension of the region in the current thermal image, For the The mean of the cumulative deviation values ​​of the fractal dimension of the region in all historical thermal images, For the The Pearson correlation coefficient of temperature variation coefficient and fractal dimension in all historical thermal images of a region.

5. The automobile parts processing monitoring method according to claim 2, characterized in that: The overlap duration of the abnormal trends in each region is determined based on the following method: The first segment is the abnormal trend starting point of the temperature variation coefficient of each area to the current thermal image, and the second segment is the abnormal trend starting point of the fractal dimension of each area to the current thermal image; The duration corresponding to the overlapping portion of the first segment and the second segment is used as the overlapping duration of the abnormal trend in the area.

6. The automobile parts processing monitoring method according to claim 2, characterized in that: The temperature variation coefficient is used to measure the intensity of temperature fluctuation. The temperature variation coefficient of each region in each thermal image is the ratio of the standard deviation of the temperature values ​​of all pixels corresponding to the region in the thermal image to the mean.

7. The automobile parts processing monitoring method according to claim 2, characterized in that: The fractal dimension is used to measure the morphological complexity of the temperature distribution. The fractal dimension of each region in each thermal image is determined by the difference box dimension method.

8. The automobile parts processing monitoring method according to claim 1, characterized in that: Monitor the machining process of automotive parts based on abnormal risk assessment results, including: Presetting a first abnormal risk index threshold and a second abnormal risk index threshold, wherein the second abnormal risk index threshold is greater than the first abnormal risk index threshold; If the abnormal risk index of the current thermal image at the corresponding moment is less than the first abnormal risk index threshold, the processing is judged to be normal; if the abnormal risk index of the current thermal image at the corresponding moment is greater than or equal to the first abnormal risk index threshold and less than the second abnormal risk index threshold, the processing is judged to have a mild abnormality; if the abnormal risk index of the current thermal image at the corresponding moment is greater than or equal to the second abnormal risk index threshold, the processing is judged to have a serious abnormality.

Images