Crankshaft surface defect online detection system and method based on machine vision

Abstract

The invention relates to the technical field of online detection, in particular to a crankshaft surface defect online detection system and method based on machine vision, and the system comprises an acquisition module, a judgment module, a screening module, a determination module, an early warning module and an adjustment module. According to the method, the production environment parameters are synchronously collected, the key parameters of the crankshafts are collected according to machine vision, the first abnormal crankshaft and the second abnormal crankshaft are obtained through rapid preliminary screening based on the environment abnormal result, and the target crankshaft with the composite defect is locked through the two types of abnormal crankshafts; according to the method, thermal oxygen coupling and high-temperature oxygen poor coupling defect type discrimination is realized for a target crankshaft, and dynamic adjustment of a threshold value according to the number change characteristic of the target crankshaft is introduced, so that optimal setting of a preset temperature threshold value and a preset oxygen content threshold value is realized; the problems of low detection efficiency and high crankshaft defect omission ratio caused by insufficient capability of identifying composite defects and process relevance of the traditional method are effectively solved.

Description

Technical Field

[0001] This invention relates to the field of online inspection technology, and in particular to an online inspection system and method for crankshaft surface defects based on machine vision. Background Technology

[0002] With the rapid development of industrial automation technology, crankshafts, as core components of critical equipment such as engines, directly affect the performance and service life of the entire machine due to their surface quality. During heat treatment, crankshafts are prone to various surface defects such as cracks and oxide layers. In particular, the combined defects of cracks and oxide layers significantly impact the mechanical properties and long-term reliability of the crankshaft. However, the complex production environment and numerous processing parameters of crankshafts result in complex and unevenly distributed defect types. Against this backdrop, how to accurately and in real-time identify complex surface defects on crankshafts in complex production environments and provide production managers with targeted process adjustment guidelines has become a critical technical problem that urgently needs to be solved to ensure crankshaft quality, improve production efficiency, and reduce defect rates.

[0003] Chinese Patent Application Publication No. CN117571954A discloses a crankshaft machining surface defect detection device and system. The device includes: a base, characterized in that a frame is provided on the surface of the base, a support plate is provided on the side of the frame, a groove is opened on the surface of the support plate, a vertical rod is provided in the groove, a second lifting plate is sleeved on the surface of the vertical rod, a hole is opened on the surface of the frame, a bearing is provided in the hole, a rotating rod is provided in the bearing, a three-jaw chuck is provided at one end of the rotating rod, and a drive motor is provided in the frame; a bottom plate, a hole is opened on the surface of the bottom plate, a rotating bearing is provided in the hole, a lifting screw is inserted into the rotating bearing, a lifting plate is screwed to the surface of the lifting screw, a detection device is provided on the side of the lifting plate, a display module is provided on the surface of the frame, a motor base is provided on the side of the frame, and a first motor is provided on the surface of the motor base.

[0004] Therefore, the crankshaft machining surface defect detection device and system have the following problems: the device can not achieve automatic judgment by mechanical lifting and fixing, and there is a risk of missing or false detection due to relying on manual observation; the position of the detection device is fixed on the side of the lifting plate, which may not cover the complex structure of the crankshaft surface, resulting in blind spots in the detection; the device relies on a single lifting screw and a single-sided detection device for movement and detection, and the single-point scanning method is inefficient and difficult to meet the needs of high-speed production. Summary of the Invention

[0005] To address this, the present invention provides an online crankshaft surface defect detection system and method based on machine vision. This system overcomes the problems of low detection efficiency and high missed detection rate of crankshaft defects caused by the insufficient ability of traditional methods to identify composite defects and their process correlations through multi-dimensional data fusion and dynamic threshold adjustment mechanisms.

[0006] To achieve the above objectives, in one aspect, the present invention provides an online crankshaft surface defect detection system based on machine vision, comprising: The acquisition module is used to acquire the quenching temperature, oxygen content in the heat treatment furnace, and surface light reflectance, surface roughness, crack depth, and oxide layer thickness of each crankshaft under test in the heat treatment process of the crankshaft industrial production line, based on the acquired images. The determination module is used to determine whether the production environment is abnormal based on the quenching temperature, the preset temperature threshold, the oxygen content, and the preset oxygen content threshold, so as to obtain an abnormal environment result. A screening module is used to determine a number of crankshafts of interest based on the surface light reflectivity and a preset reflection threshold of each crankshaft under test, based on the abnormal environmental results; to screen out a number of first abnormal crankshafts based on the correlation characteristics between the surface roughness and the crack depth of the crankshafts of interest within a preset screening time; and to screen out a number of second abnormal crankshafts based on the correlation characteristics between the surface roughness and the oxide layer thickness of the crankshafts of interest within a preset screening time. The determination module is used to determine several target crankshafts based on the overlap characteristics of the first abnormal crankshaft and the second abnormal crankshaft within a next preset determination time period; The early warning module is used to determine the defect type of the target crankshaft based on the quenching temperature, the oxygen content, the crack depth and the oxide layer thickness of the target crankshaft, and to issue an early warning prompt based on the defect type. An adjustment module is used to adjust the preset temperature threshold or the preset oxygen content threshold according to the change in the number of target crankshafts within the next preset adjustment period.

[0007] Furthermore, the determination module includes: A temperature deviation calculation unit is used to calculate the relative deviation between the quenching temperature and the preset temperature threshold to obtain the temperature deviation. An oxygen content deviation calculation unit is used to calculate the relative deviation between the oxygen content and the preset oxygen content threshold to obtain the oxygen content deviation. The determination unit is connected to the temperature deviation calculation unit and the oxygen content deviation calculation unit respectively, and is used to perform a weighted summation of the temperature deviation and the oxygen content deviation to obtain a joint deviation, and to determine that the production environment is abnormal when the joint deviation is greater than a preset joint threshold, so as to obtain an environmental abnormality result.

[0008] Furthermore, the filtering module includes: The first determination unit is used to determine that the surface of the crankshaft under test is abnormal when the surface light reflectance is lower than the preset reflection threshold, so as to obtain several crankshafts of interest. A roughness fluctuation calculation unit, which is connected to the first determination unit, is used to calculate the standard deviation of the surface roughness of each of the crankshafts of interest within the preset screening time, so as to obtain several roughness fluctuation values. A screening unit, connected to the roughness fluctuation calculation unit, is configured to: when the roughness fluctuation value of each crankshaft of interest is greater than a preset fluctuation threshold, select a number of first abnormal crankshafts from all the crankshafts of interest based on the crack depth variation characteristics within the preset screening time; and when the roughness fluctuation value of each crankshaft of interest is less than the preset fluctuation threshold, select a number of second abnormal crankshafts from all the crankshafts of interest based on the oxide layer thickness variation characteristics within the preset screening time.

[0009] Furthermore, the filtering unit includes: The depth growth rate calculation subunit is used to calculate the relative deviation of the crack depth corresponding to any two adjacent moments within the preset screening time to obtain several instantaneous depth growth rates, and to calculate the average value of all instantaneous depth growth rates to obtain the average depth growth rate. The average thickness calculation subunit is used to calculate the average thickness of all oxide layers within the preset screening time to obtain the average thickness. A filtering subunit, which is connected to the depth growth rate calculation subunit and the thickness average calculation subunit respectively, is used to determine the crankshaft of interest with the average depth growth rate greater than a preset growth rate threshold as the first abnormal crankshaft, and to determine the crankshaft of interest with the average thickness greater than a preset thickness threshold as the second abnormal crankshaft.

[0010] Furthermore, the determining module includes: The acquisition unit is used to acquire the crankshafts of interest that are simultaneously identified as the first abnormal crankshaft and the second abnormal crankshaft at each moment within the preset time period, so as to obtain a number of candidate crankshafts. A statistics unit, connected to the acquisition unit, is used to count the frequency of occurrence of each candidate crankshaft within the preset time period to obtain several overlap frequencies. A determining unit, which is connected to the statistical unit, is used to determine the candidate crankshafts whose overlap frequency is greater than a preset frequency threshold as the target crankshafts.

[0011] Furthermore, the early warning module includes: The variation calculation unit is used to calculate the difference in crack depth of each target crankshaft corresponding to any two adjacent moments within the previous preset historical defect duration, to obtain several depth variation rates; and to calculate the difference in oxide layer thickness of each target crankshaft corresponding to any two adjacent moments within the previous preset historical defect duration, to obtain several thickness variation rates; and to calculate the difference in quenching temperature corresponding to any two adjacent moments within the previous preset historical defect duration, to obtain several temperature variation rates; and to calculate the difference in oxygen content corresponding to any two adjacent moments within the previous preset historical defect duration, to obtain several oxygen content variation rates. A normalization processing unit, which is connected to the change calculation unit, is used to normalize the depth change rate, the thickness change rate, the temperature change rate and the oxygen content change rate respectively to obtain a depth normalized dataset, a thickness normalized dataset, a temperature normalized dataset and an oxygen content normalized dataset. A type diagnosis unit, connected to the normalization processing unit, is used to determine the defect type based on the depth normalized dataset, the thickness normalized dataset, the temperature normalized dataset, and the oxygen content normalized dataset. An early warning unit, which is connected to the type diagnostic unit, is used to issue an early warning based on the defect type.

[0012] Furthermore, the type diagnostic unit includes: A similarity calculation subunit is used to calculate the Pearson correlation coefficient between the depth-normalized dataset and the temperature-normalized dataset to obtain a first defect similarity, and to calculate the Pearson correlation coefficient between the thickness-normalized dataset and the oxygen content-normalized dataset to obtain a second defect similarity, and to calculate the Pearson correlation coefficient between the depth-normalized dataset and the thickness-normalized dataset to obtain a third defect similarity. A type diagnosis subunit, connected to the similarity calculation subunit, is used to determine that the defect type is thermo-oxygen coupling anomaly when the similarity of the first defect is greater than a preset first similarity threshold, the similarity of the second defect is greater than a preset second similarity threshold, and the similarity of the third defect is greater than a preset third similarity threshold; and to determine that the defect type is high-temperature oxygen-deficient coupling anomaly when the similarity of the first defect is greater than a preset first similarity threshold, the similarity of the second defect is less than a preset second similarity threshold, and the similarity of the third defect is less than a preset third similarity threshold.

[0013] Furthermore, the adjustment module includes: The growth rate calculation unit is used to calculate the difference in the number of the target crankshafts at any two adjacent moments within the preset adjustment time to obtain a number of speed changes, and to sum the number of speed changes to obtain a total number of speed changes. The quantity fluctuation calculation unit is used to calculate the standard deviation of the quantity of the target crankshaft within the preset adjustment time when the total value of the quantity changes is greater than the preset speed threshold, so as to obtain the quantity fluctuation value. An adjustment unit, connected to the quantity fluctuation calculation unit, is used to adjust the preset temperature threshold or the preset oxygen content threshold according to the quantity fluctuation value and the preset quantity fluctuation threshold.

[0014] Furthermore, the adjustment unit includes: The first adjustment subunit is used to reduce the preset temperature threshold according to the quantity fluctuation value and the preset quantity fluctuation threshold when the quantity fluctuation value is greater than the preset quantity fluctuation threshold. The first adjustment subunit is used to reduce the preset oxygen content threshold according to the quantity fluctuation value and the preset quantity fluctuation threshold when the quantity fluctuation value is less than the preset quantity fluctuation threshold.

[0015] On the other hand, the present invention also provides an online detection method for crankshaft surface defects based on machine vision, comprising: The quenching temperature, oxygen content in the heat treatment furnace, and surface light reflectance, surface roughness, crack depth, and oxide layer thickness of each crankshaft under test are obtained during the heat treatment process of the crankshaft industrial production line. The production environment is determined to be abnormal based on the quenching temperature, the preset temperature threshold, the oxygen content, and the preset oxygen content threshold, so as to obtain the result of environmental abnormality. Based on the results of environmental anomalies, several crankshafts of interest are determined according to the surface light reflectivity and preset reflection threshold of each crankshaft under test. Several first abnormal crankshafts are selected according to the correlation characteristics of the surface roughness and crack depth of the crankshafts of interest within a preset screening time. Several second abnormal crankshafts are selected according to the correlation characteristics of the surface roughness and oxide layer thickness of the crankshafts of interest within a preset screening time. Based on the overlap characteristics of the first abnormal crankshaft and the second abnormal crankshaft within the next preset time period, several target crankshafts are determined; The defect type of the target crankshaft is determined based on the quenching temperature, the oxygen content, the crack depth, and the oxide layer thickness of the target crankshaft, and an early warning is issued based on the defect type. The preset temperature threshold or the preset oxygen content threshold is adjusted based on the change in the number of target crankshafts within the next preset adjustment period.

[0016] Compared with existing technologies, the advantages of this invention lie in its ability to construct a comprehensive perception system for the process environment and surface condition by simultaneously collecting multi-dimensional parameters such as quenching temperature, oxygen content in the furnace, and surface light reflectivity, surface roughness, crack depth, and oxide layer thickness during the crankshaft heat treatment process, based on machine vision. Firstly, rapid initial screening is achieved using surface light reflectivity anomalies. Then, by combining the correlation analysis of surface roughness fluctuation characteristics and crack depth growth trends, the first type of abnormal crankshaft is accurately identified. Simultaneously, the second type of abnormal crankshaft is screened through the synergistic judgment of roughness stability and the average oxide layer thickness. Based on this, the system locks target crankshafts with composite defect characteristics based on the overlap frequency of the two types of abnormal crankshafts within a preset time period. Furthermore, by analyzing the dynamic correlation between crack depth, oxide layer thickness, quenching temperature, and the rate of change of oxygen content, the system achieves accurate differentiation of defect types such as thermo-oxygen coupling anomalies and high-temperature oxygen-deficient coupling anomalies. In addition, a threshold adaptive mechanism based on the temporal fluctuation of the target crankshaft quantity is introduced, which enables the temperature and oxygen content judgment thresholds to be dynamically optimized according to the production status, significantly improving the reliability of the system's early warning. This forms a closed-loop quality control system from multi-parameter perception, composite defect identification, defect type diagnosis to threshold self-adjustment, effectively solving the problems of low detection efficiency and high crankshaft defect false negative rate caused by the insufficient ability of traditional methods to identify composite defects and their process correlation.

[0017] Furthermore, by jointly analyzing the relative deviations of quenching temperature and preset temperature thresholds, and the relative deviations of oxygen content and preset oxygen content thresholds, accurate determination of production environment anomalies was achieved. Quantifying the deviations of quenching temperature and oxygen content during crankshaft heat treatment provides an objective physical basis for environmental anomaly determination. By weighted summation of temperature and oxygen content deviations to obtain a joint deviation, and comparing this joint deviation with a preset joint threshold, timely identification can be achieved when either of these two production parameters, quenching temperature or oxygen content, is abnormal. Even if the deviation of one parameter is within an acceptable range, if the deviation of the other parameter is large enough to cause the joint deviation to exceed the threshold, the system can still keenly detect potential environmental risks, avoiding the neglect of quality hazards caused by combined factors due to a single normal parameter. This not only improves the comprehensiveness of environmental anomaly determination but also, by transforming the two key influencing factors, temperature and oxygen content, into quantifiable numerical indicators, makes the determination process more scientific and standardized, effectively reducing errors that may arise from human experience-based judgment. This lays a solid foundation for subsequent accurate identification and timely warning of target crankshafts.

[0018] Furthermore, by jointly analyzing the surface reflectivity, roughness fluctuation, crack depth, and oxide layer thickness of the crankshaft, accurate identification of different types of surface defects was achieved. By rapidly identifying crankshafts of interest with potential anomalies based on surface reflectivity below a preset reflection threshold, screening targets were provided for subsequent analysis, improving detection efficiency. Calculating the standard deviation of surface roughness of the crankshaft of interest within a preset screening time quantifies the fluctuation range of roughness. This not only reflects the stability of the crankshaft surface texture but also serves as a key basis for judging crack propagation activity or oxide layer growth stability. When cracks propagate, the crankshaft surface will exhibit irregular texture changes due to stress concentration, leading to a significant increase in roughness fluctuation values. If the oxide layer is in an unstable growth state, its uneven thickness changes will also be reflected in the roughness fluctuation values ​​through changes in the surface microstructure. However, because the change in oxide layer thickness is relatively slower and more uniform than crack propagation, the resulting roughness fluctuation range is usually smaller than that caused by crack propagation. Therefore, by setting a preset fluctuation threshold, the two types of roughness fluctuations caused by cracks and oxide layers can be effectively distinguished, providing a clear quantitative basis for subsequent targeted screening of the first and second abnormal crankshafts. When the roughness fluctuation value is greater than the preset fluctuation threshold, the system determines that the surface defect is more likely to be related to active crack propagation, and then accurately screens the first abnormal crankshaft from the crankshafts of interest based on the crack depth change characteristics within the preset screening time. When the roughness fluctuation value is less than the preset fluctuation threshold, it indicates that the surface defect originates from an abnormal oxide layer thickness. At this time, the second abnormal crankshaft is screened based on the oxide layer thickness change characteristics within the preset screening time, thereby achieving the classification, detection, and identification of different types of surface defects.

[0019] Furthermore, by calculating the instantaneous and average growth rates of crack depth and the average value of oxide layer thickness, accurate identification of different types of defects on the crankshaft surface was achieved. Specifically, due to the suddenness and rapid propagation of abnormal crack depth changes, by calculating the difference in crack depth between any two adjacent moments within a preset screening period, the dynamic trend of crack changes in a short period of time can be captured. The obtained instantaneous growth rate of depth reflects the immediate state of crack propagation, while the average growth rate of depth obtained by averaging these instantaneous growth rates further smooths out instantaneous fluctuations and better reflects the overall propagation rate of the crack within the preset screening period. When the average growth rate of depth is greater than the preset growth rate threshold, it indicates that the crack is in a rapid propagation stage, and such crankshafts of concern are identified as the first abnormal crankshafts. Regarding oxide layer thickness, anomalies typically manifest as continuous accumulation over a certain period. Therefore, calculating the average oxide layer thickness of each crankshaft of interest within a preset screening time can comprehensively reflect the overall thickness level of the oxide layer during that period. When the average thickness is greater than the preset thickness threshold, it indicates that the oxide layer is too thick. Such crankshafts of interest are identified as the second abnormal crankshafts, which can effectively identify oxide layer thickness anomalies and provide accurate screening results for subsequent defect type judgment and early warning.

[0020] Furthermore, through precise multi-stage data screening and statistical analysis, abnormal crankshafts that may occur in production can be effectively screened and confirmed, thereby improving the system's accuracy and stability. Specifically, by acquiring crankshafts that are simultaneously identified as having both primary and secondary anomalies, it ensures that detected defective crankshafts truly exhibit dual defects, thus avoiding false positives and false negatives. Based on this, by statistically analyzing the occurrence frequency of candidate crankshafts and calculating the overlap frequency, the stability of the defect pattern is further analyzed. This ensures that only crankshafts exhibiting multiple anomalies with statistical significance are marked as target crankshafts, helping to improve the system's diagnostic accuracy for abnormal defects and reduce interference from sporadic failures. A final screening of candidate crankshafts using preset frequency thresholds ensures that only crankshafts conforming to high-frequency occurrence patterns are selected as target crankshafts. This meticulous screening process not only accurately captures potential defective crankshafts from massive amounts of data but also reduces false alarms and invalid operations, ensuring the smooth operation of the production line, reducing resource waste caused by misjudgments, improving the automated monitoring capabilities of the production process, and enhancing the accuracy and efficiency of defect detection.

[0021] Furthermore, by integrating a series of intelligent functions such as change calculation, normalization processing, type determination, and early warning prompts, the system achieves dynamic tracking and multi-parameter coupled analysis of the crankshaft surface defect formation process. This elevates the traditional single, static threshold judgment to early warning and root cause identification based on multi-parameter change trends, significantly improving the timeliness and accuracy of defect warnings. By calculating and normalizing the change rates of crack depth, oxide layer thickness, quenching temperature, and oxygen content over historical periods, the system can effectively identify abnormal fluctuation patterns of various parameters and their correlation with defect types. This allows for accurate determination of defect types and the issuance of targeted warnings, which not only helps in early intervention to prevent the expansion of defect scale but also provides data-driven decision-making basis for process parameter optimization and production line adjustment, comprehensively enhancing the predictive maintenance capabilities and intelligence level of the detection system.

[0022] Furthermore, by monitoring key indicators such as crack depth, oxide layer thickness, quenching temperature, and oxygen content, the system can capture potential anomalies, promptly identify possible defect sources, and enhance its diagnostic capabilities, enabling more accurate identification of trends that may lead to failures. Normalization processing allows for effective comparison of data with different dimensions and scales on the same platform, eliminating data inconsistencies. It converts the rate of change of each indicator into standardized data, thus avoiding data bias caused by differences in dimensions and improving the accuracy and reliability of data analysis. Through the calculation of the Pearson correlation coefficient, the interrelationships between various indicators are precisely quantified, and by setting similarity thresholds to determine defect types, not only is the sensitivity of the system's diagnosis improved, but its accuracy is also enhanced. Especially when multiple parameters change synergistically, it can effectively distinguish between thermo-oxygen coupling anomalies and high-temperature oxygen-deficient coupling anomalies, facilitating targeted measures and avoiding blind intervention, thus improving the scientific rigor and specificity of early warning. The introduction of automated early warning functions greatly improves the speed of fault diagnosis and response. By automatically analyzing and issuing early warnings, the system avoids delays caused by manual inspection and response, enabling preventative measures to be taken before malfunctions occur. This reduces the risk of equipment damage or production downtime. The system not only issues alerts at the initial stage of a fault but also provides valuable decision-making support to maintenance personnel through multi-parameter comparison and analysis, thereby improving the overall system's safety and reliability. Intelligent data analysis and diagnostic methods enhance the system's real-time monitoring capabilities, improve the accuracy and efficiency of fault diagnosis, and reduce manual intervention and response time. This results in significant benefits in improving equipment safety, extending service life, and reducing maintenance costs.

[0023] Furthermore, by calculating the growth rate of the target crankshaft quantity, the change in the number of target crankshafts at each moment is captured, thereby identifying potential quality anomalies. By calculating the total growth rate of the quantity, the changing trend of target crankshafts that simultaneously exhibit abnormal crack depth and oxide layer thickness is identified. When the total growth rate of the quantity exceeds the preset speed threshold, it indicates that the number of such composite abnormal crankshafts is on the rise. At this time, the standard deviation of the number of target crankshafts within the preset adjustment period is calculated to quantify the data dispersion during this rising process, that is, to obtain the quantity fluctuation value. The quantity fluctuation value can reflect the stability of the abnormal crankshaft quantity fluctuation, and it is compared with the preset quantity fluctuation threshold, thereby achieving dynamic and precise adjustment of key parameters in the production process.

[0024] Furthermore, by intelligently analyzing the fluctuations in the number of target crankshafts, the system effectively adjusts preset thresholds in real time. When the fluctuation in the number of target crankshafts exceeds the preset threshold, the system assumes that the quenching temperature control may be unstable. This is because the impact of abnormal quenching temperature on crankshaft quality is usually significant in the short term, especially during temperature control instability or adjustment. The number of target crankshafts can vary greatly within the preset adjustment period. Therefore, by reducing the preset temperature threshold, the system helps the production system reduce the amplitude of temperature fluctuations, ensuring that the quenching temperature remains within the optimal range and avoiding quality defects caused by excessively high temperatures. Conversely, when the fluctuation is less than the preset threshold, the system suspects that the oxygen concentration may be uneven. This is because the impact of abnormal oxygen content is often continuous, resulting in a more gradual fluctuation in crankshaft quality. If abnormal oxygen content causes surface defects on the crankshaft, there will not be a large difference in the number of target crankshafts at different times within the preset adjustment period. Through this dynamic adjustment mechanism based on the fluctuations in the number of target crankshafts, the system can accurately respond to environmental changes during the production process, automatically adjust key parameters, improve the stability of the production process, reduce the defect rate, and ensure high consistency and high quality of the final product.

[0025] Furthermore, by employing an online crankshaft surface defect detection method based on machine vision and multi-source parameter fusion, high-precision identification and intelligent early warning of crankshaft defects in industrial production processes have been achieved. By utilizing quenching temperature to reflect the stability of energy input during heat treatment, oxygen content to reveal changes in the oxidation environment within the furnace, surface reflectivity to reflect the density and oxidation degree of the material surface, and surface roughness to characterize the microscopic morphology after processing and thermal stress, potential anomalies in crankshafts of interest can be preliminarily identified. Combining the dynamic trends of crack depth and oxide layer thickness, abnormal regions affected by the coupling of high-temperature oxidation and thermal stress are further refined. The correlation between reflectivity and roughness characterizes abnormal patterns of surface energy reflection behavior, and the linkage rate of change between crack and oxide layer thickness reflects the evolution trend of surface structure degradation. The sensitivity of the screening threshold is then adjusted based on fluctuations in temperature and oxygen content, dynamically adapting to different batches and production conditions. Finally, the defect type of the target crankshaft is determined based on the multi-parameter fusion results, and a warning command with defect characteristics is generated. This ensures the real-time, accurate, and adaptive nature of the defect identification process, guaranteeing continuous monitoring and reliable early warning of crankshaft surface quality in complex production environments, effectively improving product consistency and production safety. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of the machine vision-based online crankshaft surface defect detection system in this embodiment; Figure 2 This is a logic diagram for the determination module in this embodiment to determine abnormalities in the production environment; Figure 3 This is a logic diagram for determining the target crankshaft in the module of this embodiment. Detailed Implementation

[0027] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0028] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0029] Please see Figure 1 As shown, this is a schematic diagram of the online crankshaft surface defect detection system based on machine vision according to this embodiment. On one hand, this embodiment provides an online crankshaft surface defect detection system based on machine vision, including: The acquisition module is used to acquire the quenching temperature, oxygen content in the heat treatment furnace, and surface light reflectance, surface roughness, crack depth, and oxide layer thickness of each crankshaft under test in the heat treatment process of the crankshaft industrial production line, based on the acquired images. The determination module, which is connected to the acquisition module, is used to determine whether the production environment is abnormal based on the quenching temperature, the preset temperature threshold, the oxygen content, and the preset oxygen content threshold, so as to obtain an abnormal environment result. A screening module, which is connected to the acquisition module and the determination module respectively, is used to determine a number of crankshafts of interest based on the environmental anomaly results, according to the surface light reflectivity and preset reflection threshold of each crankshaft to be tested, and to screen out a number of first abnormal crankshafts based on the correlation characteristics of the surface roughness and crack depth of the crankshafts of interest within a preset screening time, and to screen out a number of second abnormal crankshafts based on the correlation characteristics of the surface roughness and oxide layer thickness of the crankshafts of interest within a preset screening time. A determination module, which is connected to the filtering module, is used to determine several target crankshafts based on the overlap characteristics of the first abnormal crankshaft and the second abnormal crankshaft within a next preset determination time period; An early warning module, which is connected to the acquisition module and the determination module respectively, is used to determine the defect type of the target crankshaft based on the quenching temperature, the oxygen content, the crack depth and the oxide layer thickness of the target crankshaft, and to issue an early warning prompt based on the defect type. An adjustment module, which is connected to the determination module and the determination module respectively, is used to adjust the preset temperature threshold or the preset oxygen content threshold according to the change in the number of target crankshafts within the next preset adjustment time.

[0030] In this embodiment, a machine vision-based online crankshaft surface defect detection system is applied to an automated crankshaft production line. It primarily detects surface defects to determine the defect type and issues early warnings to assist management in adjusting the production environment. During the heat treatment process, excessively high quenching temperatures can easily cause surface cracks, while excessive oxygen content leads to an overly thick oxide layer, affecting crankshaft quality. Furthermore, the oxide layer promotes crack formation, which in turn accelerates internal oxidation. This composite defect—abnormal crack depth combined with abnormal oxide layer thickness—has a more significant impact on crankshaft quality than a single defect, greatly increasing the risk of crankshaft breakage. This embodiment uses machine vision to acquire crankshaft images and obtain relevant parameters such as surface reflectivity, surface roughness, crack depth, and oxide layer thickness, as well as production environment parameters such as quenching temperature and oxygen content. Crankshafts exhibiting this composite defect of abnormal crack depth and abnormal oxide layer thickness are identified as target crankshafts, and early warnings are issued for timely intervention, thereby ensuring crankshaft quality.

[0031] In this embodiment, the acquisition module integrates heat treatment process parameters with machine vision features to construct a complete crankshaft surface quality monitoring data system. Quenching temperature refers to the real-time process temperature of the crankshaft during the quenching process in the heat treatment furnace. It is a key parameter affecting the degree of martensitic transformation and residual stress. It is measured directly using a K-type thermocouple or non-contactly using an infrared thermometer, with the signal transmitted to the PLC system via a temperature transmitter. Oxygen content in the heat treatment furnace reflects the process index of the protective atmosphere and directly affects the surface oxidation degree and decarburization layer depth. It is measured directly inside the furnace using a zirconia oxygen probe or by online monitoring of exhaust gas components using a laser gas analyzer. High-speed industrial cameras deployed on the production line acquire crankshaft surface images. Based on these images, the surface reflectivity, surface roughness, crack depth, and oxide layer thickness are obtained. Surface reflectivity is an optical parameter characterizing surface smoothness and cleanliness; an abnormal decrease indicates surface defects such as abnormal crack depth or oxide layer thickness. The pixel grayscale values ​​of each region in the image are analyzed under standard lighting conditions. The reflectance distribution is calculated by comparing it with a pre-calibrated reference reflectance, directly reflecting the surface cleanliness and microscopic smoothness. Surface roughness is a geometric parameter describing the surface micromorphology; abnormal fluctuations indicate defects on the crankshaft surface. Using photometric stereo vision technology, the surface microscopic three-dimensional morphology is reconstructed by analyzing the changes in brightness under multiple light sources, and then quantitative indicators such as the arithmetic mean deviation of the profile are calculated. Crack depth is a key indicator for quantifying the severity of surface cracks and directly affects the fatigue life of the crankshaft. After accurately locating the crack area using a deep learning segmentation algorithm, the three-dimensional geometric information of the crack is obtained through parallax calculation using binocular stereo vision or structured light three-dimensional imaging principles. Oxide layer thickness is a direct parameter characterizing the degree of surface oxidation; excessive thickness will affect dimensional accuracy and service performance. Accurate measurement is achieved by analyzing the intensity characteristics of the interference colors generated by the surface oxide film in specific spectral channels and establishing a mapping model between color space and thickness values. The acquisition module constructs a real-time acquisition channel for multi-source heterogeneous data, providing a complete and reliable data foundation for subsequent intelligent diagnosis.

[0032] The preset temperature threshold is the benchmark value for determining whether the temperature in the quenching furnace process zone is qualified. It depends on the type of crankshaft material, the effective thickness of the parts, and the characteristics of the quenching medium. It is usually set between 850°C and 880°C. In this embodiment, it is set to 860°C, which can ensure that the material completes austenitization while avoiding grain coarsening or quenching cracks due to overheating. The preset oxygen content threshold is the critical concentration value for controlling the protective atmosphere of the heat treatment furnace. It depends on the type of atmosphere system and the oxidation resistance of the crankshaft material. It is usually set between 0.8% and 1.2%. In this embodiment, it is set to 1.0%, which can provide a clear judgment basis for the oxygen content deviation calculation unit and effectively control the degree of surface oxidation. The preset reflection threshold is the lowest gray value for identifying surface optical anomalies through machine vision. It depends on the basic reflectivity of the crankshaft, the illumination intensity, and the camera parameters. It is usually set between 150 and 180. In this embodiment, it is set to 165, which can quickly identify the crankshaft under test with abnormal surface light reflectivity. To improve defect screening efficiency, the preset screening time is a short-term data window used to analyze the dynamic trend of surface roughness. It depends on the production line cycle and defect development speed, and is usually set between 30 and 60 minutes. In this embodiment, it is set to 45 minutes to ensure that sufficient data samples are obtained for surface roughness standard deviation calculation and accurate identification of abnormal development trends. The preset determination time is the analysis cycle for the statistical analysis of target crankshaft overlap characteristics. It depends on the production batch size and defect occurrence frequency, and is usually set between 4 and 8 hours. In this embodiment, it is set to 6 hours to ensure the reliability of candidate crankshaft occurrence frequency statistics and improve the accuracy of target crankshaft determination. The preset adjustment time is the decision cycle for adaptive optimization of system parameters. It depends on process stability and quality control requirements, and is usually set between 7 and 14 days. In this embodiment, it is set to 10 days to evaluate the trend of target crankshaft quantity changes based on sufficient historical data and ensure the scientific nature of threshold adjustment decisions.

[0033] By simultaneously collecting quenching temperature, furnace oxygen content, and multi-dimensional parameters such as surface light reflectivity, surface roughness, crack depth, and oxide layer thickness during the crankshaft heat treatment process, a comprehensive perception system for the process environment and surface condition was constructed. First, rapid initial screening was achieved using surface light reflectivity anomalies. Then, by combining the correlation analysis of surface roughness fluctuation characteristics and crack depth growth trends, the first type of abnormal crankshaft was accurately identified. Simultaneously, the second type of abnormal crankshaft was screened through the collaborative judgment of roughness stability and the average oxide layer thickness. Based on this, the system locked target crankshafts with composite defect characteristics based on the overlap frequency of the two types of abnormal crankshafts within a preset time period. Furthermore, by analyzing the dynamic correlation between crack depth, oxide layer thickness, quenching temperature, and the rate of change of oxygen content, the system accurately distinguished defect types such as thermo-oxygen coupling anomalies and high-temperature oxygen-deficient coupling anomalies. In addition, a threshold adaptive mechanism based on the temporal fluctuation of the target crankshaft quantity is introduced, which enables the temperature and oxygen content judgment thresholds to be dynamically optimized according to the production status, significantly improving the reliability of the system's early warning. This forms a closed-loop quality control system from multi-parameter perception, composite defect identification, defect type diagnosis to threshold self-adjustment, effectively solving the problems of low detection efficiency and high crankshaft defect false negative rate caused by the insufficient ability of traditional methods to identify composite defects and their process correlation.

[0034] Please continue reading. Figure 2 As shown, this is a logic diagram for determining an anomaly in the production environment using the determination module in this embodiment. In this embodiment, the determination module includes: A temperature deviation calculation unit is used to calculate the relative deviation between the quenching temperature and the preset temperature threshold to obtain the temperature deviation. An oxygen content deviation calculation unit is used to calculate the relative deviation between the oxygen content and the preset oxygen content threshold to obtain the oxygen content deviation. The determination unit is connected to the temperature deviation calculation unit and the oxygen content deviation calculation unit respectively, and is used to perform a weighted summation of the temperature deviation and the oxygen content deviation to obtain a joint deviation, and to determine that the production environment is abnormal when the joint deviation is greater than a preset joint threshold, so as to obtain an environmental abnormality result.

[0035] In the process of weighted summation of the temperature deviation and the oxygen content deviation to obtain the combined deviation, the weight corresponding to the temperature deviation is used to measure the importance of the temperature parameter in the determination of environmental anomalies. It depends on the sensitivity of temperature fluctuations to crankshaft crack defects and the priority of process control, and is usually set between 0.6 and 0.8. In this embodiment, it is set to 0.7, which can highlight the dominant role of temperature control defects in the formation of composite defects. The weight corresponding to the oxygen content deviation is used to measure the influence of the oxygen content parameter in the determination of environmental anomalies. It depends on the contribution of oxygen content changes to oxide layer growth and the stability of the atmosphere system, and is usually set between 0.2 and 0.4. In this embodiment, it is set to 0.3, which can ensure effective monitoring of oxidation defects while avoiding excessive sensitivity of the system due to normal fluctuations in atmosphere parameters.

[0036] By jointly analyzing the relative deviations of quenching temperature and preset temperature thresholds, and the relative deviations of oxygen content and preset oxygen content thresholds, accurate determination of production environment anomalies is achieved. Quantifying the deviations of quenching temperature and oxygen content during crankshaft heat treatment provides an objective physical basis for environmental anomaly determination. By weighted summation of temperature and oxygen content deviations to obtain a joint deviation, and comparing this joint deviation with a preset joint threshold, timely identification can be achieved when either of these two production parameters, quenching temperature or oxygen content, is abnormal. Even if the deviation of one parameter is within an acceptable range, if the deviation of the other parameter is large enough to cause the joint deviation to exceed the threshold, the system can still keenly detect potential environmental risks, avoiding the neglect of quality hazards caused by combined factors due to a single normal parameter. This not only improves the comprehensiveness of environmental anomaly determination but also, by transforming the two key influencing factors, temperature and oxygen content, into quantifiable numerical indicators, makes the determination process more scientific and standardized, effectively reducing errors that may arise from human experience-based judgment. This lays a solid foundation for subsequent accurate identification and timely warning of target crankshafts.

[0037] Specifically, the filtering module includes: The first determination unit is used to determine that the surface of the crankshaft under test is abnormal when the surface light reflectance is lower than the preset reflection threshold, so as to obtain several crankshafts of interest. A roughness fluctuation calculation unit, which is connected to the first determination unit, is used to calculate the standard deviation of the surface roughness of each of the crankshafts of interest within the preset screening time, so as to obtain several roughness fluctuation values. A screening unit, connected to the roughness fluctuation calculation unit, is configured to: when the roughness fluctuation value of each crankshaft of interest is greater than a preset fluctuation threshold, select a number of first abnormal crankshafts from all the crankshafts of interest based on the crack depth variation characteristics within the preset screening time; and when the roughness fluctuation value of each crankshaft of interest is less than the preset fluctuation threshold, select a number of second abnormal crankshafts from all the crankshafts of interest based on the oxide layer thickness variation characteristics within the preset screening time.

[0038] The preset fluctuation threshold is a critical value for distinguishing the dynamic stability of surface roughness. It depends on the material processing specifications and the surface morphology evolution mechanism, and is usually set between 0.10 micrometers and 0.25 micrometers. In this embodiment, it is set to 0.18 micrometers, which can effectively distinguish between active crack defects caused by mechanical stress and gradual surface degradation dominated by oxidation reaction.

[0039] By jointly analyzing the surface reflectivity, roughness fluctuation, crack depth, and oxide layer thickness of crankshafts, accurate identification of different types of surface defects was achieved. Crankshafts of interest with potential anomalies were quickly identified based on surface reflectivity falling below a preset reflection threshold, providing screening targets for subsequent analysis and improving detection efficiency. Calculating the standard deviation of surface roughness of the crankshafts of interest within a preset screening time quantifies the fluctuation range of roughness. This not only reflects the stability of the crankshaft surface texture but also serves as a key basis for judging crack propagation activity or oxide layer growth stability. When cracks propagate, the crankshaft surface exhibits irregular texture changes due to stress concentration, leading to a significant increase in roughness fluctuation values. Similarly, if the oxide layer is in an unstable growth state, its uneven thickness changes will also be reflected in the roughness fluctuation values ​​through changes in the surface microstructure. However, because the change in oxide layer thickness is relatively slower and more uniform than crack propagation, the resulting roughness fluctuation range is usually smaller than that caused by crack propagation. Therefore, by setting a preset fluctuation threshold, the two types of roughness fluctuations caused by cracks and oxide layers can be effectively distinguished, providing a clear quantitative basis for subsequent targeted screening of the first and second abnormal crankshafts. When the roughness fluctuation value is greater than the preset fluctuation threshold, the system determines that the surface defect is more likely to be related to active crack propagation, and then accurately screens the first abnormal crankshaft from the crankshafts of interest based on the crack depth change characteristics within the preset screening time. When the roughness fluctuation value is less than the preset fluctuation threshold, it indicates that the surface defect originates from an abnormal oxide layer thickness. At this time, the second abnormal crankshaft is screened based on the oxide layer thickness change characteristics within the preset screening time, thereby achieving the classification, detection, and identification of different types of surface defects.

[0040] Specifically, the filtering unit includes: The depth growth rate calculation subunit is used to calculate the relative deviation of the crack depth corresponding to any two adjacent moments within the preset screening time to obtain several instantaneous depth growth rates, and to calculate the average value of all instantaneous depth growth rates to obtain the average depth growth rate. The average thickness calculation subunit is used to calculate the average thickness of all oxide layers within the preset screening time to obtain the average thickness. A filtering subunit, which is connected to the depth growth rate calculation subunit and the thickness average calculation subunit respectively, is used to determine the crankshaft of interest with the average depth growth rate greater than a preset growth rate threshold as the first abnormal crankshaft, and to determine the crankshaft of interest with the average thickness greater than a preset thickness threshold as the second abnormal crankshaft.

[0041] The preset growth rate threshold is a critical rate indicator for judging whether the crack depth growth has entered the dangerous range. It depends on the crack resistance of the material and the stability requirements of the heat treatment process, and is usually set between 0.8 μm / min and 1.5 μm / min. In this embodiment, it is set to 1.2 μm / min, which can effectively capture the accelerated crack propagation phenomenon caused by quenching stress concentration and prevent potential crack sources from developing into structural defects. The preset thickness threshold is a legal quality boundary value for defining whether the oxide layer thickness exceeds the standard. It depends on the oxidation resistance of the base material and the subsequent processing allowance requirements, and is usually set between 12 μm and 18 μm. In this embodiment, it is set to 15 μm, which can accurately distinguish between normal oxidation phenomena and abnormal oxidation defects, and ensure that the crankshaft surface quality meets the precision assembly standards.

[0042] By calculating the instantaneous and average growth rates of crack depth and the average value of oxide layer thickness, accurate identification of different types of defects on the crankshaft surface was achieved. Specifically, due to the sudden and rapid propagation of abnormal crack depth changes, the dynamic trend of crack changes in a short period of time can be captured by calculating the difference in crack depth between any two adjacent moments within a preset screening period. The instantaneous growth rate of depth reflects the immediate state of crack propagation, while the average growth rate of depth obtained by averaging these instantaneous growth rates further smooths out instantaneous fluctuations and better reflects the overall propagation rate of the crack within the preset screening period. When the average growth rate of depth is greater than the preset growth rate threshold, it indicates that the crack is in a rapid propagation stage, and such crankshafts of concern are identified as the first abnormal crankshafts. Regarding oxide layer thickness, anomalies typically manifest as continuous accumulation over a certain period. Therefore, calculating the average oxide layer thickness of each crankshaft of interest within a preset screening time can comprehensively reflect the overall thickness level of the oxide layer during that period. When the average thickness is greater than the preset thickness threshold, it indicates that the oxide layer is too thick. Such crankshafts of interest are identified as the second abnormal crankshafts, which can effectively identify oxide layer thickness anomalies and provide accurate screening results for subsequent defect type judgment and early warning.

[0043] Please continue reading. Figure 3 As shown, this is a logic diagram for determining the target crankshaft by the determining module in this embodiment. In this embodiment, the determining module includes: The acquisition unit is used to acquire the crankshafts of interest that are simultaneously identified as the first abnormal crankshaft and the second abnormal crankshaft at each moment within the preset time period, so as to obtain a number of candidate crankshafts. A statistics unit, connected to the acquisition unit, is used to count the frequency of occurrence of each candidate crankshaft within the preset time period to obtain several overlap frequencies. A determining unit, which is connected to the statistical unit, is used to determine the candidate crankshafts whose overlap frequency is greater than a preset frequency threshold as the target crankshafts.

[0044] The preset frequency threshold is a statistical significance boundary used to determine whether a candidate crankshaft constitutes a systematic defect. It depends on the consistency requirements of production quality and the acceptable risk level of defect occurrence. It is usually set between 60% and 80%, and in this embodiment it is set to 70%. This can effectively eliminate the interference of accidental compound defects and ensure that the target crankshaft finally locked represents a persistent process abnormality, providing a statistically reliable data basis for subsequent defect type diagnosis and threshold adaptive adjustment.

[0045] Through precise multi-stage data screening and statistical analysis, abnormal crankshafts that may occur in production can be effectively screened and confirmed, thereby improving the system's accuracy and stability. Specifically, by acquiring crankshafts that are simultaneously identified as having both primary and secondary anomalies, it ensures that detected defective crankshafts truly exhibit dual defects, thus avoiding false positives and false negatives. Based on this, the stability of defect patterns is further analyzed by statistically analyzing the occurrence frequency of candidate crankshafts and calculating the overlap frequency. This ensures that only crankshafts exhibiting multiple anomalies with statistical significance are marked as target crankshafts, helping to improve the system's diagnostic accuracy for abnormal defects and reduce interference from sporadic failures. A final screening of candidate crankshafts using preset frequency thresholds ensures that only crankshafts conforming to high-frequency occurrence patterns are selected as target crankshafts. This meticulous screening process not only accurately captures potential defective crankshafts from massive amounts of data but also reduces false alarms and invalid operations, ensuring the smooth operation of the production line, reducing resource waste caused by misjudgments, improving the automated monitoring capabilities of the production process, and enhancing the accuracy and efficiency of defect detection.

[0046] Specifically, the early warning module includes: The variation calculation unit is used to calculate the difference in crack depth of each target crankshaft corresponding to any two adjacent moments within the previous preset historical defect duration, to obtain several depth variation rates; and to calculate the difference in oxide layer thickness of each target crankshaft corresponding to any two adjacent moments within the previous preset historical defect duration, to obtain several thickness variation rates; and to calculate the difference in quenching temperature corresponding to any two adjacent moments within the previous preset historical defect duration, to obtain several temperature variation rates; and to calculate the difference in oxygen content corresponding to any two adjacent moments within the previous preset historical defect duration, to obtain several oxygen content variation rates. A normalization processing unit, which is connected to the change calculation unit, is used to normalize the depth change rate, the thickness change rate, the temperature change rate and the oxygen content change rate respectively to obtain a depth normalized dataset, a thickness normalized dataset, a temperature normalized dataset and an oxygen content normalized dataset. A type diagnosis unit, connected to the normalization processing unit, is used to determine the defect type based on the depth normalized dataset, the thickness normalized dataset, the temperature normalized dataset, and the oxygen content normalized dataset. An early warning unit, which is connected to the type diagnostic unit, is used to issue an early warning based on the defect type.

[0047] The preset historical defect duration is an observation time window used to establish a benchmark for defect evolution trend analysis. Its value depends on the production line quality assessment cycle and defect development dynamics. It is usually set between 4 and 8 hours. In this embodiment, it is set to 6 hours, which can effectively capture the dynamic change law of key parameters such as crack depth and oxide layer thickness, and provide a data foundation with temporal continuity for subsequent change rate calculation and defect type diagnosis.

[0048] By integrating a series of intelligent functions such as change calculation, normalization processing, type determination, and early warning prompts, the system achieves dynamic tracking and multi-parameter coupled analysis of the crankshaft surface defect formation process. This elevates the traditional single, static threshold judgment to early warning and root cause identification based on multi-parameter change trends, significantly improving the timeliness and accuracy of defect early warning. By calculating and normalizing the change rates of crack depth, oxide layer thickness, quenching temperature, and oxygen content over historical periods, the system can effectively identify abnormal fluctuation patterns of various parameters and their correlation with defect types. This allows for accurate determination of defect types and the issuance of targeted early warnings. This not only helps in early intervention to prevent the expansion of defect scale but also provides data-driven decision-making basis for process parameter optimization and production line adjustment, comprehensively enhancing the predictive maintenance capabilities and intelligence level of the detection system.

[0049] Specifically, the type diagnostic unit includes: A similarity calculation subunit is used to calculate the Pearson correlation coefficient between the depth-normalized dataset and the temperature-normalized dataset to obtain a first defect similarity, and to calculate the Pearson correlation coefficient between the thickness-normalized dataset and the oxygen content-normalized dataset to obtain a second defect similarity, and to calculate the Pearson correlation coefficient between the depth-normalized dataset and the thickness-normalized dataset to obtain a third defect similarity. A type diagnosis subunit, connected to the similarity calculation subunit, is used to determine that the defect type is thermo-oxygen coupling anomaly when the similarity of the first defect is greater than a preset first similarity threshold, the similarity of the second defect is greater than a preset second similarity threshold, and the similarity of the third defect is greater than a preset third similarity threshold; and to determine that the defect type is high-temperature oxygen-deficient coupling anomaly when the similarity of the first defect is greater than a preset first similarity threshold, the similarity of the second defect is less than a preset second similarity threshold, and the similarity of the third defect is less than a preset third similarity threshold.

[0050] The preset first similarity threshold is the statistical significance boundary for judging the synergy between crack depth and temperature change. Its value depends on the theoretical influence of temperature stress on crack propagation and the distribution characteristics of actual production line data, and is usually set between 0.85 and 0.95. In this embodiment, it is set to 0.90, which can effectively capture the strong correlation between quenching temperature fluctuations and crack depth growth, providing a quantitative basis for judging heat-induced crack defects. The preset second similarity threshold is a key criterion for measuring the synchronicity between oxide layer thickness and oxygen content changes. Its value is based on the theoretical correlation between concentration and thickness in the oxidation reaction kinetic model. The value is usually set between 0.80 and 0.90. In this embodiment, it is set to 0.85, which can accurately identify the oxide layer thickening phenomenon caused by abnormal atmosphere and ensure the reliability of the causal relationship between oxygen concentration control failure and surface oxidation. The preset third similarity threshold is a critical index characterizing the coupling effect between crack and oxide layer. It is set according to the bidirectional mechanism of crack accelerating oxidation and oxidation promoting crack. It is usually set between 0.75 and 0.85. In this embodiment, it is set to 0.80, which can sensitively detect the synergistic effect of abnormal crack and oxide layer development and provide statistical support for the analysis of the interaction mechanism of composite defects.

[0051] In this embodiment, the actual production environment of crankshafts is complex and variable, and the formation of surface defects on crankshafts is affected by the coupling effect of multiple factors such as quenching temperature, oxygen content in the heat treatment furnace, crack depth, and oxide layer thickness. To accurately identify composite defects on the crankshaft surface, this embodiment defines two types of composite defects by analyzing the rate of change of each parameter within a preset historical defect duration and the correlation between them and their normalized datasets. Thermo-oxygen coupling anomaly occurs because rising temperatures cause uneven thermal expansion within the crankshaft metal, generating thermal stress. When this stress concentrates in localized areas, microcracks propagate rapidly, increasing their depth. Simultaneously, high oxygen concentration accelerates oxidation on the metal surface, leading to a thicker oxide layer. Furthermore, high-temperature thermal stress not only causes crack propagation but also increases the surface atomic migration rate, accelerating the oxidation reaction. As cracks form or propagate, the crack tip and micropores expose more surface for oxygen reaction, further thickening the oxide layer. When crack depth and oxide layer thickness increase synchronously, it indicates that thermal stress and oxidation reaction act simultaneously on the crankshaft surface. Another scenario involves cracks also driven by high-temperature thermal stress, but in this case, the total oxygen content in the furnace decreases while the oxide layer thickness still increases, exhibiting an inverse relationship with changes in oxygen content. This is because the crankshaft metal contains trace amounts of dissolved oxygen or existing oxide films. At high temperatures, this oxygen can migrate to the surface and participate in the oxidation reaction. Even if the oxygen content in the furnace decreases, the surface oxide layer can still thicken, thus forming a high-temperature oxygen-deficient coupling anomaly.

[0052] By monitoring key indicators such as crack depth, oxide layer thickness, quenching temperature, and oxygen content, the system can capture potential anomalies, promptly identify possible defect sources, and enhance its diagnostic capabilities, enabling more accurate identification of trends that may lead to failures. Normalization processing allows for effective comparison of data with different dimensions and scales on the same platform, eliminating data inconsistencies. It converts the rate of change of each indicator into standardized data, thus avoiding data bias caused by differences in dimensions and improving the accuracy and reliability of data analysis. Through the calculation of the Pearson correlation coefficient, the interrelationships between various indicators are precisely quantified, and by setting similarity thresholds to determine defect types, not only is the sensitivity of the system's diagnosis improved, but its accuracy is also enhanced. Especially when multiple parameters change synergistically, it can effectively distinguish between thermo-oxygen coupling anomalies and high-temperature oxygen-deficient coupling anomalies, facilitating targeted measures and avoiding blind intervention, thus improving the scientific rigor and specificity of early warning. The introduction of automated early warning functions greatly improves the speed of fault diagnosis and response. By automatically analyzing and issuing early warnings, the system avoids delays caused by manual inspection and response, enabling preventative measures to be taken before malfunctions occur. This reduces the risk of equipment damage or production downtime. The system not only issues alerts at the initial stage of a fault but also provides valuable decision-making support to maintenance personnel through multi-parameter comparison and analysis, thereby improving the overall system's safety and reliability. Intelligent data analysis and diagnostic methods enhance the system's real-time monitoring capabilities, improve the accuracy and efficiency of fault diagnosis, and reduce manual intervention and response time. This results in significant benefits in improving equipment safety, extending service life, and reducing maintenance costs.

[0053] Specifically, the adjustment module includes: The growth rate calculation unit is used to calculate the difference in the number of the target crankshafts at any two adjacent moments within the preset adjustment time to obtain a number of speed changes, and to sum the number of speed changes to obtain a total number of speed changes. The quantity fluctuation calculation unit is used to calculate the standard deviation of the quantity of the target crankshaft within the preset adjustment time when the total value of the quantity changes is greater than the preset speed threshold, so as to obtain the quantity fluctuation value. An adjustment unit, connected to the quantity fluctuation calculation unit, is used to adjust the preset temperature threshold or the preset oxygen content threshold according to the quantity fluctuation value and the preset quantity fluctuation threshold.

[0054] The preset speed change threshold is a threshold parameter used to determine whether the rate of change of the number of target crankshafts within a preset adjustment period reaches a significant level. It depends on the cycle stability of the production line, the output rate of the target crankshaft, and the fluctuation characteristics between production batches. It is usually set between 5 and 15. In this embodiment, it is set to 10. It can trigger the subsequent quantity fluctuation calculation process when the total change value of the number of target crankshafts exceeds this value, so as to respond in time when abnormal fluctuation trends occur in production. The preset quantity fluctuation threshold is a threshold parameter used to determine whether the fluctuation of the number of target crankshafts within a preset adjustment period is abnormal. It depends on the production capacity stability, sampling period, and historical statistical fluctuation range of the number of target crankshafts. It is usually set between 2.0 and 5.0. In this embodiment, it is set to 3.5. It can trigger the system's parameter adaptive adjustment mechanism when the fluctuation of the number of target crankshafts exceeds the normal range, so as to automatically reduce the preset temperature threshold or oxygen content threshold to stabilize the production environment and reduce the probability of defective crankshafts.

[0055] By calculating the growth rate of the target crankshaft quantity, the change in the number of target crankshafts at each moment is captured, thereby identifying potential quality anomalies. By calculating the total growth rate of the quantity, the changing trend of target crankshafts with simultaneous abnormal crack depth and oxide layer thickness is identified. When the total growth rate of the quantity is greater than the preset speed threshold, it indicates that the number of such composite abnormal crankshafts is on the rise. At this time, the standard deviation of the number of target crankshafts within the preset adjustment time is calculated to quantify the data dispersion in this rising process, that is, to obtain the quantity fluctuation value. The quantity fluctuation value can reflect the stability of the abnormal crankshaft quantity fluctuation, and it is compared with the preset quantity fluctuation threshold, thereby realizing the dynamic and precise adjustment of key parameters in the production process.

[0056] Specifically, the adjustment unit includes: The first adjustment subunit is used to reduce the preset temperature threshold according to the quantity fluctuation value and the preset quantity fluctuation threshold when the quantity fluctuation value is greater than the preset quantity fluctuation threshold, M'=M×(1-s×│A-A0│ / A0), where M' is the adjusted preset temperature threshold, M is the preset temperature threshold before adjustment, s is the preset temperature adjustment coefficient, A is the quantity fluctuation value, and A0 is the preset quantity fluctuation threshold. The first adjustment subunit is used to reduce the preset oxygen content threshold according to the quantity fluctuation value and the preset quantity fluctuation threshold when the quantity fluctuation value is less than the preset quantity fluctuation threshold, N'=N×(1-m×│A-A0│ / A0), where N' is the adjusted preset oxygen content threshold, N is the original preset oxygen content threshold, and m is the preset oxygen content adjustment coefficient.

[0057] The preset temperature adjustment coefficient is a proportional parameter used to control the adjustment of the preset temperature threshold when the quantity fluctuation value exceeds the preset quantity fluctuation threshold. It depends on the sensitivity of the quenching temperature to crack-like defects during the production process and the response accuracy of the temperature control system. It is usually set between 0.05 and 0.20. In this embodiment, it is set to 0.10, which can make the system temperature threshold decrease at a relatively stable rate when a large fluctuation in the number of target crankshafts is detected, avoiding over-adjustment that leads to unstable temperature control. The preset oxygen content adjustment coefficient is a proportional parameter used to control the adjustment of the preset oxygen content threshold when the quantity fluctuation value is lower than the preset quantity fluctuation threshold. It depends on the degree of influence of changes in oxygen content in the heat treatment furnace on the formation of the crankshaft oxide layer and the adjustment sensitivity of the oxygen control system. It is usually set between 0.02 and 0.10. In this embodiment, it is set to 0.05, which can make the oxygen content threshold decrease at a gradual rate when a small fluctuation in the number of target crankshafts is detected and the proportion of oxidation defects increases.

[0058] By intelligently analyzing the fluctuations in the number of target crankshafts, the system effectively adjusts preset thresholds in real time. When the fluctuation in the number of target crankshafts exceeds the preset threshold, the system assumes that the quenching temperature control may be unstable. This is because the impact of abnormal quenching temperature on crankshaft quality is usually significant in the short term, especially during temperature control instability or adjustment. The number of target crankshafts can vary greatly within a preset adjustment period. Therefore, by reducing the preset temperature threshold, the system helps the production system reduce the amplitude of temperature fluctuations, ensuring that the quenching temperature remains within the optimal range and avoiding quality defects caused by excessive temperature. Conversely, when the fluctuation is less than the preset threshold, the system suspects that the oxygen concentration may be uneven. This is because the impact of abnormal oxygen content is often continuous, resulting in a more gradual fluctuation in crankshaft quality. If the abnormal oxygen content causes surface defects on the crankshaft, there will not be a large difference in the number of target crankshafts at different times within the preset adjustment period. Through this dynamic adjustment mechanism based on the fluctuations in the number of target crankshafts, the system can accurately respond to environmental changes during the production process, automatically adjust key parameters, improve the stability of the production process, reduce the defect rate, and ensure high consistency and high quality of the final product.

[0059] On the other hand, this embodiment also provides a machine vision-based online detection method for crankshaft surface defects, including: The quenching temperature, oxygen content in the heat treatment furnace, and surface light reflectance, surface roughness, crack depth, and oxide layer thickness of each crankshaft under test are obtained during the heat treatment process of the crankshaft industrial production line. The production environment is determined to be abnormal based on the quenching temperature, the preset temperature threshold, the oxygen content, and the preset oxygen content threshold, so as to obtain the result of environmental abnormality. Based on the results of environmental anomalies, several crankshafts of interest are determined according to the surface light reflectivity and preset reflection threshold of each crankshaft under test. Several first abnormal crankshafts are selected according to the correlation characteristics of the surface roughness and crack depth of the crankshafts of interest within a preset screening time. Several second abnormal crankshafts are selected according to the correlation characteristics of the surface roughness and oxide layer thickness of the crankshafts of interest within a preset screening time. Based on the overlap characteristics of the first abnormal crankshaft and the second abnormal crankshaft within the next preset time period, several target crankshafts are determined; The defect type of the target crankshaft is determined based on the quenching temperature, the oxygen content, the crack depth, and the oxide layer thickness of the target crankshaft, and an early warning is issued based on the defect type. The preset temperature threshold or the preset oxygen content threshold is adjusted based on the change in the number of target crankshafts within the next preset adjustment period.

[0060] Furthermore, by employing an online crankshaft surface defect detection method based on machine vision and multi-source parameter fusion, high-precision identification and intelligent early warning of crankshaft defects in industrial production processes have been achieved. By utilizing quenching temperature to reflect the stability of energy input during heat treatment, oxygen content to reveal changes in the oxidation environment within the furnace, surface reflectivity to reflect the density and oxidation degree of the material surface, and surface roughness to characterize the microscopic morphology after processing and thermal stress, potential anomalies in crankshafts of interest can be preliminarily identified. Combining the dynamic trends of crack depth and oxide layer thickness, abnormal regions affected by the coupling of high-temperature oxidation and thermal stress are further refined. The correlation between reflectivity and roughness characterizes abnormal patterns of surface energy reflection behavior, and the linkage rate of change between crack and oxide layer thickness reflects the evolution trend of surface structure degradation. The sensitivity of the screening threshold is then adjusted based on fluctuations in temperature and oxygen content, dynamically adapting to different batches and production conditions. Finally, the defect type of the target crankshaft is determined based on the multi-parameter fusion results, and a warning command with defect characteristics is generated. This ensures the real-time, accurate, and adaptive nature of the defect identification process, guaranteeing continuous monitoring and reliable early warning of crankshaft surface quality in complex production environments, effectively improving product consistency and production safety.

[0061] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A machine vision-based online detection system for crankshaft surface defects, characterized in that, The acquisition module is used to acquire the quenching temperature, oxygen content in the heat treatment furnace, and surface light reflectance, surface roughness, crack depth, and oxide layer thickness of each crankshaft under test in the heat treatment process of the crankshaft industrial production line, based on the acquired images. The determination module is used to determine whether the production environment is abnormal based on the quenching temperature, the preset temperature threshold, the oxygen content, and the preset oxygen content threshold, so as to obtain an abnormal environment result. A screening module is used to determine a number of crankshafts of interest based on the surface light reflectivity and a preset reflection threshold of each crankshaft under test, based on the abnormal environmental results; to screen out a number of first abnormal crankshafts based on the correlation characteristics between the surface roughness and the crack depth of the crankshafts of interest within a preset screening time; and to screen out a number of second abnormal crankshafts based on the correlation characteristics between the surface roughness and the oxide layer thickness of the crankshafts of interest within a preset screening time. The determination module is used to determine several target crankshafts based on the overlap characteristics of the first abnormal crankshaft and the second abnormal crankshaft within a next preset determination time period; The early warning module is used to determine the defect type of the target crankshaft based on the quenching temperature, the oxygen content, the crack depth and the oxide layer thickness of the target crankshaft, and to issue an early warning prompt based on the defect type. An adjustment module is used to adjust the preset temperature threshold or the preset oxygen content threshold according to the change in the number of target crankshafts within the next preset adjustment period.

2. The online crankshaft surface defect detection system based on machine vision according to claim 1, characterized in that, The determination module includes: A temperature deviation calculation unit is used to calculate the relative deviation between the quenching temperature and the preset temperature threshold to obtain the temperature deviation. An oxygen content deviation calculation unit is used to calculate the relative deviation between the oxygen content and the preset oxygen content threshold to obtain the oxygen content deviation. The determination unit is connected to the temperature deviation calculation unit and the oxygen content deviation calculation unit respectively, and is used to perform a weighted summation of the temperature deviation and the oxygen content deviation to obtain a joint deviation, and to determine that the production environment is abnormal when the joint deviation is greater than a preset joint threshold, so as to obtain an environmental abnormality result.

3. The online crankshaft surface defect detection system based on machine vision according to claim 2, characterized in that, The filtering module includes: The first determination unit is used to determine that the surface of the crankshaft under test is abnormal when the surface light reflectance is lower than the preset reflection threshold, so as to obtain several crankshafts of interest. A roughness fluctuation calculation unit, which is connected to the first determination unit, is used to calculate the standard deviation of the surface roughness of each of the crankshafts of interest within the preset screening time, so as to obtain several roughness fluctuation values. A screening unit, connected to the roughness fluctuation calculation unit, is configured to: when the roughness fluctuation value of each crankshaft of interest is greater than a preset fluctuation threshold, select a number of first abnormal crankshafts from all the crankshafts of interest based on the crack depth variation characteristics within the preset screening time; and when the roughness fluctuation value of each crankshaft of interest is less than the preset fluctuation threshold, select a number of second abnormal crankshafts from all the crankshafts of interest based on the oxide layer thickness variation characteristics within the preset screening time.

4. The online crankshaft surface defect detection system based on machine vision according to claim 3, characterized in that, The filtering unit includes: The depth growth rate calculation subunit is used to calculate the relative deviation of the crack depth corresponding to any two adjacent moments within the preset screening time to obtain several instantaneous depth growth rates, and to calculate the average value of all instantaneous depth growth rates to obtain the average depth growth rate. The average thickness calculation subunit is used to calculate the average thickness of all oxide layers within the preset screening time to obtain the average thickness. A filtering subunit, which is connected to the depth growth rate calculation subunit and the thickness average calculation subunit respectively, is used to determine the crankshaft of interest with the average depth growth rate greater than a preset growth rate threshold as the first abnormal crankshaft, and to determine the crankshaft of interest with the average thickness greater than a preset thickness threshold as the second abnormal crankshaft.

5. The online crankshaft surface defect detection system based on machine vision according to claim 4, characterized in that, The determining module includes: The acquisition unit is used to acquire the crankshafts of interest that are simultaneously identified as the first abnormal crankshaft and the second abnormal crankshaft at each moment within the preset time period, so as to obtain a number of candidate crankshafts. A statistics unit, connected to the acquisition unit, is used to count the frequency of occurrence of each candidate crankshaft within the preset time period to obtain several overlap frequencies. A determining unit, which is connected to the statistical unit, is used to determine the candidate crankshafts whose overlap frequency is greater than a preset frequency threshold as the target crankshafts.

6. The online crankshaft surface defect detection system based on machine vision according to claim 5, characterized in that, The early warning module includes: The variation calculation unit is used to calculate the difference in crack depth of each target crankshaft corresponding to any two adjacent moments within the previous preset historical defect duration, to obtain several depth variation rates; and to calculate the difference in oxide layer thickness of each target crankshaft corresponding to any two adjacent moments within the previous preset historical defect duration, to obtain several thickness variation rates; and to calculate the difference in quenching temperature corresponding to any two adjacent moments within the previous preset historical defect duration, to obtain several temperature variation rates; and to calculate the difference in oxygen content corresponding to any two adjacent moments within the previous preset historical defect duration, to obtain several oxygen content variation rates. A normalization processing unit, which is connected to the change calculation unit, is used to normalize the depth change rate, the thickness change rate, the temperature change rate and the oxygen content change rate respectively to obtain a depth normalized dataset, a thickness normalized dataset, a temperature normalized dataset and an oxygen content normalized dataset. A type diagnosis unit, connected to the normalization processing unit, is used to determine the defect type based on the depth normalized dataset, the thickness normalized dataset, the temperature normalized dataset, and the oxygen content normalized dataset. An early warning unit, which is connected to the type diagnostic unit, is used to issue an early warning based on the defect type.

7. The online crankshaft surface defect detection system based on machine vision according to claim 6, characterized in that, The diagnostic unit of this type includes: A similarity calculation subunit is used to calculate the Pearson correlation coefficient between the depth-normalized dataset and the temperature-normalized dataset to obtain a first defect similarity, and to calculate the Pearson correlation coefficient between the thickness-normalized dataset and the oxygen content-normalized dataset to obtain a second defect similarity, and to calculate the Pearson correlation coefficient between the depth-normalized dataset and the thickness-normalized dataset to obtain a third defect similarity. A type diagnosis subunit, connected to the similarity calculation subunit, is used to determine that the defect type is thermo-oxygen coupling anomaly when the similarity of the first defect is greater than a preset first similarity threshold, the similarity of the second defect is greater than a preset second similarity threshold, and the similarity of the third defect is greater than a preset third similarity threshold; and to determine that the defect type is high-temperature oxygen-deficient coupling anomaly when the similarity of the first defect is greater than a preset first similarity threshold, the similarity of the second defect is less than a preset second similarity threshold, and the similarity of the third defect is less than a preset third similarity threshold.

8. The online crankshaft surface defect detection system based on machine vision according to claim 7, characterized in that, The adjustment module includes: The growth rate calculation unit is used to calculate the difference in the number of the target crankshafts at any two adjacent moments within the preset adjustment time to obtain a number of speed changes, and to sum the number of speed changes to obtain a total number of speed changes. The quantity fluctuation calculation unit is used to calculate the standard deviation of the quantity of the target crankshaft within the preset adjustment time when the total value of the quantity changes is greater than the preset speed threshold, so as to obtain the quantity fluctuation value. An adjustment unit, connected to the quantity fluctuation calculation unit, is used to adjust the preset temperature threshold or the preset oxygen content threshold according to the quantity fluctuation value and the preset quantity fluctuation threshold.

9. The online crankshaft surface defect detection system based on machine vision according to claim 8, characterized in that, The adjustment unit includes: The first adjustment subunit is used to reduce the preset temperature threshold according to the quantity fluctuation value and the preset quantity fluctuation threshold when the quantity fluctuation value is greater than the preset quantity fluctuation threshold. The first adjustment subunit is used to reduce the preset oxygen content threshold according to the quantity fluctuation value and the preset quantity fluctuation threshold when the quantity fluctuation value is less than the preset quantity fluctuation threshold.

10. A machine vision-based online detection method for crankshaft surface defects, applied to the machine vision-based online detection system for crankshaft surface defects as described in any one of claims 1-9, characterized in that, include: The quenching temperature, oxygen content in the heat treatment furnace, and surface light reflectance, surface roughness, crack depth, and oxide layer thickness of each crankshaft under test are obtained during the heat treatment process of the crankshaft industrial production line. The production environment is determined to be abnormal based on the quenching temperature, the preset temperature threshold, the oxygen content, and the preset oxygen content threshold, so as to obtain the result of environmental abnormality. Based on the results of environmental anomalies, several crankshafts of interest are determined according to the surface light reflectivity and preset reflection threshold of each crankshaft under test. Several first abnormal crankshafts are selected according to the correlation characteristics of the surface roughness and crack depth of the crankshafts of interest within a preset screening time. Several second abnormal crankshafts are selected according to the correlation characteristics of the surface roughness and oxide layer thickness of the crankshafts of interest within a preset screening time. Based on the overlap characteristics of the first abnormal crankshaft and the second abnormal crankshaft within the next preset time period, several target crankshafts are determined; The defect type of the target crankshaft is determined based on the quenching temperature, the oxygen content, the crack depth, and the oxide layer thickness of the target crankshaft, and an early warning is issued based on the defect type. The preset temperature threshold or the preset oxygen content threshold is adjusted based on the change in the number of target crankshafts within the next preset adjustment period.

Images