A continuous skin sorting and collecting system for a swaging punch die

Abstract

This invention relates to the field of continuous strip classification and recycling technology for stamping equipment, and particularly to a continuous strip classification and collection system for ring forging punching dies. The system includes an acquisition module, a first judgment module, a second judgment module, a classification and recycling module, an execution module, and a calibration module. This invention collects multi-source data through the acquisition module and uses a joint evaluation of strength deviation and flatness deviation to quickly identify continuous strips with small deviations from the normal state. These strips are then finely classified and recycled according to material and morphological value. By identifying the presence of multi-station mixed production or systematic die deviations, the system enhances its ability to perceive equipment anomalies and process deviations. Furthermore, by identifying raw material quality fluctuations, the system can infer the progressive wear of the die, enabling timely equipment maintenance or calibration. This effectively solves the problem of overlooked die health risks and reduced overall quality of ring forgings due to the difficulty in inferring die health from the continuous strip state.

Description

Technical Field

[0001] This invention relates to the field of stamping equipment peel sorting and recycling technology, and in particular to a peel sorting and collection system for ring forging punching dies. Background Technology

[0002] In the existing large hydraulic press ring forging punching process, the simple structure of the mold and its limited heat and wear resistance often require frequent cleaning or replacement. The resulting scabbard is not only typically collected and cleaned manually using tools like hooks and pliers from under the punching table, but also poses safety hazards such as burns and mechanical injuries. The quality of the scabbard varies depending on the condition of the punching mold. The recycling value of different scabbards is related not only to their material but also to their thickness, surface smoothness, and warping. For example, scabbards made of high-temperature alloys, when the punching mold is in good condition, have uniform thickness, a smooth surface, and no warping. They can be directly used as high-quality furnace charge for smelting and, where technically permissible, can be used to produce parts with lower strength requirements, thus commanding a higher recycling price. However, when the punching mold has problems such as wear, the scabbards become less valuable. When issues such as decreased precision occur, the resulting sheet material will also exhibit irregular shapes or warping deformations. Not only can it not be used directly, but even if it is used as furnace charge for remelting, it needs to undergo secondary processing such as briquetting and cutting before it can be put into the furnace, which increases processing costs. Otherwise, due to unstable stacking in the furnace or uneven heating of different parts, different melting rates may occur, resulting in severe oxidation and burning of some sheet material. This leads to a significant decrease in smelting recovery rate, an increase in slag, and a huge waste of resources. Moreover, existing equipment usually only has the opportunity to detect quality problems in the sheet material when it is transferred from the collection point. At this time, the precision of the mold has already decreased significantly. Even if the appearance of the ring forging is not obvious, there are likely to be small errors or defects inside, which ultimately leads to a decrease in the yield of ring forgings. This seriously restricts the overall efficiency of the ring forging production line, the quality of ring forgings, and the recycling rate of sheet material.

[0003] Chinese Patent Application Publication No. CN106807801A discloses an intelligent online detection method and device for waste material on the surface of a high-speed fine blanking machine mold. The method includes: S1, after the fine blanking machine completes the stamping process, an air blowing device removes the waste material from the surface to be inspected, and a thermal imager is activated to scan the surface to be inspected to obtain an image of the surface and effective area temperature gradient information; S2, the thermal imager transmits the obtained image and temperature gradient information to a computer, the computer calculates the difference between the highest and lowest temperatures based on the temperature gradient information, and compares the difference with a preset value; if the difference is greater than the preset value, it indicates that there is waste material on the surface to be inspected, and the machine needs to be stopped for inspection; if the difference is not greater than the preset value, it indicates that there is no waste material on the surface to be inspected, and the image and temperature gradient information are deleted.

[0004] Therefore, the aforementioned comparative document has at least the following problems: It uses a thermal imager to scan the surface of the fine stamping machine mold and combines this with temperature gradient information to determine the presence of processing waste. However, this method only focuses on the presence or absence of waste and triggers a shutdown inspection upon detection, without addressing the material-based recycling of the punched slabs. All detected waste is mixed, and the recycling prices of slabs of different materials vary greatly, leading to high-value alloy materials being treated as ordinary waste, resulting in serious resource waste. Furthermore, it uses waste detection as a simple quality judgment and shutdown basis, only making a binary determination of the existence of waste without considering the value of waste of the same material and subsequent processing methods. It struggles to distinguish between slabs of the same material but different values, as this affects the differentiation between ordinary scrap, regular slabs, and deformed slabs. Mixed stacking of materials leads to different melting rates during subsequent smelting, ultimately resulting in a significant decrease in the overall smelting recovery rate. The system employs a fixed detection logic based on temperature threshold comparison, deleting images and temperature information after each detection if no waste is found. This only enables immediate judgment and shutdown response for waste on the surface of a single die after a single stamping operation. This necessitates individual configuration and monitoring for each machine, making it impossible to continuously statistically analyze and retrospectively examine the actual operating conditions of multiple production lines running simultaneously. Furthermore, it is difficult to identify the evolution of health risks such as progressive wear and alignment deviations at different punching stations and their dies from scattered information, hindering the formation of long-term trend perception and early warning of die health status. This easily leads to hidden defects in the quality of ring forgings, reducing the overall quality rate. Summary of the Invention

[0005] To address this issue, the present invention provides a system for classifying and collecting the connecting skin of punching dies for ring forgings. This system overcomes the problem in the prior art where the health risks of the die are overlooked due to the difficulty in inferring the health status of the die from the state of the connecting skin, which leads to a decrease in the overall quality rate of ring forgings by monitoring the quality and temporal characteristics of the connecting skin in real time.

[0006] To achieve the above objectives, the present invention provides a skin sorting and collection system for punching dies of ring forgings, comprising: The acquisition module is used to acquire in real time the specific strength coefficient and flatness of the ring forging skin at each punching station, as well as the key element ratio and skin thickness of the ring forging skin at the classification station. The first determination module is used to determine whether there is a suspected classification anomaly based on the strength deviation and flatness deviation of the ring forging skin, wherein the strength deviation and flatness deviation of the ring forging skin are determined based on the specific strength coefficient and the flatness. The second determination module is used to determine whether a material anomaly has occurred based on the material anomaly degree and a preset material threshold. The material anomaly degree is determined based on the determination result of the suspected classification anomaly, the maximum value of the absolute values ​​of all the intensity deviations within the monitoring period, and the key element ratio. The classification and recycling module is used to classify and recycle the ring forging skin at the classification station based on the determination result of the occurrence of shape abnormality. The determination result of the occurrence of shape abnormality is determined based on the determination result of the absence of the material abnormality, the maximum value of all the flatness deviations within the monitoring period, and the thickness of the skin. The execution module is used to perform corresponding quality calibration operations on each punching station based on the frequency of the material abnormalities and shape abnormalities occurring within the detection cycle, as well as the specific strength coefficient, flatness deviation, and skin thickness within the detection cycle. A calibration module is used to calibrate the preset material threshold based on the frequency of performing the quality calibration operation within the specified number of testing cycles and the material anomaly within each testing cycle.

[0007] Furthermore, the first determination module includes: The data determination unit is used to determine the strength deviation and the flatness deviation based on the specific strength coefficient, the flatness, the preset strength mean, the preset flatness mean, the preset strength standard deviation, and the preset flatness standard deviation; The suspected classification unit is used to determine whether the suspected classification anomaly occurs based on the strength deviation and the flatness deviation of the ring forging skin.

[0008] Furthermore, the suspected determination unit includes: A deviation determination subunit is used to determine the Euclidean distance deviation based on the strength deviation and the flatness deviation of the ring forging skin; The suspected classification subunit is used to determine whether the suspected classification anomaly has occurred based on the Euclidean distance deviation of the ring forging skin and a preset deviation threshold.

[0009] Furthermore, the second determination module includes: The material matching determination unit is used to determine the material anomaly degree based on the maximum value of the absolute values ​​of all the intensity deviations and the key element deviation degree when the judgment result of the suspected classification anomaly occurs, wherein the key element deviation degree is determined based on the key element ratio and the preset element ratio. The material anomaly determination unit is used to determine whether the material anomaly has occurred based on the material anomaly degree and the preset material threshold.

[0010] Furthermore, the sorting and recycling module includes: The shape matching determination unit is used to determine the shape abnormality based on the determination result that no material abnormality has occurred, according to the maximum value of all the flatness deviations and the skin thickness deviation within the monitoring period, wherein the skin thickness deviation is determined based on the skin thickness and a preset thickness, and the maximum value of all flatness deviations is the flatness deviation with the largest absolute value among all flatness deviations. A shape anomaly determination unit is used to determine whether the shape anomaly has occurred based on the shape anomaly degree and a preset shape threshold. The shape sorting and recycling unit is used to recycle the corresponding ring forging skin at the sorting station to the shape collection station based on the determination result of the occurrence of the shape abnormality.

[0011] Furthermore, the shape matching determination unit includes: The shape matching calculation subunit is used to determine the same-direction matching value based on the judgment result that no material abnormality has occurred, the maximum value of all the flatness deviations within the monitoring period and the skin thickness deviation. A shape feature calculation subunit is used to determine the shape anomaly degree based on the maximum value of all the flatness deviations, the skin thickness deviation, and the same-direction matching value.

[0012] Furthermore, the sorting and recycling module also includes: The material sorting and recycling unit is used to recycle the corresponding ring forging skin at the sorting station to the material collection station based on the determination result of the occurrence of the material abnormality.

[0013] Furthermore, the execution module includes: The proportion determination unit is used to determine the classification abnormality proportion and the shape abnormality proportion based on the frequency of the material abnormality occurring within the detection period, the frequency of the shape abnormality, and the total number of times the material abnormality has occurred. The workstation determination unit is used to determine the workstation abnormality degree based on the current strength standard deviation and the classification abnormality ratio, and to determine whether a workstation abnormality has occurred in combination with a preset workstation threshold. The current strength standard deviation is determined based on the standard deviation of all the specific strength coefficients within the detection cycle. The stacking determination unit is used to determine the stacking anomaly degree based on the shape anomaly ratio, the current flatness deviation, and the current thickness offset, and to determine whether a stacking anomaly has occurred in combination with a preset stacking threshold. The current flatness deviation is determined based on the average of all the flatness deviations within the detection period, and the current thickness offset is determined based on the average of all the skin thickness deviations within the detection period. The execution unit is used to perform corresponding quality calibration operations on each punching station when a station abnormality or stacking abnormality occurs.

[0014] Verify whether there is a situation where different products are produced in multiple stations. Otherwise, check each punching station and raw materials. If stacking abnormalities occur, adjust the reducer at the end of each punching table slide.

[0015] Furthermore, the ratio determination unit includes: The classification anomaly determination subunit is used to determine the classification anomaly ratio based on the frequency of the material anomaly occurring within the detection period, the frequency of the shape anomaly, and the total number of times the material anomaly has been determined. The shape anomaly determination subunit is used to determine the shape anomaly ratio based on the frequency of the shape anomaly occurring within the detection period and the total number of times the material anomaly has occurred.

[0016] Furthermore, the calibration module includes: The relaxed determination unit is used to determine the representative value of the material based on the average value and standard deviation of all material anomalies when no material anomalies occur in each detection cycle within the calibration number of detection cycles and a preset normal proportion; A calibration judgment unit is used to determine whether to increase or decrease the preset material threshold based on the frequency of the quality calibration operation corresponding to the workstation abnormality in the quality calibration operation performed within the specified number of detection cycles, a preset high-frequency threshold, the material representative value, and a preset stable interval, wherein the preset stable interval is determined based on the preset material threshold and a preset error. The calibration determination unit is used to increase a new preset material threshold based on the determination result of increasing the preset material threshold, based on the material representative value and the previous preset material threshold, and to decrease a new preset material threshold based on the determination result of decreasing the preset material threshold, based on the material representative value and the previous preset material threshold.

[0017] Compared with existing technologies, the beneficial effects of this invention are as follows: by acquiring multi-source data through the acquisition module, it quickly identifies connected skins with small deviations from the normal state using strength deviation and flatness deviation, reducing the system's computational burden. When suspected classification anomalies occur, it performs refined classification and recycling based on material value and morphological value. By using the maximum strength deviation, it enhances the sensitivity to the ratio of key elements related to material composition, making the detection of deviation states of connected skin materials more sensitive, reducing the probability of low-quality connected skins being mixed with high-quality connected skins due to their inability to be identified. Furthermore, it identifies the morphological value of connected skins by recognizing their warping and deformation states, and simultaneously identifies whether multi-station mixing exists. Systemic deviations in production or molds can be promptly alerted to staff and prompt for manual verification, enhancing the system's ability to detect equipment anomalies and process deviations. Alternatively, by identifying abnormal stacking during transport, timely adjustments can be made to stagger the transport distance and arrival time at the sorting station of the connecting sheets, reducing the probability of misjudgments caused by the stacking of connecting sheets. Finally, the statistical distribution of material anomalies over multiple inspection cycles can be used to identify raw material quality fluctuations, inferring the progressive wear of the mold, and enabling timely equipment maintenance or calibration. This effectively solves the problem of overlooking mold health risks and reducing the overall quality rate of ring forgings due to the difficulty in inferring mold health from the condition of connecting sheets. Attached Figure Description

[0018] Figure 1 This is a system diagram of the skin sorting and collection system for ring forging punching dies in this embodiment; Figure 2 This is a judgment diagram used in this embodiment to determine whether a suspected classification anomaly has occurred; Figure 3 This is a judgment diagram showing the classification of ring forging skin at the classification station in this embodiment; Figure 4 This is a diagram showing the determination of workstation abnormalities and stacking abnormalities in this embodiment. Detailed Implementation

[0019] 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.

[0020] 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.

[0021] Please see Figure 1 As shown, this is a system diagram of the skin sorting and collection system for ring forging punching dies in this embodiment, including: The acquisition module is used to acquire in real time the specific strength coefficient and flatness of the ring forging skin at each punching station, as well as the key element ratio and skin thickness of the ring forging skin at the classification station. The first judgment module is used to determine whether there is a suspected classification anomaly based on the strength deviation and flatness deviation of the ring forging skin. The strength deviation and flatness deviation of the ring forging skin are determined based on the specific strength coefficient and flatness. The second determination module is used to determine whether a material anomaly has occurred based on the material anomaly degree and a preset material threshold. The material anomaly degree is determined based on the determination result of suspected classification anomaly, the maximum value of the absolute value of all intensity deviations within the monitoring period, and the key element ratio. The classification and recycling module is used to classify and recycle the ring forging skin at the classification station based on the judgment result of the occurrence of shape abnormality. The judgment result of the occurrence of shape abnormality is determined based on the judgment result of the absence of material abnormality, the maximum value of all flatness deviations within the monitoring period, and the thickness of the skin. The execution module is used to perform corresponding quality calibration operations on each punching station based on the frequency of material and shape abnormalities occurring within the inspection cycle, as well as the specific strength coefficient, flatness deviation, and skin thickness within the inspection cycle. The calibration module is used to calibrate a preset material threshold based on the frequency of collection operations performed within a calibration number of detection cycles and the material anomaly degree within each detection cycle.

[0022] This solution is implemented in a large workshop equipped with a large hydraulic press ring forging punching production line. This production line includes several punching stations, conveyor chains transporting the ring forging parts to each punching station, and sorting stations. It primarily performs ring forging punching operations based on the production requirements of various materials or different workpieces. Each punching station includes hydraulic punching equipment, and the main cylinder hydraulic circuit of the hydraulic punching equipment is equipped with a pressure sensor to collect the peak punching force of the corresponding punching station. Each punching station is equipped with a punch connected to the hydraulic equipment and a punching table with quick-release punching dies. The punching dies include a final punch die, connecting bolts, T-bolts, upper and lower modules, etc. The corresponding tooling is made of hot work die steel 5CrNiMo to increase its heat resistance and wear resistance. A hydraulic ejection cylinder is installed under the punching table. After punching, the hydraulic ejection cylinder ejects the ring forging skin, causing it to separate from the cavity of the punching die and be pushed into the slide corresponding to the punching station. The slide is equipped with at least a dynamic scale for measuring the weight of the ring forging skin, a laser displacement sensor for measuring the height distribution of sampling points on the skin surface and calculating the flatness, and a discharge conveyor belt for moving the ring forging skin. The discharge conveyor belt is used to transport the ring forging skin from the corresponding punching station to the conveyor chain for unified collection of the ring forging skin. The conveyor chain is used to transport the ring forging skin to the sorting station.

[0023] The sorting station is equipped with a Laser-Induced Breakdown Spectroscopy (LIBS) analyzer and a laser displacement sensor mounted on top. It also features several electrically controlled push rods and sorting collection channels connecting to various collection boxes, as well as a default collection channel for each box. Each push rod corresponds to a specific collection channel. The system controls different push rods to push different types of forged steel sheets into different collection channels, thus completing the sorting and collection of the forged steel sheets. The LIBS analyzer is used to excite a plasma spectrum on the sheet surface to calculate the critical element ratios of the forged steel sheets. The laser displacement sensor measures the height of the sheet surface and, combined with the known height of the sorting station reference plane, determines the sheet thickness.

[0024] In this embodiment, the sorting station connects two sorting collection channels and one default collection channel. During actual deployment, the number of collection channels and the sorting criteria can be changed based on actual deployment needs. In this embodiment, the default collection channel corresponds to the high-value furnace material collection box, and the two sorting collection channels correspond to the abnormal shape collection box and the low-material mixed material collection box, respectively. Each electrically controlled push rod is electrically connected to the sorting and recycling module, and each data acquisition device is electrically connected to the acquisition module. This system is deployed based on an existing edge computing system to achieve timely control and accurate sorting. In this embodiment, the data acquired by the acquisition module is obtained by the aforementioned existing device, and the specific acquisition principle is as follows: The specific strength coefficient of the ring forging sheet at each punching station is determined by obtaining the peak punching force through pressure sensors installed in the main cylinder oil circuit of each punching station, and the sheet weight through a dynamic scale installed in the slide of the same punching station. The ratio of the two data points determines the specific strength coefficient, which reflects the shear strength characteristics per unit mass of the material. The pressure sensors in the main cylinder oil circuit of each punching station collect oil pressure at a sampling rate of not less than 1kHz during the punching process. After statistical analysis, the peak punching force representing the maximum punching force in a single operation is obtained, and its unit is kilonewtons. The dynamic scale in the slide of the same punching station collects the individual mass of the sheet as it passes through, and its unit is kilograms. The specific strength coefficient is obtained by dividing the peak punching force by the sheet weight, and its unit is kilonewtons per kilogram, i.e., kN / kg. In actual production, its value range is usually [0.5, 5.0] kN / kg.

[0025] The flatness of the ring forging skin at each punching station is directly obtained by laser displacement sensors installed in the slide rails of each punching station. The laser displacement sensors measure the height of three different radial positions on the surface of the skin, which is usually the average height of each sampling point on the inner diameter, middle diameter, and outer diameter. Then, the thickness values ​​at the three different positions are calculated by combining the height of the reference surface of the conveyor belt in the slide rail. The thickness unit is millimeters. Finally, the range of these three thickness values ​​is taken as the flatness, which reflects the degree to which the thickness of the skin deviates from the mold reference. Its value range is generally [0, 3] mm.

[0026] The critical element ratio of the ring forging skin at the sorting station is determined by using a LIBS analyzer installed above the sorting station. The analyzer generates plasma by ablating the skin surface with a pulsed laser, collecting the characteristic spectral intensities of chromium around 267.7 nm and iron around 271.4 nm, and calculating the ratio of these two intensities. This critical element ratio is a dimensionless constant commonly used in engineering applications to distinguish different base materials, such as carbon steel, alloy steel, stainless steel, and high-temperature alloys. The specific value depends on the instrument model, parameter settings, and surface pretreatment. Typical value ranges are as follows: carbon steel approximately [0.1, 0.3], alloy steel approximately [0.3, 0.8], stainless steel approximately [0.8, 2.5], iron-based high-temperature alloys approximately [0.2, 0.6], and nickel-based high-temperature alloys approximately [0.1, 0.3]. 5], cobalt-based superalloys are approximately [0.08, 0.25], while the ratio of key elements in the same batch of alloys is relatively stable. For example, the chromium-iron spectral line intensity ratio commonly used in steel. According to the national standard GB / T222, the chemical composition error of steel products of the same heat number is limited to the specified error range. For example, the difference in carbon content between steel products of the same heat number shall not exceed 0.02%, and the difference in Mn content shall not exceed 0.15%. That is, the chromium-iron spectral line intensity ratio of different parts of steel products of the same heat number only has an error fluctuation of ±0.03%. In complex industrial conditions, industrial LIBS analyzers can control the RSD below 5%. In this embodiment, the current mass-produced products are based on nickel-based superalloys, so the value range is usually [0.1, 0.35].

[0027] The thickness of the ring forging skin at the sorting station is obtained by a laser displacement sensor installed above the sorting station. The laser displacement sensor measures the distance from each sampling point on the upper surface of the skin to the sensor, and combines this with the known fixed height difference between the sensor and the reference surface of the sorting station, i.e. the conveyor chain surface, to calculate the average thickness of the skin. The unit is millimeters, which is used for joint analysis with the flatness data collected from each punching station. The value range is generally [10, 50] mm.

[0028] In this embodiment, a preset material threshold is used to compare with the current material anomaly degree of the connected skin to determine whether a material anomaly has occurred, i.e. whether low-value connected skins may be mixed into the conveyor chain. Its initial value is obtained by collecting at least 30 pieces of connected skins that have been manually confirmed as high-value during the system debugging phase and statistically analyzing their data. The material anomaly degree of each connected skin is calculated according to the material anomaly degree formula of this embodiment. 95% of the highest value among these values ​​is taken as the final preset material threshold, which is a dimensionless constant. The value range is usually set to [1.0, 2.5] based on different materials. In this embodiment, it is considered that the material anomaly degree of high-value connected skins in stable production is usually distributed between 0.8 and 1.2, while the low-value mixed connected skins are mostly higher than 1.5. Therefore, the value is set to 1.5.

[0029] The monitoring duration is used to define a continuous sliding time window. Within this window, the maximum deviation of the specific strength coefficient and flatness at each punching station is recorded. These maximum values ​​are used to calculate material and shape anomalies when the current sheet arrives at the sorting station, increasing the sensitivity to distinguish abnormal data. Based on the principle that if a station has ever recorded a value that triggers a suspected classification anomaly, it indicates that low-quality sheets may have been mixed into the conveyor belt, the large number of punching stations on the production line makes it impractical to set up a sorting line for each punching station. Typically, the sorting process... The number of punching stations is much smaller than the number of punching stations, and the punching gap varies depending on the feeding speed. When low-quality strips are mixed into the conveyor belt, it is usually difficult to trace the origin of the strip. At this time, regardless of whether the deviation corresponds to the current strip or other strips within the window, it may be a low-quality strip. Therefore, a continuously sliding window is needed to cover the anomaly observation time of multiple punching stations. Thus, the monitoring duration needs to be determined based on the transmission time from the farthest punching station to the sorting station after punching the strip, and should be greater than 1.2 times that time. The monitoring time is increased by 1.5 times to ensure that any abnormal skin defects generated at any workstation are still within the window when they reach the classification workstation. The monitoring time slides forward continuously until no suspected classification defects are detected within a consecutive monitoring time. At this point, the system can determine that the current production status has returned to stability, and the monitoring time stops sliding forward. During the monitoring time slide, the maximum value of all flatness deviation and the maximum value of all strength deviation are updated in real time based on the current coverage of the monitoring time. Each step forward in the window eliminates expired items. According to the new data, the reference extreme value always represents the extreme situation of the skin quality of each workstation in the most recent period. The monitoring time is in seconds, and the single monitoring time is usually taken as [60, 180] s. In this embodiment, the transmission time of the farthest workstation measured on site is set to 70s. Therefore, considering the full coverage of the 70-second transmission time of the farthest workstation and the additional reserve of about 70%, the value is taken as 120s. This ensures that the deviation data of any abnormal skin generated by any workstation is still within the sliding window when it reaches the classification workstation, thereby avoiding the loss of abnormal signals due to the time window being too short.

[0030] The detection cycle is the time interval for performing short-cycle adaptive statistics. That is, every detection cycle, the system counts the number of material anomalies, the number of shape anomalies, the standard deviation of the strength coefficient of each punching station, the average flatness deviation, and the average thickness deviation within that cycle. It calculates the station anomaly degree and stacking anomaly degree, and determines whether there are differences in multi-station mixed production or conveying stacking anomalies. This provides a basis for checking or inspecting equipment inspection, mold status verification, and slide conveying parameter adjustment. The detection cycle is usually determined based on the production line cycle time and the frequency of continuous skin production. Generally, a fixed time or a fixed number of pieces is used as the statistical window. Usually, a certain number of continuous skins are collected, or the default collection box is filled as a detection cycle. The unit is hours, and the value range is usually [1, 4]h. In this embodiment, based on the consideration of accumulating more than 200 continuous skin samples to ensure the universality of the statistics, it is taken as 2h.

[0031] The calibration quantity is used in long-cycle adaptive testing to determine how many test cycles of data to accumulate to calibrate the preset material threshold. The system collects the mean and standard deviation of the material anomaly in each test cycle within the calibration quantity, calculates the average of these mean values ​​and the average of the standard deviations, and then obtains the representative material value. This value is compared with the current preset material threshold to determine whether to adjust the threshold. The value of the calibration quantity must ensure the statistical representativeness of the above calculated values. It usually corresponds to the number of test cycles included in one day to one week of production. The calibration quantity is a dimensionless integer, and its value range is usually [10, 120]. This embodiment is based on historical maintenance data. Usually, maintenance is carried out once every 3 days for slow-changing issues such as mold wear and raw material batch fluctuations. There are 36 cycles in 3 days, which is sufficient to cover most samples and conforms to maintenance experience. Therefore, the value is 36.

[0032] By acquiring multi-source data through the acquisition module, and using strength deviation and flatness deviation to quickly identify skins with small deviations from the normal state, the system's computational burden is reduced. When suspected classification anomalies occur, the skins are recycled according to refined classification based on material value and morphological value. The maximum strength deviation enhances the sensitivity to the ratio of key elements related to material composition, making the detection of deviation states of skin materials more sensitive. This reduces the probability of low-quality skins being mixed with high-quality skins due to failure to be identified. Furthermore, the morphological value of skins is identified by recognizing their warping and deformation states. At the same time, the presence of multi-station mixed production or mold system deviations is also identified. The system promptly alerts staff and prompts for manual verification, enhancing its ability to detect equipment anomalies and process deviations. Alternatively, by identifying abnormal stacking during transport, it can promptly adjust and intervene to stagger the transport distance and arrival time at the sorting station of the connected sheets, reducing the probability of misjudgment caused by the stacking of connected sheets. Finally, by using the statistical distribution of material anomalies over multiple inspection cycles, it identifies the quality fluctuations of raw materials, infers the progressive wear of the mold, and promptly repairs or calibrates the equipment. This effectively solves the problem of overlooking mold health risks and reducing the overall quality rate of ring forgings due to the difficulty in inferring mold health from the connected sheet state.

[0033] Specifically, the first determination module includes: The data determination unit is used to determine the strength deviation by performing standard fraction calculations with the specific strength coefficient, the preset strength mean, and the preset strength standard deviation to identify the degree of deviation of the properties of the skin material; and to determine the flatness deviation by performing standard fraction calculations with the flatness, the preset flatness mean, and the preset strength flatness standard deviation to identify the degree of deviation of the shape properties of the skin. The suspected judgment unit is used to determine whether there is a suspected classification anomaly based on the strength deviation and flatness deviation of the ring forging skin.

[0034] The preset average strength value is used as a benchmark value for the specific strength coefficient to determine the strength deviation, so as to determine whether the specific strength coefficient of the current skin has deviated from the state of the stable production period. Its initial value is determined based on the arithmetic mean of the specific strength coefficients of 30 consecutive high-quality skins of the same material that have been manually confirmed during the system debugging phase. In subsequent operation, for example, if 30 skins that have been manually confirmed to be qualified are produced continuously, and there is no large-scale change in raw material batches during the period, whenever the system detects that no suspected classification abnormality has occurred within several consecutive detection periods, it is considered that the current period is stable production period. The specific strength coefficients of all skins in that period are included in the statistics for calibration and update. Moreover, the staff can manually reset it during any maintenance period. Its dimension is the same as that of the specific strength coefficient. It is usually based on the material variation and the value range is between [0.5, 3.0] kN / kg. In this embodiment, considering that the current large-scale production products use nickel-based high-temperature alloys as high-value base materials, the average specific strength coefficient of this material is taken as 1.2 kN / kg.

[0035] The preset flatness average value is used as the benchmark value of the flatness index to determine the flatness deviation, so as to determine whether the flatness of the current skin has deviated from the state of the stable production period. Its initial value, subsequent calibration update method and manual setting method are determined in the same way as the preset strength average value. Its dimension is the same as that of flatness, and it is usually taken as [0.2, 1.0] mm. In this embodiment, considering that the flatness of nickel-based high temperature alloy skin is usually distributed between 0.3 and 0.7 mm during stable production, the median value of its common distribution is taken as 0.5 mm.

[0036] The preset strength standard deviation is used to measure the fluctuation of the specific strength coefficient during a stable production period. It is used in conjunction with the preset strength mean to convert the absolute deviation of the specific strength coefficient into a dimensionless strength deviation. Its initial value, subsequent calibration update method, and manual setting method are similar to those of the preset strength mean. It is determined by calculating the standard deviation of the specific strength coefficient of all connected skins during the stable production period. Its dimension is the same as that of the specific strength coefficient, and its value range is usually [0.1, 0.6] kN / kg. In this embodiment, considering the stable production of nickel-based superalloys, the typical fluctuation range of the specific strength coefficient is about 20%. Based on its mean of 1.2 kN / kg, the preset strength standard deviation of 0.24 kN / kg is calculated.

[0037] The preset flatness standard deviation is used to measure the fluctuation of the flatness index during a stable production period. It is used in conjunction with the preset flatness mean to convert the absolute deviation of flatness into a dimensionless flatness deviation. Its initial value, subsequent calibration update method, and manual setting method are all the same as those of the preset strength standard deviation. Its dimension is the same as that of flatness, and its value range is usually [0.05, 0.3] mm. In this embodiment, we consider taking the fluctuation range of the preset flatness mean during the stable production of nickel-based high-temperature alloys as 20%, and the final value is 0.1 mm.

[0038] By selecting strength deviation and flatness deviation as the judgment criteria, based on the strength characteristics and geometric uniformity of the connecting material, the system identifies the changes in the type of connecting material and whether the shape is uniform. The two are independent and complementary, which weakens the absolute value fluctuations caused by the difference in mold specifications at each punching station. This allows connecting materials that are clearly within the error range to directly enter the high-value channel, saving computing power, and timely identifying suspicious connecting materials that require further in-depth judgment.

[0039] Please see Figure 2 As shown, this is a judgment diagram for determining whether a suspected classification anomaly has occurred in this embodiment. The suspected anomaly judgment unit includes: The deviation determination sub-unit is used to take the strength deviation as the first coordinate component and the flatness deviation as the second coordinate component, and substitute them into the Euclidean distance formula to obtain the Euclidean distance deviation, so as to quantify the comprehensive deviation of the skin from the stable benchmark in the two dimensions of specific strength coefficient and flatness. The suspected judgment subunit is used to determine whether a suspected classification anomaly has occurred based on the Euclidean distance deviation of the ring forging skin and a preset deviation threshold. Specifically, when the Euclidean distance deviation exceeds a preset deviation threshold, a suspected classification anomaly is identified.

[0040] The preset deviation threshold is used to compare with the calculated Euclidean distance deviation to determine whether the current skin has a suspected classification anomaly. During a stable production period, the strength deviation and flatness deviation of a batch of high-quality skins of the same material are collected. Since the strength deviation and flatness deviation approximately follow a standard normal distribution, the comprehensive deviation calculated from them follows a Rayleigh distribution. The value is taken between the 90th and 100th quantiles of the Rayleigh distribution. In actual production, there may be slight deviations due to factors such as equipment and materials. It can be fine-tuned according to the on-site debugging results. The preset deviation threshold is a dimensionless constant, and its value range is usually [2, 5]. In this embodiment, the 99th quantile of the Rayleigh distribution is taken, that is, under normal circumstances, only 1% of the skins will exceed this value, specifically 3.0.

[0041] The intensity deviation and flatness deviation are combined into a comprehensive anomaly score by using the Euclidean distance formula. The deviation contributions in the two directions are equivalent and do not cancel each other out, which can comprehensively and sensitively capture single-dimensional or two-dimensional anomalies. Finally, anomaly identification is performed by comparing with a preset threshold, which increases the sensitivity of capturing abnormal skin connections.

[0042] Specifically, the second determination module includes: The material matching and determination unit, based on the judgment result of suspected classification anomalies, determines the material anomaly degree by multiplying the maximum absolute value of all strength deviations within the monitoring period by the deviation degree of key elements. This allows the maximum absolute value of strength deviations across all workstations to cover potential material anomalies at all workstations, thereby amplifying the anomaly signal of the key element ratio. The deviation of key elements is determined based on the absolute value of the natural logarithm of the deviation multiple of the element ratio, so that the deviation multiple of one times higher than the preset material composition and half lower than the preset material composition have the same deviation, while compressing the range of large ratio values ​​and reducing the impact of extreme values ​​on the material anomaly. The element ratio deviation factor is determined based on the ratio of the key element ratio to the preset element ratio, in order to quantify the relative deviation between the current skin alloy composition and the preset material composition; The material anomaly determination unit is used to determine whether a material anomaly has occurred based on the material anomaly degree and a preset material threshold. Specifically, a material anomaly is determined to occur when the material anomaly degree exceeds a preset material threshold.

[0043] The preset element ratio refers to the standard critical element ratio corresponding to the current production material. For steel, it usually refers to the chromium-iron ratio. It is used to compare with the measured critical element ratio to calculate the critical element deviation. The greater the deviation of the measured critical element ratio from the preset element ratio, the higher the possibility of material abnormality. Its initial value is determined based on the arithmetic mean of the critical element ratios of 30 consecutive high-quality pieces of the same material with connected skin, which were manually confirmed during the system debugging phase. In subsequent production, when a batch of materials is used up and a large-scale replacement of new materials begins, the preset element ratio needs to be updated to the standard value of the new material to ensure the accuracy of system operation. In addition, during the stable production phase and when the original materials have not been replaced on a large scale, staff can also perform manual calibration based on the statistical data of the stable production period. The preset element ratio is a dimensionless constant, and the value range for different materials is usually the same as that of the critical element ratio. In this embodiment, the current large-scale production product is set to be based on nickel-based high-temperature alloy as the base material, so the value is 0.2.

[0044] By amplifying the abnormal signal of the key element ratio based on the strength deviation during special periods, when it is difficult to accurately trace the source of the skin, the signal of material abnormality is captured by measuring the deviation of the alloy composition. This ensures that the product will only increase significantly when both the strength and composition deviate significantly at the same time. Timely screening is then carried out when material abnormalities occur to increase the accuracy of identifying low-quality skin in the mixture.

[0045] Please see Figure 3 As shown, this is a judgment diagram of the classification station for classifying the ring forging skin in this embodiment. The classification and recycling module includes: The shape matching determination unit is used to determine the shape anomaly degree based on the judgment result that no material anomalies have occurred, according to the maximum value of all flatness deviations and skin thickness deviations within the monitoring period. Among them, the thickness deviation of the skin is determined by standard score calculation of the thickness of the skin and the preset average thickness and preset standard deviation of the thickness, so as to identify the degree of deviation of the thickness of the skin. The maximum value of all flatness deviations is the flatness deviation with the largest absolute value among all flatness deviations, so as to reflect the deviation of flatness. The shape anomaly detection unit is used to determine whether a shape anomaly has occurred based on the shape anomaly degree and a preset shape threshold. Specifically, when the shape anomaly degree is greater than a preset shape threshold, a shape anomaly is determined to have occurred. The shape sorting and recycling unit is used to recycle the corresponding ring forging skin at the sorting station to the shape collection station based on the judgment result of the occurrence of shape abnormality.

[0046] In this embodiment, when the shape anomaly judgment unit determines that the current piece of skin has a shape anomaly, the system sends an action command to the electrically controlled push rod at the entrance of the corresponding shape collection channel. This command is synchronized with the speed of the conveyor chain and the position of the skin: by detecting the arrival time of the skin by the photoelectric sensor set at the entrance of the sorting station, combined with the speed of the conveyor belt and the distance between the push rod and the sensor, when the skin runs to the front of the push rod, the push rod extends instantly, pushing the skin away from the main conveyor chain, so that it slides into the shape collection box along the shape collection channel. After the action is completed, the push rod automatically retracts back to its original position, waiting for the judgment result of the next piece of skin. If it is determined that no shape anomaly has occurred, the push rod does not move, and the skin continues to move along the main conveyor chain, enters the high-value furnace material collection box through the default collection channel. The entire push rod action process is linked with the conveying speed to ensure that each piece of skin is accurately diverted and does not affect the normal passage of subsequent skins.

[0047] The preset average thickness is used as a reference value for the thickness of the connecting skin. It is used to calculate the current thickness deviation of the connecting skin, thereby determining whether the thickness of the connecting skin has deviated relative to the stable production state. Its initial value is determined based on the arithmetic mean of the thickness data of 30 consecutive high-quality connecting skins of the same material, which were manually confirmed during the system debugging phase. Subsequently, during the stable production period, the system will automatically update this average value using the connecting skin data that has not triggered the abnormal judgment. Operators can also manually reset it during maintenance. Its unit is millimeters, and the value range is usually [10, 50] mm. In this embodiment, considering that the thickness of nickel-based high-temperature alloy connecting skin is usually distributed between 20 mm and 30 mm in stable production, the middle value of 25 mm is taken as the preset average thickness to make the thickness deviation more symmetrical.

[0048] The preset thickness standard deviation is used to measure the degree of fluctuation in the thickness of the skin during a stable production period. It is used in conjunction with the preset thickness mean to convert the absolute deviation of the thickness into a dimensionless thickness deviation. Its initial value is determined by calculating the standard deviation of the sample based on the skin thickness data of 30 pieces of high-quality skin of the same material in the same batch. The subsequent calibration update and manual setting methods are the same as the preset thickness mean. Its value range is usually [1, 5] mm. In this embodiment, considering that the standard deviation of the thickness of nickel-based high-temperature alloy skin is usually distributed between 1 mm and 3 mm in stable production, the average of the two, 2 mm, is taken as the preset thickness standard deviation to more reasonably reflect the normal fluctuation range.

[0049] A preset shape threshold is used to compare with the calculated shape anomaly degree to determine whether the current skin has a shape anomaly. The initial value of the preset shape threshold is determined by collecting data of a batch of skins with qualified shapes during a stable production period, calculating their shape anomaly degree and taking the 95th to 100th percentile. The preset shape threshold can be fine-tuned according to different molds and materials under actual production conditions. Its value range is usually [1.5, 3.0]. In this embodiment, considering that the shape anomaly degree of regular skins of nickel-based high temperature alloys is usually between 1.2 and 1.8, the maximum value of the above range is taken and after being amplified by a safety factor of 1.1 and rounded, 2.0 is taken as the preset shape threshold to reduce the probability of misjudging regular skins with normal fluctuations as abnormal.

[0050] By amplifying the maximum flatness deviation of all workstations within the monitoring period, potential geometric anomalies can be amplified. This allows for the capture of obvious deformation signals when it is difficult to accurately trace the source of the connected skin. The changes in flatness and thickness can be identified to distinguish between uniformly thick, regular connected skin and connected skin with warping deformation. Finally, the shape classification and recycling unit controls the push rod to send the connected skin into the shape processing channel, improving the identification efficiency of warped and deformed connected skin.

[0051] Specifically, the shape matching determination unit includes: The shape matching calculation subunit is used to determine the same-direction matching value based on the judgment result that no material abnormality has occurred, and the product of the maximum value of all flatness deviations and the thickness deviation of the skin during the monitoring period. This amplifies the abnormal signal of the skin thickness while reflecting whether the deviation directions of the two deviations are the same. The shape feature calculation subunit is used to determine the shape anomaly degree based on the square root of the Euclidean correction degree of the shape. The shape Euclidean correction is determined based on the difference between the shape Euclidean square distance and the same-direction matching value. It corrects the overall deviation based on the directional relationship between the two deviations. When the two have the same sign, a positive value is subtracted to compress the distance, thereby reducing the shape anomaly and lowering the probability of misjudging a uniformly thick and regular skin as an anomaly. When the two have opposite signs, a negative value is subtracted to expand the distance, thereby increasing the shape anomaly and making it more sensitive to the warping deformation of the skin. The shape Euclidean squared distance is determined by the sum of the squares of the maximum values ​​of all flatness deviations and the squares of the thickness deviations of the skin, in order to quantify the overall deviation of the skin in two dimensions: thickness uniformity and single-point thickness offset.

[0052] By determining whether the two deviations are in the same or opposite directions, the accuracy of identifying typical features of skin warping deformation is increased. Without increasing the cost of the sensor, the Euclidean distance is corrected into a direction-sensitive asymmetric metric using directional information, so that the shape anomaly can distinguish between uniform size offset and non-uniform warping deformation, improve the detection sensitivity of warping deformation and reduce the false judgment rate of regular skin, so as to reliably separate regular skin from deformed skin.

[0053] Specifically, the recycling module also includes: The material sorting and recycling unit is used to recover the corresponding ring forging skin from the sorting station to the material collection station based on the judgment result of the occurrence of material abnormality.

[0054] When the material anomaly judgment unit determines that the current piece of fabric has a material anomaly, the system sends an action command to the electrically controlled push rod at the entrance of the corresponding material collection channel. When the fabric piece moves to the front of the push rod, the push rod extends instantly, pushing the fabric piece away from the main conveyor chain and allowing it to slide along the material collection channel into the collection box corresponding to the low-value mixed material. After the action is completed, the push rod automatically retracts back to its original position, waiting for the judgment result of the next piece of fabric piece. The entire push rod action is synchronized with the press PLC and MES to ensure rapid response during batch mixing or material switching without affecting the production cycle. If no material anomaly is determined, the push rod does not move, and the fabric piece continues to move forward into the subsequent shape judgment or default channel.

[0055] When a material anomaly is detected, the push rod extends instantaneously as the connecting skin moves directly in front of it, pushing the connecting skin away from the main conveyor chain to the material collection channel. The push rod then resets and waits for the next judgment result. If no material anomaly is detected, the push rod does not move, and the connecting skin continues to move into the subsequent shape judgment or default channel. The entire process is synchronized with the press PLC and MES to ensure that the action rules are updated in real time when changing production or materials without affecting the production cycle. This allows for reliable screening of low-value mixed connecting skins, increases the purity guarantee of the high-value channel, and avoids interference from low-value mixed connecting skins in the subsequent shape judgment process, enabling the shape classification and recycling unit to focus on connecting skins with qualified materials.

[0056] Please see Figure 4 As shown, this is a judgment diagram for determining workstation abnormalities and stacking abnormalities in this embodiment. The execution module includes: The proportion determination unit is used to determine the classification anomaly proportion and the shape anomaly proportion based on the frequency of material anomalies, the frequency of shape anomalies, and the total number of times a material anomaly is determined to occur within the detection cycle. The workstation determination unit determines the workstation anomaly degree based on the product of the current intensity standard deviation and the classification anomaly ratio, and combines this with a preset workstation threshold to determine whether a workstation anomaly has occurred. The current strength standard deviation is determined based on the standard deviation of all specific strength coefficients within the testing period; When the anomaly level of a workstation exceeds the preset workstation threshold, a workstation anomaly is determined to have occurred. The stacking determination unit determines the stacking anomaly degree based on the product of the shape anomaly ratio and the absolute value of the flatness thickness difference. This quantifies the severity of warping deformation caused by conveying and stacking, and, in conjunction with a preset stacking threshold, determines whether a stacking anomaly has occurred. The flatness thickness difference value is determined based on the difference between the absolute value of the current flatness deviation and the absolute value of the current thickness offset. It is used to amplify the stacking features in the proportion of abnormal shapes that are dominated by flatness abnormalities but have small thickness offsets, thereby distinguishing between the situation of skin-on stacking and the situation where both increase simultaneously due to mold wear. Among them, the current flatness deviation is determined based on the average of all flatness deviations within the detection cycle, and the current thickness offset is determined based on the average of all skin thickness deviations within the detection cycle. When the stacking anomaly degree exceeds the preset stacking threshold, a stacking anomaly is determined to have occurred. The execution unit is used to perform corresponding quality calibration operations on each punching station when a station abnormality or stacking abnormality occurs.

[0057] The corresponding data collection operations include: When the execution unit determines that a workstation abnormality has occurred, it indicates that there is a significant difference in the specific strength coefficient between the punching workstations. This means that different punching workstations may be processing ring forgings of different materials or specifications, or it may indicate that the mold of a certain workstation is worn, or that there is a systematic deviation in the quality of the billet. In this case, the system first issues an alarm to the production management system or the field operation station, prompting the operator to verify whether there is a situation where different products are being produced in multiple workstations. If it is confirmed that it is a normal mixed production arranged in the production plan, no equipment adjustment is required. If it is confirmed that it is not a mixed production in the plan, it is necessary to further check the mold status, raw material batch and punching parameter settings of the corresponding punching workstation, and check whether there are problems such as mold wear, centering deviation or incoming material error, and take corresponding maintenance or adjustment measures. The investigation process can shorten the corresponding investigation time by reading and screening whether there are frequent abnormalities in the specific strength coefficient and flatness of each punching workstation in the current detection cycle.

[0058] When the execution unit detects a stacking anomaly, it may indicate a systematic deviation in the billet quality or that the sheets produced at each punching station are being squeezed and stacked upon reaching the conveyor chain. This leads to repeated checks showing warping and deformation of the sheets. The latter situation is usually caused by mismatched conveyor speeds at different stations or excessively concentrated arrival times of the sheets at the convergence point. To address this, the system automatically acquires the current rotation parameters of the discharge conveyor belt at the end of each punching station's chute. Based on the differences in punching gaps and conveyor intervals at each station, the system re-plans the distribution of the sheets produced at each punching station on the conveyor chain through simulation. Based on this distribution, the system simulates the preset rotation parameters of the discharge conveyor belts at two punching stations that are too close together, thus staggering the arrival times of the sheets at the convergence point and reducing the probability of stacking and collisions. Simultaneously, the system records the adjusted parameters for adaptive optimization in subsequent cycles. If stacking anomalies occur frequently, the system will issue a warning, suggesting that maintenance personnel check for jamming points on the chute surface or malfunctions in the reducer.

[0059] The preset workstation threshold is used to compare with the calculated workstation anomaly degree to determine whether there is an anomaly in multiple workstations. The initial value of the preset workstation threshold is determined by collecting workstation anomaly degree data from multiple consecutive stable production periods, such as five consecutive stable production cycles, and calculating its 95th percentile. Since the number of workstations and equipment status vary in different production lines, the value range is usually [0.05, 0.5] kN / kg based on the different production lines. In this embodiment, the 95th percentile of the workstation anomaly degree data in the nickel-based high-temperature alloy production line is taken as 0.15 kN / kg.

[0060] A preset stacking threshold is used to compare with the calculated stacking anomaly degree to determine whether a conveying stacking anomaly has occurred. The initial value of the preset stacking threshold is determined by collecting stacking anomaly degree data from multiple consecutive stable production periods, such as five consecutive stable production cycles, and taking the 95th percentile. The typical value range of the preset stacking threshold is a dimensionless constant of [0.02, 0.2]. In this embodiment, the upper limit of the stacking anomaly degree under normal conditions is considered to be about 0.02. When conveying stacking occurs, the proportion of shape anomalies increases to above 0.1, while the flatness deviation increases significantly and the thickness deviation remains basically unchanged. The difference can reach above 0.8. Therefore, the stacking anomaly degree is at least 0.08. Thus, setting the preset stacking threshold to 0.08 can effectively distinguish conveying stacking anomalies.

[0061] By triggering station anomaly alarms only when there are large intensity fluctuations between workstations and a high overall anomaly rate, the system can distinguish between systemic deviations such as mixed production and normal process fluctuations. This allows for timely adjustments to production plans or equipment maintenance, preventing economic losses caused by high-value materials being continuously misclassified as low-value mixed materials due to mixed production. It also reduces the probability of decreased ring forging quality due to undetected equipment failures. Furthermore, by identifying stacking, the system adjusts the slide reducer to stagger the arrival time of connecting strips when stacking anomalies occur, reducing the shape misjudgment rate caused by stacking and improving classification accuracy. This enables automatic detection and short-term process intervention for two common production anomalies: multi-station mixed production and conveyor congestion. It also increases the redundancy of perceiving the overall status of the production line through statistical features of connecting strip data.

[0062] Specifically, the proportion determination unit includes: The anomaly determination subunit is used to determine the anomaly ratio based on the ratio of the total number of anomaly classifications within the detection cycle to the total number of determinations of whether a material anomaly has occurred. Wherein, The total number of classification anomalies is determined based on the sum of the frequency of material anomalies and the frequency of shape anomalies within the detection period; The shape anomaly determination subunit is used to determine the shape anomaly ratio based on the ratio of the frequency of shape anomalies occurring within the detection cycle to the total number of times a material anomaly is determined.

[0063] By comprehensively reflecting the total occurrence rate of all valuable anomalies throughout the entire inspection cycle through the classification anomaly ratio, the probability of overlooking systemic deviations in the production line due to missed detection of a single anomaly type is reduced. At the same time, by calculating the shape anomaly ratio, the frequency of occurrence of geometric defects is specifically tracked, adding a purer morphological signal to the stacking anomaly detection and reducing the probability of interference from material anomaly fluctuations. The two ratios decouple and quantify the health status of the production line from different dimensions, enabling the system to distinguish between material-related problems and morphological problems at the macro level, extract actionable production diagnostic information from fragmented inspection data, and implement differentiated intervention measures in a timely manner.

[0064] Specifically, the calibration module includes: The relaxed determination unit is used to determine the representative value of the material based on the average value and quantile fluctuation boundary of all material anomalies when no material anomalies are found in each detection cycle within the calibration number of detection cycles. This quantifies the statistical boundary representing the vast majority of normal fluctuation ranges. The quantile fluctuation boundary is determined based on the product of the standard deviation of all material anomalies and the preset normal proportion. It is used to amplify the dispersion of material anomalies according to the specified quantile, thereby forming a statistical boundary that represents the vast majority of normal fluctuation ranges. The calibration judgment unit is used to determine whether to increase or decrease the preset material threshold based on the frequency of the quality calibration operation corresponding to the workstation abnormality in the quality calibration operation performed within the calibration number of detection cycles, the preset high-frequency threshold, the material representative value, and the preset stable interval. The preset stable interval is determined based on the preset material threshold and the preset error. Among them, when the frequency of performing collection operations exceeds the preset high-frequency threshold, an alarm is triggered and staff are prompted to optimize the production plan; otherwise, the preset material threshold is determined and adjusted by combining the material representative value and the preset stable range. When a judgment result is obtained that combines the material representative value and the preset stable range to determine and adjust the preset material threshold, if the material representative value is greater than the highest value of the preset stable range, the preset material threshold is increased; if the material representative value is less than the minimum value of the preset stable range, the preset material threshold is decreased. The calibration determination unit is used to increase the preset material threshold based on the determination result of increasing the preset material threshold and by adjusting the step size. Based on the determination result of reducing the preset material threshold, the new preset material threshold is reduced based on the adjustment step size, wherein, The adjustment step size is proportional to the ratio of the material representative value to the previous preset material threshold, and will not exceed the preset maximum step size.

[0065] In this embodiment, based on the consideration of balancing false positive rate and false negative rate, the final range of the preset material threshold will not be calibrated to outside [0.8, 2.5].

[0066] The preset normal proportion is used as a representative coefficient when calculating the representative value of the material. It is usually taken as a preset high quantile of the normal distribution or chi-square distribution, so that the material anomaly of most skins will be lower than the upper limit during the stable production period. Based on actual production experience, its value range generally corresponds to the [90%, 99] quantile of the normal distribution, that is, [1.28, 2.33]. In this embodiment, the 95th quantile is taken, which is the statistical control limit commonly used in industrial process control, and its corresponding calculated value is 1.645.

[0067] The preset high-frequency threshold is used to determine whether multiple workstation anomalies occur within a certain number of calibration detection cycles, i.e., whether the frequency of workstation anomalies is too high. In actual production, the value of the preset high-frequency threshold is related to the acceptable number of calibrations, and can usually be set to [2%, 10%] of the number of calibrations. In this embodiment, the number of calibrations is 36, so the preset high-frequency threshold is 2, which is about 5%. This can ensure that the state of frequent anomalies is sensitively identified, while avoiding false alarms due to a single occasional workstation anomaly if the threshold is set to 1.

[0068] The preset stable interval is used to determine whether the currently calculated material representative value is within the acceptable fluctuation range near the preset material threshold. It is determined based on the preset material threshold and the preset error. That is, the interval is from the preset material threshold minus the preset error to the preset material threshold plus the preset error. The initial value of the preset material threshold is determined by collecting the material anomaly data of a batch of qualified connected skins during a stable production period and taking the 95th or 99th percentile. Its value range is usually [0.8, 2.5]. In this embodiment, based on the typical distribution of material anomaly of nickel-based high-temperature alloys, the 95th percentile is taken as 1.5. The preset error is used together with the preset material threshold to define the preset stable interval. It is usually taken as [5%, 15%] of the preset material threshold. In this embodiment, considering that the fluctuation of the material representative value relative to the threshold during stable production usually does not exceed ±0.1, 10% is taken as 0.15 to leave a margin. Therefore, the preset stable interval is [1.35, 1.65].

[0069] The preset maximum step size is used to control the maximum amplitude when adjusting the preset material threshold during each calibration. It is a dimensionless constant and is the upper limit absolute value allowed for a single adjustment. It is set based on production experience and the allowable fluctuation range of the threshold. It is usually taken as [10%, 20%] of the initial value of the preset material threshold, i.e., [0.15, 0.3], to ensure that each adjustment does not deviate excessively from the original threshold. In this embodiment, considering the long-term calibration interval of 3 days during the historical stable generation stage, an additional margin of about 33% is considered to deal with extreme cases. The preset maximum step size is set to 0.2.

[0070] By evolving the material judgment threshold from a static fixed value to a dynamic adaptive value, and by accumulating the mean and standard deviation of material anomalies over multiple detection cycles, a statistical comparison is made between the upper limit of typical anomalies under the current production state and the current threshold. This is combined with monitoring the frequency of anomalies at workstations to eliminate interference from abnormal working conditions such as mixed production, and the material judgment threshold is calibrated. This approach can sensitively respond to the slow drift of production conditions, prevent over-adjustment through proportional control, prevent oscillation through step size limitation, and reduce adjustments triggered by meaningless small fluctuations by utilizing a stable range, thereby reducing the probability of misjudging high-value materials and missing low-value mixed materials.

[0071] 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 system for sorting and collecting the outer skin of a punching die for ring forgings, characterized in that, include: The acquisition module is used to acquire in real time the specific strength coefficient and flatness of the ring forging skin at each punching station, as well as the key element ratio and skin thickness of the ring forging skin at the classification station. The first determination module is used to determine whether there is a suspected classification anomaly based on the strength deviation and flatness deviation of the ring forging skin, wherein the strength deviation and flatness deviation of the ring forging skin are determined based on the specific strength coefficient and the flatness. The second determination module is used to determine whether a material anomaly has occurred based on the material anomaly degree and a preset material threshold. The material anomaly degree is determined based on the determination result of the suspected classification anomaly, the maximum value of the absolute values ​​of all the intensity deviations within the monitoring period, and the key element ratio. The classification and recycling module is used to classify and recycle the ring forging skin at the classification station based on the determination result of the occurrence of shape abnormality. The determination result of the occurrence of shape abnormality is determined based on the determination result of the absence of the material abnormality, the maximum value of all the flatness deviations within the monitoring period, and the thickness of the skin. The execution module is used to perform corresponding quality calibration operations on each punching station based on the frequency of the material abnormalities and shape abnormalities occurring within the detection cycle, as well as the specific strength coefficient, flatness deviation, and skin thickness within the detection cycle. A calibration module is used to calibrate the preset material threshold based on the frequency of performing the quality calibration operation within the specified number of testing cycles and the material anomaly within each testing cycle.

2. The skin-separated sorting and collection system for ring forging punching dies according to claim 1, characterized in that, The first determination module includes: The data determination unit is used to determine the strength deviation and the flatness deviation based on the specific strength coefficient, the flatness, the preset strength mean, the preset flatness mean, the preset strength standard deviation, and the preset flatness standard deviation; The suspected classification unit is used to determine whether the suspected classification anomaly occurs based on the strength deviation and the flatness deviation of the ring forging skin.

3. The skin-separated collection system for punching dies of ring forgings according to claim 2, characterized in that, The suspected determination unit includes: A deviation determination subunit is used to determine the Euclidean distance deviation based on the strength deviation and the flatness deviation of the ring forging skin; The suspected classification subunit is used to determine whether the suspected classification anomaly has occurred based on the Euclidean distance deviation of the ring forging skin and a preset deviation threshold.

4. The skin-separated sorting and collection system for ring forging punching dies according to claim 3, characterized in that, The second determination module includes: The material matching determination unit is used to determine the material anomaly degree based on the maximum value of the absolute values ​​of all the intensity deviations and the key element deviation degree when the judgment result of the suspected classification anomaly occurs, wherein the key element deviation degree is determined based on the key element ratio and the preset element ratio. The material anomaly determination unit is used to determine whether the material anomaly has occurred based on the material anomaly degree and the preset material threshold.

5. The skin-separated sorting and collection system for ring forging punching dies according to claim 4, characterized in that, The sorting and recycling module includes: The shape matching determination unit is used to determine the shape abnormality based on the determination result that no material abnormality has occurred, according to the maximum value of all the flatness deviations and the skin thickness deviation within the monitoring period, wherein the skin thickness deviation is determined based on the skin thickness and a preset thickness, and the maximum value of all flatness deviations is the flatness deviation with the largest absolute value among all flatness deviations. A shape anomaly determination unit is used to determine whether the shape anomaly has occurred based on the shape anomaly degree and a preset shape threshold. The shape sorting and recycling unit is used to recycle the corresponding ring forging skin at the sorting station to the shape collection station based on the determination result of the occurrence of the shape abnormality.

6. The skin-separated collection system for punching dies of ring forgings according to claim 5, characterized in that, The shape matching determination unit includes: The shape matching calculation subunit is used to determine the same-direction matching value based on the judgment result that no material abnormality has occurred, the maximum value of all the flatness deviations within the monitoring period and the skin thickness deviation. A shape feature calculation subunit is used to determine the shape anomaly degree based on the maximum value of all the flatness deviations, the skin thickness deviation, and the same-direction matching value.

7. The skin-separated collection system for punching dies of ring forgings according to claim 6, characterized in that, The sorting and recycling module also includes: The material sorting and recycling unit is used to recycle the corresponding ring forging skin at the sorting station to the material collection station based on the determination result of the occurrence of the material abnormality.

8. The skin-separated collection system for punching dies of ring forgings according to claim 7, characterized in that, The execution module includes: The proportion determination unit is used to determine the classification abnormality proportion and the shape abnormality proportion based on the frequency of the material abnormality occurring within the detection period, the frequency of the shape abnormality, and the total number of times the material abnormality has occurred. The workstation determination unit is used to determine the workstation abnormality degree based on the current strength standard deviation and the classification abnormality ratio, and to determine whether a workstation abnormality has occurred in combination with a preset workstation threshold. The current strength standard deviation is determined based on the standard deviation of all the specific strength coefficients within the detection cycle. The stacking determination unit is used to determine the stacking anomaly degree based on the shape anomaly ratio, the current flatness deviation, and the current thickness offset, and to determine whether a stacking anomaly has occurred in combination with a preset stacking threshold. The current flatness deviation is determined based on the average of all the flatness deviations within the detection period, and the current thickness offset is determined based on the average of all the skin thickness deviations within the detection period. The execution unit is used to perform corresponding quality calibration operations on each punching station when a station abnormality or stacking abnormality occurs.

9. The skin-separated sorting and collection system for ring forging punching dies according to claim 8, characterized in that, The ratio determination unit includes: The classification anomaly determination subunit is used to determine the classification anomaly ratio based on the frequency of the material anomaly occurring within the detection period, the frequency of the shape anomaly, and the total number of times the material anomaly has been determined. The shape anomaly determination subunit is used to determine the shape anomaly ratio based on the frequency of the shape anomaly occurring within the detection period and the total number of times the material anomaly has occurred.

10. The skin-separated sorting and collection system for ring forging punching dies according to claim 9, characterized in that, The calibration module includes: The relaxed determination unit is used to determine the representative value of the material based on the average value and standard deviation of all material anomalies when no material anomalies occur in each detection cycle within the calibration number of detection cycles and a preset normal proportion; A calibration judgment unit is used to determine whether to increase or decrease the preset material threshold based on the frequency of the quality calibration operation corresponding to the workstation abnormality in the quality calibration operation performed within the specified number of detection cycles, a preset high-frequency threshold, the material representative value, and a preset stable interval, wherein the preset stable interval is determined based on the preset material threshold and a preset error. The calibration determination unit is used to increase a new preset material threshold based on the determination result of increasing the preset material threshold, based on the material representative value and the previous preset material threshold, and to decrease a new preset material threshold based on the determination result of decreasing the preset material threshold, based on the material representative value and the previous preset material threshold.

Images