Aluminum alloy melt quality detection method, device and system

Abstract

The present application relates to the technical field of heat analysis, in particular to a kind of aluminum alloy melt quality detection method, device and system.The method is based on temperature time series data to calculate the instantaneous cooling rate of each monitoring point at each collection time;According to the instantaneous cooling rate of each monitoring point, the current process risk index is determined;According to the preset hydrogen sensitivity weight, the current hydrogen content, the current process risk index and the preset risk baseline value are fused, and the comprehensive risk score of each monitoring point is determined;Wherein, hydrogen sensitivity weight and risk baseline value are obtained by the following offline analysis process: according to the correlation between the historical process risk index of each monitoring point and corresponding historical hydrogen content, the hydrogen sensitivity weight of each monitoring point is determined, and the risk baseline value is determined based on the historical process risk index of each monitoring point, so as to accurately locate the potential pore risk area.

Description

Technical Field

[0001] This invention relates to the field of thermal analysis technology, specifically to a method, apparatus, and system for detecting the quality of aluminum alloy melt. Background Technology

[0002] Porosity defects in aluminum alloy melt cooling are a common quality problem in aluminum alloy casting. During the casting process, as the temperature of the aluminum alloy melt continuously decreases, the solubility of gases in the melt decreases, causing gases to be released in the form of bubbles, forming pores. Porosity defects reduce the strength and toughness of the casting, leading to cracks and failures during use, and affecting the mechanical properties and appearance of the casting. Therefore, detecting or controlling the quality of aluminum alloy melt is an essential part of the casting process.

[0003] Currently, existing technologies often use a single sensor to measure the overall gas content of the melt and set a uniform threshold for alarms. However, this method has significant shortcomings: First, the formation of porosity defects is not solely determined by the overall gas content of the melt, but is also closely related to the cooling uniformity and stability of different areas of the mold during the casting process. Local overheating or cooling turbulence can lead to porosity in specific areas. Moreover, different areas of the casting have different sensitivities to porosity; for example, hot spots are extremely sensitive to hydrogen content. This results in a lack of detailed analysis of temperature changes during the casting process and fails to consider the impact of hydrogen content on porosity defects in different areas of the casting. Consequently, it affects the accuracy of quality risk assessment for different spatial locations of the casting, leading to insufficient targeting of risk assessments. Summary of the Invention

[0004] To address the technical problem, this invention provides a method, apparatus, and system for detecting the quality of aluminum alloy melt, the specific technical solution of which is as follows:

[0005] This invention proposes a method for detecting the quality of aluminum alloy melt, the method comprising:

[0006] Collect temperature time-series data from multiple monitoring points within the mold cavity during the current casting process, and obtain the current hydrogen content of the melt;

[0007] The instantaneous cooling rate of each monitoring point at each acquisition time is calculated based on temperature time series data; the local temperature difference index is determined based on the difference in instantaneous cooling rate between each monitoring point and its preset neighboring monitoring points; the stability index is determined based on the fluctuation trend of the instantaneous cooling rate of each monitoring point; and the current process risk index is determined based on the local temperature difference index and the stability index.

[0008] Based on the preset hydrogen sensitivity weights, the current hydrogen content, current process risk indicators, and preset risk baseline values ​​are integrated to determine the comprehensive risk score for each monitoring point. The hydrogen sensitivity weights and risk baseline values ​​are obtained through the following offline analysis process: historical temperature time-series data and corresponding historical hydrogen content are collected from multiple monitoring points within the mold cavity during various historical casting processes; historical process risk indicators are determined based on the historical temperature time-series data; the hydrogen sensitivity weights for each monitoring point are determined based on the correlation between the historical process risk indicators and the corresponding historical hydrogen content, and the risk baseline values ​​are determined based on the historical process risk indicators for each monitoring point.

[0009] Process control decision information is generated based on the comprehensive risk score.

[0010] Furthermore, the temperature time-series data is continuously acquired by multiple temperature sensors arranged on the surface of the mold cavity at preset acquisition time intervals; for each monitoring point, the temperature time-series data constitutes a sequence of temperature values ​​arranged in chronological order with the acquisition time interval as the step size, wherein each temperature value corresponds to an acquisition time; the calculation of the instantaneous cooling rate of each monitoring point at each time based on the temperature time-series data includes:

[0011] For each monitoring point at each collection time, the temperature value at that time and the temperature value at the immediately preceding collection time are obtained;

[0012] Calculate the difference between the temperature value at the previous acquisition time and the temperature value at the current acquisition time, and use it as the temperature difference value;

[0013] Divide the temperature difference by the preset data acquisition time interval to obtain the instantaneous cooling rate of the monitoring point at the time of acquisition.

[0014] Furthermore, the process for determining the local temperature difference index includes:

[0015] For each monitoring point, at each data acquisition time, the absolute difference between the instantaneous cooling rate of the monitoring point at the acquisition time and the instantaneous cooling rate of each of its preset neighboring monitoring points at the same time is calculated as the rate difference.

[0016] Sum all the calculated rate differences and then divide by the number of neighboring monitoring points to obtain the instantaneous difference value;

[0017] The arithmetic mean of the instantaneous differences at all collection times of the monitoring point is normalized to obtain the local temperature difference index of the monitoring point.

[0018] Furthermore, the process of determining the stationarity index includes:

[0019] For each monitoring point, starting from the second acquisition time, the absolute difference between the instantaneous cooling rate of the monitoring point at each acquisition time and the instantaneous cooling rate of the immediately preceding acquisition time is calculated as the adjacent rate change.

[0020] Calculate the arithmetic mean of all adjacent rate changes at the monitoring point from the second acquisition time to the current acquisition time, and use it as the mean of the fluctuation intensity.

[0021] The value of an exponential function with the natural constant as the base and the negative of the mean fluctuation intensity as the exponent is calculated as an indicator of the stability of the monitoring point.

[0022] Furthermore, the determination of current process risk indicators based on local temperature difference and stability indicators includes:

[0023] For each monitoring point, an initial risk value is obtained based on the local temperature difference index and stability index;

[0024] The initial risk values ​​calculated for each monitoring point are normalized to obtain the current process risk indicators for each monitoring point.

[0025] Furthermore, the hydrogen sensitivity weight determination process includes:

[0026] For each monitoring point, multiple sets of historical data are constructed based on the historical process risk indicators and corresponding historical hydrogen content obtained from the monitoring points in multiple historical casting processes.

[0027] Based on multiple sets of historical data from monitoring points, a fitted straight line reflecting the trend of historical process risk indicators with historical hydrogen content changes was constructed through linear fitting; and the slope of the fitted straight line was statistically analyzed.

[0028] For each set of historical data, obtain the predicted historical process risk index of the fitted straight line under the corresponding historical hydrogen content; divide the historical process risk index value in the historical data set by the sum of the corresponding predicted historical process risk index and the preset minimum positive number to obtain the historical risk ratio; calculate the absolute difference between the positive integer 1 and the historical risk ratio as the relative deviation.

[0029] Calculate the arithmetic mean of the relative deviations of all historical data sets of the monitoring points as the average fitting deviation; input the average fitting deviation into a preset monotonically decreasing weight function to obtain a confidence coefficient with a value between zero and one, wherein the confidence coefficient decreases as the average fitting deviation increases.

[0030] The hydrogen sensitivity weight of the monitoring point is obtained by normalizing the product of the absolute value of the slope and the confidence coefficient.

[0031] Furthermore, each historical casting process is associated with a defect outcome label, which indicates whether the casting produced by the historical casting process ultimately has porosity defects; the determination of the risk baseline value based on the historical process risk indicators of each monitoring point includes:

[0032] For each monitoring point, select the target historical casting process with the defect outcome label indicating no defects;

[0033] Obtain the target historical process risk indicators determined by the monitoring points in each target historical casting process;

[0034] Calculate the arithmetic mean of all historical process risk indicators for the target, and use it as the risk baseline value for the monitoring points.

[0035] Furthermore, the comprehensive risk scoring determination process includes:

[0036] For each monitoring point, calculate the first product, which is equal to the product of the hydrogen sensitivity weight of the monitoring point, the current hydrogen content, and the current process risk index of the monitoring point.

[0037] Calculate the second product, which is equal to the difference between the positive integer 1 and the hydrogen sensitivity weight of the monitoring point, and then multiply it by the risk baseline value of the monitoring point.

[0038] Add the first product to the second product to obtain the comprehensive risk score of the monitoring point.

[0039] An aluminum alloy melt quality inspection system, the system comprising functional modules for implementing each step of an aluminum alloy melt quality inspection method, including:

[0040] The data acquisition module is used to collect temperature time-series data from multiple monitoring points within the mold cavity during the current casting process, and to obtain the current hydrogen content of the melt.

[0041] The determination module is used to calculate the instantaneous cooling rate of each monitoring point at each acquisition time based on temperature time series data; determine the local temperature difference index based on the difference in instantaneous cooling rate between each monitoring point and its preset neighboring monitoring points; determine the stability index based on the fluctuation trend of the instantaneous cooling rate of each monitoring point; and determine the current process risk index based on the local temperature difference index and the stability index.

[0042] The fusion module is used to fuse the current hydrogen content, current process risk indicators, and preset risk baseline values ​​according to preset hydrogen sensitivity weights to determine the comprehensive risk score for each monitoring point. The hydrogen sensitivity weights and risk baseline values ​​are obtained through the following offline analysis process: collecting historical temperature time-series data and corresponding historical hydrogen content from multiple monitoring points within the mold cavity during multiple historical casting processes; determining historical process risk indicators based on the historical temperature time-series data; determining the hydrogen sensitivity weights for each monitoring point based on the correlation between the historical process risk indicators and corresponding historical hydrogen content; and determining the risk baseline value based on the historical process risk indicators for each monitoring point.

[0043] The decision-making module is used to generate process control decision information based on the comprehensive risk score.

[0044] An aluminum alloy melt quality testing device includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of an aluminum alloy melt quality testing method.

[0045] The present invention has the following beneficial effects:

[0046] This invention analyzes the time-series temperature data of multiple monitoring points on the mold cavity surface to determine local temperature difference indices. These indices reflect the spatial uniformity of the cooling process at each monitoring point. It also identifies stability indices, reflecting the stability of the cooling process over time, thus yielding current process risk indices for each monitoring point. This allows for precise identification of specific locations on the casting (i.e., the locations of the monitoring points) experiencing abnormal cooling processes, accurately pinpointing potential porosity risk areas. Through offline analysis, based on multiple historical casting data, and combining the correlation between historical process risk indices and hydrogen content at each monitoring point, a hydrogen sensitivity weight is determined. This helps uncover the correlation between hydrogen content at different monitoring points and their location characteristics. Furthermore, historical process risk indices determined during historical casting processes replace traditional uniform threshold standards to establish a risk baseline. This makes risk assessment more closely aligned with specific casting scenarios and the location characteristics of the monitoring points, significantly improving adaptability to different casting structures and mold types. Attached Figure Description

[0047] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1 This is a flowchart of a method for detecting the quality of aluminum alloy melt provided in one embodiment of the present invention;

[0049] Figure 2 This is an example diagram illustrating the hydrogen sensitivity weight determination process provided in one embodiment of the present invention. Detailed Implementation

[0050] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of an aluminum alloy melt quality detection method, apparatus, and system proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0051] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0052] The following description, in conjunction with the accompanying drawings, details the specific scheme of the aluminum alloy melt quality detection method, apparatus, and system provided by the present invention.

[0053] Please see Figure 1 The diagram illustrates a flowchart of a method for detecting the quality of aluminum alloy melt according to an embodiment of the present invention. The method includes:

[0054] S101: Collect temperature time-series data from multiple monitoring points within the mold cavity during the current casting process, and obtain the current hydrogen content of the melt.

[0055] To ensure that monitoring points cover key areas of the mold and that most monitoring points have neighboring monitoring points, a series of temperature monitoring points can be pre-determined and arranged. The arrangement of monitoring points can follow the following principles: at least cover known hot spots, thick sections, final solidification areas, and high-risk areas for porosity defects such as the end of molten metal flow; the monitoring points can be arranged with non-uniform density, with denser points in high-risk areas.

[0056] It should be noted that the specific selection of high-risk areas for porosity defects can be determined based on industry common sense, and this embodiment does not impose specific limitations.

[0057] In this embodiment, the temperature time series data is continuously collected by multiple temperature sensors arranged on the surface of the mold cavity according to a preset collection time interval. For each monitoring point, the temperature time series data constitutes a temperature value sequence arranged in chronological order with the collection time interval as the step size, wherein each temperature value corresponds to a collection time.

[0058] It should be noted that the specific value of the preset acquisition time interval can be determined according to the actual situation, and this embodiment does not impose a specific limitation. For example, in aluminum alloy casting, especially in the process of die casting or gravity casting, the changes in molten metal filling and initial solidification are drastic, and the thermal process is dynamic and rapid. In order to effectively capture the thermal process, the preset acquisition time interval is usually set in the range of 0.05 seconds to 0.5 seconds, and is usually taken as 0.1 seconds (i.e., 10Hz). For example, for a casting cycle that lasts for 30 seconds, acquiring data at a 0.1-second interval will obtain about 300 consecutive temperature values.

[0059] It should be noted that the current hydrogen content is obtained by measuring the dissolved hydrogen content inside the aluminum alloy molten metal before it enters the mold cavity during the current casting process. The specific measurement method is a common technique used by those skilled in the art, and will not be described in detail in this embodiment. For example, the industry-standard "Reduced Pressure Test (RPT) method" can be used to measure the hydrogen content.

[0060] It should be noted that the current hydrogen content of the melt is uniquely linked to the current batch of casting process.

[0061] S102: Calculate the instantaneous cooling rate of each monitoring point at each acquisition time based on temperature time series data; determine the local temperature difference index based on the difference in instantaneous cooling rate between each monitoring point and its preset neighboring monitoring points; determine the stability index based on the fluctuation trend of the instantaneous cooling rate of each monitoring point; determine the current process risk index based on the local temperature difference index and the stability index.

[0062] In this embodiment, for each monitoring point at each acquisition time, the temperature value at the acquisition time and the temperature value at the immediately preceding acquisition time are obtained; the difference between the temperature value at the preceding acquisition time and the temperature value at the acquisition time is calculated as the temperature difference; the temperature difference is divided by the preset acquisition time interval to obtain the instantaneous cooling rate of the monitoring point at the acquisition time.

[0063] Instantaneous cooling rate quantifies the average rate at which the temperature in the area of ​​a monitoring point on the surface of the mold cavity decreases within a preset data acquisition time interval. If the instantaneous cooling rate of a monitoring point at a given acquisition time is greater than 0, it indicates that the temperature has decreased (cooled) during the preset data acquisition time interval, which is a normal phenomenon in the casting process. If the instantaneous cooling rate of a monitoring point at a given acquisition time is less than 0, it indicates that the temperature at that monitoring point has increased at that acquisition time, which may be due to new high-temperature molten metal flowing to that monitoring point, or a measurement anomaly (a rare phenomenon).

[0064] It should be noted that for the first data acquisition moment after the casting process begins, the instantaneous cooling rate cannot be directly calculated because there is no temperature value from the previous acquisition moment. As a preferred implementation method, the instantaneous cooling rate of all monitoring points at the first acquisition moment is initialized to zero or considered an invalid value, and the calculation officially begins from the second acquisition moment.

[0065] It should be noted that after calculating the instantaneous cooling rate of the monitoring point at the time of acquisition, since a negative instantaneous cooling rate usually corresponds to normal scenarios such as molten metal replenishment and new liquid flow replenishment, and non-cooling processes are abnormal, the instantaneous cooling rate can also be non-negatively processed (if the value is negative, it is corrected to 0) to avoid interfering with subsequent risk assessment.

[0066] It is important to understand that the formation of porosity, especially precipitated porosity, is closely related to the local accumulation of hydrogen in the melt. This local accumulation is often driven by uneven cooling rates between different areas of the casting. For example, slower-cooling areas can become "collection zones" for hydrogen. Therefore, the degree of difference in cooling rate between each monitoring point and its surrounding area can be quantified by determining a local temperature difference index. Specifically, a larger local temperature difference index at a monitoring point indicates a greater asynchrony in the thermal state between that monitoring point and its surrounding area, making it more likely that this asymmetry will lead to local hydrogen enrichment, thus forming porosity defects.

[0067] In this embodiment, for each monitoring point, at each acquisition time, the absolute difference between the instantaneous cooling rate of the monitoring point at the acquisition time and the instantaneous cooling rate of each of its preset neighboring monitoring points at the same time is calculated as the rate difference; all the calculated rate differences are summed and then divided by the number of neighboring monitoring points to obtain the instantaneous difference value; the arithmetic mean of the instantaneous difference values ​​of the monitoring point at all acquisition times is normalized to obtain the local temperature difference index of the monitoring point.

[0068] Preset neighbor monitoring points refer to a set of other monitoring points that are predefined in the system deployment for each monitoring point on the surface of the mold cavity.

[0069] It should be noted that the determination of the preset neighborhood monitoring points can be based on the three-dimensional spatial geometric distance between the monitoring points. For example, for each monitoring point, the N other monitoring points with the closest Euclidean distance are selected as the neighborhood monitoring points of the monitoring point, where N is usually between 3 and 8.

[0070] It should be noted that the arithmetic mean of the instantaneous differences of the monitoring points at all collection times can be normalized using the minimum-maximum normalization method. The specific process is well known to those skilled in the art and will not be described in detail in this embodiment.

[0071] The rate difference quantifies the degree of "asynchrony" in cooling rates between two spatially adjacent points on the mold cavity surface—a monitoring point and a preset neighboring monitoring point—at the same acquisition time. Specifically, the larger the rate difference between a monitoring point and its preset neighboring monitoring point at a given acquisition time, the more significant the divergence in cooling behavior between these two adjacent monitoring points at that acquisition time. For example, one monitoring point might cool rapidly due to the flow of molten metal, while an adjacent monitoring point might remain at a high temperature and cool slowly due to being in a dead zone. A larger rate difference indicates more localized uneven cooling at that acquisition time, and is a more concrete manifestation of spatial inhomogeneity risk.

[0072] It is important to understand that a stable and continuous cooling process is conducive to the orderly precipitation and discharge of gas. However, drastic fluctuations in the cooling rate (such as sudden cooling and heating) can disrupt the flow of the melt, intensify the turbulence of the molten metal, and may entrain gas. At the same time, it can disrupt the balanced distribution of hydrogen and easily lead to the formation of gas entrapment or irregular pores. Therefore, the degree of fluctuation of the instantaneous cooling rate at each monitoring point over time can be quantified by calculating the stability index. The smaller the stability index of a monitoring point, the more unstable and turbulent the cooling process at that point is, and the more likely it is to be entrained by gas or generate disturbing defects due to such temporal instability.

[0073] In this embodiment, for each monitoring point, starting from the second acquisition time, the absolute difference between the instantaneous cooling rate of the monitoring point at each acquisition time and the instantaneous cooling rate of the immediately preceding acquisition time is calculated as the adjacent rate change; the arithmetic mean of all adjacent rate changes of the monitoring point from the second acquisition time to the current acquisition time is calculated as the mean fluctuation intensity; and the exponential function value with the natural constant as the base and the negative of the mean fluctuation intensity as the exponent is calculated as the stability index of the monitoring point.

[0074] It is understandable that, since the "instantaneous cooling rate" is calculated from the temperature values ​​of two adjacent acquisition times, in order to calculate the change in the instantaneous cooling rate between adjacent acquisition times, it is necessary to ensure that the instantaneous cooling rate at adjacent acquisition times is a known valid value. However, the calculation of the instantaneous cooling rate at the first acquisition time depends on the instantaneous cooling rate at the adjacent zeroth acquisition time, and the instantaneous cooling rate at the zeroth acquisition time does not exist. Therefore, the starting point of the sequence of changes in adjacent rates naturally starts from the second acquisition time.

[0075] The adjacent rate change quantifies the degree of drastic change in cooling rate between two adjacent acquisition times at a given monitoring point, reflecting the magnitude of instantaneous acceleration in the cooling process. Specifically, a larger adjacent rate change at a given monitoring point between two adjacent acquisition times indicates a more drastic jump or abrupt change in cooling rate at that point during the time interval between those two acquisition times. This typically corresponds to an unstable physical process, such as violent disturbances or impacts in molten metal flow, or unstable advancement of the solidification front. Furthermore, this time interval is more prone to entrainment of air or cavity gas, disrupting the normal distribution of hydrogen in the melt and leading to the formation of irregular pores such as air entrapment holes and turbulent holes. Therefore, a larger adjacent rate change is a clear signal of instantaneous disturbances in the cooling process.

[0076] Since a larger average fluctuation intensity at a monitoring point indicates more severe and unstable overall fluctuations in the cooling process at that point, and a smaller average fluctuation intensity among adjacent monitoring points indicates a more stable and continuous cooling process, the closer the stability index value is to 1, the better the cooling stability at that monitoring point and the lower the risk of irregular porosity forming over time. Therefore, the stability index can be expressed by the following formula:

[0077]

[0078] in, The stationarity index of the k-th monitoring point is represented by T; T represents the total number of data collection times at the k-th monitoring point. This represents the change in the rate between adjacent data points at time t, starting from the second data collection time. This represents an exponential function with the natural constant as its base.

[0079] It is understandable that during the system deployment phase, after the temperature sensors are arranged on the surface of the mold cavity, each monitoring point in the network will be assigned a unique logical number, such as 1, 2, 3, ..., S (S is the total number of monitoring points). This number can be generated in a pre-set order (such as from left to right or from top to bottom) or randomly generated, as long as the uniqueness of the monitoring point number is guaranteed. Thus, the kth monitoring point refers to the monitoring point with the unique number k.

[0080] It is important to understand that the formation of porosity defects is the result of the combined effects of uneven spatial cooling and temporal instability. That is, uneven spatial cooling will lead to hydrogen accumulation, and temporal instability will easily lead to gas entrainment. The occurrence of either one or the combination of the two will significantly increase the risk of porosity defects. Therefore, the process risk index can be determined by coupling the local temperature difference index with the stability index. The process risk index comprehensively reflects the risk level of porosity defects caused by the cooling process performance of a certain monitoring point during a single casting process.

[0081] In this embodiment, for each monitoring point, the difference between the positive integer 1 and the stability index is calculated as the first difference value. The local temperature difference index is added to the first difference value as the initial risk value. The initial risk values ​​calculated for each monitoring point are normalized to obtain the current process risk index for each monitoring point.

[0082] The initial risk value reflects the degree of risk at a certain monitoring point in a single casting process caused by the coupled effect of "uniform spatial cooling" and "unstable temporal cooling".

[0083] It should be noted that a higher initial risk value at a monitoring point indicates a stronger coupling of risk characteristics, namely, spatially highly uneven cooling and / or temporally extremely unstable cooling process. For example, a larger local temperature difference index and a higher stability index indicate more significant spatial differences in the cooling process, but a relatively stable cooling process over time, with the main risk stemming from hydrogen accumulation. Conversely, a smaller local temperature difference index and a very small stability index indicate less spatial variation in the cooling process, but violent fluctuations over time, with the main risk stemming from gas entrainment or strong disturbances. Finally, a larger local temperature difference index and a very small stability index indicate the most dangerous situation, where the cooling process is severely out of sync with surrounding monitoring points, and the monitoring point's own cooling process is extremely unstable over time, making it more prone to local hydrogen enrichment while being entrained by other gases or experiencing violent disturbances, potentially leading to severe porosity defects.

[0084] It should be noted that the minimum-maximum normalization method can be used to normalize the initial risk values ​​calculated for each monitoring point. The specific method will not be described in this embodiment.

[0085] S103: Based on the preset hydrogen sensitivity weight, the current hydrogen content, current process risk indicators, and preset risk baseline values ​​are integrated to determine the comprehensive risk score for each monitoring point. The hydrogen sensitivity weight and risk baseline value are obtained through the following offline analysis process: historical temperature time series data and corresponding historical hydrogen content of multiple monitoring points in the mold cavity are collected during multiple historical casting processes; historical process risk indicators are determined based on the historical temperature time series data; hydrogen sensitivity weight of each monitoring point is determined based on the correlation between the historical process risk indicators and the corresponding historical hydrogen content of each monitoring point, and risk baseline values ​​are determined based on the historical process risk indicators of each monitoring point.

[0086] It should be noted that the method used to determine the historical process risk indicators of each monitoring point is the same as the method used to determine the current process risk indicators. For details, please refer to the relevant description of step S102. This embodiment will not repeat the description.

[0087] It should be noted that before each historical casting begins, the melt used is sampled and measured using the same method as in step S101 (such as the reduced pressure solidification method), and a unique hydrogen content value is obtained and recorded, which is the historical hydrogen content.

[0088] It should be noted that, in order to ensure that historical data and real-time data have the same time domain resolution and to avoid the calculated instantaneous cooling rate, stability index and other characteristics losing comparability on the time scale, the preset acquisition time interval for collecting historical temperature time series data from multiple monitoring points during the historical casting process is exactly the same as the preset acquisition time interval used for collecting real-time temperature time series data during the current casting process.

[0089] It is important to understand that, since different areas on the casting (such as the center of the hot spot and the far end of the gate) have fundamentally different sensitivities to the hydrogen content of the melt, the porosity risk in some areas is mainly driven by the current cleanliness of the melt, while in others it is mainly determined by inherent factors such as mold design. Therefore, by analyzing historical data generated from the historical casting process offline, the preset hydrogen sensitivity weight of each monitoring point can be determined, which can be used to quantify the strength of the dependence between the risk of porosity defects at each monitoring point and the hydrogen content of the melt.

[0090] The process of determining hydrogen sensitivity weights is as follows: Figure 2 As shown, it includes:

[0091] S103-1: For each monitoring point, based on the historical process risk indicators and corresponding historical hydrogen content obtained by the monitoring point in multiple historical casting processes, multiple sets of historical data are constructed.

[0092] It should be noted that, in order to avoid the situation of "the historical hydrogen content range being too narrow" (for example, all historical hydrogen contents are concentrated in 0.1-0.2 mL / 100g, and the fitting slope is not statistically significant), when collecting historical process risk indicators and corresponding historical hydrogen contents, it is necessary to ensure that the historical hydrogen contents cover the typical range of normal production, and the sample size of historical casting processes is not less than 30 groups; outlier detection is performed on the historical data set before fitting (the 3σ criterion can be used); if the sample size is less than 20 groups after outlier detection, or the determination coefficient of the fitted line is less than 0.3 (indicating extremely weak linear correlation), then the hydrogen sensitivity weight of that monitoring point is set to the default value of 0.5 (i.e., representing moderate sensitivity).

[0093] It can be understood that the historical data set consists of data pairs composed of historical hydrogen content and historical process risk indicators corresponding to a certain monitoring point in a certain historical casting.

[0094] S103-2: Based on multiple sets of historical data from monitoring points, a fitted straight line reflecting the trend of historical process risk indicators with historical hydrogen content changes is constructed through linear fitting; and the slope of the fitted straight line is statistically analyzed.

[0095] It should be noted that the method of constructing a straight line through linear fitting is a well-known technique in the art, and will not be described in detail in this embodiment. For example, using the historical hydrogen content of a certain monitoring point as the abscissa and the historical process risk index of a certain monitoring point as the ordinate, multiple sets of data pairs of the monitoring point are plotted on a scatter plot, and the linear regression method is used to fit these data pairs to obtain an optimally fitted straight line.

[0096] It should be noted that the method for calculating the slope is a common technique, and will not be described in detail in this embodiment.

[0097] S103-3: For each set of historical data, obtain the predicted historical process risk index of the fitted straight line under the corresponding historical hydrogen content; divide the historical process risk index value in the historical data set by the sum of the corresponding predicted historical process risk index and the preset minimum positive number to obtain the historical risk ratio; calculate the absolute difference between the positive integer 1 and the historical risk ratio as the relative deviation.

[0098] The predicted historical process risk index refers to the predicted value of the "historical process risk index" corresponding to the actual "historical hydrogen content" value obtained by substituting the actual "historical hydrogen content" value from a set of historical data of a certain monitoring point into the fitted straight line fitted to that monitoring point.

[0099] It should be noted that the specific value of the preset minimum positive number is determined based on industry experience, and this embodiment does not impose a specific limitation. For example, the preset minimum positive number is a very small constant used to prevent division by zero errors in mathematical calculations, and its typical value range is 1e-6.

[0100] The historical risk ratio quantifies the degree to which the "actual risk performance" of a historical casting process at a certain monitoring point deviates from the "risk level predicted based on the overall historical trend" at a certain historical hydrogen content level.

[0101] Specifically, if the historical risk ratio of a certain monitoring point in a certain historical casting process is larger (e.g., significantly greater than 1), it indicates that in that historical casting, given the corresponding historical hydrogen content, the actual risk performance of the monitoring point exceeds the expected value based on historical trends. That is, the predicted historical process risk index indicates that in addition to the historical hydrogen content, there are other stronger factors (such as abnormally severe cooling unevenness or disturbance in that historical casting) that lead to a higher cooling process risk. If the historical risk ratio of a certain monitoring point in a certain historical casting process is closer to 1, it indicates that in that historical casting, the actual risk performance of the monitoring point is consistent with the risk performance expected based on historical trends.

[0102] Similarly, if the relative deviation of a certain monitoring point in a certain historical casting process is smaller, it means that in that historical casting, given the corresponding historical hydrogen content, the actual risk performance of the monitoring point is closer to the risk performance expected based on historical trends.

[0103] S103-4: Calculate the arithmetic mean of the relative deviations of all historical data sets of the monitoring points as the average fitting deviation; input the average fitting deviation into a preset monotonically decreasing weight function to obtain a confidence coefficient with a value between zero and one, wherein the confidence coefficient decreases as the average fitting deviation increases.

[0104] Since a larger average fitting deviation at a monitoring point indicates a higher degree of overall dispersion of the historical data around the fitted line, and a less reliable linear relationship, the confidence coefficient will significantly decrease from 1 and approach 0. This means that the slope learned from the historical data of that monitoring point (i.e., the sensitivity of different monitoring points to the cooling process) has very low reliability, and the influence of the slope should be weakened in subsequent calculations. Therefore, the confidence coefficient can be represented by the following pre-defined monotonically decreasing weighting function:

[0105]

[0106] in, represents the reliability coefficient of the k-th monitoring point; J represents the number of historical data sets for the k-th monitoring point; This represents the relative deviation of the j-th historical data set at the k-th monitoring point; This represents the normalization function.

[0107] It should be clearly stated that, in order to ensure seamless integration between offline modeling and online application, the number, spatial location, and logical number of the monitoring points on the mold cavity surface used in the historical data collection phase (i.e., the definition of the kth monitoring point) are completely consistent with the number, spatial location, and logical number of the monitoring points used in the current real-time casting process.

[0108] It should be noted that the following is adopted: The specific process by which the function maps the average fitting deviation to the interval (0, 1) is a well-known technique to those skilled in the art, and will not be described in detail in this embodiment.

[0109] S103-5: Normalize the product of the absolute value of the slope and the confidence coefficient to obtain the hydrogen sensitivity weight of the monitoring point.

[0110] It is important to understand that if the absolute value of the slope of a certain monitoring point is larger and the confidence coefficient is larger, it means that the monitoring point is not only highly sensitive to hydrogen content, but also more consistent with the cooling process pattern of historical data, and therefore more reliable. The product of the two will produce a larger original score, and after normalization, its hydrogen sensitivity weight will be higher. This means that in the real-time early warning of melt quality, the risk of the monitoring point is more highly dependent on the current hydrogen content.

[0111] It should be noted that if the slope of the fitted line is less than or equal to 0, the hydrogen sensitivity weight of the corresponding monitoring point is directly reset to 0. Only for monitoring points with a slope greater than 0, the hydrogen sensitivity weight is obtained by normalization calculation.

[0112] It should be noted that the product of the absolute value of the slope and the confidence coefficient can be normalized using the minimum-maximum normalization method. The specific method will not be described in this embodiment.

[0113] It should be noted that each historical casting process is associated with a defect outcome label, which indicates whether the casting produced by the historical casting process ultimately has porosity defects.

[0114] For example, each recorded historical casting process must be associated with a defect outcome label. This label can be a binary or categorical metadata tag, indicating whether the casting produced after the historical casting process was ultimately determined to have porosity defects during quality inspection. For instance, a "no defects" label is typically marked "0" or "False." This label is associated with a historical casting process if no porosity defects are found after a specified non-destructive testing (such as X-ray or ultrasonic testing). A "defective" label is typically marked "1" or "True." This label is associated with a historical casting process if the casting produced after inspection is found to have one or more porosity defects.

[0115] In this embodiment, for each monitoring point, target historical casting processes with defect outcome labels indicating no defects are selected; the target historical process risk indicators determined by the monitoring point in each target historical casting process are obtained; and the arithmetic mean of all target historical process risk indicators is calculated as the risk baseline value of the monitoring point.

[0116] It should be noted that since the risk baseline value is intended to characterize the inherent risk level of a monitoring point under normal and stable qualified production conditions, and the calculation of the arithmetic mean is the best estimate of the central tendency of a set of homogeneous data, the calculation of the arithmetic mean of all target historical process risk indicators can most effectively filter out accidental fluctuations and random noise in a single historical casting, and can also extract the repeatable risk center value determined by inherent factors such as the location of the monitoring point and the mold design. Thus, the risk baseline value represents the risk performance of the monitoring point under normal conditions after excluding abnormal interference.

[0117] In this embodiment, for each monitoring point, a first product is calculated, where the first product is equal to the product of the hydrogen sensitivity weight of the monitoring point, the current hydrogen content, and the current process risk index of the monitoring point; a second product is calculated, where the second product is equal to the difference between the positive integer 1 and the hydrogen sensitivity weight of the monitoring point, and then multiplied by the risk baseline value of the monitoring point; the first product and the second product are added together to obtain the comprehensive risk score of the monitoring point.

[0118] It should be noted that before calculating the first product, the current hydrogen content is normalized to the interval [0, 1] (normalization formula: normalized hydrogen content = (current hydrogen content - minimum hydrogen content) ÷ (maximum hydrogen content - minimum hydrogen content), where the minimum and maximum hydrogen content are the statistical extreme values ​​of historical hydrogen content).

[0119] It should be noted that, since the formation of porosity defects in hydrogen-sensitive regions mainly follows a coupling mechanism of "the higher the hydrogen content, the higher the risk of porosity defect formation; at the same time, the more abnormal the cooling process, the more this risk is aggravated," the hydrogen sensitivity weight can be directly used as an adjustment coefficient. The current hydrogen content is multiplied by the current process risk index to obtain the first product. The first product represents the real-time risk component driven entirely by the current melt state and the current cooling process. The first product is significant only at points with high hydrogen sensitivity and is amplified by both high hydrogen content (i.e., the higher the current hydrogen content) and high-risk process (i.e., the higher the current process risk index).

[0120] It should be noted that the risk of porosity defects in non-hydrogen-sensitive areas (such as process dead zones) on the casting is mainly determined by inherent process factors such as mold design and venting system, and is not sensitive to changes in current hydrogen content and current process risk indicators. Therefore, 1-hydrogen sensitivity weight can be used as an adjustment coefficient for the risk of this part of the monitoring point. Multiplying it by the risk baseline value yields a second product, which represents the risk component determined entirely by the inherent process defect tendency of the monitoring point. The second product has a higher proportion at points with lower hydrogen sensitivity and is not affected by fluctuations in current hydrogen content and real-time cooling process.

[0121] S104: Generate process control decision information based on comprehensive risk score.

[0122] For example, the comprehensive risk score can be compared with a preset risk threshold. If the comprehensive risk score of a certain monitoring point exceeds the preset risk threshold, the monitoring point is determined to be at high risk. Subsequently, based on the hydrogen sensitivity weight of the monitoring point that triggered the alarm and the composition characteristics of the current process risk indicators, intelligent matching can be performed from a preset process control strategy library to generate clear instructions containing specific control measures (such as adjusting the pouring rate and cooling intensity), target location (i.e., the location of the monitoring point that triggered the alarm), and priority (such as "high risk" or "low risk"), which serve as the final output decision information.

[0123] It should be noted that the specific value of the preset risk threshold can be obtained based on the comprehensive risk score analysis calculated from historical data, and this embodiment does not impose specific limitations. For example, the average and standard deviation of the comprehensive risk scores of all monitoring points in the target's historical casting process can be calculated, and the warning threshold can be set near the average value + 2 × standard deviation to ensure that warnings and interventions are only triggered for abnormal risks that significantly deviate from the normal fluctuation range.

[0124] An aluminum alloy melt quality inspection system, the system comprising:

[0125] The data acquisition module is used to collect temperature time-series data from multiple monitoring points within the mold cavity during the current casting process, and to obtain the current hydrogen content of the melt.

[0126] The determination module is used to calculate the instantaneous cooling rate of each monitoring point at each acquisition time based on temperature time series data; determine the local temperature difference index based on the difference in instantaneous cooling rate between each monitoring point and its preset neighboring monitoring points; determine the stability index based on the fluctuation trend of the instantaneous cooling rate of each monitoring point; and determine the current process risk index based on the local temperature difference index and the stability index.

[0127] The fusion module is used to fuse the current hydrogen content, current process risk indicators, and preset risk baseline values ​​according to preset hydrogen sensitivity weights to determine the comprehensive risk score for each monitoring point. The hydrogen sensitivity weights and risk baseline values ​​are obtained through the following offline analysis process: collecting historical temperature time-series data and corresponding historical hydrogen content from multiple monitoring points within the mold cavity during multiple historical casting processes; determining historical process risk indicators based on the historical temperature time-series data; determining the hydrogen sensitivity weights for each monitoring point based on the correlation between the historical process risk indicators and corresponding historical hydrogen content; and determining the risk baseline value based on the historical process risk indicators for each monitoring point.

[0128] The decision-making module is used to generate process control decision information based on the comprehensive risk score.

[0129] An aluminum alloy melt quality inspection device includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of an aluminum alloy melt quality inspection method.

[0130] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0131] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A method for quality inspection of aluminum alloy melt, characterized in that, The method includes: Collect temperature time-series data from multiple monitoring points within the mold cavity during the current casting process, and obtain the current hydrogen content of the melt; The instantaneous cooling rate of each monitoring point at each acquisition time is calculated based on temperature time series data; the local temperature difference index is determined based on the difference in instantaneous cooling rate between each monitoring point and its preset neighboring monitoring points; and the stability index is determined based on the fluctuation trend of the instantaneous cooling rate of each monitoring point. For each monitoring point, the difference between the positive integer 1 and the stability index is calculated as the first difference value. The local temperature difference index is added to the first difference value as the initial risk value. The initial risk values ​​calculated for each monitoring point are normalized to obtain the current process risk index for each monitoring point. Based on a preset hydrogen sensitivity weight, the current hydrogen content, current process risk index, and preset risk baseline value are integrated to determine the comprehensive risk score for each monitoring point. The integration includes: for each monitoring point, calculating a first product, where the first product equals the product of the monitoring point's hydrogen sensitivity weight, current hydrogen content, and current process risk index; calculating a second product, where the second product equals the difference between the positive integer 1 and the monitoring point's hydrogen sensitivity weight, multiplied by the monitoring point's risk baseline value; and adding the first and second products to obtain the comprehensive risk score for the monitoring point. The hydrogen sensitivity weight and risk baseline value were obtained through the following offline analysis process: historical temperature time series data and corresponding historical hydrogen content of multiple monitoring points in the mold cavity during multiple historical casting processes were collected; historical process risk indicators were determined based on the historical temperature time series data; hydrogen sensitivity weight of each monitoring point was determined based on the correlation between the historical process risk indicators of each monitoring point and the corresponding historical hydrogen content, and risk baseline value was determined based on the historical process risk indicators of each monitoring point. Process control decision information is generated based on the comprehensive risk score.

2. The method of claim 1, wherein The temperature time-series data is continuously acquired by multiple temperature sensors arranged on the surface of the mold cavity at preset acquisition time intervals. For each monitoring point, the temperature time-series data constitutes a sequence of temperature values ​​arranged in chronological order with the acquisition time interval as the step size, wherein each temperature value corresponds to a acquisition time. The calculation of the instantaneous cooling rate of each monitoring point at each time based on the temperature time-series data includes: For each monitoring point at each collection time, the temperature value at that time and the temperature value at the immediately preceding collection time are obtained; Calculate the difference between the temperature value at the previous acquisition time and the temperature value at the current acquisition time, and use it as the temperature difference value; Divide the temperature difference by the preset data acquisition time interval to obtain the instantaneous cooling rate of the monitoring point at the time of acquisition.

3. The method of claim 2, wherein the aluminum alloy melt quality is determined by measuring the concentration of the at least one element in the aluminum alloy melt. The process for determining the local temperature difference index includes: For each monitoring point, at each data acquisition time, the absolute difference between the instantaneous cooling rate of the monitoring point at the acquisition time and the instantaneous cooling rate of each of its preset neighboring monitoring points at the same time is calculated as the rate difference. Sum all the calculated rate differences and then divide by the number of neighboring monitoring points to obtain the instantaneous difference value. The arithmetic mean of the instantaneous differences at all collection times of the monitoring point is normalized to obtain the local temperature difference index of the monitoring point.

4. The method of claim 2, wherein the aluminum alloy melt quality is determined by measuring the concentration of the at least one element in the aluminum alloy melt. The process of determining the stationarity index includes: For each monitoring point, starting from the second acquisition time, the absolute difference between the instantaneous cooling rate of the monitoring point at each acquisition time and the instantaneous cooling rate of the immediately preceding acquisition time is calculated as the adjacent rate change. Calculate the arithmetic mean of all adjacent rate changes at the monitoring point from the second acquisition time to the current acquisition time, and use it as the mean of the fluctuation intensity. The value of an exponential function with the natural constant as the base and the negative of the mean fluctuation intensity as the exponent is calculated as an indicator of the stability of the monitoring point.

5. The method of claim 1, wherein The process for determining the hydrogen sensitivity weight includes: For each monitoring point, multiple sets of historical data are constructed based on the historical process risk indicators and corresponding historical hydrogen content obtained from the monitoring points in multiple historical casting processes. Based on multiple sets of historical data from monitoring points, a fitted straight line reflecting the trend of historical process risk indicators with historical hydrogen content changes was constructed through linear fitting; and the slope of the fitted straight line was statistically analyzed. For each set of historical data, obtain the predicted historical process risk index of the fitted straight line under the corresponding historical hydrogen content; divide the historical process risk index value in the historical data set by the sum of the corresponding predicted historical process risk index and the preset minimum positive number to obtain the historical risk ratio; calculate the absolute difference between the positive integer 1 and the historical risk ratio as the relative deviation. Calculate the arithmetic mean of the relative deviations of all historical data sets of the monitoring points as the average fitting deviation; input the average fitting deviation into a preset monotonically decreasing weight function to obtain a confidence coefficient with a value between zero and one, wherein the confidence coefficient decreases as the average fitting deviation increases. The hydrogen sensitivity weight of the monitoring point is obtained by normalizing the product of the absolute value of the slope and the confidence coefficient.

6. The method of claim 5, wherein the step of detecting the quality of the aluminum alloy melt is performed by a method comprising: Each historical casting process is associated with a defect outcome label, which indicates whether the casting produced by the historical casting process ultimately has porosity defects. The determination of the risk baseline value based on historical process risk indicators of each monitoring point includes: For each monitoring point, select the target historical casting process with the defect outcome label indicating no defects; Obtain the target historical process risk indicators determined by the monitoring points in each target historical casting process; Calculate the arithmetic mean of all historical process risk indicators for the target, and use it as the risk baseline value for the monitoring points.

7. An aluminum alloy melt quality detection system characterized by, The system is used to implement the functional modules of each step in the aluminum alloy melt quality detection method as described in any one of claims 1-6, including: The data acquisition module is used to collect temperature time-series data from multiple monitoring points within the mold cavity during the current casting process, and to obtain the current hydrogen content of the melt. The determination module is used to calculate the instantaneous cooling rate of each monitoring point at each acquisition time based on temperature time series data; determine the local temperature difference index based on the difference in instantaneous cooling rate between each monitoring point and its preset neighboring monitoring points; determine the stability index based on the fluctuation trend of the instantaneous cooling rate of each monitoring point; and determine the current process risk index based on the local temperature difference index and the stability index. The fusion module is used to fuse the current hydrogen content, current process risk indicators, and preset risk baseline values ​​according to preset hydrogen sensitivity weights to determine the comprehensive risk score for each monitoring point. The hydrogen sensitivity weights and risk baseline values ​​are obtained through the following offline analysis process: collecting historical temperature time-series data and corresponding historical hydrogen content from multiple monitoring points within the mold cavity during multiple historical casting processes; determining historical process risk indicators based on the historical temperature time-series data; determining the hydrogen sensitivity weights for each monitoring point based on the correlation between the historical process risk indicators and corresponding historical hydrogen content; and determining the risk baseline value based on the historical process risk indicators for each monitoring point. The decision-making module is used to generate process control decision information based on the comprehensive risk score.

8. An apparatus for detecting the quality of an aluminum alloy melt, characterized by comprising: The apparatus includes a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that the processor executes the computer program to implement the steps of the method as described in any one of claims 1 to 6.

Images