A casting quality abnormality identification method based on process parameter correlation analysis

Abstract

The application discloses a casting quality abnormality identification method based on process parameter correlation analysis, relates to the technical field of casting identification, and is used for solving the problem that potential quality abnormality cannot be identified in advance in the defect formation stage. The method comprises the following steps: detecting molten iron liquid level data of a pouring cup when a pouring machine is used to pour a current sand box at a fixed point, extracting liquid level change solidification characteristics, collecting time sequence of molten iron surface fluctuation in a thermal analysis cup, calculating fluctuation attenuation time, simultaneously calling current sand box mixing and feeding records, evaluating sand mold activity state and obtaining a sand mold activity index, performing correlation deviation analysis on the fluctuation attenuation time and the sand mold activity index, generating abnormal correlation factors, counting the abnormal correlation factors in a judgment window to identify the number of risk sand boxes, comprehensively judging a process state in combination with historical database molten iron cooling curve characteristics, generating an early warning signal and outputting a process parameter adjustment strategy when a warning condition is met, and early warning of potential quality abnormality.

Description

Technical Field

[0001] This invention relates to the field of casting identification technology, and more specifically, to a method for identifying casting quality anomalies based on process parameter correlation analysis. Background Technology

[0002] In the casting production process, sand casting usually involves multiple process steps such as sand mixing, molding, pouring and solidification. Among them, the solidification behavior and sand mold state during the pouring of molten iron directly affect the formation of the internal structure of the casting and the final quality state. In the existing casting production process, the casting process is usually adjusted empirically by controlling process conditions such as molten iron temperature, pouring speed, sand mold ratio and sand mixing parameters.

[0003] The existing technology has the following shortcomings: Currently, existing quality monitoring methods typically rely on visual inspection, non-destructive testing, and sampling analysis after casting is formed. These methods can only judge the quality results that have already been formed and are difficult to obtain the dynamic correlation between the solidification behavior of molten iron and the changes in the state of the sand mold during the pouring process. As a result, potential quality anomalies cannot be identified in advance during the defect formation stage, and the adjustment of the casting process depends on post-process feedback, which reduces the quality control capability in the continuous production process. Therefore, a casting quality anomaly identification method based on process parameter correlation analysis is proposed.

[0004] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] To overcome the aforementioned deficiencies in the prior art, embodiments of the present invention provide a casting quality anomaly identification method based on process parameter correlation analysis. This method addresses the problems mentioned in the background art by employing features extraction of molten iron surface changes and solidification state analysis, detection of fluctuation attenuation features in the thermal analysis cup, correlation evaluation of sand mixing and feeding records, construction of sand mold activity index, and process parameter anomaly correlation identification methods.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for identifying casting quality anomalies based on process parameter correlation analysis, comprising the following steps: Step S1: When the casting machine is casting at a fixed point in the current sand box, it detects the molten iron level data in the casting cup, performs feature analysis on the molten iron level data to generate solidification features, and determines whether to activate the same box query mechanism based on the solidification features. Step S2: In the same box query mechanism, collect the surface fluctuation time sequence of molten iron in the thermal analysis cup and calculate the fluctuation decay time, obtain the sand mixing and feeding record corresponding to the current sand box, and evaluate the sand mold activity index of the current sand box based on the sand mixing and feeding record; Step S3: Perform deviation product analysis on the fluctuation decay time and sand mold activity index to generate the abnormal correlation factor of the current sand box, set a judgment counting window, traverse and count the abnormal correlation factors in the judgment counting window, and identify the number of risky sand boxes. Step S4: Access the historical database to retrieve the molten iron cooling curve in the thermal analysis cup and mark the inflection point slope characteristics. Combine the number of risk sand boxes to determine whether to generate an early warning signal and output the process parameter adjustment strategy.

[0007] In a preferred embodiment, in step S1, when the casting machine performs fixed-point casting on the current sand box, it continuously monitors the height of the molten iron level in the casting cup to obtain molten iron level data. For the molten iron level height at two adjacent sampling times in the molten iron level data, calculate the rate of descent of the molten iron level, and calculate the rate of change of the descent rate based on the rate of descent of the molten iron level. Calculate the average rate of liquid level descent to obtain the average descent rate; The standard deviation of the liquid level drop rate is obtained by calculating the standard deviation of the drop rate. Calculate the mean of the absolute values ​​of the rates of change of the rate of descent to obtain the mean of the absolute values ​​of the rates of change; The solidification feature vector is formed by combining the mean descent rate, the standard deviation of the descent rate, and the mean absolute value of the rate of change.

[0008] In a preferred embodiment, in step S1, the solidification feature vector is input into the pre-trained logistic regression model, and the linear response value is calculated based on the solidification feature vector. The probability of solidification anomalies is calculated using a logic function based on the linear response value. Read the preset probability threshold. When the probability of solidification anomaly is greater than or equal to the probability threshold, determine that the current sand box has a tendency to solidify anomaly and start the same box query mechanism. When the probability of solidification anomaly is less than the probability threshold, the current sand box solidification status is determined to be normal, and the same box query mechanism is not activated.

[0009] In a preferred embodiment, in step S2, after the same box query mechanism is activated, the current sand box number is used as the index identifier to access the sand mixing section database and retrieve the molten iron surface fluctuation time sequence and sand mixing and feeding record corresponding to the current sand box. The liquid level detection device installed inside the thermal analysis cup collects the surface height of molten iron at a fixed sampling period to form the time sequence of molten iron surface fluctuations. The fluctuation offset is calculated based on the fluctuation time sequence of the molten iron surface, and the fluctuation envelope amplitude is calculated based on the fluctuation offset. The maximum value in the fluctuation envelope amplitude is read as the initial fluctuation amplitude, and the attenuation judgment ratio is multiplied by the initial fluctuation amplitude as the attenuation end threshold. When the amplitude of the fluctuation envelope is less than or equal to the decay end threshold, the corresponding sampling time is determined and recorded as the decay end time; The interval between the occurrence of the initial fluctuation amplitude and the end of the decay is taken as the fluctuation decay time.

[0010] In a preferred embodiment, in step S2, the sand mixing and feeding record corresponding to the current sand box is obtained. The sand mixer feeding record is the original process record generated when the current sand box is completed in the sand mixing process. The driving torque value of the bentonite feeding screw scale was extracted from the feeding record of the sand mixer, a torque data sequence was constructed, and the torque change between adjacent sampling times was calculated. Calculate the torque fluctuation value based on all torque changes; Extract the phase offset data output by the microwave humidity meter from the sand mixer feeding record, calculate the average value of the phase offset data at all sampling times, and obtain the microwave phase offset amount of the sand mixing batch corresponding to the current sand box. The sand mold activity index is calculated based on the microwave phase offset of the current sand box corresponding to the mixed sand batch and the torque fluctuation value of the bentonite feeding screw scale.

[0011] In a preferred embodiment, in step S3, the fluctuation decay time and sand mold activity index are read, with reference to the historical normal process batch fluctuation decay time benchmark value and sand mold activity index benchmark value, respectively. The fluctuation decay time deviation rate is calculated based on the fluctuation decay time and the fluctuation decay time benchmark value, and the sand mold activity deviation rate is calculated based on the sand mold activity index and the sand mold activity index benchmark value. Perform deviation product analysis on the fluctuation decay time deviation rate and sand mold activity deviation rate to calculate the abnormal correlation factor of the current sand box.

[0012] In a preferred embodiment, in step S3, a determination counting window is established, and the determination counting window continuously records the abnormal correlation factors of the most recent sand box according to the sand box pouring sequence; Read the threshold for judging abnormal correlation factors, traverse and count all abnormal correlation factors in the judgment counting window, and mark the corresponding sandbox as a risk sandbox when the abnormal correlation factor is greater than or equal to the threshold for judging abnormal correlation factors. After completing all iterations, count the number of risk sandboxes within the counting window.

[0013] In a preferred embodiment, in step S4, the historical database is accessed to retrieve the molten iron cooling curve data in the thermal analysis cup corresponding to the current risk sand box batch; A temperature acquisition device is installed inside the thermal analysis cup. The temperature acquisition device collects the temperature of the molten iron according to a fixed sampling period to form a cooling curve of the molten iron. The slope of the cooling curve is obtained by performing a first-order difference calculation on the molten iron cooling curve. The sampling point where the slope of the cooling curve changes from increasing to decreasing is marked as the inflection point; For each inflection point, the slope of the corresponding cooling curve is recorded as the inflection point slope feature. Read the benchmark value of the inflection point slope corresponding to the historical normal batch and calculate the inflection point slope deviation rate; Read the length of the judgment counting window, divide the number of risk sandboxes by the length of the judgment counting window, and calculate the proportion of risk sandboxes.

[0014] In a preferred embodiment, in step S4, an early warning judgment index is constructed based on the inflection point slope deviation rate and the proportion of risky sandboxes; Read the warning judgment threshold. When the warning judgment index is greater than or equal to the warning judgment threshold, generate a casting quality abnormality warning signal. When the warning judgment indicator is less than the warning judgment threshold, no warning signal is generated; After generating the early warning signal, the process parameter adjustment strategy is output based on the inflection point slope deviation rate and the proportion of risky sandboxes. The process parameter adjustment strategy is a feedforward process compensation scheme generated based on the early warning judgment indicators, which is used to correct the key process parameters of the smelting section and the sand mixing section.

[0015] The technical effects and advantages of this invention are as follows: This invention detects the molten iron level data in the casting cup during fixed-point casting of the current sand box using a casting machine. Solidification characteristics are extracted based on the changes in the molten iron level. The solidification characteristics determine whether to activate the same-box query mechanism. Once activated, the timing of surface fluctuations in the molten iron within the thermal analysis cup is collected, and the fluctuation decay time is calculated. Simultaneously, the sand mixing and feeding records corresponding to the current sand box are retrieved to evaluate the sand mold activity state and obtain the sand mold activity index. A correlation deviation analysis is performed between the fluctuation decay time and the sand mold activity index to generate abnormal correlation factors. These abnormal correlation factors are continuously counted within a judgment counting window to identify the number of risky sand boxes. Combined with the molten iron cooling curve characteristics in the historical database, the process status of the current sand box is comprehensively judged. When warning conditions are met, a warning signal is generated, and a corresponding process parameter adjustment strategy is output. This achieves the correlation identification between molten iron solidification behavior and sand mold activity state during casting, providing early warning of potential quality anomalies and a feedforward adjustment basis for smelting and sand mixing processes. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the implementation of a casting quality anomaly identification method based on process parameter correlation analysis according to the present invention.

[0017] Figure 2 This is a schematic diagram illustrating the steps of a casting quality anomaly identification method based on process parameter correlation analysis according to the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] This invention detects the molten iron level data in the casting cup during fixed-point casting of the current sand box using a casting machine. Solidification characteristics are extracted based on the changes in the molten iron level. The solidification characteristics determine whether to activate the same-box query mechanism. Once activated, the timing of surface fluctuations in the molten iron within the thermal analysis cup is collected, and the fluctuation decay time is calculated. Simultaneously, the mixing and feeding records corresponding to the current sand box are retrieved to evaluate the sand mold activity state and obtain the sand mold activity index. A correlation deviation analysis is performed between the fluctuation decay time and the sand mold activity index to generate abnormal correlation factors. These abnormal correlation factors are continuously counted within a judgment counting window to identify the number of risky sand boxes. Combined with the molten iron cooling curve characteristics in the historical database, the process status of the current sand box is comprehensively judged. When warning conditions are met, a warning signal is generated, and a corresponding process parameter adjustment strategy is output. This achieves the correlation identification between molten iron solidification behavior and sand mold activity state during the casting process.

[0020] Example 1, as Figures 1 to 2 As shown, a method for identifying casting quality anomalies based on process parameter correlation analysis includes the following steps: Step S1: When the casting machine is casting at a fixed point in the current sand box, it detects the molten iron level data in the casting cup, performs feature analysis on the molten iron level data to generate solidification features, and determines whether to activate the same box query mechanism based on the solidification features. Step S2: In the same box query mechanism, collect the surface fluctuation time sequence of molten iron in the thermal analysis cup and calculate the fluctuation decay time, obtain the sand mixing and feeding record corresponding to the current sand box, and evaluate the sand mold activity index of the current sand box based on the sand mixing and feeding record; Step S3: Perform deviation product analysis on the fluctuation decay time and sand mold activity index to generate the abnormal correlation factor of the current sand box, set a judgment counting window, traverse and count the abnormal correlation factors in the judgment counting window, and identify the number of risky sand boxes. Step S4: Access the historical database to retrieve the molten iron cooling curve in the thermal analysis cup and mark the inflection point slope characteristics. Combine the number of risk sand boxes to determine whether to generate an early warning signal and output the process parameter adjustment strategy.

[0021] The specific implementation is as follows: In step S1, when the casting machine performs fixed-point casting on the current sand box, it continuously monitors the molten iron level in the casting cup. A liquid level detection device is installed above the casting cup, which collects the molten iron level height according to a fixed sampling period to obtain molten iron level data. The molten iron level height reflects the real-time change in the amount of molten iron in the casting cup during the casting process; a higher value indicates a higher amount of molten iron in the casting cup, while a lower value indicates a higher degree of molten iron consumption.

[0022] It should be noted that the liquid level detection device is a non-contact liquid level measurement device installed above the casting cup, used to continuously collect data on the height of the molten iron level in the casting cup according to a fixed sampling cycle.

[0023] Feature analysis is performed on the molten iron level data. Specifically, the rate of descent of the molten iron level is calculated. For the molten iron level height at two adjacent sampling times, the rate of descent is calculated: ; in, Let be the rate of liquid level descent at the i-th sampling time. Let be the height of the molten iron surface at the (i-1)th sampling time. Let be the height of the molten iron surface at the i-th sampling time. For the i-th sampling time, For the (i-1)th sampling time, This is the index value at the sampling time.

[0024] The rate of molten iron level drop reflects the magnitude of the drop in molten iron level per unit time. A larger value indicates a faster pouring speed, while a smaller value indicates a slower pouring speed. After calculating the rate of molten iron level drop for all sampling intervals, a sequence of molten iron level drop rates is obtained.

[0025] Calculate the rate of change of descent velocity based on the liquid level drop velocity: ; in, Let be the rate of change of the descent velocity at the i-th sampling time. Let be the rate of liquid level descent at the i-th sampling time. Let be the rate of liquid level descent at the (i-1)th sampling time. For the i-th sampling time, For the (i-1)th sampling time, This is the index value at the sampling time.

[0026] The rate of change of descent speed reflects the degree of change in the rate at which the molten iron surface descends. A positive value indicates an increase in the rate of descent, while a negative value indicates a decrease. The larger the absolute value, the more drastic the change in the rate of descent.

[0027] To quantify the overall fluctuation of the molten iron level drop process, the average rate of molten iron level drop is calculated. The average rate of molten iron level drop reflects the average rate of molten iron level drop throughout the entire pouring stage. A larger value indicates a higher overall pouring speed, and a smaller value indicates a lower overall pouring speed.

[0028] Further calculation of the standard deviation of the liquid level descent rate yields the standard deviation of the descent rate. The standard deviation of the descent rate reflects the dispersion of the liquid level descent rate during the pouring process. The larger the value, the more obvious the fluctuation of the liquid level descent rate; the smaller the value, the more stable the liquid level descent rate.

[0029] Simultaneously, the mean of the absolute values ​​of the rate of change of the descent speed is calculated, yielding the mean absolute value of the rate of change. The mean absolute value of the rate of change reflects the average intensity of the change in the rate of descent of the liquid surface; the larger the value, the more drastic the change in the rate of descent, and the smaller the value, the more gradual the change in the rate of descent.

[0030] The solidification characteristic vector is formed by combining the mean descent rate, the standard deviation of the descent rate, and the mean absolute value of the rate of change. This solidification characteristic vector reflects the comprehensive dynamic characteristics of the molten iron level drop during the current sand box casting stage, and is used to characterize the flow and solidification tendency of the molten iron in the initial stage of casting. The mean descent rate reflects the overall flow velocity, the standard deviation of the descent rate reflects the degree of flow fluctuation, and the mean absolute value of the rate of change reflects the intensity of changes in the flow state.

[0031] The solidification feature vector is input into a pre-trained logistic regression model. The logistic regression model is trained based on liquid surface feature data from historical normal sandboxes and historical abnormal sandboxes. The model parameters include the intercept term and feature weight coefficients.

[0032] First, calculate the linear response value based on the solidification eigenvector: ; in, For linear response values, For the intercept term, These are the feature weight coefficients corresponding to the solidification feature vectors. The mean descent velocity in the solidification eigenvector. The standard deviation of the descent rate in the solidification eigenvector. This represents the mean absolute value of the rate of change in the solidification eigenvector.

[0033] The linear response value reflects the overall response of the current sandbox liquid level characteristics to historical anomaly patterns. The larger the value, the closer the current liquid level characteristics are to the characteristics of historical anomaly samples.

[0034] Furthermore, the probability of solidification anomalies is calculated using a logistic function based on the linear response value: ; in, This represents the probability of solidification anomalies. For linear response values, It is a natural constant.

[0035] The probability of solidification anomaly ranges from 0 to 1, reflecting the likelihood of solidification anomaly occurring in the current sand box. The higher the value, the higher the degree of matching between the liquid surface characteristics and historical abnormal samples; the lower the value, the higher the degree of matching between the liquid surface characteristics and historical normal samples.

[0036] A pre-set probability threshold is read, which is determined based on the analysis of the subject operating characteristic curves of historical samples.

[0037] When the probability of solidification anomaly is greater than or equal to the probability threshold, it is determined that the current sand box has a tendency to solidify anomaly, and the same box query mechanism is activated. When the probability of solidification anomaly is less than the probability threshold, the current sand box solidification status is determined to be normal, and the same box query mechanism is not activated.

[0038] In step S2, after the same box query mechanism is started, the current sand box number is used as the index identifier to access the sand mixing section database and retrieve the iron surface fluctuation time sequence and sand mixing and feeding record corresponding to the current sand box.

[0039] The liquid level detection device installed inside the thermal analysis cup collects the surface height of the molten iron at a fixed sampling period to form a time sequence of molten iron surface fluctuations. The fluctuation offset is then calculated based on the time sequence of molten iron surface fluctuations. ; in, This is the fluctuation offset. Let be the height of the molten iron surface at the i-th sampling time. This represents the average liquid level height during the initial sampling phase of the thermal analysis cup. For the i-th sampling time, This is the index value at the sampling time.

[0040] It should be noted that the sand mixing section database is a data management unit used to store the entire process record of the sand mixing process for each sand box; the thermal analysis cup is a standardized testing container used to collect the thermal characteristics and liquid level fluctuation characteristics of molten iron during solidification, and is connected to the casting cup through a sampling channel. When the casting machine starts to cast the current sand box at a fixed point, the molten iron in the casting cup flows along the main gating channel to the sand mold cavity under the action of gravity, and at the same time, a portion of the molten iron enters the thermal analysis cup through the preset sampling channel.

[0041] The fluctuation offset reflects the degree of deviation of the molten iron surface from its initial stable state. The larger the absolute value, the more violent the fluctuation of the liquid surface; the smaller the absolute value, the more stable the liquid surface.

[0042] Further calculate the amplitude of the fluctuation envelope based on the fluctuation offset: ; in, For the amplitude of the fluctuation envelope, This represents the fluctuation offset.

[0043] The maximum value in the fluctuation envelope amplitude is read as the initial fluctuation amplitude. The attenuation judgment ratio is multiplied by the initial fluctuation amplitude and used as the attenuation end threshold.

[0044] When the amplitude of the fluctuation envelope is less than or equal to the attenuation end threshold, the corresponding sampling time is determined and recorded as the attenuation end time. The interval between the initial fluctuation amplitude occurrence time and the attenuation end time is taken as the fluctuation attenuation time. The fluctuation attenuation time reflects the duration for the fluctuation on the molten iron surface to return to a stable state. The larger the value, the slower the energy dissipation rate of the molten iron fluctuation; the smaller the value, the faster the energy dissipation rate of the molten iron fluctuation.

[0045] After calculating the fluctuation decay time, the mixing and feeding record corresponding to the current sand box is obtained. The mixing machine feeding record is the original process record generated when the current sand box completes the mixing process. It includes the spiral scale operation data during the bentonite feeding process and the detection data of the microwave moisture meter at the mixing machine outlet. A one-to-one correspondence between the casting sand box and the mixing batch is established through the sand box number to ensure that the mixing parameters used in subsequent analyses are consistent with the source of the current casting sand box.

[0046] The driving torque value of the bentonite feeding screw scale was extracted from the sand mixer's feeding records to construct a torque data sequence. The driving torque value of the screw scale reflects the resistance change of bentonite during the conveying process, and its value is affected by both the degree of water absorption and swelling of bentonite and the uniformity of mixing.

[0047] It should be noted that the bentonite feed screw scale is a metering and conveying device used in sand mixers to quantitatively convey bentonite. The output torque of its drive motor can reflect the changes in mechanical load of bentonite during the conveying process in real time.

[0048] Calculate the torque change between adjacent sampling times based on torque data sequences: ; in, This represents the change in torque. Let J be the driving torque value at the j-th sampling time. The driving torque value at the (j-1)th sampling time. For the j-th sampling time of the torque data sequence, For the (j-1)th sampling time of the torque data sequence, This is the index value of the sampling time in the torque data sequence.

[0049] Calculate the torque fluctuation value based on the total torque change: ; in, This represents the torque fluctuation value. This represents the number of sampling times in the torque data sequence. This represents the change in torque. This is the index value of the sampling time in the torque data sequence.

[0050] Torque fluctuation value reflects the uniformity of bentonite's water absorption and expansion. The larger the value, the more uneven the water absorption and expansion process of bentonite; the smaller the value, the more uniform the water absorption and expansion process of bentonite.

[0051] Simultaneously, phase shift data output by the microwave moisture meter is extracted from the feeding record of the sand mixer. The phase shift data reflects the phase change of microwave propagation in molding sand, and its value is affected by the ratio of free water to bound water in the molding sand.

[0052] It should be noted that the microwave moisture meter is an online moisture detection device installed at the outlet of the sand mixer. It is used to emit microwave signals and detect the phase change of the microwaves after passing through the molding sand.

[0053] The mean value of the phase shift data at all sampling times is calculated to obtain the microwave phase shift of the sand batch corresponding to the current sand box. A larger microwave phase shift indicates a higher proportion of bound water, while a smaller value indicates a higher proportion of free water.

[0054] The sand mold activity index is calculated based on the microwave phase offset of the current sand box corresponding to the mixed sand batch and the torque fluctuation value of the bentonite feeding screw scale: ; in, The activity index of the sand mold. This is the microwave phase offset. This represents the torque fluctuation value.

[0055] The sand activity index reflects the comprehensive relationship between the bound water state in the current sand box molding sand and the uniformity of bentonite water absorption and expansion. The larger the value, the higher the bound water ratio and the more uniform the bentonite water absorption and expansion, and the better the sand activity state. The smaller the value, the lower the bound water ratio or the lower the uniformity of bentonite water absorption and expansion, and the worse the sand activity state.

[0056] In step S3, the fluctuation decay time and sand mold activity index are read. The deviation product analysis is performed on the solidification state of the molten iron in the current sand box and the sand mold activity state, respectively, with reference to the historical normal process batch fluctuation decay time benchmark value and sand mold activity index benchmark value.

[0057] The baseline values ​​for the fluctuation decay time and sand mold activity index of historical normal process batches are obtained from the statistics of corresponding batches of historical normal castings and are used to characterize the standard state under normal process conditions.

[0058] First, calculate the fluctuation decay time deviation rate based on the fluctuation decay time and the fluctuation decay time benchmark value; ; in, The fluctuation decay time deviation rate For fluctuation decay time, This serves as the baseline value for fluctuation decay time.

[0059] The fluctuation decay time deviation rate reflects the degree of deviation of the current surface fluctuation recovery speed of molten iron in the sand box from the normal process state. The larger the value, the slower the molten iron fluctuation decays and the weaker the solidification recovery ability; the smaller the value, the faster the molten iron fluctuation decays and the stronger the solidification recovery ability.

[0060] Further, the sand mold activity deviation rate is calculated based on the sand mold activity index and the benchmark value of the sand mold activity index: ; in, The sand mold activity deviation rate. This is the benchmark value for the sand mold activity index. It represents the activity index of the sand type.

[0061] The sand mold activity deviation rate reflects the degree of deviation of the current sand activity state from the normal process state. The larger the value, the more obvious the decrease in sand mold activity; the smaller the value, the closer the sand mold activity is to the normal process state.

[0062] To identify the coupling anomaly between the solidification state of molten iron and the active state of the sand mold, a deviation product analysis was performed on the fluctuation decay time deviation rate and the sand mold activity deviation rate to calculate the anomaly correlation factor of the current sand box: ; in, It is an abnormal correlation factor. The sand mold activity deviation rate. This represents the fluctuation decay time deviation rate.

[0063] The anomaly correlation factor reflects the degree of synchronous coupling between deviations in molten iron solidification state and sand mold activity deviations. When both the fluctuation decay time deviation rate and the sand mold activity deviation rate increase simultaneously, the anomaly correlation factor increases accordingly, indicating that the decrease in molten iron solidification recovery ability and the decrease in sand mold activity occur simultaneously, and their coupling anomaly degree is enhanced. When either deviation is small or the two change in opposite directions, the anomaly correlation factor decreases, indicating that the two types of process states do not form a significant correlation anomaly.

[0064] After calculating the anomaly correlation factor of the current sand box, a decision counting window is established. The decision counting window continuously records the anomaly correlation factors of the most recent sand box according to the sand box pouring order. When a new sand box enters, its anomaly correlation factor is written into the decision counting window, and the earliest entering anomaly correlation factor in the window is removed, so that the decision counting window always maintains a fixed length.

[0065] The threshold for judging abnormal correlation factors is read, which is obtained by statistical analysis of historical normal batches and historical abnormal batches. All abnormal correlation factors within the judgment counting window are traversed and statistically analyzed. When a sandbox satisfies that its abnormal correlation factor is greater than or equal to the threshold for judging abnormal correlation factors, the corresponding sandbox is marked as a risk sandbox.

[0066] After completing the entire traversal, the number of risk sand boxes within the counting window is counted. The number of risk sand boxes reflects the cumulative number of sand boxes that have experienced associated anomalies during continuous casting. The larger the value, the greater the range of deviation from the normal state of the current process. The smaller the value, the more likely the associated anomaly has occurred in a single sand box.

[0067] In step S4, the historical database is accessed to retrieve the molten iron cooling curve from the thermal analysis cup corresponding to the current risk sand box batch. A temperature acquisition device is installed inside the thermal analysis cup, which collects the molten iron temperature according to a fixed sampling period to form the molten iron cooling curve. The molten iron cooling curve reflects the solidification and cooling process of the molten iron in the thermal analysis cup, and its temperature change process can characterize the heat release characteristics of the graphite precipitation and eutectic solidification stages of the molten iron.

[0068] It should be noted that the historical database is a data storage unit used to store process data and quality results of historical normal batches and historical abnormal batches; the temperature acquisition device is a temperature measuring device installed inside the thermal analysis cup, used to continuously collect molten iron temperature data according to a fixed sampling period and form molten iron cooling curves.

[0069] The slope of the cooling curve is obtained by performing a first-order difference calculation on the molten iron cooling curve. The slope of the cooling curve reflects the rate at which the molten iron temperature decreases; the smaller the value, the faster the cooling rate, and the larger the value, the slower the cooling rate.

[0070] The sampling point where the slope of the cooling curve changes from increasing to decreasing is marked as an inflection point. For each inflection point, the corresponding slope of the cooling curve is recorded as the inflection point slope characteristic. The inflection point slope characteristic reflects the heat release capacity during the eutectic solidification stage; the larger the value, the more sufficient the heat release during solidification, and the smaller the value, the weaker the heat release during solidification.

[0071] Read the benchmark value of the inflection point slope corresponding to the historical normal batch and calculate the inflection point slope deviation rate: ; in, The inflection point slope deviation rate. The inflection point slope benchmark value, This represents the slope characteristic at the inflection point.

[0072] The inflection point slope deviation rate reflects the degree of deviation of the current solidification heat release capacity of molten iron from the normal process state. The larger the value, the more obvious the decrease in solidification heat release capacity; the smaller the value, the closer the solidification heat release capacity is to the normal process state.

[0073] After extracting the inflection point slope feature, the number of risk sand boxes and the length of the judgment counting window are read. The proportion of risk sand boxes is calculated by dividing the number of risk sand boxes by the length of the judgment counting window. The proportion of risk sand boxes reflects the distribution density of associated abnormal sand boxes during continuous casting. The larger the value, the more likely the associated abnormality is to spread continuously; the smaller the value, the more likely the associated abnormality is to occur only locally.

[0074] Further, early warning judgment indicators are constructed based on the inflection point slope deviation rate and the proportion of risky sandboxes: ; in, As an indicator for early warning judgment, The inflection point slope deviation rate. This represents the percentage of risky sandboxes.

[0075] The early warning judgment index comprehensively reflects the degree of continuous sand box correlation anomaly and the degree of decrease in the solidification heat release capacity of molten iron. The larger the value, the more likely the current batch has both continuous correlation anomaly and decreased solidification heat release capacity. The smaller the value, the smaller the scope of the anomaly and the degree of solidification deviation.

[0076] The system reads the warning judgment threshold, which is obtained based on statistical analysis of historical normal batches and historical defective batches. When the warning judgment index is greater than or equal to the warning judgment threshold, a casting quality abnormality warning signal is generated; when the warning judgment index is less than the warning judgment threshold, no warning signal is generated.

[0077] After generating the early warning signal, the process parameter adjustment strategy is output based on the inflection point slope deviation rate and the proportion of risky sand boxes.

[0078] The process parameter adjustment strategy is a feedforward process compensation scheme generated based on the early warning judgment indicators, which is used to correct the key process parameters of the smelting section and the sand mixing section.

[0079] When the inflection point slope deviation rate increases, the iron carbon equivalent setpoint is increased; when the proportion of risky sand boxes increases, the water addition correction in the sand mixing section is increased and the bentonite addition ratio is adjusted; when both indicators increase simultaneously, the iron carbon equivalent compensation parameters, sand mixing water addition compensation parameters, and bentonite addition correction parameters are output simultaneously. The process parameter adjustment strategy is synchronously associated with and stored with the risky sand box number and the corresponding sand mixing batch, used to locate the anomaly to the specific sand box batch and the corresponding sand mixing time period, and serves as the feedforward adjustment basis for the smelting section and the sand mixing section.

[0080] Finally, it should be noted that in this paper, relational terms such as first and second are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations.

[0081] Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0082] In this document, the singular forms “a,” “an,” and “the” may also include the plural forms unless the context clearly indicates otherwise. It should also be understood that terms such as “comprising / including” or “having” specify the presence of the stated features, integrals, steps, operations, components, parts, or combinations thereof, but do not preclude the possibility of the presence or addition of one or more other features, integrals, steps, operations, components, parts, or combinations thereof. Meanwhile, the term “and / or” as used in this specification includes any and all combinations of the associated listed items.

[0083] The various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The various embodiments can be combined as needed, and the same or similar parts can be referred to each other.

[0084] The above description of the disclosed embodiments will enable those skilled in the art to make or use various modifications to these embodiments. It will be readily apparent to those skilled in the art that the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for identifying casting quality anomalies based on process parameter correlation analysis, characterized in that: Includes the following steps: Step S1: When the casting machine is casting at a fixed point in the current sand box, it detects the molten iron level data in the casting cup, performs feature analysis on the molten iron level data to generate solidification features, and determines whether to activate the same box query mechanism based on the solidification features. Step S2: In the same box query mechanism, collect the surface fluctuation time sequence of molten iron in the thermal analysis cup and calculate the fluctuation decay time, obtain the sand mixing and feeding record corresponding to the current sand box, and evaluate the sand mold activity index of the current sand box based on the sand mixing and feeding record; Step S3: Perform deviation product analysis on the fluctuation decay time and sand mold activity index to generate the abnormal correlation factor of the current sand box, set a judgment counting window, traverse and count the abnormal correlation factors in the judgment counting window, and identify the number of risky sand boxes. Step S4: Access the historical database to retrieve the molten iron cooling curve in the thermal analysis cup and mark the inflection point slope characteristics. Combine the number of risk sand boxes to determine whether to generate an early warning signal and output the process parameter adjustment strategy.

2. The casting quality anomaly identification method based on process parameter correlation analysis according to claim 1, characterized in that: In step S1, when the casting machine is casting the current sand box at a fixed point, the height of the molten iron in the casting cup is continuously monitored to obtain the molten iron level data. For the molten iron level height at two adjacent sampling times in the molten iron level data, calculate the rate of descent of the molten iron level, and calculate the rate of change of the descent rate based on the rate of descent of the molten iron level. Calculate the average rate of liquid level descent to obtain the average descent rate; The standard deviation of the liquid level drop rate is obtained by calculating the standard deviation of the drop rate. Calculate the mean of the absolute values ​​of the rates of change of the rate of descent to obtain the mean of the absolute values ​​of the rates of change; The solidification feature vector is formed by combining the mean descent rate, the standard deviation of the descent rate, and the mean absolute value of the rate of change.

3. The casting quality anomaly identification method based on process parameter correlation analysis according to claim 2, characterized in that: In step S1, the solidification feature vector is input into the pre-trained logistic regression model, and the linear response value is calculated based on the solidification feature vector. The probability of solidification anomalies is calculated using a logic function based on the linear response value. Read the preset probability threshold. When the probability of solidification anomaly is greater than or equal to the probability threshold, determine that the current sand box has a tendency to solidify anomaly and start the same box query mechanism. When the probability of solidification anomaly is less than the probability threshold, the current sand box solidification status is determined to be normal, and the same box query mechanism is not activated.

4. The casting quality anomaly identification method based on process parameter correlation analysis according to claim 1, characterized in that: In step S2, after the same box query mechanism is started, the current sand box number is used as the index identifier to access the sand mixing section database and retrieve the molten iron surface fluctuation time sequence and sand mixing and feeding record corresponding to the current sand box. The liquid level detection device installed inside the thermal analysis cup collects the surface height of molten iron at a fixed sampling period to form the time sequence of molten iron surface fluctuations. The fluctuation offset is calculated based on the fluctuation time sequence of the molten iron surface, and the fluctuation envelope amplitude is calculated based on the fluctuation offset. The maximum value in the fluctuation envelope amplitude is read as the initial fluctuation amplitude, and the attenuation judgment ratio is multiplied by the initial fluctuation amplitude as the attenuation end threshold. When the amplitude of the fluctuation envelope is less than or equal to the decay end threshold, the corresponding sampling time is determined and recorded as the decay end time; The interval between the occurrence of the initial fluctuation amplitude and the end of the decay is taken as the fluctuation decay time.

5. The casting quality anomaly identification method based on process parameter correlation analysis according to claim 4, characterized in that: In step S2, the sand mixing and feeding record corresponding to the current sand box is obtained. The sand mixer feeding record is the original process record generated when the current sand box is completed in the sand mixing process. The driving torque value of the bentonite feeding screw scale was extracted from the feeding record of the sand mixer, a torque data sequence was constructed, and the torque change between adjacent sampling times was calculated. Calculate the torque fluctuation value based on all torque changes; Extract the phase offset data output by the microwave humidity meter from the sand mixer feeding record, calculate the average value of the phase offset data at all sampling times, and obtain the microwave phase offset amount of the sand mixing batch corresponding to the current sand box. The sand mold activity index is calculated based on the microwave phase offset of the current sand box corresponding to the mixed sand batch and the torque fluctuation value of the bentonite feeding screw scale.

6. The casting quality anomaly identification method based on process parameter correlation analysis according to claim 1, characterized in that: In step S3, the fluctuation decay time and sand mold activity index are read, with reference to the historical normal process batch fluctuation decay time benchmark value and sand mold activity index benchmark value, respectively. The fluctuation decay time deviation rate is calculated based on the fluctuation decay time and the fluctuation decay time benchmark value, and the sand mold activity deviation rate is calculated based on the sand mold activity index and the sand mold activity index benchmark value. Perform deviation product analysis on the fluctuation decay time deviation rate and sand mold activity deviation rate to calculate the abnormal correlation factor of the current sand box.

7. The casting quality anomaly identification method based on process parameter correlation analysis according to claim 6, characterized in that: In step S3, a judgment counting window is established, and the judgment counting window continuously records the abnormal correlation factors of the most recent sand box according to the sand box pouring order; Read the threshold for judging abnormal correlation factors, traverse and count all abnormal correlation factors in the judgment counting window, and mark the corresponding sandbox as a risk sandbox when the abnormal correlation factor is greater than or equal to the threshold for judging abnormal correlation factors. After completing all iterations, count the number of risk sandboxes within the counting window.

8. The casting quality anomaly identification method based on process parameter correlation analysis according to claim 1, characterized in that: In step S4, the historical database is accessed to retrieve the cooling curve data of molten iron in the thermal analysis cup corresponding to the current risk sand box batch; A temperature acquisition device is installed inside the thermal analysis cup. The temperature acquisition device collects the temperature of the molten iron according to a fixed sampling period to form a cooling curve of the molten iron. The slope of the cooling curve is obtained by performing a first-order difference calculation on the molten iron cooling curve. The sampling point where the slope of the cooling curve changes from increasing to decreasing is marked as the inflection point; For each inflection point, the slope of the corresponding cooling curve is recorded as the inflection point slope feature. Read the benchmark value of the inflection point slope corresponding to the historical normal batch and calculate the inflection point slope deviation rate; Read the length of the judgment counting window, divide the number of risk sandboxes by the length of the judgment counting window, and calculate the proportion of risk sandboxes.

9. The casting quality anomaly identification method based on process parameter correlation analysis according to claim 8, characterized in that: In step S4, an early warning judgment index is constructed based on the inflection point slope deviation rate and the proportion of risk sandboxes; Read the warning judgment threshold. When the warning judgment index is greater than or equal to the warning judgment threshold, generate a casting quality abnormality warning signal. When the warning judgment indicator is less than the warning judgment threshold, no warning signal is generated; After generating the early warning signal, the process parameter adjustment strategy is output based on the inflection point slope deviation rate and the proportion of risky sand boxes. The process parameter adjustment strategy is a feedforward process compensation scheme generated based on the early warning judgment indicators, which is used to correct the key process parameters of the smelting section and the sand mixing section.

Images