A Smart Analysis Method and System for Titanium Alloy Melting Effect

Abstract

The present application relates to the technical field of titanium alloy smelting analysis, in particular to a titanium alloy smelting effect intelligent analysis method and system, comprising: recording smelting events of smelting equipment in a preset time period; recording abnormal events of each smelting time in a single parameter and parameter pair combination mode, forming a correlation parameter chain; performing mode clustering on the correlation parameter chain according to the combined single parameter and parameter pair form, and determining the propagation path after clustering; based on the propagation path, scoring each batch of abnormal events for a single event, and constructing a batch-index correlation matrix; according to the change trend of each key index in the batch-index correlation matrix in multiple smeltings, obtaining the transmission relationship of abnormal events in multiple smeltings; based on the transmission relationship of abnormal events in multiple smeltings, determining the target parameter of each batch, and outputting the target parameter; the accuracy of smelting analysis and the efficiency of dynamic perception are realized.

Description

Technical Field

[0001] This invention relates to the field of titanium alloy smelting analysis technology, specifically to an intelligent analysis method and system for titanium alloy smelting effects. Background Technology

[0002] Titanium alloys are metallic materials composed of titanium and other metallic elements, possessing high strength, low density, and corrosion resistance. They are widely used in aerospace, aviation, chemical, and other fields. Titanium alloy smelting technologies mainly include vacuum consumable electric arc furnace smelting, electron beam cold hearth furnace smelting, plasma cold hearth furnace smelting, and vacuum solidification furnace smelting. The smelting effect of titanium alloys is affected by the adjustment of smelting parameters. It is necessary to deduce the current titanium alloy smelting effect based on the specific process of each parameter during the smelting process to prevent low smelting efficiency and resource waste.

[0003] For example, Chinese Patent Publication No. CN117910886A discloses an intelligent analysis method and system for melting effects applied to titanium alloy melting. This method includes: acquiring element information and attribute characteristic information of batches of titanium alloy melting, obtaining sample groups of multiple melting batches by sampling corresponding products, detecting each product sample in the sample group to obtain detection result data, processing the detection result data of each sample to obtain a product sample melting effectiveness index, extracting element characteristic data of each batch based on melting element information and processing it to obtain interference coefficients, then correcting and aggregating the melting effectiveness indices of each batch of product samples to obtain melting quality and efficiency evaluation data, and evaluating the melting quality of all batches of titanium alloy by comparing the results with a threshold; thus, based on big data, performing effect detection and full batch melting quality effect evaluation on titanium alloy melting batch product samples to obtain melting quality evaluation results for titanium alloy melting products.

[0004] For example, Chinese Patent Publication No. CN120598431A discloses a method and system for quality management in titanium alloy smelting. This invention includes: collecting real-time monitoring data during the titanium alloy smelting process; classifying and storing the real-time monitoring data to obtain historical process data; issuing equipment anomaly warnings based on the equipment operation data and a preset first warning rule, and generating first warning information; obtaining a process capability index based on the process control node data; issuing process data deviation warnings based on the obtained process capability index and a preset second warning rule, and generating second warning information; issuing product manufacturing and testing warnings based on the ingot product manufacturing and testing data and a preset third warning rule, and generating third warning information; and displaying and processing the first, second, and third warning information through a human-computer interaction method.

[0005] Existing technologies analyze aspects such as porosity, grain size, and mechanical properties during smelting to illustrate the effect of titanium alloy smelting; and use operational node early warning to indicate abnormal situations during smelting. However, the smelting process involves complex parameters, and relying solely on single-point abnormality early warning makes it difficult to identify scenarios that trigger abnormalities. Furthermore, analyzing the physical and mechanical properties of titanium alloys after smelting can easily overlook real-time perception of the smelting process, dynamic tracking of abnormal propagation paths, and prediction and analysis of long-term trends. This leads to a significant decrease in the quality consistency and equipment utilization rate during titanium alloy smelting, resulting in reduced efficiency in identifying the effect of titanium alloy smelting. Summary of the Invention

[0006] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: a smart analysis method for titanium alloy melting effect, including: S1, acquiring melting process parameters, mapping the melting process parameters to the melting equipment, and recording melting events of the melting equipment within a preset time period.

[0007] S2 records the abnormal events during each smelting process using a combination of single parameters and parameter pairs, forming a chain of related parameters.

[0008] S3. Cluster the associated parameter chains according to the combined single parameter and parameter pair forms to determine the propagation path after clustering.

[0009] S4. Based on the propagation path, score the abnormal events in each batch. Calculate the transmission efficiency, single event score, and path characteristics of each batch based on the correlation parameter chain, and construct a batch-indicator correlation matrix. Record the changing trends of each key indicator in the batch-indicator correlation matrix during multiple smelting processes. By analyzing the correlation of the changing trends of each key indicator, obtain the transmission relationship of abnormal events in multiple smelting processes.

[0010] S5 traces the melting events based on the transmission relationship of abnormal events in multiple melting processes, determines the target parameters for each batch, and outputs the target parameters.

[0011] A smart analysis system for titanium alloy smelting effect includes: a smelting event module, used to acquire smelting process parameters, map the smelting process parameters to the smelting equipment, and record smelting events of the smelting equipment within a preset time period.

[0012] The smelting association module is used to record abnormal events during each smelting process in the form of single parameters and parameter pairs, forming an association parameter chain.

[0013] The propagation path module is used to perform pattern clustering on the associated parameter chains according to the combination of single parameters and parameter pairs, and to determine the propagation path after clustering.

[0014] The transmission analysis module is used to score each batch of abnormal events based on the propagation path, calculate the transmission efficiency, single event score and path characteristics of each batch based on the correlation parameter chain, and construct a batch-indicator correlation matrix. It records the changing trends of each key indicator in the batch-indicator correlation matrix during multiple smelting processes, and obtains the transmission relationship of abnormal events in multiple smelting processes by analyzing the correlation of the changing trends of each key indicator.

[0015] The parameter output module is used to trace melting events based on the transmission relationship of abnormal events in multiple melting processes, determine the target parameters for each batch, and output the target parameters.

[0016] The beneficial effects of this invention are as follows: First, by acquiring smelting process parameters and mapping them to equipment, this invention achieves standardized recording of smelting events, providing a data foundation for subsequent analysis. Then, by forming a chain of related parameters using single parameters and parameter pairs, it enhances the identification of the correlation of abnormal events. Next, through pattern clustering, the related parameter chain is transformed into a propagation path, realizing the abstraction and classification of abnormal patterns and reducing analytical complexity. By constructing a batch-index correlation matrix and analyzing trends, dynamic monitoring of abnormal propagation efficiency is achieved. Finally, by tracing target parameters based on the transmission relationship, a closed-loop data processing is realized, completing the overall identification and analysis of the smelting process and ensuring that abnormal events are handled promptly and accurately.

[0017] Second, this invention determines the melting process parameters identified in the current scenario by determining the melting boundary, risk exposure time, and risk mitigation time, and determines the currently acquired data based on the stable and unstable forms of the parameters, thereby enhancing the form of parameter dependence and improving the accuracy of abnormal event identification and processing.

[0018] Third, this invention abstracts abnormal events into abnormal nodes and time intervals, and constructs a multi-level associated parameter chain by checking the intersection of time intervals and recursively searching for chain nodes: parameter pairs are preferentially set as root nodes to realize the transmission and combination of abnormal events under multiple parameter combinations, thus avoiding the problem of delayed early warning in the smelting process. Attached Figure Description

[0019] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0020] Figure 1 This is a flowchart illustrating an intelligent analysis method for titanium alloy melting effects.

[0021] Figure 2 This is a flowchart illustrating step S1 of an intelligent analysis method for titanium alloy melting effects.

[0022] Figure 3 This is a flowchart illustrating step S2 of an intelligent analysis method for titanium alloy melting effects.

[0023] Figure 4 This is a flowchart illustrating step S4 of an intelligent analysis method for titanium alloy melting effects.

[0024] Figure 5 This is a system framework diagram of an intelligent analysis system for titanium alloy melting effects. Detailed Implementation

[0025] The embodiments of the present invention are described in detail below. The embodiments described below are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention. Where specific techniques or conditions are not specified in the embodiments, they shall be performed in accordance with the techniques or conditions described in the literature in the art or in accordance with the product manual.

[0026] See Figure 1 A smart analysis method for titanium alloy smelting effect includes: S1, acquiring smelting process parameters, mapping the smelting process parameters to the smelting equipment, and recording smelting events of the smelting equipment within a preset time period.

[0027] S2 records the abnormal events during each smelting process using a combination of single parameters and parameter pairs, forming a chain of related parameters.

[0028] S3. Cluster the associated parameter chains according to the combined single parameter and parameter pair forms to determine the propagation path after clustering.

[0029] S4. Based on the propagation path, score the abnormal events in each batch. Calculate the transmission efficiency, single event score, and path characteristics of each batch based on the correlation parameter chain, and construct a batch-indicator correlation matrix. Record the changing trends of each key indicator in the batch-indicator correlation matrix during multiple smelting processes. By analyzing the correlation of the changing trends of each key indicator, obtain the transmission relationship of abnormal events in multiple smelting processes.

[0030] S5 traces the melting events based on the transmission relationship of abnormal events in multiple melting processes, determines the target parameters for each batch, and outputs the target parameters.

[0031] In step S1, melting process parameters are acquired, including melting power, current, voltage, vacuum degree, leakage rate, cooling water flow / temperature, crucible rotation speed, and melting time at each stage. These data are combined with timestamps to form melting process parameters. The preset time period is the length of time required to complete the production of one batch of titanium alloy. Melting events refer to specific behaviors that occur during the entire melting process, are monitorable and identifiable, and have a direct or indirect impact on the stability of the production process or the quality of the final product. These include raw material melting events such as raw material ratio deviation events, raw material particle size inhomogeneity events, and raw material impurity exceeding standards; equipment-related melting events such as VAR electrode short circuit, unstable electron gun beam in the EB furnace, substandard melting chamber cleanliness, and vacuum system leakage; melting events with abnormal process parameters such as molten pool overheating, abnormal melting atmosphere compatibility, melting rate fluctuation, and abnormal amplitude and frequency of overheating and overpressure; and quality-related melting events such as ingot inclusions, porosity, abnormal grain size, and substandard performance.

[0032] In conventional titanium alloy smelting, it is usually necessary to smelt 2-3 times when using titanium alloy electric arc melting to improve the effect. At this time, it is also crucial to judge the difference in the effect of each batch. The purpose of obtaining the smelting process parameters is to identify the smelting abnormalities that exist in the process before and after titanium alloy production, and to emphasize the consistency and transmission process of smelting abnormalities, so as to explain the conditions under which smelting abnormalities occur and improve the efficiency of subsequent titanium alloy smelting production.

[0033] like Figure 2 As shown, step S1 is implemented as follows: S11, based on the mapped smelting equipment, the smelting boundary of the current smelting process parameters is determined, as well as the risk exposure time when the smelting process parameters on each smelting equipment reach the smelting boundary and the risk mitigation time when the smelting boundary decreases to the normal value. When calculating the smelting boundary, the arithmetic mean of each parameter (such as vacuum degree and current) is calculated based on 100 batches of normal data from the current equipment smelting; then the standard deviation is calculated based on the aforementioned 100 batches of data. If the currently identified titanium alloy type changes, normal data for the corresponding model is re-collected.

[0034] S12, calculate the average and standard deviation of the smelting process parameters during the risk exposure time and risk mitigation time, determine the stability of the smelting process parameters during each smelting, and extract the smelting event at the smelting boundary when the stability of the smelting process parameters meets the requirements.

[0035] S13, when the stability of the smelting process parameters does not meet the requirements, the smelting process parameters are divided into multiple stages, the smelting process parameters are correlated according to each stage, the correlation of the smelting process parameters under different stages is determined, and the smelting process parameter with the greatest correlation is output as the extracted smelting event.

[0036] The system, based on pre-loaded safety boundaries for various parameters of the current equipment and continuously monitoring the values ​​of these smelting process parameters, starts timing when a parameter begins to deviate from its normal value and moves towards the boundary. It calculates the time taken for the parameter to return from its normal range to the boundary value, which is recorded as the risk exposure time. At this time, the safety boundary represents the confidence interval of each parameter under normal equipment operation. The data within this confidence interval is taken as the normal value, that is, the current smelting boundary is set in the form of mean ± three standard deviations. Similarly, when the parameter returns from exceeding the smelting boundary to the normal range, it is recorded as the risk mitigation time. These two times represent the process of risk increase and decrease during current production, and the corresponding times are recorded.

[0037] At this point, the recorded time can be recorded using a continuous 7-point increase method. That is, when the current value is gradually increasing and approaching the upper limit of the confidence interval, the system starts timing at the 7th time point or the 6th time point to record the process and time when the current smelting process parameter exceeds the upper limit of the confidence interval, so as to obtain the risk exposure time of deviation from the increase. Similarly, the handling method is the same when the value approaches the lower limit and the upper limit of the confidence interval, using a continuous 7-point decrease method to obtain the risk exposure time of deviation from the decrease.

[0038] As for the risk mitigation time, it is timed when the smelting process parameters exceed the upper and lower limits of the confidence interval and show a reverse change, and the time it takes to decrease to within the confidence interval is recorded.

[0039] The system does not process only a single event that approaches the melting boundary. Instead, it calculates the average and standard deviation of these events over time, evaluating stability using the standard deviation and average. If the stability meets the requirements, events that significantly deviate from the current data are marked as anomalous events.

[0040] If the stability does not meet the requirements, these values ​​will be relatively scattered, making it difficult to predict and compare the process during smelting evaluation. At this point, it is necessary to find the root cause of the instability. The time when the parameters reach the boundary is related to different stages of smelting, such as the preheating stage, melting stage, refining stage, and casting stage. Focus on which time period is more unstable and find a time period with obvious problems.

[0041] Therefore, the implementation of step S12 also includes: S121, based on the smelting demand information of titanium alloy, analyzing the current smelting process parameters using a preset smelting stability threshold; and comparing the preset smelting stability threshold with the average value and standard deviation of the current smelting process parameters according to the titanium alloy smelting process flow.

[0042] S122, if any value of the average or standard deviation of the current smelting process parameters exceeds the upper or lower limit of the preset smelting stability threshold, the stability of the smelting process parameters is considered to be unsatisfactory.

[0043] S123. Only when the average value and standard deviation of the current smelting process parameters are both within the upper and lower limits of the preset smelting stability threshold are the stability of the smelting process parameters considered to meet the requirements.

[0044] It should be noted that the current preset smelting stability threshold is based on the average and standard deviation of smelting process parameters that show a continuous distribution in historical data. The average and standard deviation of multiple batches or multiple time periods are sorted from smallest to largest. The average and standard deviation at the 75th percentile of the data after sorting are used as the upper limit of the preset smelting stability threshold, and the smallest average and standard deviation at the time of sorting are used as the lower limit of the preset smelting stability threshold. The average and standard deviation of the smelting process parameters currently statistically analyzed according to the risk exposure time and risk mitigation time are compared with the preset smelting stability threshold. When the values ​​exceed the upper or lower limits, the data is considered unstable, indicating that the data fluctuates greatly in the corresponding time period and the average value is large. In this case, identifying smelting events solely through smelting boundaries has relative limitations. The focus should be on the time period in which the anomaly occurred to identify the abnormal situations in the smelting event.

[0045] As for the processing method in step S13, it focuses on processing the parameters of the melting process at different stages. The melting process is divided into multiple stages. By calculating the Pearson correlation coefficient of the parameters in each stage, the parameter pair with the highest correlation is found. The correlation results of all stages are compared to find the parameter pair with the highest global correlation. This parameter pair represents the key melting event in the current melting process, indicating that the abnormal event has been located from a vague time period to a specific melting stage. The specific stage time of the abnormality is directly output according to the correlation coefficient of the parameter pairs in each stage. For example, if the melting power and vacuum degree have a high correlation, such as a correlation coefficient greater than 0.8, it indicates a vacuum system failure, leading to gas inclusion. Based on the currently obtained correlation coefficient, the possible melting events in each stage are defined, and these parameter pairs are used as the output melting events. This can locate abnormal situations in multiple stages. At this time, the calculated data will first undergo normalization standardization processing to eliminate the dimensions between multiple sets of data to complete the calculation of the corresponding data. It should be noted that the calculated melting process parameters will be calculated in multiple stages according to different parameters in melting power, current, voltage, vacuum degree, leakage rate, cooling water flow / temperature, crucible rotation speed and melting time of each stage.

[0046] The implementation of step S13 also includes: pairing the smelting process parameters in each stage into parameter pairs, calculating the Pearson correlation coefficient of all parameter pairs, and obtaining the correlation of parameter pairs in each stage.

[0047] For each stage, the parameter pair with the highest correlation is identified and regarded as the correlation result for that stage.

[0048] Compare the correlation results across all stages to determine the globally most correlated parameter pair, and output it as a smelting event. This output smelting event will record the corresponding parameter pair and its timestamp to illustrate the most significantly correlated parameter pair under the current smelting process analysis. It represents the most severe process imbalance and illustrates the main anomaly under extremely discrete data scenarios, serving as the main part of subsequent processing.

[0049] In one embodiment of the present invention, in step S2, it is necessary to centrally associate the abnormal events existing during the smelting event processing to form a chain-like parameter chain. The abnormal events described here refer to the smelting events extracted in step S13 and the smelting events that exceed the smelting boundary in step S12. These events belong to the abnormal events existing in the current smelting process.

[0050] like Figure 3 As shown, the implementation of step S2 includes: S21, based on the abnormal events during each smelting, extracting the smelting process parameters associated with the abnormal events in the form of single parameters and parameter pairs, and treating each monitored single parameter and parameter pair as an abnormal node.

[0051] S22, for each abnormal node, the time period when the abnormal event occurred is statistically analyzed, and consecutively occurring abnormal nodes are merged into one abnormal time interval, which must contain at least one abnormal node. When merging abnormal time intervals, it is necessary to determine the time interval between the time periods covered by two abnormal events. For example, the currently monitored abnormal node is displayed in the form of a fluctuation curve, and the time value of the first trough when the fluctuation occurs is used as the basis for this setting. When identifying consecutively occurring abnormal events, the time value corresponding to this trough will represent the time interval between the current consecutive abnormal events. The interval for merging abnormal time intervals is set to only 1.2 or other multiples of the time value corresponding to this trough. As long as it is slightly larger than this trough, a relatively comprehensive abnormal time interval will be identified. When identifying the trough, the inflection point is selected such as the first 5 sampling points decreasing and the last 5 sampling points increasing, and the inflection point value is lower than the adjacent 10 sampling points. At this time, the corresponding trough is obtained. Starting from the first time the parameter exceeds the boundary, the time value corresponding to the first trough is T, and the merging interval is 1.2T. If the interval between two abnormal intervals is ≤1.2T, they are merged into one interval. As for cases where no merger is involved, we can examine the average abnormal time of the current smelting process parameters when abnormalities occur by looking at historical data, and set a time value of 5-10 times to illustrate isolated abnormal events within the current abnormal time interval.

[0052] S23, for each abnormal time interval, check if there is any intersection between the abnormal time intervals formed by all other abnormal nodes and the current abnormal time interval. That is, if there is an overlap in time intervals or a containment of time points, it indicates that there is a relative overlap between the two time intervals. If there is an intersection, the abnormal node in the current abnormal time interval is considered the root node, and the other abnormal nodes are considered secondary chain nodes of the abnormal node in the current abnormal time interval. It should be noted that the secondary chain nodes are set up gradually in chronological order. For example, if the vacuum degree has an abnormal time interval [t1, t2], and the current has an abnormal time interval [t1.5, t2.5], then the abnormal time interval of the current is contained within the abnormal time interval of the vacuum degree, which is a chain structure of vacuum degree → current, with current as its secondary chain node.

[0053] S24. If there is no intersection, the current abnormal time interval is composed into an output associated parameter chain in the form of timestamp progression.

[0054] S25. Count all second-level chain nodes, and find the third-level chain nodes under the second-level chain nodes according to the method that there is an intersection in the abnormal time intervals. The search is carried out recursively until there is no intersection in the abnormal time intervals. The chain nodes formed and the corresponding abnormal nodes are used as the output association parameter chain.

[0055] The process described above involves associating abnormal parameters during the smelting process according to their time periods, and then outputting the associated parameters. It should be noted that if multi-level chained associations exist, a maximum recursion depth needs to be set, such as setting it to 5 levels, to prevent the structure from becoming overly complex due to some parameters gradually looping.

[0056] Preferably, when comparing abnormal time intervals, if the corresponding abnormal node is a single-parameter abnormal node and a parameter pair abnormal node, it is also necessary to identify relationship abnormal precedence and single-parameter abnormal precedence.

[0057] The relationship anomaly precedes the individual anomaly of its constituent parameters. If the anomaly of a parameter pair occurs earlier than the anomaly of any of its constituent parameters, the system determines it as an early fault symptom. In this case, the anomaly of the combined parameter is taken as the root node, and the individual parameter is taken as its chain node.

[0058] The "single parameter anomaly first" means that if a single parameter anomaly occurs first, then the parameters affected by it are traced back to locate the initial fault point; that is, the single parameter anomaly occurs first, and then the common anomalies of the two parameters occur. Here, it is only necessary to explain it according to the original time sequence.

[0059] The implementation method of setting the secondary chain node in step S23 also includes: determining whether the abnormal time interval of the current parameter pair is earlier than the single abnormal time interval of its constituent parameters. If it is earlier, the abnormal time interval of the current parameter pair is taken as the root node, and its constituent parameters are regarded as secondary chain nodes.

[0060] If it is not earlier than that, the formation of secondary chain nodes will not be processed, and the output associated parameter chain will still be formed in the form of time sequence according to the original processing order.

[0061] The restriction operation in step S23 is for the comparison of parameter pairs and single parameters. In step S1, parameters with abnormal fluctuations that are too large or significantly exceed the normal abnormal fluctuations are combined and output with the parameter pairs with the greatest correlation. Here, it is necessary to verify whether the abnormal combination of parameter pairs causes the continuous abnormality of subsequent single parameters, and to explain the main parameters that cause the subsequent abnormalities. The associated parameter chain composed in chronological order is corrected in the form of a relative causal chain.

[0062] In one embodiment of the present invention, when performing pattern clustering, the generated association parameter chains ignore the timestamps and specific data, and only retain the structural topology and node types of the chains. The resulting chains are then clustered, and each cluster represents a recurring abnormal pattern, such as a cooling system failure mode, a vacuum leak mode, or other abnormal patterns.

[0063] At this point, a chain structure identified, such as the chain structure of cooling water flow rate → flow rate-temperature relationship → molten pool temperature, will be encoded. Using graph similarity or machine learning methods, it will be clustered with regular abnormal chain structures to identify recurring abnormal patterns in the currently output correlation parameter chain. The corresponding abnormal patterns will be output as propagation paths after clustering, illustrating the root cause, typical propagation path, and impact of each abnormal pattern. These specific clustered contents will serve as paths. Subsequently, single-event analysis and cumulative correlation analysis can be performed on these paths to explain the further correlations between abnormal events during smelting. The final output propagation path will interpret each node based on the originally formed correlation parameter chain, forming multiple sets of chain structures related to abnormal patterns such as cooling system failure modes and vacuum system leakage modes. These chain structures are considered as the output propagation paths.

[0064] The implementation of step S3 includes: setting a node type for each chain node in the associated parameter chain and recording the structural topology of the current associated parameter chain.

[0065] Anomaly pattern clustering is performed based on the node type and structural topology corresponding to the association parameter chain. Each chain structure in the association parameter chain is regarded as input data and converted into a vector in encoded form. The vector contains node type and structural topology information. Clustering uses cosine similarity to calculate the similarity between any two chain structures in the current association parameter chain. Hierarchical clustering is used. First, each chain structure is regarded as an initial cluster. Then, the two clusters with the highest similarity are merged to form a new cluster. The similarity between the new cluster and other clusters is calculated again. The merging is repeated until the preset number of clusters is reached or the similarity between clusters is lower than the threshold. The multiple chain structures that have completed pattern clustering are used as the output propagation path.

[0066] At this point, the corresponding chain structure is converted into vector form to calculate the similarity of node types and structural topology of different chains. Using this similarity, clusters are gradually merged. At this point, the number of clusters can be set to 5 to identify typical abnormal patterns output by the associated parameter chain in the current scenario. As for the part with similarity below the threshold, the threshold can be selected as 0.65, which is a commonly used threshold that balances the accuracy and completeness of clustering and can form relatively clustered abnormal patterns.

[0067] Currently, the abnormal pattern corresponding to the current propagation path is determined by treating the chain structure as input data and comparing the clustering results of anomaly patterns labeled in historical data with the current chain structure.

[0068] In one embodiment of the present invention, step S4 calculates a score for each propagation path formed in each batch, and for each propagation path generated in one melting process, quantifying the impact of the anomaly. It should be noted that when performing correlation calculations on current key indicators, the analysis is based on the changing trends of multiple propagation paths under the same anomaly pattern cluster across multiple batches.

[0069] like Figure 4 As shown, the implementation of step S4 includes: S41, performing statistics on the propagation path for each batch, setting a weight for each node in the propagation path based on the abnormal mode to which the propagation path belongs, and setting the single event score corresponding to the propagation path in a weighted summation manner according to the degree of abnormality of each node in the propagation path.

[0070] At this point, the total number of nodes in the propagation path is counted. The longer the path, the wider the scope of the fault propagation and the higher the severity. Based on prior knowledge or historical data, weights are assigned to different types of nodes. For example, based on historical fault statistics, the weight of each node in the propagation path is assigned according to the ratio of the frequency of the corresponding type of fault to the frequency of all faults. The value represented by each node in the propagation path is calculated by subtracting the upper limit of the corresponding parameter under normal control from its maximum value during the abnormal period and dividing by the upper limit to indicate the degree of abnormality of each node in the propagation path. Then, a weighted sum is used to indicate the single event score of the entire propagation path.

[0071] S42 integrates the propagation paths of multiple batches, taking the average length of the propagation path under the same abnormal mode, the maximum abnormality value of each propagation path during the current melting, and the abnormal mode label corresponding to the propagation path as path features; and using the ratio of the difference in abnormality between the root node and the end node of the propagation path to the time length corresponding to the propagation path to set the transmission efficiency, so as to complete the construction of the batch-index correlation matrix.

[0072] The scores of each single event from multiple melting processes are then integrated, feature values ​​are extracted for each propagation path, and the long-term trend of the propagation path during multiple integrations is analyzed.

[0073] It's important to note that propagation efficiency is calculated by subtracting the anomaly severity of the last node from the anomaly severity of the root node in the propagation path, and then dividing by the corresponding time length of the propagation path. Propagation efficiency measures the efficiency or ability of an anomaly to propagate from its root cause to its final effect. A higher value indicates a stronger ability of the system to amplify faults, which is unhealthy. A trend of increasing propagation efficiency is a significant risk signal that requires immediate attention and action. This value is normalized beforehand, and the calculated propagation efficiency is generally set relatively low to emphasize the deterioration of the overall system when severe anomalies exist.

[0074] It should be noted that the anomaly pattern label is not used for subsequent trend analysis, but is only used to describe the label of each parameter in the currently formed matrix; the average length of the propagation path reflects the time and related parameters involved in each anomaly, and can measure the horizontal risk impact caused by the anomaly event; the maximum anomaly degree value measures the situation of the largest anomaly in a single node, which is to analyze a single node to judge the overall relative change.

[0075] S43 uses transmission efficiency, path characteristics, and single event score as key indicator values ​​for the batch-indicator correlation matrix, and calculates the trend along the time series to obtain the slope of the trend line corresponding to each key indicator.

[0076] S44 calculates the Pearson correlation coefficient based on the slope of the trend line corresponding to multiple key indicators. The calculated value is regarded as the correlation of the changing trends of each key indicator. When the correlation value of the changing trends of each key indicator is consistent with the slope of the trend line corresponding to the key indicator, the corresponding abnormal event is used as the output transmission relationship.

[0077] The rows of the batch-index correlation matrix represent the batch ID of each melting process, and the columns represent the indices that need to be monitored, such as the single event score of the current melting process, the average length of the propagation path under the same abnormal mode during the current melting process, the maximum abnormality value of each propagation path during the current melting process, and the corresponding abnormality transmission efficiency, etc. These indices will be used as key indicators for subsequent analysis.

[0078] For the key indicator values ​​in each column of the batch-indicator correlation matrix, the changing trend is calculated along the time series, the trend line is calculated using linear regression, and the changing trend of the corresponding titanium alloy during the corresponding abnormal time is explained according to the slope of the corresponding indicator. Finally, the changing trend is subjected to correlation analysis. By calculating the Pearson correlation coefficient, the key indicators are analyzed one by one, and their correlation values ​​are obtained.

[0079] As for consistency, it refers to the consistency of the trend of the current key indicators, the consistency of the path of the propagation path, and the stability of the transmission efficiency. Data that conforms to these three aspects of analysis will be used as the transmission relationship of the output at this time.

[0080] Path consistency indicates whether the shape of the propagation path is stable when the same anomaly pattern occurs multiple times. Specifically, by comparing the graph edit distance of each node in the propagation path, if multiple propagation paths under the current anomaly pattern have small graph edit distances and low standard deviations, it indicates that the anomaly transmission in the current propagation path is relatively stable and belongs to the relatively stable anomaly events obtained from the current analysis. The criteria for judging small graph edit distances and low standard deviations are based on the average values ​​calculated in historical data under the current anomaly pattern. That is, when the graph edit distance and standard deviation are less than the average value in historical data, it is considered a high consistency in path consistency; if these values ​​are greater than the average value in historical data, it indicates a low consistency in path consistency. After labeling these specific values, the corresponding propagation relationships are output according to their conformity.

[0081] Regarding trend consistency, it indicates that the current correlation value is highly correlated, and the changing trends of the two key indicators are consistent, such as both being greater than 0 or both being less than 0. A high correlation value is indicated by a Pearson correlation coefficient greater than 0.7, requiring relatively few output data points with high correlation. Key indicators conforming to trend consistency represent a clear direction of change in the current abnormal event, allowing for a relatively accurate result based on trend changes. Otherwise, the trend changes will be relatively chaotic, indicating low consistency, and relevant data will be labeled. For example, a slope of transmission efficiency less than 0 and a slope of single event score less than 0 indicate a possible shortening of the duration of abnormal events and a decrease in the relative value of a single abnormal event. These data represent the relative relationship of abnormal events under multiple transmissions, illustrating the specific situation when an anomaly exists in the current smelting event.

[0082] Regarding the stability of transmission efficiency, it is required that the standard deviation and coefficient of variation of the transmission efficiency of the corresponding key indicators be small. The coefficient of variation is expressed as the ratio of the standard deviation to the mean. When the standard deviation and coefficient of variation are small, it indicates that the characteristics of each fault propagation are very consistent, which is a high-consistency anomaly. If these two values ​​are large, it indicates that the anomaly is difficult to predict and belongs to a part with extremely high risk, which is a low-consistency anomaly. These data are output according to the consistency condition as the transmission relationship at this time.

[0083] In the current processing, the input single parameters and parameter pairs are processed in the same way, according to the key indicators in the batch-indicator correlation matrix. Compared with single parameters, the abnormality value of parameter pairs is calculated by vector modulo to calculate the abnormality value of the two parameters in the parameter pair, so as to more accurately reflect the abnormality of the corresponding parameter pair under the current smelting event.

[0084] Step S5 obtains the parameter combination-event sequence-quality result data through the transmission relationship of abnormal events. This data will represent the parameter combination under relevant abnormalities during melting, which will facilitate subsequent monitoring of the occurrence of melting events and improve the processing effect for titanium alloy melting.

[0085] The implementation of step S5 includes: determining the combination form of single parameters and parameter pairs corresponding to the melting process parameters based on the transmission relationship of abnormal events in multiple melting processes, and taking the corresponding single parameters and parameter pairs as the target parameters for output according to the abnormal events pointed to by the melting process parameters.

[0086] At this point, a simple backtracking method is used to find the single parameter and parameter pair corresponding to the current abnormal event, and these data are used as target parameters for subsequent monitoring to complete the intelligent analysis and monitoring of the titanium alloy melting effect in multiple batches, so as to improve the efficiency of subsequent processing of titanium alloys.

[0087] like Figure 5 As shown, the present invention also provides an intelligent analysis system for titanium alloy smelting effect, including: a smelting event module, a smelting correlation module, a propagation path module, a transmission analysis module, and a parameter output module; wherein, the output end of the smelting event module is connected to the smelting correlation module, the output end of the smelting correlation module is connected to the propagation path module, the output end of the propagation path module is connected to the transmission analysis module, and the output end of the transmission analysis module is connected to the parameter output module.

[0088] The smelting event module is used to acquire smelting process parameters, map the smelting process parameters to the smelting equipment, and record smelting events of the smelting equipment within a preset time period.

[0089] The smelting association module is used to record abnormal events during each smelting process in the form of single parameters and parameter pairs, forming an association parameter chain.

[0090] The propagation path module is used to perform pattern clustering on the associated parameter chains according to the combination of single parameters and parameter pairs, and to determine the propagation path after clustering.

[0091] The transmission analysis module is used to score each batch of abnormal events based on the propagation path, calculate the transmission efficiency, single event score and path characteristics of each batch based on the correlation parameter chain, and construct a batch-indicator correlation matrix. It records the changing trends of each key indicator in the batch-indicator correlation matrix during multiple smelting processes, and obtains the transmission relationship of abnormal events in multiple smelting processes by analyzing the correlation of the changing trends of each key indicator.

[0092] The parameter output module is used to trace melting events based on the transmission relationship of abnormal events in multiple melting processes, determine the target parameters for each batch, and output the target parameters.

[0093] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention, which are still covered within the protection scope of the present invention.

Claims

1. A titanium alloy smelting effect intelligent analysis method, characterized in that, The method comprises the following steps: S1, obtaining smelting process parameters, mapping the smelting process parameters to smelting equipment, and recording smelting events of the smelting equipment in a preset time period; The implementation of step S1 comprises the following steps: S11, determining a smelting boundary of the current smelting process parameter based on the mapped smelting equipment, and determining a risk exposure time of the smelting process parameter reaching the smelting boundary and a risk mitigation time of the smelting process parameter decreasing from the smelting boundary to a normal value on each smelting equipment; S12, calculating the average value and the standard deviation of the smelting process parameter in the risk exposure time and the risk mitigation time, determining the stability of the smelting process parameter in each smelting, and extracting the smelting event by using the smelting boundary when the stability of the smelting process parameter meets the requirements; S13, when the stability of the smelting process parameter does not meet the requirements, dividing the smelting process parameter into multiple stages, associating the smelting process parameter according to the stages, determining the correlation of the smelting process parameter in different stages, and outputting the smelting process parameter with the maximum correlation as the extracted smelting event; S2, recording abnormal events in each smelting by using single parameters and parameter pairs, and forming an associated parameter chain; The implementation of step S2 comprises the following steps: S21, based on the abnormal events in each smelting, extracting the smelting process parameters associated with the abnormal events in the form of single parameters and parameter pairs, and regarding each monitored single parameter and parameter pair as an abnormal node; S22, for each abnormal node, calculating the time period when the abnormal event occurs, and merging the continuously occurring abnormal nodes into an abnormal time interval, wherein the abnormal time interval contains at least one abnormal node; S23, for each abnormal time interval, checking whether there is an intersection between the abnormal time interval and the abnormal time interval formed by all other abnormal nodes, if there is an intersection, regarding the abnormal nodes in the current abnormal time interval as root nodes, and regarding other abnormal nodes as secondary chain nodes of the abnormal nodes in the current abnormal time interval; S24, if there is no intersection, forming the output associated parameter chain in the form of time stamp advancement; S25, calculating all secondary chain nodes, and finding tertiary chain nodes under the secondary chain nodes in the form of the existence of the intersection of the abnormal time intervals, recursively finding step by step until there is no intersection of the abnormal time intervals, and regarding the formed chain nodes and corresponding abnormal nodes as the output associated parameter chain; S3, performing mode clustering on the associated parameter chain in the form of combined single parameters and parameter pairs, and determining a propagation path after clustering; S4, based on the propagation path, performing single event scoring on each batch of abnormal events, calculating the transmission efficiency, single event scoring and path characteristics of each batch based on the associated parameter chain, and constructing a batch-index association matrix; recording the change trend of each key indicator in the batch-index association matrix in multiple smeltings, and obtaining the transmission relationship of the abnormal events in multiple smeltings by analyzing the correlation of the change trend of each key indicator; S5, based on the transmission relationship of the abnormal events in multiple smeltings, tracing the smelting events, determining target parameters of each batch, and outputting the target parameters.

2. The method for intelligent analysis of titanium alloy smelting effect according to claim 1, characterized in that, The implementation manner of step S12 further includes: S121, based on the smelting demand information of the titanium alloy, analyzing the current smelting process parameters by using the preset smelting stability threshold; and comparing the preset smelting stability threshold with the average value and the standard deviation of the current smelting process parameters according to the processing flow of the titanium alloy smelting; S122, if any one of the average value or the standard deviation of the current smelting process parameters is out of the upper and lower limit ranges of the preset smelting stability threshold, the stability of the smelting process parameters is considered to be not in line with the requirements; S123, only when both the average value and the standard deviation of the current smelting process parameters are within the upper and lower limit ranges of the preset smelting stability threshold, the stability of the smelting process parameters is considered to be in line with the requirements.

3. The method for intelligent analysis of titanium alloy smelting effect according to claim 1, characterized in that, The implementation manner of step S13 further includes: The smelting process parameters in each stage are paired to form parameter pairs, the Pearson correlation coefficients of all the parameter pairs are calculated, and the correlation of the parameter pairs in each stage is obtained; For each stage, the parameter pair with the maximum correlation is found, and the correlation result in the corresponding stage is considered. The correlation results of all stages are compared, the global maximum correlation parameter pair is determined, and the smelting event is output.

4. The method for intelligent analysis of titanium alloy smelting effect according to claim 1, characterized in that, The implementation manner of setting the secondary chain node in step S23 further includes: It is judged whether the abnormal time interval of the current parameter pair is earlier than the single abnormal time interval of the parameter constituting the current parameter pair, if so, the abnormal time interval of the current parameter pair is taken as the root node, and the parameter constituting the current parameter pair is taken as the secondary chain node.

5. The method for intelligent analysis of titanium alloy smelting effect according to claim 1, characterized in that, The implementation manner of step S3 includes: A node type is set for each chain node in the associated parameter chain, and the structure topology of the current associated parameter chain is recorded; The node types and the structure topologies corresponding to the associated parameter chain are used for abnormal mode clustering, and the multiple sets of chain structures after the mode clustering are taken as the output propagation paths.

6. The method for intelligent analysis of titanium alloy smelting effect according to claim 1, characterized in that, The implementation manner of step S4 includes: S41, the propagation paths in each batch are counted, the weights of each node in the propagation paths are set based on the abnormal modes to which the propagation paths belong, and the single event score corresponding to the propagation path is set in a weighted summation manner according to the abnormal degree of each node on the propagation path; S42, the propagation paths of multiple batches are integrated, the average length of the propagation paths under the same abnormal mode, the maximum abnormal degree value of each propagation path at the time of the smelting, and the abnormal mode label corresponding to the propagation path are taken as the path features; and the transfer efficiency is set by using the ratio of the abnormal degree difference between the root node and the terminal node of the propagation path to the time length of the propagation path, so as to complete the construction of the batch-index association matrix; S43, the transfer efficiency, the path features and the single event score are used as the key indicator values of the batch-index association matrix, and the change trend is calculated along the time sequence to obtain the trend line slope corresponding to each key indicator; S44, the Pearson correlation coefficient of the trend line slopes corresponding to multiple key indicators is calculated, and the calculated value is taken as the correlation of the change trends of the key indicators; when the correlation value of the change trends of the key indicators and the trend line slope corresponding to the corresponding key indicator are consistent, the corresponding abnormal event is taken as the output transfer relationship.

7. The method for intelligent analysis of titanium alloy smelting effect according to claim 1, characterized in that, The implementation manner of step S5 includes: Based on the transmission relationship of abnormal events in multiple smelting, the single parameter and parameter pair combination form corresponding to the smelting process parameters are determined, and according to the abnormal events pointed by the smelting process parameters, the corresponding single parameter and parameter pair are taken as the output target parameters.

8. A titanium alloy smelting effect intelligent analysis system for performing the steps in the titanium alloy smelting effect intelligent analysis method according to any one of claims 1-7, characterized in that, Comprise: Smelting event module, for acquiring smelting process parameters, mapping smelting process parameters to smelting equipment, recording smelting events of smelting equipment in a preset time period; Smelting correlation module, for recording abnormal events in each smelting in the form of single parameter and parameter pair combination of smelting events in each smelting, forming a correlation parameter chain; Propagation path module, for mode clustering of the correlation parameter chain in the form of combined single parameter and parameter pair, and determining the propagation path after clustering; Transmission analysis module, for single event scoring of each batch of abnormal events based on the propagation path, calculating the transmission efficiency, single event scoring and path characteristics of each batch based on the correlation parameter chain, and constructing a batch-index correlation matrix; recording the change trend of each key index in the batch-index correlation matrix in multiple smeltings, and obtaining the transmission relationship of abnormal events in multiple smeltings by analyzing the correlation of the change trend of each key index; Parameter output module, for tracing smelting events based on the transmission relationship of abnormal events in multiple smeltings, determining the target parameters of each batch, and outputting the target parameters.

Images