A method for precise temperature control of flexible molds

Abstract

This invention relates to the field of injection molding temperature control technology, specifically a method for precise temperature control of flexible molds. The method includes: acquiring the logical regions of the current mold; collecting temperature time-series datasets for each logical region; analyzing the temperature characteristics of each logical region at stage boundaries based on the number of offset points in each logical region; defining the event types and association rules corresponding to the temperature characteristics; calculating the temperature similarity of each logical region using its temperature characteristics as input; marking the logical regions requiring adjustment as regions to be adjusted; generating temperature change signals based on the duration and control cycle of the regions to be adjusted at each stage boundary; and extracting anomaly labels from the temperature change signals; using the temperature change signals as the basis for strategy comparison; using the data of the temperature change signals at each stage boundary as nodes in a strategy tree; and determining the target temperature control strategy according to the path parameter values ​​of the strategy tree. This improves the response rate and accuracy of temperature control.

Description

Technical Field

[0001] This invention relates to the field of injection molding temperature control technology, specifically a method for precise temperature control of flexible molds. Background Technology

[0002] In hot processing technologies such as injection molding and die casting, precise control of mold temperature is crucial for ensuring product quality and improving production efficiency. However, in mold temperature control scenarios, uneven temperature control due to fixed partitioning can lead to lag in the temperature control strategy's response, ultimately reducing mold production efficiency.

[0003] For example, Chinese Patent Publication No. CN118809988A discloses an automatic injection molding machine operation monitoring system and method. During the injection molding process, which includes multiple stages such as mold filling, pressure holding, and cooling, the mold temperature and injection pressure fluctuate with each stage. To accurately monitor abnormal situations, real-time and historical time-series data of mold temperature and injection pressure were acquired. By analyzing the similarities between the real-time and historical time-series data of injection pressure, the current injection molding process is divided into stage time periods. Considering the impact of changes in production conditions on the duration of each stage, it is necessary to analyze the differences in duration between real-time and historical data within each stage time period to accurately pinpoint the comparison point.

[0004] For example, Chinese Patent Publication No. CN114734604A discloses an online mold temperature control method for the injection molding process. To overcome the problems of existing technologies that adjust temperature by real-time temperature monitoring, which cannot eliminate the effects of time lag in temperature changes and result in poor temperature control stability, this invention includes the following steps: dividing the cavity from the gate to the end of the cavity into several detection areas; determining the temperature control mode of each detection area based on the temperature change rate per unit time, the real-time temperature value, and the current stage of the injection molding process; executing the corresponding temperature control mode in each detection area; training a model based on the inlet and outlet water temperatures, water flow rate, and water pressure of the cooling / heating water circuit; using the model to predict the temperature changes in the detection area and control the water supply of the cooling / heating water circuit; and completing the mold injection process.

[0005] Existing technologies identify temperature trends by comparing temperature differences at different times during the injection molding process and by using water flow loops to monitor temperature changes during the current injection molding process. However, existing technologies tend to overlook local anomalies on the mold surface when identifying temperature. This makes it difficult to dynamically adjust the mold zones and promptly check the event types in each zone when the mold has a complex shape, thus hindering precise control of local areas on the mold. Summary of the Invention

[0006] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: a method for precise temperature control of flexible mold, including: S1, obtaining a three-dimensional model of the current mold, dividing the current mold into feature partitions according to shape features to form various logical regions.

[0007] S2 collects temperature time-series datasets for each logical region, analyzes the temperature characteristics of each logical region under the stage boundary based on the number of offset points in each logical region, and defines the event types and association rules corresponding to the temperature characteristics.

[0008] S3 takes the temperature characteristics of each logical region as input, calculates the temperature similarity of each logical region, determines the relationship between each logical region and the temperature characteristics, and marks the logical regions that need to be adjusted as regions to be adjusted.

[0009] S4. Perform temperature operation status analysis on the area to be adjusted, generate temperature change signals based on the duration and control cycle of the area to be adjusted at each stage boundary, and extract anomaly labels from the temperature change signals.

[0010] S5 uses the temperature change signal as the basis for strategy comparison, uses the data of the temperature change signal under the boundary of each stage as the node of the strategy tree, and determines the target temperature control strategy according to the path parameter value of the strategy tree.

[0011] The beneficial effects of this invention are as follows: First, by dividing the mold surface into multiple logical regions and identifying data points with offsets based on the temperature change rate in each logical region, this invention takes any temperature feature as a starting point, aligns the temperature deviation with the process stage boundary, calculates the time offset, and generates conditional probability association rules to describe the spatiotemporal correlation of the current temperature feature under the process stage, thus providing data support for subsequent strategy decisions.

[0012] Second, this invention performs temperature similarity calculation and classifies outlier regions based on the calculated temperature similarity. It distinguishes the anomaly type by the difference between mold opening temperature and cooling rate, thereby defining the regions with temperature anomalies in the current processing. Then, it queries the problematic regions to determine the anomaly labels present in the current region clustering, thus realizing the local anomaly location and identification.

[0013] Third, this invention uses temperature values, anomaly labels, and logical regions as input nodes, associates stage boundaries with temperature control strategies, and selects the optimal strategy through path weight optimization to obtain the target strategy in the current scenario, thereby improving the response speed and accuracy of temperature control. Attached Figure Description

[0014] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0015] Figure 1This is a flowchart illustrating a method for precise temperature control of flexible molds.

[0016] Figure 2 This is a flowchart illustrating step S2 of a method for precise temperature control of a flexible mold.

[0017] Figure 3 This is a flowchart illustrating step S3 of a method for precise temperature control of a flexible mold.

[0018] Figure 4 This is a flowchart illustrating step S4 of a method for precise temperature control of a flexible mold.

[0019] Figure 5 This is a flowchart illustrating step S5 of a method for precise temperature control of a flexible mold. Detailed Implementation

[0020] 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.

[0021] See Figure 1 A method for precise temperature control of a flexible mold includes: S1, acquiring a three-dimensional model of the current mold, dividing the current mold into feature regions according to its shape characteristics, and forming various logical regions.

[0022] S2 collects temperature time-series datasets for each logical region, analyzes the temperature characteristics of each logical region under the stage boundary based on the number of offset points in each logical region, and defines the event types and association rules corresponding to the temperature characteristics.

[0023] S3 takes the temperature characteristics of each logical region as input, calculates the temperature similarity of each logical region, determines the relationship between each logical region and the temperature characteristics, and marks the logical regions that need to be adjusted as regions to be adjusted.

[0024] S4. Perform temperature operation status analysis on the area to be adjusted, generate temperature change signals based on the duration and control cycle of the area to be adjusted at each stage boundary, and extract anomaly labels from the temperature change signals.

[0025] S5 uses the temperature change signal as the basis for strategy comparison, uses the data of the temperature change signal under the boundary of each stage as the node of the strategy tree, and determines the target temperature control strategy according to the path parameter value of the strategy tree.

[0026] When forming logical regions, factors such as wall thickness variations at multiple locations, cavity complexity, gate location, hot runner distribution, and cooling channel layout are considered. The area represented by the mold surface is then divided into logical regions for temperature identification. This division is based on shape features and distance, creating multiple densely packed grids covering the mold to form regions with different descriptive forms. Region division can be based on pre-defined areas. Temperature sensors are then deployed in each logical region to ensure sufficient temperature data. The temperature of the mold can be obtained from points with the largest temperature gradient, points near the mold cavity surface, and points near the cooling channels.

[0027] The implementation of step S1 includes: S11, using the shape features of the current model as input to perform geometric topology analysis, and obtaining the key features of the mold. The key features will represent multiple regions that are similar in shape, such as the wall thickness change zone, surface curvature, gate / hot runner position, and cooling water channel layout. If there is a conflict in the description of the regions, the larger region in the conflict will be used as the standard for division to divide the mold surface into multiple regions. Then, the temperature change pattern of each region will be queried to find out whether the current mold is under uneven heating.

[0028] S12, based on the key features of the mold, divides the mold surface into multiple logical partitions. The logical partitions are determined by the different representations of the key features; each key feature's area is circled and considered a logical partition. For more precise control of the mold, the area corresponding to each key feature can be divided into multiple equally sized directional grids to form the currently monitored logical partition. The temperature of the logical partitions can be determined using an array of temperature sensors to identify temperature values ​​at multiple locations on the mold surface.

[0029] In one embodiment of the present invention, the temperature time series dataset will contain all the data of each region during the completion of a mold production cycle. At this time, sufficient temperature data will be obtained according to the batch of mold production to complete the adjustment and control of the current mold.

[0030] The aforementioned stage boundaries include the temperature values ​​at the start and end points of the preheating, injection, holding, and cooling stages. This information describes whether the mold requires heating to compensate for temperature changes or adjusting the cooling temperature during production. It also indicates whether there are any abnormal temperature changes at the start and end points of each stage. The event types and association rules for temperature characteristics under each stage are labeled using the temperature values ​​from the start to the end point. The extracted temperature features represent the start temperature, end temperature, average temperature, temperature change, peak temperature, and cooling rate for each stage. These data are combined to form the extracted temperature features, which are then used to identify and process the temperature change trends during mold production.

[0031] The cooling rate here is used to describe the temperature reduction of the mold at various stages of production. This prevents problems such as excessive or slow temperature reduction. In this case, water at 25°C or other temperatures will be injected to adjust the temperature by cooling the water channels and reducing the heating rate.

[0032] Therefore, the implementation of step S2 also includes: using the temperature characteristics of the temperature time series dataset under the stage boundary, performing feature recognition on the temperature values ​​within each stage boundary, and obtaining the temperature change rate of each data point.

[0033] Based on the temperature change rate of each data point, cluster analysis is performed on each data point of the temperature time series dataset to obtain at least one set of temperature change rate offset points and neighborhood temperature change offset points.

[0034] When the number of temperature change rate offset points and the number of neighboring temperature change offset points exceed the preset threshold, the corresponding logical region will be output to obtain the number of offset points in each logical region.

[0035] At this point, check whether the temperature value in each logical region has shifted across multiple stages, and output the portion where the shift point is greater than the preset threshold to illustrate the shift situation; use the temperature change rate shift point and the neighboring temperature change shift point as the shift points for each logical region.

[0036] The aforementioned temperature change rate offset point is the average of historical temperature values ​​in the same period plus three times the standard deviation, which serves as the threshold for the temperature change rate offset point at this time. This is biased towards judging whether the temperature change rate per unit time exceeds the threshold of the temperature change rate in historical data, in order to obtain the current temperature accumulation situation.

[0037] The neighborhood temperature change offset point indicates that there is a part where the temperature change rate of a data point is inconsistent with that of the adjacent logical region data points. The corresponding set of data points is regarded as the offset point identified at this time to illustrate the current temperature change rate.

[0038] As for the preset number threshold of temperature change rate offset points, the average number of offset points used in historical data when making offset point alarms is used as the preset number threshold at this time. For the preset number threshold of neighboring temperature change offset points, the average number of consecutive neighboring temperature change offset points when making alarms in historical data is used as its preset number threshold.

[0039] When there are too many temperature change rate offset points and neighboring temperature change offset points, it can be determined that there is an anomaly in the current logical region. Then, the output is based on the temperature change rate offset points and neighboring temperature change offset points identified at this time. The output logical region is used as the part for further analysis of temperature features to explain what event types exist in the logical region with heating fluctuations and abnormal temperature changes, and what association rules need to be set for the relevant logical regions to check for possible anomalies in the corresponding regions.

[0040] Preferably, when obtaining the number of offset points in each logical region using the number of temperature change rate offset points and the number of neighboring temperature change offset points, the method further includes: extracting the time points corresponding to the number of offset points, viewing the distribution time periods of each offset point in the temperature time series dataset, and outputting the temperature time series dataset in sequence based on the number of offset points contained within the distribution time period and taking the time period with the most offset points as the center.

[0041] The extracted time points are used to describe the distribution of offset points, indicating which time periods are more likely to experience abnormal temperature changes. This helps to understand the trends and patterns of abnormal temperature. Then, focusing on the time period with the most offset points, we can concentrate on the time period when problems are most likely to occur. The temperature time series datasets can be analyzed and processed sequentially according to the number of offset points, from largest to smallest.

[0042] In addition, when describing its temperature characteristics, it is also necessary to use sensors to view the duration of cooling and heating of the mold under the same historical data, and associate these data with the currently acquired temperature characteristics to facilitate subsequent retrieval and query.

[0043] like Figure 2As shown, the implementation of step S2 also includes: S21, based on any temperature feature under the stage boundary, using the data where the temperature feature has deviations at each time point, clustering the temperature values ​​corresponding to each data point, setting temperature labels for the temperature values ​​of each cluster after clustering, identifying the event type of each temperature feature under the temperature value, the temperature labels set at this time are used to indicate the temperature value taken by the temperature feature within the cluster after clustering, and then labeling the temperature features of these clusters to indicate whether there is a corresponding event type, such as the cluster showing a sudden increase in temperature in a certain area, or a sudden decrease in temperature value. These event types are used to indicate whether the temperature feature can represent whether there are potential abnormal events within the logical area divided by the current mold after clustering; as for the clustering method, K-means, DBSCAN and other clustering methods can be used to cluster any temperature feature under the current stage boundary to identify the temperature pattern existing under the current mold production process.

[0044] The aforementioned event types include, but are not limited to, sudden temperature rise, sudden temperature drop, abnormal periodic fluctuations, abnormal temperature deviation from the baseline value, local overheating, and abnormal uniformity, which are used to indicate whether there are corresponding abnormal temperature conditions in the multiple logical regions divided by the current mold.

[0045] S22, align the first trigger time of each event type with the time series corresponding to the stage boundary, calculate the time offset of the trigger time of each event type relative to the stage boundary, and use it as the trigger time of each event type. At this time, the trigger time is used to represent the time point when the temperature characteristics have deviations. The time point when the corresponding time type can appear is associated with the corresponding stage during the current mold production. For example, the first trigger time of the event type is associated with the time of injection start, holding pressure end, etc. Then, describe the time length difference between the corresponding event type and the corresponding stage to explain how long the corresponding abnormality may occur when the corresponding temperature value is maintained for a certain time, delayed for a certain time, or when the corresponding temperature value changes.

[0046] The trigger time mentioned above represents the time point during which each event type can continue and be included after it is first identified by the system within the current time period.

[0047] S23, examine the conditional probabilities of each temperature feature and event type. Based on the conditional probabilities and trigger times corresponding to each temperature feature and event type, generate association rules corresponding to each temperature feature and add the association rules of each temperature feature to the boundary of each stage. For example, the association rule can be expressed as: if the temperature value of the pressure holding stage is A and the trigger time is <5 seconds, then there is a 90% probability of a temperature rise mutation. At this time, the possible event types and temperature values, the trigger time of temperature value changes, the duration of reaching the temperature value, etc. are comprehensively described to express what form an anomaly will be represented when there is an anomaly in the current logical partition.

[0048] The conditional probability of temperature features and event types is used to describe the probability of satisfying a current temperature value with a deviation and the corresponding event type. When calculating the conditional probability, the frequency of the temperature feature is calculated, and the frequency of the related event type is also calculated. Then, the probabilities of these frequencies and the total sample size are calculated, and the conditional probability of each temperature feature and event type is expressed using the formula for conditional probability.

[0049] In one embodiment of the present invention, the calculation of temperature similarity will mainly focus on identifying the starting temperature in the preset stage, the amount of temperature change in the injection stage, the peak and average temperature in the holding stage, and the corresponding cooling rate in the cooling stage.

[0050] By combining the temperature characteristics of the logical regions with the differences in the content identified under the boundaries of each stage, the temperature similarity of each logical region under the boundary conditions of different stages is calculated. The temperature similarity can be obtained by using the cosine similarity calculation method to obtain the temperature similarity of the current corresponding logical region. At the same time, it is necessary to determine whether there are any abnormal parts that need to be adjusted among the temperature characteristics, and mark these abnormal parts as the parts that need to be adjusted, so as to regulate the temperature of the flexible mold at various positions.

[0051] In step S3, based on the event types and association rules identified after analyzing temperature features, multiple logical regions are correlated to illustrate the interrelationships between these regions when an anomaly exists in a single region. Based on these interrelationships, the regions requiring joint adjustment are identified to complete the specific temperature identification process for the current mold production. This mainly involves using the obtained event types and association rules to map the acquired temperature features using labels, and then comparing the mapped data to determine if the current logical region needs to be marked as an area requiring adjustment.

[0052] like Figure 3 As shown, the implementation of step S3 includes: S31, calculating the temperature similarity of the temperature features of the current logical partition based on the event type and association rules corresponding to the temperature features, and calculating the temperature similarity sequentially using the temperature mean, standard deviation, and range within each logical partition; at this time, similarity processing is performed on the static features of each logical region, and similar patterns under different time sequences are retained in a time sequence form; as for the event type and association rules, when processing the dynamic features related to the temperature change rate, the analysis content under the static features and dynamic features is combined and processed to obtain more comprehensive analysis data for the current logical region, thereby improving the identification of relevant data within the region.

[0053] It should be noted that after performing temperature-related calculations, if there are different dimensions or large differences in values, a standardization process can be used to analyze all temperature values ​​as normalized values.

[0054] The temperature similarity, obtained from the mean, standard deviation, and range, will be used to represent the correlation between the current logical partition and other partitions under certain temperature features corresponding to specific event types and association rules.

[0055] S32. After the temperature similarity calculation is completed, hierarchical clustering is performed based on the temperature similarity between each logical region to obtain multiple clusters. At this time, a bottom-up merging strategy is adopted during clustering. The data after temperature similarity calculation is set as a similarity matrix. The data of each logical region in which temperature similarity is calculated is used as an initial cluster. The two clusters with the highest similarity are gradually merged to complete the hierarchical clustering and obtain multiple clusters after the current clustering is completed.

[0056] S33, outlier detection is performed on the logical regions within each cluster. Outliers that exist after clustering indicate that the temperature behavior pattern of the region is significantly different from the overall temperature change of the mold, usually indicating local circulation abnormalities or cooling failures. Based on the results of outlier detection, the corresponding logical regions are marked as regions to be adjusted.

[0057] When performing outlier detection on logical regions within each cluster, the implementation method also includes: if outliers exist, classifying them based on mold opening temperature difference and cooling rate, dividing them into abnormal mold opening temperature difference regions and abnormal cooling rate regions corresponding to the current temperature characteristics; at the same time, it also checks the relevant type of the current logical region. For example, if the current logical region is a cooling channel, it is not necessary to directly check its outlier regions, and the abnormal mold opening temperature difference regions and abnormal cooling rate regions are marked as regions to be adjusted.

[0058] An abnormal temperature difference region during mold opening indicates that a certain cluster corresponds to an area where the temperature exceeds the allowable deviation during mold opening, which will lead to product warping or demolding damage. This area is considered an abnormal temperature difference region during mold opening. An abnormal cooling rate region indicates low cooling efficiency, with any phenomenon indicating a reduced cooling rate, such as excessively low water flow rate, excessively high water temperature, or poor contact. In this case, by describing the rate of change of the temperature value of the cluster after clustering during the cooling stage, outliers that meet the criteria for insufficient cooling rate are identified, and the areas corresponding to these data are considered abnormal cooling rate regions.

[0059] For example, outlier detection methods use a local outlier factor algorithm for identification. For each data point p in a cluster, k nearest neighbor data points are identified. The Euclidean distance between each data point is calculated by normalizing the temperature value of each data point in the cluster. The Euclidean distance of the kth nearest neighbor is used as the k-distance of data point p. Then, the reachability distance from the data point to other data points is calculated using the k-distance, and the local reachability density of each data point is determined. Finally, the local outlier factor of each data point is obtained. Outlier regions are then filtered according to the value of the local outlier factor. Regions with a local outlier factor greater than 1.5 are considered as outliers to represent differences relative to the overall region. Alternatively, other forms of local outlier factor values ​​can be used, which can be set based on current computational needs. Specific calculation formulas can be found in existing technologies and will not be elaborated upon here. As for the nearest neighbor data points obtained, they can be selected based on 3 to obtain data points that are relatively close to the current data point. Then, using the outlier detection calculation method, the local outlier factor represented by each data point is identified, thereby selecting the local data that are obviously different.

[0060] In one embodiment of the present invention, when it is determined that the area to be adjusted has temperature anomalies related to mold opening temperature difference and cooling rate, temperature operation analysis is performed. This analysis combines the average temperature, standard deviation, lag time, overshoot, and number of abnormal event types corresponding to the stage boundaries of the area to be adjusted. These data are then combined into a sequence at different stage boundaries to illustrate the specific stage at which the temperature anomaly occurs and the relative changes observed. Lag time is the time required for the current temperature to reach the target value after receiving a command or at the start of a stage; it reflects the system's response speed. The shorter the lag time, the faster the system response. Overshoot indicates when the system response exceeds the target value; it is calculated by taking the ratio of the lag time to the preset response time. The number of abnormal event types is based on the event types corresponding to the temperature characteristics. The number of abnormal event types in the current area to be adjusted is statistically analyzed to describe the relative situation of the area to be adjusted.

[0061] like Figure 4 As shown, the implementation of step S4 includes: S41, extracting the average temperature, standard deviation, lag time, overshoot, and number of abnormal event types of the area to be adjusted based on the duration of each stage boundary and the control cycle; the duration of the stage boundary describes the duration of the preheating stage, injection stage, pressure holding stage, and cooling stage; the control cycle determines how often the system checks the current temperature and adjusts its temperature control according to the current temperature.

[0062] At this point, the extracted values ​​will be based on the duration and the temperature value under the corresponding control period, and will include the average temperature, standard deviation, lag time, overshoot, and number of abnormal event types. If the control period is a short value, these values ​​will be obtained based on the duration.

[0063] S42, the average temperature, standard deviation, lag time, overshoot, and number of abnormal event types of the area to be adjusted are spliced ​​together according to the boundaries of each stage to form a temperature change signal.

[0064] At this point, the main task is to splice the data to be processed, continuously update the values ​​in the temperature change signal, add the updated values ​​to the temperature change signal, record the number of abnormal event types currently identified to describe the abnormal pattern of the current temperature change signal, perform threshold detection on the temperature change signal, and identify the abnormal pattern of the temperature change signal.

[0065] S43 uses the average temperature, standard deviation, lag time, overshoot, and number of abnormal event types of the area to be adjusted as query rules, and combines and searches each rule in turn. If there is an abnormal pattern in the database that corresponds to the value of the current rule, the area to be adjusted is marked and output as an abnormal label of the temperature change signal.

[0066] The above-mentioned abnormal patterns can be represented as follows: when the average temperature is greater than 185℃ or less than 17℃, a temperature over-limit alarm is triggered; or when the standard deviation is greater than 1℃, a temperature fluctuation abnormal alarm is triggered. At this time, the five values ​​extracted from the area to be adjusted will be queried in a one-to-one or many-to-one manner to find out whether there are related abnormal patterns in the current area to be adjusted. At this time, the event type obtained is more specific in describing the abnormal situation that may occur at the boundary of each stage of the area to be adjusted, and these abnormalities are marked with the most specific labels possible to indicate whether the data contained in the temperature change signal will affect the overall mold production.

[0067] In one embodiment of the present invention, step S5 is used to set up a policy tree by taking the data queried in the temperature change signal as nodes. The nodes of the policy tree include not only anomaly labels and temperature values, but also the location of the logical region corresponding to the temperature value, event type, association rules, and other processing content. Each node of the policy tree can perform policy matching based on the anomaly label and the specific temperature value. The sequence data covered by the temperature change signal is used as the root node of the decision tree, and the anomaly label and current stage of the temperature change signal are used as its branch nodes. After describing the policies adopted by the leaf nodes under multiple branch nodes, it indicates the part of the current temperature change signal that needs to be responded to and processed first. A path is formed: temperature change signal - anomaly label - logical region - stage boundary - temperature value - temperature control strategy, to realize the temperature control strategy for handling specific temperature anomalies. At this time, weights are assigned to the policy tree nodes according to the content contained in the nodes. For example, weights are set according to the content described by the anomaly label, and then weights are sequentially set for the problematic parts of the logical region, stage boundary, and temperature value. The path with the highest priority or weight is selected as the priority temperature control strategy. Regarding the weight of anomaly labels for each node in the strategy tree, it is set based on the severity values ​​of different anomaly labels in the database. These anomaly labels will exhibit different degrees of impact depending on their descriptions; therefore, the weight is primarily set based on the relevant anomaly flags in the database. For nodes related to logical regions, the weight is the ratio of the frequency of anomalies occurring in the logical region to the total frequency of anomalies. For nodes describing the current stage, such as stage boundaries, the weight is the ratio of the frequency of anomalies occurring in each stage to the total frequency of anomalies. For nodes related to temperature values, the weight is the ratio of their value to the total value in the currently sampled data. This indicates the path corresponding to the currently executed temperature control strategy. Temperature values ​​include multiple values ​​such as temperature, label difference, and variance in the temperature change signal. All current temperature-related values ​​are considered as temperature values, and the content of each temperature value combination is adapted to the content in the stable control strategy.

[0068] like Figure 5 As shown, the implementation of step S5 includes: S51, using the temperature value, abnormal label and corresponding logical region in the temperature change signal as the decision basis for strategy comparison; using the stage boundary associated with the temperature change signal as the decision constraint; using the control parameters, control strategy pointer and strategy coordination weight of the temperature control strategy in the database as the decision mapping; and constructing a strategy tree with the decision basis, decision constraints and decision mapping.

[0069] S52 connects the nodes in the strategy tree to form multiple strategy paths. The path parameter values ​​of each strategy path are compared based on the temperature control strategy that the strategy path ultimately points to, and the path parameter value with the largest value is taken as the target temperature control strategy output.

[0070] The target temperature control strategy represents a temperature control strategy corresponding to the maximum path parameter value under the current anomaly label. This temperature control strategy is adjusted according to the control strategy pointers it contains. The path parameter value is represented as the sum of the weights of the nodes contained in the strategy path. It should be noted that the weights obtained here do not include the strategy coordination weights corresponding to the temperature control strategy.

[0071] The control parameters of the aforementioned temperature control strategy represent the target temperature value and the allowable temperature fluctuation range adjusted in response to the current temperature change signal. These data describe the values ​​that the strategy tree needs to adjust and achieve in response to corresponding anomalies in the temperature change signal. The control strategy pointer indicates the identifier pointing to the specific strategy execution, such as PID parameter adjustment, heating power limiting, and cooling water flow compensation, which directly point to the content of the temperature control strategy execution method. The strategy coordination weight represents the weight set for the specific content executed by the current temperature control strategy, which is used to explain the execution order under the current temperature control strategy that needs to execute multiple control strategy policies. This weight is set in advance in the database.

[0072] Preferably, the implementation of the strategy tree further includes: taking the current input temperature change signal as the root node of the strategy tree, taking the abnormal label, logical region and stage boundary of the temperature change signal as the branch nodes of the root node in sequence, taking the temperature value of the temperature change signal and the temperature control strategy as the secondary nodes of the branch nodes, and taking the control strategy pointer and strategy coordination weight of the temperature control strategy as the leaf nodes to form the strategy tree.

[0073] Preferably, when forming the strategy tree, it is also necessary to perform decision mapping between the temperature value of the temperature change signal and the temperature control strategy. The temperature value of the current temperature change signal is compared with the control parameters of the temperature control strategy in turn. When the control parameters of the temperature control strategy are satisfied, the control strategy pointer and strategy coordination weight of the current temperature control strategy are used as the output target temperature control strategy.

[0074] When comparing the temperature change signal with the control parameters of the temperature control strategy, the temperature value and the temperature fluctuation range are compared. Only when the current temperature control strategy meets the abnormal label, logical region and stage boundary of the temperature change signal and corresponds to the temperature value of the temperature change signal, will the relevant content in the temperature control strategy be executed according to the control strategy pointer it contains. The execution is gradually achieved from large to small by the size of the strategy coordination weight.

[0075] If the conditions are not met, the nearest neighbor temperature control strategy in the historical data will be used as the current target temperature control strategy. When querying using the current temperature change signal, if it does not correspond to the temperature fluctuation range of the current temperature control strategy, the strategy implemented in the historical decision-making process for this situation will be adopted. This means that when a strategy cannot completely match the current situation, the strategy path corresponding to the current temperature change signal in the strategy tree is converted into a vector form. This vector form is then used to calculate the cosine similarity with other temperature control strategies in the historical data. The strategy with the highest cosine similarity is chosen as its nearest neighbor temperature control strategy, or the temperature control strategy closest in time to the current time sequence is chosen as the target temperature control strategy. This strategy rollback process is used to achieve dynamic strategy control for the current mold production.

[0076] 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 method for precisely controlling the temperature of a flexible mold, characterized in that, include: S1, obtain the three-dimensional model of the current mold, divide the current mold into feature regions according to shape features, and form each logical region. Shape features include wall thickness variation, cavity complexity, gate location, hot runner distribution, and cooling water channel layout. S2, collect temperature time series datasets for each logical region, analyze the temperature characteristics of each logical region under the stage boundary based on the number of offset points in each logical region, and define the event types and association rules corresponding to the temperature characteristics; The number of offset points is determined by clustering the temperature time series data to obtain the temperature change rate offset points and the neighborhood temperature change offset points. The number of offset points is counted and compared with a preset threshold to obtain the number of offset points in each logical region; Extract the time period of the offset point distribution, and sort and output the time period with the most offset points as the center. Offset points include temperature change rate offset points and neighborhood temperature change offset points; stage boundaries include temperature values ​​at the start and end positions of the preheating stage, injection stage, pressure holding stage, and cooling stage; Temperature characteristics represent the starting temperature, ending temperature, average temperature, temperature change, peak temperature, and cooling rate for each stage; event types include sudden temperature rise, sudden temperature drop, abnormal periodic fluctuations, abnormal temperature deviation from the baseline value, local overheating, and abnormal uniformity; association rules represent a comprehensive description of the current event type, temperature value, the trigger time of the temperature value change, and the duration of reaching the temperature value, forming an association rule form. S3, taking the temperature characteristics of each logical region as input, calculates the temperature similarity of each logical region, determines the relationship between each logical region and the temperature characteristics, and marks the logical regions that need to be adjusted as regions to be adjusted. Temperature similarity is calculated by extracting the mean, standard deviation, and range of the temperature of the logical region. Cosine similarity is used to calculate the temperature similarity between regions; Hierarchical clustering based on similarity yields clusters; The outlier detection and analysis method is as follows: the local outlier factor algorithm is used to detect outlier regions within clusters; Outlier regions were classified based on mold opening temperature difference and cooling rate. Mark areas with abnormal mold opening temperature differences and abnormal cooling rates as areas to be adjusted. S4. Perform temperature operation status analysis on the area to be adjusted, generate temperature change signals based on the duration and control cycle of the area to be adjusted at each stage boundary, and extract anomaly labels from the temperature change signals. The average temperature, standard deviation, lag time, overshoot, and number of abnormal event types of the region to be adjusted are spliced ​​together according to the boundaries of each stage to form a temperature change signal. The method for extracting anomaly labels is as follows: extract the average temperature, standard deviation, lag time, overshoot, and number of anomaly event types of the region to be adjusted; Combine the above parameters to retrieve abnormal patterns in the database; If an abnormal pattern is matched, it is labeled as an abnormal temperature change signal. S5 uses the temperature change signal as the basis for strategy comparison, uses the data of the temperature change signal under the boundary of each stage as the node of the strategy tree, and determines the target temperature control strategy according to the path parameter value of the strategy tree. The root node contains the sequence data covered by the temperature change signal; the branch nodes contain the anomaly labels and current stage of the temperature change signal; the secondary nodes contain the temperature values ​​and temperature control strategies; and the leaf nodes contain the control strategy pointers and strategy coordination weights. Use the sum of the weights of the nodes contained in the strategy path as the path parameter value; The strategy tree is constructed and optimized by using decision basis, decision constraints, and decision mapping to build the strategy tree. Connect the nodes in the policy tree to form multiple policy paths; Calculate the parameter values ​​for each path, and take the strategy corresponding to the maximum value as the target control strategy.

2. The method for precise temperature control of a flexible mold according to claim 1, characterized in that, The implementation of step S1 includes: S11, using the shape features of the current model as input to perform geometric topology analysis and obtain the key features of the mold; S12 divides the mold surface into multiple logical partitions based on the key features of the mold; the key features refer to the geometric topological core features of the mold, such as the wall thickness change zone, surface curvature, gate / hot runner location, and cooling water channel layout.

3. The method for precise temperature control of a flexible mold according to claim 1, characterized in that, The implementation of step S2 also includes: Using the temperature characteristics of the temperature time series dataset under the stage boundaries, feature identification is performed on the temperature values ​​within each stage boundary to obtain the temperature change rate of each data point. Based on the temperature change rate of each data point, cluster analysis is performed on each data point of the temperature time series dataset to obtain at least one set of temperature change rate offset points and neighborhood temperature change offset points. When the number of temperature change rate offset points and neighborhood temperature change offset points exceeds the preset threshold, the corresponding logical region is output to obtain the number of offset points in each logical region. The temperature change rate offset points and neighborhood temperature change offset points are used as the offset points identified in each logical region. The temperature change rate offset point is the temperature change rate anomaly point where the historical temperature mean for the same period is added to three times the standard deviation. The neighborhood temperature change offset point represents a data point whose temperature change rate is inconsistent with that of the adjacent logical region.

4. The method for precise temperature control of a flexible mold according to claim 3, characterized in that, Other methods for obtaining the number of offset points for each logical region include: Extract the time points corresponding to the number of offset points, view the distribution time periods of each offset point in the temperature time series dataset, and output the temperature time series dataset in order, centered on the time period with the most offset points, based on the number of offset points contained in the distribution time period.

5. The method for precise temperature control of a flexible mold according to claim 1, characterized in that, The implementation of step S2 also includes: S21. Based on any temperature feature under the stage boundary, cluster the temperature values ​​corresponding to each data point using the data where the temperature feature has a deviation at each time point, set temperature labels using the temperature values ​​of each cluster after clustering, and identify the event type of each temperature feature under the temperature value. S22, Align the first trigger time of each event type with the time series corresponding to the stage boundary, calculate the time offset of the trigger time of each event type relative to the stage boundary, and use it as the trigger time of each event type; S23, view the conditional probabilities of each temperature feature and event type, generate association rules for each temperature feature based on the conditional probabilities and trigger times corresponding to each temperature feature and event type, and add the association rules for each temperature feature to the boundary of each stage.

6. The method for precise temperature control of a flexible mold according to claim 1, characterized in that, Step S3 can be implemented in the following ways: S31, calculate the temperature similarity of the temperature features of the current logical partition based on the event type and association rules corresponding to the temperature features, and calculate the temperature similarity in turn using the mean, standard deviation and range of the temperature in each logical partition. S32, After the temperature similarity calculation is completed, hierarchical clustering is performed based on the temperature similarity between each logical region to obtain multiple clusters; S33, perform outlier detection on the logical regions within each cluster, and mark the corresponding logical regions as regions to be adjusted based on the outlier detection results.

7. The method for precise temperature control of a flexible mold according to claim 6, characterized in that, When performing outlier detection on logical regions within each cluster, the implementation methods also include: If outlier regions exist, they are classified according to the mold opening temperature difference and cooling rate. The abnormal mold opening temperature difference region and abnormal cooling rate region corresponding to the current temperature characteristics are divided and marked as regions to be adjusted.

8. The method for precise temperature control of a flexible mold according to claim 1, characterized in that, Step S4 can be implemented in the following ways: S41, using the duration and control cycle of each stage boundary, extract the average temperature, standard deviation, lag time, overshoot, and number of abnormal event types of the region to be adjusted; S42, the average temperature, standard deviation, lag time, overshoot, and number of abnormal event types of the area to be adjusted are spliced ​​together according to the boundaries of each stage to form a temperature change signal; S43 uses the average temperature, standard deviation, lag time, overshoot, and number of abnormal event types of the area to be adjusted as query rules, and combines and searches each rule in turn. If there is an abnormal pattern in the database that corresponds to the value of the current rule, the area to be adjusted is marked and output as an abnormal label of the temperature change signal. The lag time is the time required for the current temperature to reach the target value after receiving a command or at the start of a phase; it reflects the system's response speed. The overshoot represents the situation where the system response exceeds the target value, which is the ratio of the difference between the lag time and the preset response time.

9. The method for precise temperature control of a flexible mold according to claim 1, characterized in that, Step S5 can be implemented in the following ways: S51 uses the temperature values, anomaly labels, and corresponding logical regions in the temperature change signal as the decision basis for strategy comparison; uses the stage boundaries associated with the temperature change signal as decision constraints; and uses the control parameters, control strategy pointers, and strategy coordination weights of the temperature control strategy in the database as decision mappings. A strategy tree is constructed using the decision basis, decision constraints, and decision mappings. S52 connects the nodes in the strategy tree to form multiple strategy paths. The path parameter values ​​of each strategy path are compared based on the temperature control strategy that the strategy path ultimately points to, and the path parameter value with the largest value is taken as the output target temperature control strategy. The control parameters of the temperature control strategy are the target temperature value to be adjusted in response to the current temperature change signal and the allowable temperature fluctuation range. The control strategy pointer points to an identifier of a specific control execution method, including PID parameter adjustment, heating power limiting, and cooling water flow compensation; The strategy coordination weight represents the weight set for the specific content executed by the current temperature control strategy.

10. The method for precise temperature control of a flexible mold according to claim 9, characterized in that, Other ways to implement strategy trees include: The current input temperature change signal is used as the root node of the policy tree. The abnormal label, logical region and stage boundary of the temperature change signal are used as branch nodes of the root node in sequence. The temperature value of the temperature change signal and the temperature control policy are used as secondary nodes of the branch nodes. The control policy pointer and policy coordination weight of the temperature control policy are used as leaf nodes to form the policy tree.

Images