A method and system for real-time monitoring of automotive injection mold data

Abstract

This invention provides a method and system for real-time monitoring of automotive injection mold data, relating to the field of data processing technology. The method includes: comparing the incremental changes of each parameter within adjacent sampling time periods based on raw data, and marking parameter points where the changes exceed a preset response threshold; extracting continuous parameter records within the time interval where abnormal changes occur, and generating abnormal interval analysis data based on their change trends; classifying the abnormal interval analysis data into different levels and generating multi-dimensional abnormal description data; retrieving historical normal parameter intervals, performing multi-dimensional feature matching and comparison with them, calculating parameter adjustment amounts, and generating dynamic control suggestions. This invention improves the autonomy and accuracy of real-time monitoring of automotive injection mold data.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and system for real-time monitoring of automotive injection mold data. Background Technology

[0002] In existing technologies, data monitoring for automotive injection molds primarily relies on periodically collecting mold operating parameters and recording relevant data such as temperature, pressure, and injection cycle to a local server or cloud platform using pre-set data acquisition equipment. A typical implementation involves deploying multiple sets of sensors at key locations within the mold, periodically collecting various parameters, and then uploading them uniformly. The backend system then performs statistical analysis on the historical data. Some systems also incorporate alarm thresholds, sending alerts to operators when data anomalies occur. This approach enables basic monitoring of the mold's operating status and provides data support for subsequent maintenance and optimization.

[0003] However, when applying the above technology to actual automotive bumper injection molding production lines, there is a problem of insufficient sensitivity to changes in key parameters. For example, when a partial blockage occurs in the mold's cooling circuit, leading to a decrease in cooling efficiency, the temperature change signal may not be recorded until the next data acquisition cycle, preventing operators from promptly detecting the anomaly. This data monitoring delay may result in multiple consecutive production runs of defective parts, increasing raw material waste and causing further damage to the mold, posing practical risks to quality control and equipment maintenance. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for real-time monitoring of automotive injection mold data, aiming to solve the problems mentioned in the background art.

[0005] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows:

[0006] A first aspect includes a method for real-time monitoring of automotive injection mold data, the method comprising:

[0007] Based on the raw data, the incremental changes of each parameter in adjacent sampling time are compared, and the parameter points where the change exceeds the preset response threshold are marked to obtain the change judgment result.

[0008] Based on the change determination results, continuous parameter records within the time interval of abnormal change are extracted, and abnormal interval analysis data is generated based on the change trend. The abnormal interval analysis data includes information on fluctuation direction, change rate and duration.

[0009] The abnormal interval analysis data is classified and judged. Based on the fluctuation direction, rate of change and duration of abnormal changes, combined with the amplitude of abnormal changes and the mold structure, each abnormal interval is divided into different levels and multi-dimensional abnormal description data is generated.

[0010] Based on the multidimensional abnormal description data, historical normal parameter ranges are retrieved, and multidimensional feature matching and comparison are performed with them to calculate the parameter adjustment amount and generate dynamic control suggestions.

[0011] Preferably, based on the change determination result, continuous parameter records within the time interval of the abnormal change are extracted, and abnormal interval analysis data is generated based on their change trend. The abnormal interval analysis data includes information on fluctuation direction, change rate, and duration, including:

[0012] Based on the change determination results, for each marked parameter point, a preset time window centered on that point is set, and all continuous parameter records within the time window are extracted to form the first abnormal interval dataset.

[0013] Based on the first abnormal interval dataset, the abnormal interval dataset is subjected to boundary correction processing, and the time window range is dynamically adjusted to maximize the continuity of parameter changes and exclude isolated abnormal parameters, thus obtaining the second abnormal interval dataset.

[0014] Based on the second abnormal interval dataset, multi-scale trend decomposition was performed on the data within the abnormal interval. Short-period moving average and long-period moving weighted average methods were used successively to distinguish between local abrupt changes and global trend changes. The fluctuation direction, rate of change and duration information within the interval were calculated respectively to obtain abnormal interval analysis data.

[0015] Preferably, the abnormal interval analysis data is graded and determined. Based on the fluctuation direction, rate of change, and duration of the abnormal changes, combined with the amplitude of the abnormal changes and the structural parts of the mold, each abnormal interval is divided into different levels, and multi-dimensional abnormal description data is generated, including:

[0016] Based on the abnormal interval analysis data, the fluctuation direction, rate of change and duration of each abnormal interval are extracted, and the maximum change within the abnormal interval is counted as the change amplitude of the abnormal change.

[0017] The direction, rate, duration, and magnitude of the abnormal changes are weighted according to preset weighting coefficients to obtain the first abnormality score.

[0018] Based on a pre-set 3D mold structure database, the mold parts are located for each abnormal interval, matched with pre-set importance comparison data to obtain a level correction factor, and multiplied by the first abnormal score value to obtain a second abnormal score value.

[0019] The second anomaly score is compared with the preset grading standard data to determine the anomaly level of the anomaly interval. The anomaly level, fluctuation direction, rate of change, duration, amplitude of change, and mold structure of each anomaly interval are then combined to form multidimensional anomaly description data.

[0020] Preferably, based on the multidimensional anomaly description data, historical normal parameter ranges are retrieved, and multidimensional feature matching and comparison are performed with them to calculate parameter adjustment amounts and generate dynamic control suggestions, including:

[0021] Based on the multidimensional anomaly description data, the anomaly level, fluctuation direction, rate of change, duration, magnitude of change, and mold structure part of each anomaly interval are extracted and used as a retrieval index to retrieve a preset number of historical normal parameter intervals with the highest multidimensional similarity to the current anomaly interval.

[0022] For the selected historical normal parameter range, the parameter adjustment records and output results for each range are statistically analyzed to form a parameter adjustment experience dataset showing the relationship between parameter adjustment amount and output results.

[0023] By jointly analyzing multidimensional anomaly description data and parameter adjustment experience dataset, an adjustment amount optimization objective function is established. The adjustment amount optimization objective function takes the anomaly level, fluctuation direction, change rate, duration, anomaly amplitude and mold structure part of the current anomaly interval as variables, and the historical parameter adjustment amount and its output results as references. An iterative optimization algorithm is used to calculate the parameter adjustment amount of the current anomaly interval.

[0024] Based on the parameter adjustment amount in the current abnormal range, dynamic control suggestions are generated for the current abnormal range. The control suggestions include the parameters to be adjusted, the adjustment range, the adjustment priority, and the mold structure parts to be adjusted.

[0025] Preferably, based on the first abnormal interval dataset, boundary correction processing is performed on the abnormal interval dataset to dynamically adjust the time window range, maximizing the continuity of parameter changes and excluding isolated abnormal parameters, resulting in the second abnormal interval dataset, including:

[0026] Based on the dataset of the first abnormal interval, read the start and end times of each group of continuous parameter records and the corresponding parameter change trend, and use them as windows to adjust the input data;

[0027] Adjust the input data according to the window, move the start and end boundaries of the abnormal interval forward and backward in sequence, and determine whether the change of parameter value within the interval continues to maintain the same direction or exceeds the preset change amplitude threshold. When a reverse change occurs or the fluctuation amplitude is less than the change amplitude threshold, determine the boundary of the abnormal interval and obtain the initially corrected abnormal interval.

[0028] Based on the initially corrected abnormal intervals, the time interval and difference between each parameter and its adjacent parameters are calculated. When the time interval is greater than the preset time threshold or the difference exceeds the preset trend threshold, the parameter is determined to be an isolated abnormal parameter and is removed from the interval.

[0029] Output the second outlier interval dataset after boundary correction and exclusion of isolated outlier parameters.

[0030] Preferably, based on the second abnormal interval dataset, multi-scale trend decomposition is performed on the data within the abnormal interval. Short-period moving average and long-period moving weighted average methods are used sequentially to distinguish between local abrupt changes and global trend changes. The fluctuation direction, rate of change, and duration information within the interval are calculated respectively, resulting in abnormal interval analysis data, including:

[0031] Based on the second anomaly interval dataset, time series parameters for each anomaly interval are selected as input parameters for multi-scale trend decomposition.

[0032] The input parameters are processed by a short-period moving average. The parameter at each time point is averaged with the parameters in its neighborhood to obtain a smoothed short-period data sequence.

[0033] The input parameters are processed by long-period moving weighted average, which calculates the parameter values ​​within a longer time window by weighting them according to a preset time weight, and obtains a smoothed long-period data sequence.

[0034] By comparing and analyzing short-period smoothed sequences with long-period smoothed sequences, and using methods such as difference or amplitude change detection between the two, local rapid abrupt changes are identified, and global trend changes are separated. Furthermore, the fluctuation direction, rate of change, and duration within each interval are calculated to generate abnormal interval analysis data.

[0035] Preferably, multidimensional anomaly description data and parameter adjustment experience dataset are jointly analyzed to establish an adjustment amount optimization objective function. This objective function uses the anomaly level, fluctuation direction, rate of change, duration, anomaly amplitude, and mold structure location of the current anomaly interval as variables, and historical parameter adjustment amounts and their output results as references. An iterative optimization algorithm is used to calculate the parameter adjustment amount for the current anomaly interval, including:

[0036] Based on the parameter adjustment experience dataset, extract historical normal parameter intervals with multidimensional similarity higher than the preset similarity threshold, and extract the parameter adjustment amount and output results corresponding to the interval to form a parameter adjustment candidate set for the current abnormal interval;

[0037] Based on the parameter adjustment candidate set, the characteristic parameter values ​​of the current abnormal interval are compared with the characteristic parameter values ​​in the parameter adjustment candidate set to quantify the improvement of the output results under similar anomalies by different parameter adjustment amounts.

[0038] Using the improvement of output results by historical parameter adjustment amount as the evaluation index, and combining the difference of characteristic parameter values ​​between the current abnormal range and the historical normal parameter range, an optimization objective function for adjustment amount is established. The objective function aims to minimize the difference between abnormal parameters and historical good output parameters or maximize the improvement of output results. The adjustment amount of each parameter is identified as the variable to be optimized.

[0039] An iterative optimization algorithm is used to continuously adjust the parameter adjustment amount. After each adjustment, the objective function is optimized using the adjustment amount to evaluate the expected output performance corresponding to the current parameter adjustment amount. This process continues until the objective function of the adjustment amount converges or reaches the preset convergence criterion, and the parameter adjustment amount that maximizes the output result is output.

[0040] Secondly, a real-time data monitoring system for automotive injection molds, the system comprising:

[0041] The incremental change determination module compares the incremental changes of each parameter in adjacent sampling time based on the original data, marks the parameter points where the change exceeds the preset response threshold, and obtains the change determination result.

[0042] The abnormal interval analysis module is used to extract continuous parameter records within the time interval where abnormal changes occur based on the change judgment results, and generate abnormal interval analysis data based on its change trend. The abnormal interval analysis data includes information on fluctuation direction, change rate and duration.

[0043] The grading and judgment module is used to grade and judge the abnormal interval analysis data. Based on the fluctuation direction, rate of change and duration of abnormal changes, combined with the amplitude of abnormal changes and the mold structure, each abnormal interval is divided into different levels and multi-dimensional abnormal description data is generated.

[0044] The parameter matching and control module is used to retrieve historical normal parameter ranges based on multidimensional abnormal description data, perform multidimensional feature matching and comparison with them, calculate parameter adjustment amounts, and generate dynamic control suggestions.

[0045] The above-described solution of the present invention has at least the following beneficial effects:

[0046] By immediately comparing the incremental changes of each parameter within adjacent sampling times using the collected raw data, the invention effectively overcomes the response lag problem in existing technologies. This invention can pinpoint the location and time interval of abnormal parameter fluctuations in real time, greatly improving the detection speed of abnormal states such as cooling circuit blockages and sudden temperature changes, thus facilitating early warning and preventing the escalation of anomalies and production losses.

[0047] This invention, through trend analysis and multi-dimensional feature classification of detected abnormal intervals, can quantify the direction, rate of change, and duration of abnormal fluctuations. Furthermore, by combining the maximum amplitude of change with specific mold structural locations, it achieves precise determination of the scope and severity of the abnormality's impact. This multi-dimensional description and classification method is far superior to the crude monitoring methods of existing technologies that rely solely on single threshold alarms, providing a data foundation for subsequent precise process adjustments and decisions.

[0048] Furthermore, this invention, by matching and comparing the multidimensional features of abnormal intervals with historical normal parameter intervals, can automatically calculate parameter adjustment amounts based on historical production experience and generate dynamic control suggestions, enabling operators to adjust injection molding process parameters in a targeted manner. This not only effectively shortens the response time from anomaly detection to process intervention but also significantly reduces the defect rate and mold damage risk, improving the overall intelligence level and quality stability of automotive injection molding production. Attached Figure Description

[0049] Figure 1 This is a flowchart of a real-time data monitoring method for automotive injection molds provided by an embodiment of the present invention. Detailed Implementation

[0050] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0051] like Figure 1 As shown, an embodiment of the present invention proposes a method for real-time monitoring of automotive injection mold data, the method comprising:

[0052] Acquire raw data, which are time-series operating parameters collected by sensors installed at multiple locations on the mold during the injection molding process. The operating parameters include temperature data and pressure data.

[0053] Based on the raw data, the incremental changes of each parameter in adjacent sampling time are compared, and the parameter points where the change exceeds the preset response threshold are marked to obtain the change judgment result.

[0054] Based on the change determination results, continuous parameter records within the time interval of abnormal change are extracted, and abnormal interval analysis data is generated based on the change trend. The abnormal interval analysis data includes information on fluctuation direction, change rate and duration.

[0055] The abnormal interval analysis data is classified and judged. Based on the fluctuation direction, rate of change and duration of abnormal changes, combined with the amplitude of abnormal changes and the mold structure, each abnormal interval is divided into different levels and multi-dimensional abnormal description data is generated.

[0056] Based on the multidimensional abnormal description data, historical normal parameter ranges are retrieved, and multidimensional feature matching and comparison are performed with them to calculate the parameter adjustment amount and generate dynamic control suggestions.

[0057] In this embodiment of the invention, the method first installs temperature and pressure sensors at multiple key locations on the injection mold, enabling continuous acquisition of time-series data during the mold's operation throughout the injection molding production cycle. These sensors are preferably distributed in the heat-sensitive and high-pressure zones of the mold to ensure they reflect the main influencing factors of the injection molding process. A dedicated data acquisition device stores the real-time detected temperature and pressure data as raw data along a time axis, ensuring a complete record of the original operational information.

[0058] After the raw data acquisition is completed, the data is input into the data processing module to perform incremental calculations on the changes of each parameter (including temperature and pressure) between any two adjacent sampling time points. The data after each sampling is compared with the previous sampling data to obtain the incremental change of each parameter. Points whose changes exceed a preset response threshold are automatically marked, and the change judgment result is output. This threshold setting can sensitively capture abnormal fluctuations in process parameters, such as abnormal temperature increases caused by abnormal mold cooling, or abnormal points caused by uncontrolled injection pressure.

[0059] Once the change assessment indicates an abnormal change over a certain period, all continuous parameter records within that time interval will be extracted and further combined with historical parameter trends to generate abnormal interval analysis data. This abnormal interval analysis data primarily includes: the main direction of temperature or pressure change within the abnormal interval (i.e., continuous increase or decrease), the rate of change (i.e., the speed of parameter change per unit time), and the total duration of the abnormal change. This information allows for a more detailed differentiation between sudden and chronic abnormal phenomena, effectively assisting in subsequent anomaly classification and response.

[0060] Furthermore, the abnormal interval analysis data is graded based on multiple dimensions, including the direction, rate, and duration of the abnormal changes, combined with the maximum amplitude of parameter changes within the interval and the structural location of the mold when the abnormality occurs. Each abnormal interval is then classified into different levels. For example, a high-amplitude pressure surge occurring in the cooling channel can be classified as a high-level abnormality, requiring priority for alarm activation and process intervention. The grading results are integrated with the abnormal interval analysis data to form multi-dimensional abnormality description data, providing a comprehensive basis for subsequent process control.

[0061] After abnormal intervals are identified and classified, the multidimensional anomaly description data is compared with parameter data from historical normal production cycles. By analyzing the degree of matching between the anomaly description data and historical normal parameter intervals in multidimensional features, the closest historical process conditions can be determined, thereby calculating the required parameter adjustments. Finally, based on the matching results, dynamic control suggestions are automatically generated, clearly indicating the process parameters that need adjustment, the suggested adjustment range, priority, and the affected mold structure parts. This significantly improves the mold anomaly response speed and production stability, and reduces the defect rate and reliance on manual labor.

[0062] In a preferred embodiment of the present invention, based on the change determination result, continuous parameter records within the time interval where abnormal changes occur are extracted, and abnormal interval analysis data is generated based on their change trends. The abnormal interval analysis data includes information on fluctuation direction, change rate, and duration, including:

[0063] Based on the change determination results, for each marked parameter point, a preset time window centered on that point is set, and all continuous parameter records within the time window are extracted to form the first abnormal interval dataset.

[0064] Based on the first abnormal interval dataset, the abnormal interval dataset is subjected to boundary correction processing, and the time window range is dynamically adjusted to maximize the continuity of parameter changes and exclude isolated abnormal parameters, thus obtaining the second abnormal interval dataset.

[0065] Based on the second abnormal interval dataset, multi-scale trend decomposition was performed on the data within the abnormal interval. Short-period moving average and long-period moving weighted average methods were used successively to distinguish between local abrupt changes and global trend changes. The fluctuation direction, rate of change and duration information within the interval were calculated respectively to obtain abnormal interval analysis data.

[0066] In this embodiment of the invention, after detecting an abnormal increase at a certain parameter point, a preset time window is set centered on the abnormal point, and all continuous parameter records within the window are extracted to form a first abnormal interval dataset. The size of this time window can be flexibly set according to the injection molding cycle and sampling frequency, which can both cover the complete abnormal process and take into account data processing efficiency.

[0067] To further ensure the continuity and representativeness of the outlier data intervals, boundary correction is performed on the first outlier interval dataset. The boundary correction process includes: analyzing the trend of parameter values ​​within the window to determine whether the parameter changes maintain a consistent direction or whether the magnitude of the change exceeds a set threshold. When a reverse change or a significant decrease in magnitude is detected at the interval boundary, the window boundary is automatically adjusted, removing discontinuous changes from the interval. Furthermore, the time interval and numerical differences between each parameter record within the interval and its preceding and following records are compared. If the time interval between some points and surrounding records exceeds a preset threshold, or the parameter change exceeds a trend threshold, these points are identified as isolated outliers and removed from the interval data. This ultimately results in a second outlier interval dataset with optimal continuity after removing irrelevant points.

[0068] After obtaining the second anomaly interval dataset, multi-scale trend decomposition was performed on the data. First, the parameter sequence was processed using a short-period moving average method to obtain smoothed data at the local time scale, which was used to filter high-frequency noise and reflect short-term abrupt changes. Next, the same data was processed using a long-period moving weighted average method to extract the overall trend of parameter changes, reflecting slow variables and global trends. By comparing and analyzing the short-period and long-period results, local rapid abrupt changes and global slow changes can be effectively identified. Furthermore, based on the differences between the two, the fluctuation direction, rate of change, and duration information of each segment within the interval were calculated, ultimately outputting accurate and highly resolving anomaly interval analysis data, laying a data foundation for subsequent anomaly identification and intervention.

[0069] In a preferred embodiment of the present invention, the abnormal interval analysis data is classified and determined. Based on the fluctuation direction, rate of change, and duration of the abnormal changes, combined with the amplitude of the abnormal changes and the structural parts of the mold, each abnormal interval is divided into different levels, and multidimensional abnormal description data is generated, including:

[0070] Based on the abnormal interval analysis data, the fluctuation direction, rate of change and duration of each abnormal interval are extracted, and the maximum change within the abnormal interval is counted as the change amplitude of the abnormal change.

[0071] The direction, rate, duration, and magnitude of the abnormal changes are weighted according to preset weighting coefficients to obtain the first abnormality score.

[0072] Based on a pre-set 3D mold structure database, the mold parts are located for each abnormal interval, matched with pre-set importance comparison data to obtain a level correction factor, and multiplied by the first abnormal score value to obtain a second abnormal score value.

[0073] The second anomaly score is compared with the preset grading standard data to determine the anomaly level of the anomaly interval. The anomaly level, fluctuation direction, rate of change, duration, amplitude of change, and mold structure of each anomaly interval are then combined to form multidimensional anomaly description data.

[0074] In this embodiment of the invention, the fluctuation direction, rate of change, and duration corresponding to each abnormal interval are first extracted from the obtained abnormal interval analysis data. The maximum change in temperature or pressure parameters within that interval is then used as the amplitude of the abnormal change. To quantitatively describe the severity of the abnormal interval, the aforementioned features (fluctuation direction, rate of change, duration, and amplitude of change) are weighted according to preset weighting coefficients to obtain a first abnormality score. The weighting coefficients can be adjusted according to actual process requirements and production experience; for example, anomalies with larger pressure changes are given higher weights.

[0075] Subsequently, using the mold's three-dimensional structural database, the sensor location information of the data collected within the abnormal zone is analyzed to locate the mold parts within the abnormal zone. Then, based on the importance comparison data of each structural part of the mold, the level correction factor of each structural part is multiplied by the first abnormality score value to obtain the second abnormality score value, thus fully considering the impact of different mold areas on product quality and safety.

[0076] Finally, the second anomaly score is compared with the grading standard data within different intervals to determine the anomaly level, forming a complete grading judgment. For example, when the score falls into the highest level interval, an alarm is automatically triggered, suggesting immediate inspection of relevant process parameters and mold components. Ultimately, key data such as the anomaly level, fluctuation direction, rate of change, duration, amplitude of change, and mold structure of each anomaly interval are merged to form multidimensional anomaly description data. This multidimensional anomaly description data provides a complete and accurate basis for the precise identification of anomaly intervals, process optimization, and subsequent traceability analysis, and significantly improves the level of intelligence in the injection molding production process.

[0077] The preset weighting coefficients specifically include:

[0078] Based on the actual needs and experience of automotive injection molding processes, different weights are assigned to different parameters in the anomaly analysis data, such as fluctuation direction, rate of change, duration, and amplitude of change. Generally, amplitude and duration of change have a greater impact on the anomaly level and can be assigned higher weights; while fluctuation direction and rate of change are adjusted according to the actual product quality control requirements. The weighting coefficients can be determined after multiple rounds of evaluation based on historical production data, failure cases, and process expert opinions. For example, when scoring anomalies, the system multiplies the actual value of each parameter by its corresponding weight and then sums them to obtain the anomaly score, thus reflecting the contribution ratio of each parameter to the overall anomaly judgment. In practical applications, the weighting coefficients can be optimized and continuously improved on the production floor to adapt to the sensitivity requirements of different mold structures and process parameters.

[0079] The pre-set mold 3D structure database specifically includes:

[0080] A database containing spatial distribution information of various structural parts of the mold is established. This database records the installation location of each sensor and its corresponding physical coordinates, and establishes a mapping relationship with the mold's 3D structural model. This database can be stored in the form of CAD model files, 3D point cloud data, or structural annotation tables. In actual use, the system quickly locates the corresponding mold structural parts (such as cavities, cooling channels, gates, ejector pin areas, etc.) using the collected sensor numbers or coordinate information, achieving a one-to-one correspondence between abnormal intervals and the physical location of the mold. The establishment of this database facilitates subsequent anomaly identification and risk assessment of specific mold structural parts, improving the system's traceability and responsiveness to spatial anomalies in the mold.

[0081] The preset importance comparison data specifically includes:

[0082] For each structural component in the mold's 3D structure database, different importance levels are pre-assessed based on production processes, failure mode analysis, and historical quality data. For example, components with a significant impact on product quality (such as cooling channels and cavity surfaces) are assigned higher importance weights, while components with relatively minor impact (such as auxiliary support areas) are assigned lower weights. Importance comparison data can be stored using weighted lists, grade scores, or key component identifiers. After locating mold components within anomaly zones, the system automatically retrieves the importance comparison data for those components, using it as a grade correction factor to adjust subsequent anomaly scores, ensuring that the final anomaly level assessment more closely reflects actual production risks.

[0083] The pre-set grading standard data specifically includes a series of threshold intervals for classifying anomaly scores. This grading standard data typically includes different score ranges, such as dividing anomaly scores into low-risk, medium-risk, and high-risk ranges. The boundary values ​​for each range can be set based on historical anomaly handling cases, process stability analysis results, and expert recommendations. In practice, after obtaining anomaly scores, the system compares them one by one with the ranges in the grading standard data to determine which range the score falls into, thus establishing the final grade for that anomaly range. This method enables standardized determination and automatic classification of anomaly range grades, improving the standardization and operability of anomaly management.

[0084] In a preferred embodiment of the present invention, based on multidimensional anomaly description data, historical normal parameter ranges are retrieved, and multidimensional feature matching and comparison are performed with them to calculate parameter adjustment amounts and generate dynamic control suggestions, including:

[0085] Based on the multidimensional anomaly description data, the anomaly level, fluctuation direction, rate of change, duration, magnitude of change, and mold structure part of each anomaly interval are extracted and used as a retrieval index to retrieve a preset number of historical normal parameter intervals with the highest multidimensional similarity to the current anomaly interval.

[0086] For the selected historical normal parameter range, the parameter adjustment records and output results for each range are statistically analyzed to form a parameter adjustment experience dataset showing the relationship between parameter adjustment amount and output results.

[0087] By jointly analyzing multidimensional anomaly description data and parameter adjustment experience dataset, an adjustment amount optimization objective function is established. The adjustment amount optimization objective function takes the anomaly level, fluctuation direction, change rate, duration, anomaly amplitude and mold structure part of the current anomaly interval as variables, and the historical parameter adjustment amount and its output results as references. An iterative optimization algorithm is used to calculate the parameter adjustment amount of the current anomaly interval.

[0088] Based on the parameter adjustment amount in the current abnormal range, dynamic control suggestions are generated for the current abnormal range. The control suggestions include the parameters to be adjusted, the adjustment range, the adjustment priority, and the mold structure parts to be adjusted.

[0089] In this embodiment of the invention, dynamic intelligent control of the injection molding process is achieved by matching and analyzing multidimensional anomaly description data with historical normal parameter intervals. Specifically, after detecting graded anomaly intervals during mold operation, multidimensional feature parameters such as the anomaly level, fluctuation direction, rate of change, duration, amplitude of change, and mold structural parts corresponding to each anomaly interval are automatically extracted. These parameters are used as search indexes to compare with the historical normal parameter database, prioritizing the selection of several historical normal parameter intervals most similar to the current anomaly interval. Through this multidimensional similarity retrieval method, combined with historical production experience, the differences between the current anomaly and historically excellent processes can be accurately located.

[0090] After selecting historical normal parameter ranges, further statistical analysis is performed on the parameter adjustment records and output results within these ranges to form a parameter adjustment experience dataset. This experience dataset records in detail the different parameter adjustment methods used historically under similar abnormal operating conditions and their corresponding output results. For example, for high-level anomalies in the cooling channel area, the experience dataset provides indicators such as pass rate and defect rate corresponding to different pressure or temperature adjustment ranges. In this way, not only can the best adjustment suggestions be obtained, but the process database can also be continuously accumulated and optimized.

[0091] Subsequently, using the multidimensional parameters of the current abnormal range as optimization variables and combining historical parameter adjustment experience, an objective function for adjusting the parameters was established, with the process output result as the goal. The optimization objective can be set as minimizing the difference between abnormal parameters and historical optimal output parameters, or maximizing practical indicators such as product qualification rate. An iterative optimization algorithm was used to continuously adjust the variables within the parameter adjustment space, evaluate the impact of different parameter combinations on the objective function in real time, and output the final parameter adjustment amount based on the optimal solution.

[0092] Finally, the system automatically generates dynamic control suggestions, including the parameters to be adjusted (such as specific temperature or pressure points), the adjustment range, the adjustment priority, and the recommended mold structure parts to be adjusted. After operators adjust the process according to the suggestions, they can significantly improve the efficiency of abnormal response, achieve closed-loop optimization and autonomous adaptation of the production process, reduce manual intervention and debugging costs, and improve the overall production yield.

[0093] The methods for calculating multidimensional similarity specifically include:

[0094] When using the multidimensional anomaly description data of the current anomaly interval as input for historical parameter retrieval, the system first pairs each characteristic parameter of the current anomaly interval (such as anomaly level, fluctuation direction, rate of change, duration, amplitude of change, and mold structure location) with the corresponding parameters of historical normal parameter intervals. For each pair of parameters, the system calculates the similarity score of a single parameter by comparing the absolute difference or proportional difference between the current value and the historical value. Subsequently, according to the importance of each parameter for anomaly detection, weights are assigned, and all single parameter similarity scores are weighted and summarized to obtain the final comprehensive similarity score. By setting a similarity threshold, several historical normal parameter intervals that are closest to the current anomaly interval in the multidimensional feature space are selected. This method effectively combines multidimensional factors such as parameter values, change trends, and structural positions to achieve accurate matching between anomaly intervals and historical experience data, facilitating subsequent experience transfer and the generation of control suggestions.

[0095] Specifically, for the selected historical normal parameter range, the parameter adjustment records and output results for each range are statistically analyzed to form a parameter adjustment experience dataset showing the relationship between parameter adjustment amount and output results. This dataset includes:

[0096] The system analyzes each historical normal parameter interval filtered through multi-dimensional similarity, sequentially retrieving the corresponding process parameter adjustment records within each interval, including the parameter value before adjustment, the parameter value after adjustment, and the adjustment time. Simultaneously, it extracts actual output information before and after the parameter adjustment, such as product pass rate, dimensional accuracy, and surface defects. The system pairs and stores the parameter adjustment amount (i.e., the numerical difference before and after the parameter adjustment) with the corresponding output results, forming a dataset showing the relationship between parameter adjustment amount and output results. By summarizing and statistically analyzing the parameter adjustments and output performance across multiple historical intervals, the system can identify the advantages and disadvantages of different adjustment schemes under similar abnormal conditions, providing a historical experience-based reference for control decisions in current abnormal intervals. This experience dataset can be continuously expanded and optimized with accumulated production, gradually improving the system's intelligent control capabilities.

[0097] Specifically, based on the parameter adjustment amount within the current abnormal range, dynamic control suggestions are generated for the current abnormal range, including:

[0098] After completing the optimized calculation of parameter adjustment amounts, the system will obtain a set of optimal parameter adjustment schemes for the current abnormal range. Based on the actual process characteristics of the current abnormal range, the system identifies the specific process parameters that need adjustment (such as the temperature value at a certain temperature measurement point, the pressure value at a certain pressure control point, etc.) and combines historical experience with the current production status to reasonably set the adjustment range. For example, if historical data shows that moderately increasing the cooling channel temperature can significantly reduce the defect rate in this abnormal range, the system will suggest prioritizing temperature adjustment at that location and clearly recommend the target value or range of change. Simultaneously, the system prioritizes each parameter adjustment based on its impact on anomaly correction and process stability, prioritizing adjustments for critically affected areas to reduce ineffective intervention. Finally, the system outputs dynamic control suggestions, including the name of the adjusted parameters, the adjustment range, the adjustment order, and the corresponding mold structure location, to assist operators in quickly and scientifically completing anomaly response and process optimization, effectively improving production yield and equipment operating safety.

[0099] In a preferred embodiment of the present invention, based on the first abnormal interval dataset, boundary correction processing is performed on the abnormal interval dataset, the time window range is dynamically adjusted to maximize the continuity of parameter changes, and isolated abnormal parameters are excluded to obtain a second abnormal interval dataset, including:

[0100] Based on the dataset of the first abnormal interval, read the start and end times of each group of continuous parameter records and the corresponding parameter change trend, and use them as windows to adjust the input data;

[0101] Adjust the input data according to the window, move the start and end boundaries of the abnormal interval forward and backward in sequence, and determine whether the change of parameter value within the interval continues to maintain the same direction or exceeds the preset change amplitude threshold. When a reverse change occurs or the fluctuation amplitude is less than the change amplitude threshold, determine the boundary of the abnormal interval and obtain the initially corrected abnormal interval.

[0102] Based on the initially corrected abnormal intervals, the time interval and difference between each parameter and its adjacent parameters are calculated. When the time interval is greater than the preset time threshold or the difference exceeds the preset trend threshold, the parameter is determined to be an isolated abnormal parameter and is removed from the interval.

[0103] Output the second outlier interval dataset after boundary correction and exclusion of isolated outlier parameters.

[0104] In this embodiment of the invention, the boundary correction and isolated outlier parameter exclusion process for the outlier interval dataset achieves high-quality extraction of outlier interval data. Specifically, the process involves first reading all continuous parameter records within the first outlier interval dataset, including the acquisition time, acquisition order, and trend of each parameter group. By analyzing the continuous change state of the parameter values, the start and end boundaries of the time window are extended forward and backward sequentially, and the direction and magnitude of parameter changes within the window are detected in real time. When a reversal in the direction of parameter change or a change magnitude less than a set threshold is detected at a certain moment, a new interval boundary is automatically determined, ensuring that the final interval can cover the complete outlier change process while avoiding interference from irrelevant data.

[0105] To further improve the purity and representativeness of the data, each parameter in the initially corrected outlier interval was analyzed. By calculating the time interval between the parameter and its adjacent data points and the difference in parameter values, if the sampling time interval between a parameter point and the data points before and after it exceeds a preset time threshold, or if the difference in its value exceeds a preset trend threshold, then that point is identified as an isolated outlier parameter. All parameters identified as isolated outliers are automatically removed from the outlier interval. This effectively prevents false anomalies caused by single-point noise or sampling failures from affecting subsequent analysis.

[0106] After the boundary correction and isolated outlier parameter exclusion steps described above, the final output second outlier interval dataset showed significant improvements in both data continuity and outlier representativeness. Subsequent data analysis and outlier classification were based on this high-quality interval, ensuring a true reflection of outlier trends and providing more reliable data support for multi-scale trend analysis and process adjustments. This high-standard data interval extraction method effectively improves the accuracy of outlier detection and the process stability during injection molding production.

[0107] The preset threshold for the magnitude of change specifically includes:

[0108] When correcting the boundaries of abnormal interval datasets, the system needs to determine whether parameter changes meet the criteria for identifying abnormal intervals. To this end, it is necessary to combine actual injection molding production processes, historical sampling data, and expert experience to set corresponding threshold values ​​for the variation range of key parameters such as temperature and pressure. Generally, the normal fluctuation range of each key parameter during stable mold production can be selected, and the upper limit of this range can be used as the threshold value for determining abnormal variation. For example, for temperature parameters, based on the temperature data sequence collected during normal production, the temperature difference between each two samples can be calculated, and the larger value plus a certain safety margin can be selected as the threshold. In practical applications, during boundary correction, the system continuously reads each parameter record and subtracts each data point from its previous data point. If the difference exceeds the set threshold value, it is determined to be a valid boundary of the abnormal interval; conversely, if the continuous numerical differences do not reach this threshold, it is considered to be within the normal fluctuation range, and no adjustment of the interval boundary is required. This setting method can effectively avoid false alarms caused by normal process fluctuations and ensure accurate and reliable boundary identification of abnormal intervals.

[0109] The preset time threshold specifically includes:

[0110] To ensure sufficient continuity and data quality of the parameter data selected within the abnormal range, the system needs to preset a time threshold to determine whether the time interval between data points is reasonable. This threshold is typically set based on process characteristics such as the sampling frequency of the production equipment, data upload latency, and the duration of the abnormality. For example, if the production equipment collects parameter data once per second, the time threshold is usually set to a range of one to several seconds. During processing, the system compares the collection times of any two adjacent parameter records within the same range. If the sampling time interval between two adjacent points exceeds the preset time threshold, it is considered that data continuity has been broken, possibly due to sensor malfunction or data loss. Such data points are identified as isolated abnormal parameters and removed. By introducing a time threshold judgment mechanism, it is possible to ensure that the remaining parameter records have a uniform time distribution and represent the actual production process, improving the accuracy and effectiveness of subsequent abnormal trend analysis and range correction.

[0111] In a preferred embodiment of the present invention, based on the second abnormal interval dataset, multi-scale trend decomposition is performed on the data within the abnormal interval. Short-period moving average and long-period moving weighted average methods are used sequentially to distinguish between local abrupt changes and global trend changes. The fluctuation direction, rate of change, and duration information within the interval are calculated respectively to obtain abnormal interval analysis data, including:

[0112] Based on the second anomaly interval dataset, time series parameters for each anomaly interval are selected as input parameters for multi-scale trend decomposition.

[0113] The input parameters are processed by a short-period moving average. The parameter at each time point is averaged with the parameters in its neighborhood to obtain a smoothed short-period data sequence.

[0114] The input parameters are processed by long-period moving weighted average, which calculates the parameter values ​​within a longer time window by weighting them according to a preset time weight, and obtains a smoothed long-period data sequence.

[0115] By comparing and analyzing short-period smoothed sequences with long-period smoothed sequences, and using methods such as difference or amplitude change detection between the two, local rapid abrupt changes are identified, and global trend changes are separated. Furthermore, the fluctuation direction, rate of change, and duration within each interval are calculated to generate abnormal interval analysis data.

[0116] In this embodiment of the invention, a multi-scale trend decomposition method for anomaly interval datasets achieves high-resolution feature extraction of anomalous changes. Specifically, firstly, based on the second anomaly interval dataset, time series parameters for each anomaly interval are selected and used as input parameters for multi-scale trend decomposition. By performing short-period moving average processing on the input parameters, the average result of the parameters at each time point and its neighborhood can be obtained, forming a smoothed short-period data sequence that effectively filters out measurement noise and random fluctuations. This step is extremely effective for detecting rapid changes within a short period, such as sudden pressure increases caused by instantaneous mechanical failures during injection molding.

[0117] Furthermore, a long-period moving weighted average is applied to the same input parameter, using a weighted average of parameter values ​​over a longer time window to obtain a smoothed long-period data sequence. This processing can highlight the global trend of the process, such as the gradual increase or decrease of mold temperature due to changes in ambient temperature control.

[0118] Subsequently, short-period smoothed sequences are compared and analyzed with long-period smoothed sequences. By using methods to detect the difference or amplitude changes between the two, local rapid abrupt changes and overall slow trend changes can be accurately identified. By distinguishing between these two types of abnormal characteristics, it is possible to better determine whether the anomaly is caused by instantaneous equipment disturbances or by a gradual deterioration of production conditions, which helps in selecting different response and adjustment strategies.

[0119] Building upon the multi-scale decomposition described above, the fluctuation direction, rate of change, and duration information within each anomaly interval are further calculated. The fluctuation direction reflects whether the parameter is increasing or decreasing, the rate of change quantifies the severity of the anomaly, and the duration reflects the period during which the anomaly persists. Ultimately, the output anomaly interval analysis data can accurately describe the anomaly characteristics from multiple dimensions, providing a comprehensive and high-resolution data foundation for subsequent classification and process optimization. This implementation scheme significantly improves the ability to identify and classify complex anomalies, making production monitoring more targeted and intelligent.

[0120] This involves performing a long-period moving weighted average on the input parameters, weighting the parameter values ​​over a longer time window according to a preset time weight to obtain a smoothed long-period data sequence, specifically including:

[0121] During the analysis of data within anomaly intervals, to accurately grasp the global trend of parameter changes, the system selects a long-term time window and integrates the parameter sample values ​​at each moment and several moments before and after it. Specifically, for each sampling point within this time window, a time weight is assigned based on its proximity to the current moment, sampling order, or other process-related significance. Typically, sampling points closer to the current moment have higher weights, while those farther away have lower weights. Then, the parameter values ​​of all sampling points within the time window are multiplied by their respective weights, and all products are summed and normalized to obtain the weighted average parameter value for the current moment. By continuously moving the time window and repeating the above weighted averaging process, the system generates a smoothed long-term data sequence corresponding to the entire anomaly interval. This sequence effectively reduces short-term disturbances and noise in the parameters, more realistically reflects the long-term trend of the process, and facilitates subsequent trend decomposition and anomaly identification.

[0122] This involves comparing and analyzing short-period smoothed sequences with long-period smoothed sequences, identifying local rapid abrupt changes through methods such as difference or amplitude change detection, and separating global trend changes. Further calculations are performed on the fluctuation direction, rate of change, and duration within each interval to generate abnormal interval analysis data, specifically including:

[0123] After completing short-cycle and long-cycle smoothing, the system first compares the short-cycle smoothed parameter values ​​with the long-cycle smoothed parameter values ​​at the same time point and calculates their numerical differences. If, within a certain time period, the difference between the short-cycle smoothed sequence and the long-cycle smoothed sequence suddenly increases and exceeds a pre-set discrimination threshold, it can be determined that there is a local rapid parameter mutation within this period, usually corresponding to process anomalies or sudden disturbances. When the numerical changes of the two are relatively close, or the difference changes gradually over a long period of time, it indicates that the parameter is in a stable global trend range. Based on this, the system can distinguish and separate local mutations from overall trends. Furthermore, for the detected local mutation ranges and global trend ranges, the main direction of parameter change (i.e., whether the parameter is continuously rising, falling, or fluctuating), the rate of change per unit time, and the duration of the mutation or trend are statistically analyzed, ultimately forming detailed anomaly range analysis data. This data provides a scientific basis for subsequent classification judgments and process control recommendations, effectively improving the accuracy and real-time performance of anomaly diagnosis.

[0124] In a preferred embodiment of the present invention, multidimensional anomaly description data and parameter adjustment experience dataset are jointly analyzed to establish an adjustment amount optimization objective function. The adjustment amount optimization objective function uses the anomaly level, fluctuation direction, rate of change, duration, anomaly amplitude, and mold structure location of the current anomaly interval as variables, and historical parameter adjustment amounts and their output results as references. An iterative optimization algorithm is used to calculate the parameter adjustment amount for the current anomaly interval, including:

[0125] Based on the parameter adjustment experience dataset, extract historical normal parameter intervals with multidimensional similarity higher than the preset similarity threshold, and extract the parameter adjustment amount and output results corresponding to the interval to form a parameter adjustment candidate set for the current abnormal interval;

[0126] Based on the parameter adjustment candidate set, the characteristic parameter values ​​of the current abnormal interval are compared with the characteristic parameter values ​​in the parameter adjustment candidate set to quantify the improvement of the output results under similar anomalies by different parameter adjustment amounts.

[0127] Using the improvement of output results by historical parameter adjustment amount as the evaluation index, and combining the difference of characteristic parameter values ​​between the current abnormal range and the historical normal parameter range, an optimization objective function for adjustment amount is established. The objective function aims to minimize the difference between abnormal parameters and historical good output parameters or maximize the improvement of output results. The adjustment amount of each parameter is identified as the variable to be optimized.

[0128] An iterative optimization algorithm is used to continuously adjust the parameter adjustment amount. After each adjustment, the objective function is optimized using the adjustment amount to evaluate the expected output performance corresponding to the current parameter adjustment amount. This process continues until the objective function of the adjustment amount converges or reaches the preset convergence criterion, and the parameter adjustment amount that maximizes the output result is output.

[0129] In this embodiment of the invention, the intelligent optimization process for parameter adjustment amounts in abnormal intervals achieves a closed-loop capability for adaptive parameter control. Specifically, the process involves first retrieving historical normal parameter intervals from the parameter adjustment experience dataset whose multidimensional similarity exceeds a preset threshold using multidimensional anomaly description data. Then, the parameter adjustment amounts and output results corresponding to these historical intervals are extracted to form a candidate set of parameter adjustments highly correlated with the current abnormal interval, providing data support for subsequent parameter optimization.

[0130] Next, a point-by-point difference analysis is performed between the characteristic parameter values ​​of the current abnormal interval and the historical characteristic parameter values ​​in the parameter adjustment candidate set to quantify the improvement of the output results under different parameter adjustment amounts. By comparison, it is possible to identify which parameter adjustment combinations have historically been more helpful in improving product quality or reducing defect rates.

[0131] Based on the above quantitative analysis, the improvement in output results caused by historical parameter adjustments is used as the evaluation index. Combining the differences in characteristic parameter values ​​between the current abnormal range and the historical normal parameter range, an optimization objective function for the adjustment amount is established. This objective function explicitly aims to minimize the difference between abnormal parameters and historically excellent output parameters, or to maximize the improvement in output results. In this case, the adjustment amounts of each parameter are input into the objective function as variables to be optimized.

[0132] Subsequently, an iterative optimization algorithm is employed to continuously adjust the parameter values. After each adjustment, the objective function is used to evaluate the adjustment effect in real time, continuously optimizing the adjustment strategy. When the objective function converges or reaches a preset convergence criterion, the parameter adjustment value that maximizes the output result is output. Finally, by combining the optimal adjustment value with multi-dimensional anomaly descriptions, precise dynamic control suggestions can be provided to operators or automated devices, enabling intelligent adaptive adjustment of the production process. This method effectively improves anomaly response speed and production yield, reduces trial and error and manual intervention, and has significant industrial application value and promising prospects for widespread adoption.

[0133] Specifically, based on the parameter adjustment candidate set, a difference analysis is performed on each characteristic parameter value of the current anomaly interval and the characteristic parameter values ​​in the parameter adjustment candidate set to quantify the improvement of the output results under similar anomalies. This includes:

[0134] The system first selects several historical parameter records from the parameter adjustment candidate set that have the highest similarity to the current abnormal interval. Each historical record includes the characteristic parameter value before adjustment, the corresponding parameter adjustment amount, and the output result after adjustment. The system compares the characteristic parameter values ​​of the current abnormal interval (such as abnormality level, fluctuation direction, rate of change, duration, abnormality amplitude, and mold structure location) with the same parameters in the historical records, and calculates the numerical or categorical differences between the two. Through this item-by-item comparison, the system can clarify the similarity between various features and the current abnormality under different historical parameter adjustment amounts, and, combined with historical output data, statistically analyze the improvement of output results for each parameter adjustment amount under similar abnormal operating conditions. For example, when historical data shows that a certain adjustment scheme can significantly improve the yield rate, the system correlates its corresponding parameter adjustment amount with the output improvement magnitude, providing basic data support for subsequent optimization.

[0135] Among them, the improvement of the output results by the historical parameter adjustment amount is used as the evaluation index. Combining the difference of characteristic parameter values ​​between the current abnormal range and the historical normal parameter range, an optimization objective function for the adjustment amount is established. The objective function aims to minimize the difference between the abnormal parameters and the historical excellent output parameters or maximize the improvement of the output results. The adjustment amount of each parameter is clearly defined as the variable to be optimized, specifically including:

[0136] After analyzing the differences between the current abnormal range and each historical parameter adjustment candidate record, the system defines an optimization objective based on the improvement magnitude of historical output results. This objective can be either to bring the current abnormal parameter value closer to the parameter value in the historical excellent output range (i.e., minimize the difference), or to achieve the optimal level of the adjusted process output (i.e., maximize yield, reduce defect rate, etc.). To achieve this objective, the system sets different parameter adjustment amounts (such as the adjustment range of temperature, pressure, etc.) as variables to be adjusted in the optimization process. In actual calculations, the system sequentially selects different combinations of parameter adjustment amounts, predicts the impact of each combination on process output performance, and calculates the value of the objective function. Through this iterative process, the system can filter out parameter adjustment directions that better match the actual process and significantly improve output.

[0137] The algorithm employs an iterative optimization approach, continuously adjusting the parameter adjustments. After each adjustment, the objective function is optimized using the adjusted parameter to evaluate the expected output performance corresponding to the current parameter adjustment, until the objective function converges or reaches a preset convergence criterion. The resulting parameter adjustment maximizes the output, specifically including:

[0138] Based on the aforementioned objective function and optimization variable settings, the system initiates an iterative optimization process. In each iteration, the system fine-tunes the parameter adjustments based on the optimization results of the previous round (using existing industrial optimization methods such as step adjustment, gradient search, simulated annealing, and genetic strategies), and re-evaluates the impact of the adjusted process parameter combination on the objective function, i.e., simulating or predicting the expected output. When the objective function value is found to stabilize, meaning that subsequent adjustments can no longer significantly improve output performance, or the objective function has reached the preset convergence criteria (e.g., the target improvement is below a threshold, or a set number of iterations has been reached), the system terminates the iteration process and outputs the set of parameter adjustments that yielded the best output throughout the entire optimization process. Through this method, the system can intelligently and efficiently determine the optimal process parameter adjustment scheme for the current abnormal range, achieving adaptive optimization of the production process and maximizing the yield rate.

[0139] Embodiments of the present invention also provide a real-time data monitoring system for automotive injection molds, the system comprising:

[0140] The data acquisition module is used to acquire raw data, which are time-series operating parameters collected by sensors installed at multiple locations on the mold during the injection molding process. The operating parameters include temperature data and pressure data.

[0141] The incremental change determination module compares the incremental changes of each parameter in adjacent sampling time based on the original data, marks the parameter points where the change exceeds the preset response threshold, and obtains the change determination result.

[0142] The abnormal interval analysis module is used to extract continuous parameter records within the time interval where abnormal changes occur based on the change judgment results, and generate abnormal interval analysis data based on its change trend. The abnormal interval analysis data includes information on fluctuation direction, change rate and duration.

[0143] The grading and judgment module is used to grade and judge the abnormal interval analysis data. Based on the fluctuation direction, rate of change and duration of abnormal changes, combined with the amplitude of abnormal changes and the mold structure, each abnormal interval is divided into different levels and multi-dimensional abnormal description data is generated.

[0144] The parameter matching and control module is used to retrieve historical normal parameter ranges based on multidimensional abnormal description data, perform multidimensional feature matching and comparison with them, calculate parameter adjustment amounts, and generate dynamic control suggestions.

[0145] It should be noted that this system is a system corresponding to the above method. All implementation methods in the above method embodiments are applicable to this embodiment and can achieve the same technical effect.

[0146] Embodiments of the present invention also provide a computing device, including: a processor and a memory storing a computer program, wherein the computer program, when executed by the processor, performs the method described above. All implementations in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0147] Embodiments of the present invention also provide a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the method described above. All implementations in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0148] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A real-time monitoring method for automobile injection mold data, characterized in that, The method comprises: According to the original data, the incremental changes of each parameter within adjacent sampling times are compared, and parameter points with changes exceeding a preset response threshold are marked to obtain a change determination result; According to the change determination result, continuous parameter records within a time interval where abnormal changes occur are extracted, and abnormal interval analysis data are generated according to the change trend, wherein the abnormal interval analysis data include fluctuation direction, change rate and duration information; The abnormal interval analysis data are subjected to hierarchical determination, different grades are divided for each abnormal interval based on the fluctuation direction, change rate and duration of abnormal changes, and multi-dimensional abnormal description data are generated, including: According to the abnormal interval analysis data, the fluctuation direction, change rate and duration of each abnormal interval are extracted, and the maximum change within the abnormal interval is counted as the change amplitude of abnormal changes; The fluctuation direction, change rate, duration and change amplitude of abnormal changes are weighted according to preset weight coefficients to obtain a first abnormal score value; Based on a preset mold three-dimensional structure database, the mold part of each abnormal interval is positioned, matched with preset importance comparison data to obtain a grade correction factor, and multiplied with the first abnormal score value to obtain a second abnormal score value; The second abnormal score value is compared with preset hierarchical standard data to determine the abnormal grade of the abnormal interval, and the abnormal grade, fluctuation direction, change rate, duration and change amplitude of each abnormal interval are combined to form multi-dimensional abnormal description data. According to the change determination result, continuous parameter records within a time interval where abnormal changes occur are extracted, and abnormal interval analysis data are generated according to the change trend, wherein the abnormal interval analysis data include fluctuation direction, change rate and duration information, including: According to the change determination result, for each marked parameter point, a preset time window centered on the point is set, all continuous parameter records within the time window are extracted to form a first abnormal interval data set; 2. The real-time monitoring method of automobile injection mold data according to claim 1, characterized in that, According to the first abnormal interval data set, boundary correction processing is performed on the abnormal interval data set, the time window range is dynamically adjusted to maximize the parameter change continuity, and isolated abnormal parameters are excluded to obtain a second abnormal interval data set; According to the second abnormal interval data set, the data within the abnormal interval are subjected to multi-scale trend decomposition, short-period moving average and long-period moving weighted average methods are used in sequence to distinguish local mutation changes and global trend changes, and the fluctuation direction, change rate and duration information within the interval are calculated respectively to obtain the abnormal interval analysis data. ​ ​ 3. The method of claim 1, wherein, According to the multi-dimensional abnormal description data, the historical normal parameter interval is retrieved, and multi-dimensional feature matching and comparison are performed, the parameter adjustment amount is calculated, and the dynamic control suggestion is generated, including: According to the multi-dimensional abnormal description data, the abnormal level, fluctuation direction, change rate, duration, change amplitude and mold structure part of each abnormal interval are extracted as retrieval indexes, and the historical normal parameter interval with the highest multi-dimensional similarity to the current abnormal interval is retrieved; For the selected historical normal parameter interval, the parameter adjustment record and output result of each interval are counted to form the parameter adjustment experience data set of the relationship between the parameter adjustment amount and the output result; The multi-dimensional abnormal description data and the parameter adjustment experience data set are jointly analyzed, the adjustment amount optimization objective function is established, the adjustment amount optimization objective function takes the abnormal level, fluctuation direction, change rate, duration, abnormal amplitude and mold structure part of the current abnormal interval as variables, and the historical parameter adjustment amount and its output result as reference, and the iterative optimization algorithm is adopted to calculate the parameter adjustment amount of the current abnormal interval; According to the parameter adjustment amount of the current abnormal interval, the dynamic control suggestion for the current abnormal interval is generated, and the control suggestion includes the parameters to be adjusted, the adjustment amplitude, the adjustment priority and the suggested adjustment mold structure part.

4. The real-time monitoring method of automobile injection mold data according to claim 2, characterized in that, According to the first abnormal interval data set, the boundary correction processing is performed on the abnormal interval data set, the time window range is dynamically adjusted to maximize the parameter change continuity, and the isolated abnormal parameters are excluded to obtain the second abnormal interval data set, including: According to the first abnormal interval data set, the start and end time of each set of continuous parameter records and the corresponding parameter change trend are read as window adjustment input data; According to the window adjustment input data, the start and end boundaries of the abnormal interval are moved forward and backward in turn, and it is judged whether the change of the parameter value in the interval continuously remains in the same direction or exceeds the preset change amplitude threshold value, when the reverse change or the fluctuation amplitude is less than the change amplitude threshold value, the boundary of the abnormal interval is determined, and the preliminary corrected abnormal interval is obtained; According to the preliminary corrected abnormal interval, the time interval and difference value between each parameter and the adjacent parameter are calculated, when the time interval is greater than the preset time threshold value or the difference value exceeds the preset trend threshold value, the parameter is determined as an isolated abnormal parameter, and it is excluded from the interval; The second abnormal interval data set formed after the boundary correction and the exclusion of the isolated abnormal parameters is output.

5. The method of claim 2, wherein the method further comprises: According to the second abnormal interval data set, the data in the abnormal interval is subjected to multi-scale trend decomposition, short-term moving average and long-term moving weighted average methods are adopted in turn to distinguish local mutation change and global trend change, and the fluctuation direction, change rate and duration information in the interval are calculated respectively to obtain the abnormal interval analysis data, including: According to the second abnormal interval data set, the time sequence parameters of each abnormal interval are selected as the input parameters of multi-scale trend decomposition; The input parameters are subjected to short-term moving average processing, each time point parameter and the parameters in its neighborhood are subjected to average operation to obtain the smoothed short-term data sequence; The input parameters are subjected to long-period moving weighted average processing, and the parameter values in a long time window are subjected to weighted average processing according to preset time weights, so as to obtain a smoothed long-period data sequence; The short-period smoothed sequence and the long-period smoothed sequence are compared and analyzed, local rapid mutations are recognized through difference or amplitude change detection, and global trend changes are separated, the fluctuation direction, change rate and duration in each interval are further calculated, and abnormal interval analysis data are generated.

6. The real-time monitoring method of an automobile injection mold data according to claim 3, characterized in that, The multi-dimensional abnormal description data and the parameter adjustment experience data set are jointly analyzed, an adjustment amount optimization objective function is established, the adjustment amount optimization objective function takes the abnormal level, fluctuation direction, change rate, duration, abnormal amplitude and mold structure part of the current abnormal interval as variables, takes the historical parameter adjustment amount and its output result as a reference, and adopts an iterative optimization algorithm to calculate the parameter adjustment amount of the current abnormal interval, including: According to the parameter adjustment experience data set, historical normal parameter intervals with a multi-dimensional similarity higher than a preset similarity threshold are extracted, and the parameter adjustment amount and its output result corresponding to the interval are extracted to form a parameter adjustment candidate set for the current abnormal interval; According to the parameter adjustment candidate set, the characteristic parameter values of the current abnormal interval and the characteristic parameter values in the parameter adjustment candidate set are subjected to item-by-item difference analysis, and the improvement amplitude of different parameter adjustment amounts on the output result under similar abnormalities is quantified; Taking the improvement amplitude of the historical parameter adjustment amount on the output result as an evaluation index, and combining the difference between the characteristic parameter values of the current abnormal interval and the historical normal parameter interval, an adjustment amount optimization objective function is established, the objective function takes the minimization of the difference between the abnormal parameters and the historical excellent output parameters or the maximization of the output result improvement as the optimization objective, and clearly defines each parameter adjustment amount as a to-be-optimized variable; An iterative optimization algorithm is adopted to continuously adjust the parameter adjustment amount, and the adjustment amount optimization objective function is used to evaluate the expected output performance corresponding to the current parameter adjustment amount after each adjustment, until the adjustment amount optimization objective function converges or reaches a preset convergence standard, and the parameter adjustment amount that can maximize the output result is output.

7. A real-time monitoring system for automotive injection mold data, characterized in that, The system is applied to the method of any one of claims 1-6, and the system comprises: An incremental change determination module compares the incremental changes of each parameter within adjacent sampling times according to the original data, and marks the parameter points whose changes exceed a preset response threshold to obtain a change determination result; An abnormal interval analysis module extracts continuous parameter records in a time interval where abnormal changes occur according to the change determination result, and generates abnormal interval analysis data according to the change trend, wherein the abnormal interval analysis data includes fluctuation direction, change rate and duration information; A grading determination module performs grading determination on the abnormal interval analysis data, divides each abnormal interval into different grades based on the fluctuation direction, change rate and duration of the abnormal changes, and generates multi-dimensional abnormal description data in combination with the amplitude of the abnormal changes and the mold structure part. The parameter matching and regulation module is used for searching a historical normal parameter interval according to the multi-dimensional abnormal description data, performing multi-dimensional feature matching and comparison with the historical normal parameter interval, calculating a parameter adjustment amount, and generating a dynamic regulation suggestion.

8. A computing device, comprising: The method comprises the following steps: one or more processors; a memory device for storing one or more programs, when the one or more programs are executed by the one or more processors, so that the one or more processors implement the method as claimed in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a program, and the program is executed by the processor to implement the method as claimed in any one of claims 1 to 6.

Images