Vacuum suction self-adaptive control method and system for negative pressure hopper of injection molding machine

Abstract

The present application relates to injection molding machine equipment technical field, more specifically, the present application relates to vacuum suction self-adaptive control method and system of negative pressure hopper for injection molding machine, method includes: obtaining the pressure time sequence of negative pressure hopper completing a job;The pressure time sequence is divided into three stages;History database is constructed, and the weight of three stages is obtained;The cosine similarity between the pressure time sequence corresponding to any job in the database and the pressure time sequence corresponding to the current job is calculated, and the reference value of the current job based on historical job is obtained based on the confidence weighted average of three stages;The historical reference job that is consistent with the current job mode is screened out, and the vacuum loss is calculated, and according to the comparison result of the vacuum loss of the current job and the set threshold, the vacuum pump power is adjusted.The present application can timely and accurately adjust the vacuum pump power according to the actual operation condition, and effectively deal with the sudden problems such as vacuum leakage.

Description

Technical Field

[0001] This invention relates to the field of injection molding machine technology. More specifically, this invention relates to a vacuum extraction adaptive control method and system for a negative pressure hopper in an injection molding machine. Background Technology

[0002] The vacuum hopper of an injection molding machine uses vacuum suction to convey granular materials in a sealed manner. The stability of the vacuum level plays a decisive role in feeding efficiency and workshop cleanliness. However, in actual operation, the inner wall of the hopper will wear down due to long-term friction, and the particle size and moisture content of the raw materials are uneven. These factors cause the vacuum pressure curve to exhibit a three-stage characteristic of "sudden drop-stabilization-rebound". Moreover, the sensitivity to seal failure varies in different stages. Traditional fixed threshold alarm methods have limitations; they are difficult to distinguish between normal aging of the hopper and sudden leakage, and they cannot automatically adjust the reference weight according to the characteristics of different machine models.

[0003] Therefore, there is an urgent need for an intelligent method that can adaptively assess the degree of vacuum failure by combining historical operation data and adjust the pump power online. Summary of the Invention

[0004] To address the technical problem that traditional fixed threshold alarm methods have limitations due to the specific characteristics of the vacuum pressure curve of the negative pressure hopper of the injection molding machine caused by various factors and the different sensitivities to seal failure at different stages, the present invention provides solutions in the following aspects.

[0005] In the first aspect, the vacuum pumping adaptive control method for the negative pressure hopper of the injection molding machine includes:

[0006] The pressure of the negative pressure hopper at each moment of completing one operation is collected and preprocessed to obtain a pressure time series. All pressure points in the pressure time series are traversed, and any two pressure points are selected as a set of dividing points. Each set of dividing points corresponds to a calculated dividing response value. When the dividing response value is the smallest, the corresponding set of dividing points is the optimal dividing point. The pressure time series is divided into three stages according to the optimal dividing point.

[0007] A historical database is constructed, which records the pressure of all negative pressure hoppers of the same model as the negative pressure hopper at each stage under different operation numbers. The confidence scores of the three stages of the database are calculated respectively. The confidence scores of the three stages of the database are weighted by summing them into one to obtain the weights of each of the three stages. The cosine similarity between the pressure time series corresponding to any operation in the database and the pressure time series corresponding to the current operation is calculated. The reference value of the current operation based on the historical operations is obtained by weighting the confidence scores of the three stages.

[0008] All calculated reference values ​​are sorted and categorized in descending order to obtain historical reference operations that match the current operation mode. Vacuum loss is calculated based on historical reference operations, and the vacuum pump power is adjusted according to the comparison between the vacuum loss of the current operation and the set threshold.

[0009] Preferably, when the vacuum loss is greater than the threshold, it is determined that there is a vacuum leak in the current operation, and the vacuum pump power of the negative pressure hopper is linearly increased until the vacuum loss is lower than the threshold.

[0010] Preferably, if the vacuum loss is still greater than the threshold when the vacuum pump power reaches the maximum power, it indicates that the negative pressure hopper is severely damaged and cannot meet the vacuum operation conditions. In this case, the operation should be stopped immediately for investigation.

[0011] Preferably, the process of obtaining the segmentation response value includes:

[0012] Calculate the coefficient of variation of the subsequence before the first division point, the coefficient of variation of the subsequence between the first and second division points, and the coefficient of variation of the subsequence after the second division point after dividing the pressure time series using two arbitrarily selected pressure points.

[0013] The product of the three calculated coefficients of variation is used as the segmentation response value.

[0014] Preferably, the confidence level acquisition process in the first stage includes:

[0015] Two assignments are randomly selected from the database as the first and second assignments, respectively.

[0016] Calculate the Pearson correlation coefficient between the first and second tasks in the first stage of stress; calculate the mean relative difference in stress between the first and second tasks in the first stage of stress.

[0017] The contribution values ​​of the first and second assignments are obtained by multiplying the calculated Pearson correlation coefficient by the relative difference of the mean.

[0018] By combining all job pairs, repeat the above contribution value operation and multiply all contribution values ​​consecutively to obtain the confidence level for the first stage;

[0019] The confidence levels for the second and third stages are calculated in the same way as those for the first stage.

[0020] Preferably, the confidence level acquisition process in the first stage includes:

[0021] Two assignments are randomly selected from the database as the first and second assignments, respectively.

[0022] Calculate the shape similarity of the pressure in the first stage between the first and second tasks, which is obtained using DTW distance calculation; calculate the mean relative difference of the pressure in the first stage between the first and second tasks.

[0023] Multiply the calculated shape similarity by the mean relative difference to obtain the contribution values ​​of the first and second assignments;

[0024] By combining all job pairs, repeat the above contribution value operation and multiply all contribution values ​​consecutively to obtain the confidence level for the first stage;

[0025] The confidence levels for the second and third stages are calculated in the same way as those for the first stage.

[0026] Preferably, the classification operation includes:

[0027] After sorting all the calculated reference values ​​in descending order, the Otsu method is used to classify the sorted reference value sequence, and the historical jobs corresponding to the first category of reference values ​​are taken as the historical reference jobs that match the current job mode.

[0028] Preferably, the classification operation includes:

[0029] After sorting all the calculated reference values ​​in descending order, the classification response value is calculated starting from the second element in the sorted reference value sequence. That is, any element is selected from all elements after the second element as a candidate classification point, and the value of the second element is divided by the mean of the classification sequence formed by the first element and the candidate classification point as the classification response value of the candidate classification point.

[0030] Traverse all candidate classification points. When a candidate classification point whose classification response value is less than or equal to the preset classification threshold is taken as the final classification point, all historical jobs corresponding to the classification sequence formed by the first element and the final classification point are the historical reference jobs that match the current job mode.

[0031] Preferably, the vacuum loss is the average of the historical reference values ​​of all historical reference jobs selected for the current job.

[0032] Secondly, a vacuum extraction adaptive control system for a negative pressure hopper of an injection molding machine includes: a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the vacuum extraction adaptive control method for a negative pressure hopper of an injection molding machine as described in any one of the claims is implemented.

[0033] The beneficial effects of this invention are:

[0034] First, the pressure at each moment of the negative pressure hopper operation is collected and preprocessed into a time series. The optimal segmentation point is selected and the hopper is divided into three stages. The characteristics of pressure curve changes are fully considered to accurately identify the operation stage and avoid the problem of unclear stage characteristics in traditional methods.

[0035] Secondly, a historical database is constructed to record the pressure at each stage of different operations for the same model of negative pressure hopper. The confidence levels of the three stages are calculated and weighted to obtain the weights, reflecting the importance of each stage and its impact on seal failure. This makes the reference value calculation more scientific and reasonable, and improves the accuracy of judging the operation status. Then, the cosine similarity of the pressure time series of the current operation and the historical operation is calculated. The reference value is obtained based on the weighted average of the stage confidence levels. By comprehensively considering the similarity and stage importance, a reliable basis for judging vacuum leakage is provided.

[0036] Finally, the reference values ​​are sorted and categorized to obtain historical reference operations that match the current operation mode. The most similar historical operations are selected, and their data is of significant reference value for assessing vacuum loss and adjusting vacuum pump power. Vacuum loss is calculated based on historical reference operations, and the vacuum pump power is adjusted according to the comparison with a set threshold. If the loss exceeds the threshold, it is linearly increased until it falls below the threshold. If the loss still exceeds the threshold even after reaching the maximum power, the operation is stopped for investigation. This adaptive control method can adjust the power in a timely manner to deal with vacuum leakage and improve operational stability and reliability. Attached Figure Description

[0037] Figure 1 This is a flowchart of steps S1-S3 in the vacuum pumping adaptive control method for the negative pressure hopper of the injection molding machine according to an embodiment of the present invention.

[0038] Figure 2 This is a schematic diagram of the vacuum pumping adaptive control system for the negative pressure hopper of the injection molding machine according to an embodiment of the present invention. Detailed Implementation

[0039] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0040] Reference Figure 1 The vacuum pumping adaptive control method for the negative pressure hopper of the injection molding machine includes steps S1-S3, as follows:

[0041] S1: Collect the pressure at each moment when the negative pressure hopper completes one operation, and preprocess it to obtain the pressure time series; traverse all pressure points in the pressure time series, select any two pressure points as a set of dividing points, each set of dividing points corresponds to a calculated dividing response value, when the dividing response value is the smallest, the corresponding set of dividing points is the optimal dividing point, and divide the pressure time series into three stages according to the optimal dividing point.

[0042] In one embodiment, during the process of a negative pressure hopper completing a full operation (such as feeding, vacuuming, discharging, etc.), the pressure values ​​at each moment are recorded in chronological order to form the original pressure time series.

[0043] The original pressure time series was cleaned and standardized, including preprocessing operations such as noise reduction, filling missing values ​​or unifying sampling intervals, normalization, and smoothing, to finally obtain the pressure time series.

[0044] Furthermore, two pressure points are randomly selected from the pressure time series as a set of dividing points, and the pressure time series is divided into three subsequences: the first subsequence is from the beginning to the first dividing point (excluding the point), the second subsequence is between the two dividing points (including the first and second dividing points), and the third subsequence is from the second dividing point (excluding the point) to the end.

[0045] For the first subsequence, the ratio of the standard deviation to the mean of the subsequence is calculated as the coefficient of variation of the first subsequence. The coefficients of variation for the second and third subsequences are calculated similarly based on the calculation of the coefficient of variation for the first subsequence. The product of the three calculated coefficients of variation is used as the segmented response value.

[0046] By iterating through all the group split points, the group split point corresponding to the smallest split response value is taken as the optimal split point, thus dividing the pressure time series into three stages.

[0047] The first stage corresponds to the start-up stage (such as the hopper starting to be vacuumed and the pressure dropping rapidly), the second stage corresponds to the stable operation stage (such as the pressure remaining constant and the operation running efficiently), and the third stage corresponds to the end stage (such as the pressure rising, the hopper discharging material or shutting down).

[0048] The above segmentation allows for the quantification of characteristics such as the rate of pressure change and duration at each stage. This helps to gain a deeper understanding of the operating patterns and characteristics of the negative pressure hopper in different operational stages, providing a basis for optimizing work processes and improving operational efficiency.

[0049] S2: Construct a historical database that records the pressure of all negative pressure hoppers of the same model as the negative pressure hopper at each stage under different operation numbers. Calculate the confidence level of the database for each of the three stages, and sum the confidence levels of the three stages to obtain the weights of each stage. Calculate the cosine similarity between the pressure time series corresponding to any operation in the database and the pressure time series corresponding to the current operation, and obtain the reference value of the current operation based on the historical operations by weighted averaging the confidence levels of the three stages.

[0050] In one embodiment, the pressure of all negative pressure hoppers of the same model as the negative pressure hopper is collected at various stages under different number of operations, thereby constructing a centralized and systematic historical database.

[0051] Pressure data from negative pressure hoppers at different operating times can be affected by various factors (such as equipment aging, changes in operating conditions, and measurement errors), leading to discrepancies between data points. Low data consistency can result in misjudgments if historical data is used directly as a benchmark. Therefore, by calculating confidence levels, the reliability of pressure data from all operating times in the historical database is quantified, providing a basis for subsequent equipment status adjustments.

[0052] Taking the first stage as an example, in one embodiment, a confidence level calculation based on the Pearson correlation coefficient is provided. Specifically:

[0053] Two tasks are randomly selected from the database as the first task and the second task, respectively. The Pearson correlation coefficient between the first task and the second task in the first stage of stress is calculated. The mean relative difference in stress between the first task and the second task in the first stage is calculated. The calculated Pearson correlation coefficient is multiplied by the mean relative difference to obtain the contribution value of the first task and the second task. The above contribution value operation is repeated for all task pairs, and all contribution values ​​are multiplied continuously to obtain the confidence level of the first stage.

[0054] The specific calculation formula for the average relative difference in pressure during the first stage between the first and second tasks mentioned above is provided as an example:

[0055]

[0056] In the formula, The average relative difference in pressure during the first stage mentioned above. The average of all pressure values ​​in the first stage of the selected first task. This represents the average of all pressure values ​​in the first stage of the selected second operation. The calculated relative difference between the average and the mean is used to address the limitation of the Pearson correlation coefficient, which fails to express the range of similarity in data. A larger product indicates a more consistent upward trend in pressure across the injection molding machine within the same time frame during the first stage.

[0057] It should be noted that, for example, if the first task selected contains only 50 pressure points and the second task selected contains 70 pressure points, then only the first 50 pressure points are selected for calculation in the second task.

[0058] In another embodiment, a confidence calculation based on DTW (Dynamic Time Warping) is provided. Specifically:

[0059] Two tasks are randomly selected from the database as Task 1 and Task 2, respectively. The shape similarity of the pressure in the first stage between Task 1 and Task 2 is calculated using the DTW distance. The mean relative difference of the pressure in the first stage between Task 1 and Task 2 is calculated. The calculated shape similarity is multiplied by the mean relative difference to obtain the contribution value of Task 1 and Task 2. This process is repeated for all task pairs, and all contribution values ​​are multiplied consecutively to obtain the confidence level for the first stage. The calculation method for the mean relative difference here is the same as that used in the confidence level calculation based on the Pearson correlation coefficient.

[0060] The formula for calculating the shape similarity mentioned above is given as an example:

[0061]

[0062] In the formula, For shape similarity, The pressure vector for the first task selected in the first stage. The pressure vector for the selected second task in the first stage. This represents an exponential function with the natural number e as its base.

[0063] The above approach solves the computational error problem caused by the difference in vector length by introducing DTW.

[0064] The confidence levels for the second and third stages are obtained using the same method as for the first stage. Then, the confidence levels from the three stages are summed and weighted to obtain the weights for the first, second, and third stages, respectively. This allows for the reasonable allocation of weights based on the importance of each stage in the overall task.

[0065] Then, the cosine similarity between the pressure time series corresponding to any operation in the database and the pressure time series corresponding to the current operation is calculated, and the reference value of the current operation based on the historical operations is obtained by weighted averaging the confidence levels of the three stages.

[0066] The above operations can be used to obtain all reference values ​​for the current job based on all historical jobs.

[0067] By assessing the similarity of current and historical operations in terms of pressure variation patterns, a reference value for the current operation based on historical operations is obtained by weighted averaging of confidence levels across three stages. This reference value comprehensively considers the reliability of historical data and the importance of different stages, providing a more targeted reference for assessing and adjusting the current equipment status. It helps operators more accurately determine whether the equipment is operating normally and whether corresponding adjustments are needed.

[0068] S3: Sort and classify all the calculated reference values ​​in descending order to obtain historical reference operations that match the current operation mode. Calculate the vacuum loss based on the historical reference operations, and adjust the vacuum pump power according to the comparison between the vacuum loss of the current operation and the set threshold.

[0069] In one embodiment, all reference values ​​calculated in S2 above are sorted in descending order, and then the Otsu method is used to classify the sorted reference value sequence.

[0070] Otsu's method is an image segmentation method based on a global threshold. It can also be used for data classification by finding an optimal threshold to divide data into two classes that maximize the variance between the two classes. Sort data helps Otsu's method find a suitable threshold more effectively, improving the accuracy and stability of classification. This threshold is determined based on the statistical properties of the data (such as within-class variance and between-class variance) and does not require subjective manual setting.

[0071] After the Otsu method's binary classification, the aforementioned reference value sequence was divided into two categories. It is generally believed that larger historical reference values ​​may correspond to a work process that is closer to the ideal state, or that is closer to the current work process in terms of key indicators. Therefore, the first category (the data class that is in a higher position after sorting) is determined to be the work process corresponding to the historical reference value that is similar to the current work process, and can be used as a historical reference work that matches the current work pattern.

[0072] In another embodiment, instead of simply bisecting the entire sorted sequence, starting from the second element of the sorted sequence, specific classification response values ​​are calculated to determine which historical reference values ​​correspond to job processes that match the current job process, and these can be used as reference jobs. Specifically:

[0073] After sorting all the calculated reference values ​​in descending order, the classification response value is calculated starting from the second element in the sorted reference value sequence. That is, any element after the second element is selected as a candidate classification point, and the value of the second element is divided by the mean of the classification sequence formed by the first element and the candidate classification point to obtain the classification response value of the candidate classification point.

[0074] Traverse all candidate classification points. When a candidate classification point whose classification response value is less than or equal to the preset classification threshold is taken as the final classification point, all historical jobs corresponding to the classification sequence formed by the first element and the final classification point are the historical reference jobs that match the current job mode.

[0075] In actual operation, the distribution of historical reference values ​​can be complex and diverse. The initial classification threshold of 0.8 was determined based on the analysis and statistics of a large amount of similar processing data, and it can adapt to classification needs under different data distribution conditions. When the data distribution is relatively concentrated, 0.8 can effectively filter out samples similar to the current operation process; when the data distribution is relatively dispersed, it can also ensure the rationality of the classification to a certain extent, avoiding misclassifying excessively divergent data as similar.

[0076] After obtaining historical reference jobs that match the current job mode, calculate the vacuum loss of the current job, which is one minus the average of the historical reference values ​​of all historical reference jobs selected for the current job.

[0077] Further threshold settings can be made, such as setting it to 0.5. Alternatively, the minimum historical reference value among all referenced jobs in the current job can be set as the threshold.

[0078] Compare the calculated vacuum loss of the current operation with the set threshold. If the vacuum loss of the current operation is greater than the threshold, it indicates an anomaly in the operation process and a vacuum leak. Upon determining an anomaly, the power of the vacuum pump in the negative pressure hopper should be increased linearly, for example, by 5% of the total power each time. If the vacuum pump power has been increased to its maximum value, but the vacuum loss is still greater than the threshold, it indicates that the negative pressure hopper is severely damaged and can no longer meet normal vacuum operation conditions. The vacuum pump and related equipment should be shut down immediately, and the negative pressure hopper should be thoroughly inspected.

[0079] This invention also provides a vacuum pumping adaptive control system for a negative pressure hopper in an injection molding machine. For example... Figure 2 As shown, the system includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the vacuum pumping adaptive control method for a negative pressure hopper of an injection molding machine according to the first aspect of the present invention.

[0080] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

[0081] It should be noted that those skilled in the art can make various modifications and improvements without departing from the inventive concept, and these all fall within the scope of protection of this invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A vacuum pumping adaptive control method for a negative pressure hopper in an injection molding machine, characterized in that, include: The pressure of the negative pressure hopper at each moment of completing one operation is collected and preprocessed to obtain a pressure time series. All pressure points in the pressure time series are traversed, and any two pressure points are selected as a set of dividing points. Each set of dividing points corresponds to a calculated dividing response value. When the dividing response value is the smallest, the corresponding set of dividing points is the optimal dividing point. The pressure time series is divided into three stages according to the optimal dividing point. A historical database is constructed, which records the pressure of all negative pressure hoppers of the same model as the negative pressure hopper at each stage under different operation numbers. The confidence scores of the three stages of the database are calculated respectively. The confidence scores of the three stages of the database are weighted by summing them into one to obtain the weights of each of the three stages. The cosine similarity between the pressure time series corresponding to any operation in the database and the pressure time series corresponding to the current operation is calculated. The reference value of the current operation based on the historical operations is obtained by weighting the confidence scores of the three stages. All calculated reference values ​​are sorted and categorized in descending order to obtain historical reference operations that match the current operation mode. Vacuum loss is calculated based on historical reference operations, whereby the vacuum loss is 1 minus the average of all historical reference values ​​selected for the current operation. The vacuum pump power is adjusted based on the comparison between the vacuum loss of the current operation and a set threshold. If the vacuum loss is greater than the threshold, it is determined that there is a vacuum leak in the current operation, and the vacuum pump power of the negative pressure hopper is linearly increased until the vacuum loss is lower than the threshold. If the vacuum pump power reaches the maximum power and the vacuum loss is still greater than the threshold, it indicates that the negative pressure hopper is severely damaged and cannot meet the vacuum operation conditions, so the operation is immediately stopped for investigation.

2. The vacuum pumping adaptive control method for the negative pressure hopper of an injection molding machine according to claim 1, characterized in that, The process of obtaining the segmentation response value includes: Calculate the coefficient of variation of the subsequence before the first division point, the coefficient of variation of the subsequence between the first and second division points, and the coefficient of variation of the subsequence after the second division point after dividing the pressure time series using two arbitrarily selected pressure points. The product of the three calculated coefficients of variation is used as the segmentation response value.

3. The vacuum pumping adaptive control method for the negative pressure hopper of an injection molding machine according to claim 1, characterized in that, The first stage of obtaining confidence includes: Two assignments are randomly selected from the database as the first and second assignments, respectively. Calculate the Pearson correlation coefficient between the first and second tasks in the first stage of stress; calculate the mean relative difference in stress between the first and second tasks in the first stage of stress. The contribution values ​​of the first and second assignments are obtained by multiplying the calculated Pearson correlation coefficient by the relative difference of the mean. By combining all job pairs, repeat the above contribution value operation and multiply all contribution values ​​consecutively to obtain the confidence level for the first stage; The confidence levels for the second and third stages are calculated in the same way as those for the first stage.

4. The vacuum pumping adaptive control method for the negative pressure hopper of an injection molding machine according to claim 1, characterized in that, The first stage of obtaining confidence includes: Two assignments are randomly selected from the database as the first and second assignments, respectively. Calculate the shape similarity of the pressure in the first stage between the first and second tasks, which is obtained using DTW distance calculation; calculate the mean relative difference of the pressure in the first stage between the first and second tasks. Multiply the calculated shape similarity by the mean relative difference to obtain the contribution values ​​of the first and second assignments; By combining all job pairs, repeat the above contribution value operation and multiply all contribution values ​​consecutively to obtain the confidence level for the first stage; The confidence levels for the second and third stages are calculated in the same way as those for the first stage.

5. The vacuum pumping adaptive control method for a negative pressure hopper in an injection molding machine according to claim 1, characterized in that, The classification operation includes: After sorting all the calculated reference values ​​in descending order, the Otsu method is used to classify the sorted reference value sequence, and the historical jobs corresponding to the first category of reference values ​​are taken as the historical reference jobs that match the current job mode.

6. The vacuum pumping adaptive control method for a negative pressure hopper in an injection molding machine according to claim 1, characterized in that, The classification operation includes: After sorting all the calculated reference values ​​in descending order, the classification response value is calculated starting from the second element in the sorted reference value sequence. That is, any element is selected from all elements after the second element as a candidate classification point, and the value of the second element is divided by the mean of the classification sequence formed by the first element and the candidate classification point as the classification response value of the candidate classification point. Traverse all candidate classification points. When a candidate classification point whose classification response value is less than or equal to the preset classification threshold is taken as the final classification point, all historical jobs corresponding to the classification sequence formed by the first element and the final classification point are the historical reference jobs that match the current job mode.

7. A vacuum pumping adaptive control system for a negative pressure hopper in an injection molding machine, characterized in that, include: A processor and a memory, the memory storing computer program instructions that, when executed by the processor, implement the vacuum pumping adaptive control method for a negative pressure hopper of an injection molding machine according to any one of claims 1-6.

Images