Injection molding defect correction rule learning method and terminal

Abstract

The invention discloses an injection molding defect correction rule learning method and a terminal, and relates to the field of injection molding process optimization, and the injection molding defect correction rule learning method comprises the following steps: collecting multiple groups of mold testing data of a mold testing process; subtracting the data corresponding to each former mode from the last mode data of each trial mode process to construct a training data set; screening the training data set according to a preset support degree threshold to obtain a frequent data set; screening the frequent data set according to a preset confidence threshold to obtain a rule set; and screening the rule set to obtain a defect correction rule. According to the technical scheme, the data sets of multiple mold adjustment processes can be collected, learning of injection molding defect correction knowledge is automatically completed, manual intervention is not carried out, individual difference, deviation and locality of manual knowledge summarization are avoided, and good data compatibility can be achieved.

Description

Technical Field

[0001] The present application relates to the field of injection molding process optimization, and in particular to a method and terminal for learning injection molding defect correction rules. Background Art

[0002] In the production process of injection molded plastic products, the control of injection molding machine process parameters has a huge impact on the production quality of injection molded products. Improper settings will lead to various defects such as water ripples, weld marks, flash and short shots after the product is molded. It is necessary to further adjust the process parameters to obtain products that meet the quality requirements. The traditional injection molding machine parameter setting follows the "trial injection"-"correction" mode, which requires experience or intensive multiple attempts to finally determine the appropriate parameters. Modern injection molding machine parameter setting methods can use rule-based reasoning methods to complete parameter setting and optimization, and have achieved good results in laboratory environments. However, the knowledge of these methods often comes from the experience of process personnel, because the entire process of injection molding is completed in a closed mold, and its related phenomena are difficult to capture. The method based on data statistics does not pay attention to its internal mechanism, which is a more ideal method for establishing process optimization experience. However, the process parameters and product quality are strongly nonlinear, strongly dependent on the intuition or experience of the process personnel, and the related laws are difficult to summarize manually, the construction efficiency is low, and there may be subjective deviations. Therefore, in practical applications, it is more hoped that the process optimization knowledge can be directly learned from the original data to guide the defect correction process in the actual production of injection molding. Summary of the invention

[0003] The main purpose of this application is to propose a method and terminal for learning injection molding defect correction rules, which aims to collect data sets from multiple trial mold processes, automatically complete the learning of injection molding defect correction knowledge, without manual intervention, avoid the individual differences, bias and locality of manually summarized knowledge, and achieve better data compatibility.

[0004] To achieve the above objectives, the injection molding defect correction rule learning method proposed in this application includes: Collect multiple groups of mold trial data during the mold trial process; The last mold data of each mold trial process is subtracted from the data corresponding to each previous mold trial to construct a training data set; Filter the training data set according to the preset support threshold to obtain a frequent data set; Filter the frequent data set according to the preset confidence threshold to obtain the rule set; The rule set is filtered to obtain defect correction rules.

[0005] In one embodiment, the training data set is defined as follows: ; ; ; ; in, is the training data set, Indicates The dataset of the mold trial process, Indicates During the mold trial Group data, is the product feature vector, is the material characteristic vector, is the device feature vector, Indicates the parameter change between two modules. The parameter change includes the process parameter change and the defect parameter change. Indicates the process parameters of the last mold in a single set of mold trials. Indicates During the mold trial Moulding process parameters.

[0006] In one embodiment, the step of constructing a training data set by subtracting the data corresponding to each previous mold from the last mold data of each mold trial process specifically includes: The last mold data of each mold trial process is subtracted from the data corresponding to each previous mold trial, and the parameters of the same field are normalized using the deviation standardization method; Using the equidistance dispersion method, each parameter is discretized within the normalized interval, and a training data set is constructed based on the discretized data.

[0007] In one embodiment, the normalized calculation formula is: ; in, is the minimum value of the same field parameter, The maximum value of the parameters in the same field.

[0008] In one embodiment, the step of filtering the training data set according to the preset support threshold to obtain the frequent data set specifically includes: Scan the training data set to generate the first candidate data set containing 1 element, set the variable k, and assign k=2; Scan the first candidate data set to generate a second candidate data set containing k elements, calculate the support of each data in the second candidate data set, and remove the items in the second candidate data set whose support is less than a preset support threshold to obtain a third candidate data set; If the third candidate data set is not an empty set, the third candidate data set is used as a frequent sub-data set, and k=k+1 is assigned, and the execution step is returned to scan the first candidate data set to generate a second candidate data set containing k elements, the support of each data in the second candidate data set is calculated, and the items whose support is less than a preset support threshold in the second candidate data set are removed to obtain the third candidate data set; If the third candidate data set is an empty set, all frequent sub-data sets are output to obtain a frequent data set.

[0009] In one embodiment, the calculation formula for the support of each data in the second candidate data set is: ; in, is the support of each data in the second candidate data set, is the number of sets containing the required items in the second candidate data set, is the total number of sets of the second candidate data set.

[0010] In one embodiment, the filtering of the frequent data set according to the preset confidence threshold to obtain the rule set includes: Get all frequent sub-datasets; Take every two frequent sub-data sets as rule antecedents and rule consequents, construct a defect correction sub-rule, and calculate the confidence of each defect correction sub-rule; The defect correction sub-rules whose confidence is less than the preset confidence are removed to obtain a rule set.

[0011] In one embodiment, the confidence calculation formula of each defect correction sub-rule is: ; in, is the confidence of each defect correction sub-rule, Indicates the frequent subset of data, including and The support of the two items, Indicates that it contains support.

[0012] In one embodiment, the trial mold data includes process parameters and defect parameters; The step of screening the rule set to obtain the defect correction rule specifically includes: The rule set is screened, and defect sub-rules whose rule antecedents contain defect parameters and whose rule causality only contains process parameters are retained as defect correction rules.

[0013] In one embodiment, the step of screening the rule set to obtain the defect correction rule includes: The preset artificial neural network is used to match the defect parameters and the rule antecedents, and the rule consequences are executed.

[0014] The present application also proposes a terminal, comprising: Memory; and A processor, an injection molding defect correction rule learning program stored in the memory and executed by the processor, wherein the injection molding defect correction rule learning program implements the injection molding defect correction rule learning method as described above when executed by the processor.

[0015] The technical solution of the present application proposes a method for learning injection molding defect correction rules, which includes: collecting multiple sets of trial mold data from trial mold processes; subtracting the data corresponding to each previous mold from the last mold data of each trial mold process to construct a training data set; filtering the training data set according to a preset support threshold to obtain a frequent data set; filtering the frequent data set according to a preset confidence threshold to obtain a rule set; filtering the rule set to obtain defect correction rules. The technical solution of the present application can collect data sets from multiple mold adjustment processes, automatically complete the learning of injection molding defect correction knowledge, without manual intervention, avoiding the individual differences, bias and locality of manually summarized knowledge, and can achieve better data compatibility. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0017] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0018] Figure 1 This is a flow chart of an embodiment of the injection molding defect correction rule learning method of the present application; Figure 2 A flow chart of another embodiment of the injection molding defect correction rule learning method of the present application; Figure 3 A schematic flow chart of another embodiment of the injection molding defect correction rule learning method of the present application; Figure 4 This is a flow chart of another embodiment of the injection molding defect correction rule learning method of the present application; Figure 5A flow chart of another embodiment of the injection molding defect correction rule learning method of the present application; Figure 6 A schematic flow chart of another embodiment of the injection molding defect correction rule learning method of the present application; Figure 7 This is a schematic diagram of an embodiment of the injection molding defect correction rule learning method of the present application; Figure 8 This is a schematic diagram of a circuit function module of an embodiment of a terminal of the present application; Fig. 9 A schematic diagram of a circuit function module for establishing communication between a terminal and an injection molding machine according to an embodiment of the present application.

[0019] Description of Figure Numbers: 100. Terminal; 10. Memory; 20. Processor; 200. Injection molding machine.

[0020] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0021] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0022] In the production process of injection molded plastic products, the control of injection molding machine process parameters has a huge impact on the production quality of injection molded products. Improper settings will lead to various defects such as water ripples, weld marks, flash and short shots after the product is molded. It is necessary to further adjust the process parameters to obtain products that meet the quality requirements. The traditional injection molding machine parameter setting follows the "trial injection"-"correction" mode, which requires experience or intensive multiple attempts to finally determine the appropriate parameters. Modern injection molding machine parameter setting methods can use rule-based reasoning methods to complete parameter setting and optimization, and have achieved good results in laboratory environments. However, the knowledge of these methods often comes from the experience of process personnel, because the entire process of injection molding is completed in a closed mold, and its related phenomena are difficult to capture. The method based on data statistics does not pay attention to its internal mechanism, which is a more ideal method for establishing process optimization experience. However, the process parameters and product quality are strongly nonlinear, strongly dependent on the intuition or experience of the process personnel, and the related laws are difficult to summarize manually, the construction efficiency is low, and there may be subjective deviations. Therefore, in practical applications, it is more hoped that the process optimization knowledge can be directly learned from the original data to guide the defect correction process in the actual production of injection molding.

[0023] To solve the above problems, the present invention proposes a method for learning injection molding defect correction rules, which aims to collect data sets of multiple trial mold processes, automatically complete the learning of injection molding defect correction knowledge, without human intervention, avoiding the individual differences, bias and locality of manually summarized knowledge, and can achieve better data compatibility.

[0024] In one embodiment of the present application, refer to Figure 1 , Figure 7 and Figure 8 , Figure 1 This is a flow chart of an embodiment of the injection molding defect correction rule learning method of the present application. Figure 7 This is a schematic diagram of an embodiment of the injection molding defect correction rule learning method of the present application. Figure 8 This is a schematic diagram of a circuit function module of an embodiment of the terminal 100 of the present application. The injection molding defect correction rule learning method includes: S10, collecting multiple groups of mold trial data during the mold trial process; The injection molding defect correction rule learning method of the present application can be implemented based on the terminal 100, that is, the injection molding defect correction rule learning method is mainly executed on the terminal 100, and the terminal 100 is used to establish a communication connection with the control device of the injection molding machine 200 via a communication component, and receive the trial mold data output by the control device. Figure 8 As shown, Figure 8 The circuit function module diagram of an embodiment of establishing communication between the terminal 100 of the present application and the injection molding machine 200 is shown. Optionally, the terminal 100 may refer to a host computer, a mobile phone, or a tablet computer, etc., which is not specifically limited here. The communication component may be a wired communication component or a wireless communication component. For example, when the communication component is a wireless communication component, the wireless communication component may specifically be a WiFi component, a Bluetooth component, or a 3G / 4G / 5G component, which is not specifically limited here.

[0025] In this embodiment, the process personnel set the process parameters to the control device of the injection molding machine 200 according to actual experience, and obtain the test mold data by running the injection molding machine 200 for trial molding and transmit it to the terminal 100. The trial mold data may include but are not limited to basic product parameters, material performance parameters, equipment parameters, process parameters and defect parameters. The basic product parameters include average wall thickness, maximum flow length, product weight, etc.; material performance parameters include melt density, shear line density, degradation temperature, etc.; equipment parameters include screw diameter, maximum injection stroke, etc.; process parameters include temperature parameters, injection parameters, holding pressure parameters, cooling parameters, metering parameters and ejection parameters, etc., temperature parameters include mold temperature, barrel temperature, nozzle temperature, injection parameters include injection time, injection pressure, injection speed, injection position, holding pressure parameters include holding pressure, holding pressure speed, holding pressure time, cooling parameters refer to cooling time, metering parameters include screw speed, metering back pressure, metering position, ejection parameters include pre-metering decompression distance, pre-metering decompression speed, post-metering decompression distance, post-metering decompression speed. Defects refer to the appearance defects of injection molded products that can be directly observed by process personnel, such as short shots, bubbles, shrinkage, weld lines, flow marks, flash, warping, burning, white top, demolding difficulties, etc. Defect parameters refer to the quantification degree of appearance defects, which can be divided into three levels, specifically represented by numbers 1-3. For example, a small number of bubbles indicates a low quantification degree of defects, which can be represented by the number 1; a medium number of bubbles indicates an average quantification degree of defects, which can be represented by the number 2; a large number of bubbles indicates a high quantification degree of defects, which can be represented by the number 3. It can be seen that the quantification degree of other appearance defects is similar to that of bubbles, so they will not be repeated here.

[0026] S20, subtract the data corresponding to each previous mold from the last mold data of each mold trial process to construct a training data set; It is understandable that since the basic parameters of the product, material performance parameters, and equipment parameters are fixed parameters, they will not change significantly during the mold trial process. What changes are the process parameters. Therefore, in order to consider the impact of the process parameters on the defect parameters, in this embodiment, based on the injection molding machine 200 During the first group of mold trials, the terminal 100 will subtract the process parameters of the last mold in the first group of mold trials from the process parameters corresponding to each of the previous mold trials, and then divide the result by the process parameters corresponding to each of the previous mold trials, to obtain the process parameter change amount of the first group of mold trials. Correspondingly, the terminal 100 will subtract the defect parameters of the last mold in the first group of mold trials from the defect parameters corresponding to each of the previous mold trials, and then divide the result by the defect parameters corresponding to each of the previous mold trials, to obtain the defect parameter change amount of the first group of mold trials. Repeat the above operation to obtain The terminal 100 calculates the process parameter variation and defect parameter variation of the mold group test process. The training data set is constructed by analyzing the basic product parameters, material performance parameters, equipment parameters, process parameter changes, and defect parameter changes in each trial mold process.

[0027] In order to better understand the process of constructing the training data set, the following is an explanation based on only the first group of mold trials. Assuming that five mold trials are performed in the first group of mold trials, the process parameters of the fifth mold are subtracted from the process parameters of the first mold and then divided by the process parameters of the first mold to obtain the change in the process parameters of the first mold. Repeat the operation to obtain the change in the process parameters from the second mold to the fourth mold. Correspondingly, the terminal 100 subtracts the defect parameters of the fifth mold from the defect parameters of the first mold and then divides them by the defect parameters of the first mold to obtain the change in the defect parameters of the first mold. Repeat the operation to obtain the change in the defect parameters from the second mold to the fourth mold. The training data set is constructed based on the basic product parameters, material performance parameters, equipment parameters, the change in process parameters from the first mold to the fourth mold, and the change in defect parameters.

[0028] Based on the above content, it should be particularly noted that, whether it is the process parameter change or the defect parameter change, its value can be 0 or not. Taking the process parameter change as an example, when the process parameter change is 0, it means that the process parameters of the last mold in a single set of mold trials are the same as the process parameters of a previous mold, that is, the molding process of the last mold has not changed compared with the process parameters of a previous mold; when the process parameter change is not 0, it means that the process parameters of the last mold in a single set of mold trials are different from the process parameters of a previous mold, that is, the process parameters of the last mold have changed compared with the process parameters of a previous mold. It can be seen that the situation where the defect parameter change is 0 or not 0 is similar to the situation where the process parameter change is 0 or not 0, and will not be repeated here.

[0029] In one embodiment of the present application, the training data set is defined as follows: ; ; ; ; in, is the training data set, Indicates The dataset of the mold trial process, Indicates During the mold trial Group data, is the product feature vector, is the material characteristic vector, is the device feature vector, Indicates the parameter change between two modules. The parameter change includes the process parameter change and the defect parameter change. Indicates the process parameters of the last mold in a single set of mold trials. Indicates During the mold trial Moulding process parameters.

[0030] In one embodiment of the present application, refer to Figure 2 , Figure 2 This is a flow chart of another embodiment of the injection molding defect correction rule learning method of the present application. The step of constructing a training data set by subtracting the data corresponding to each previous mold from the last mold data of each mold trial process specifically includes: S21, subtract the data corresponding to each previous mold from the last mold data of each mold trial process, and normalize the parameters of the same field using the deviation standardization method; S22. Using the equidistance dispersion method, each parameter is discretized within the normalized interval, and a training data set is constructed based on the discretized data.

[0031] In this embodiment, first, the terminal 100 subtracts the data of the last module in each trial process from the data corresponding to each previous module and divides the result by the data corresponding to each previous module to obtain a training sub-data set; secondly, the terminal 100 can use the deviation normalization method to normalize the same field parameters in the training sub-data set. In combination with the above embodiments, the same-field parameters refer to a single type of parameter variation in the training sub-dataset, such as the temperature parameter variation, injection parameter variation, holding pressure parameter variation, cooling parameter variation, metering parameter variation and ejection parameter variation in the process parameter variation, and the short shot parameter variation, bubble parameter variation, shrinkage parameter variation, weld line parameter variation, wave flow mark parameter variation, flash parameter variation, warping parameter variation, scorch parameter variation, top white parameter variation, demolding difficulty parameter variation, etc. in the defect parameter variation. By removing the minimum and maximum values ​​of the temperature parameter variation and other variations, the data error is reduced; again, the terminal 100 uses the equidistance dispersion method to discretize each parameter variation in the normalized interval [0,1] to improve the matching efficiency of the training data set; finally, the terminal 100 constructs the training data set based on the discretized data.

[0032] In one embodiment of the present application, the normalized calculation formula is: ; in, is the minimum value of the same field parameter, The maximum value of the parameters in the same field.

[0033] S30, filtering the training data set according to a preset support threshold to obtain a frequent data set; In this embodiment, the preset support threshold is a support threshold pre-stored in the terminal 100, and the preset support threshold can be set as needed. The terminal 100 screens the training data set according to the preset support threshold to obtain a frequent data set.

[0034] In one embodiment of the present application, refer to Figure 3 , Figure 3 This is a flow chart of another embodiment of the injection molding defect correction rule learning method of the present application. The step of selecting the training data set according to the preset support threshold to obtain the frequent data set specifically includes: S31, scan the training data set to generate a first candidate data set containing 1 element, set a variable k, and assign k=2; In this embodiment, the terminal 100 scans the process parameter changes and defect parameter changes of the first group of trial molds in the training data set, generates a first candidate data set containing 1 element, sets the variable k, and assigns k=2. The first candidate data set refers to a set of process parameter changes and defect parameter changes that are not 0, and the process parameter changes may include but are not limited to temperature parameter changes, injection parameter changes, holding pressure parameter changes, cooling parameter changes, metering parameter changes, and shot back parameter changes, etc. The defect parameter changes may include but are not limited to short shot parameter changes, bubble parameter changes, shrinkage parameter changes, weld line parameter changes, wave flow mark parameter changes, flash parameter changes, warp parameter changes, scorch parameter changes, top white parameter changes, demolding difficulty parameter changes, etc. It can be inferred that based on the second group of trial molds to the first The process parameter changes and defect parameter changes during the mold group trial can also generate The first candidate dataset.

[0035] S32, scanning the first candidate data set to generate a second candidate data set containing k elements, calculating the support of each data in the second candidate data set, and removing the items in the second candidate data set whose support is less than a preset support threshold, to obtain a third candidate data set; S33a, if the third candidate data set is not an empty set, the third candidate data set is used as a frequent sub-data set, and k=k+1 is assigned, and the execution step is returned to scan the first candidate data set to generate a second candidate data set containing k elements, the support of each data in the second candidate data set is calculated, and the items whose support is less than a preset support threshold in the second candidate data set are removed to obtain the third candidate data set; S33b: If the third candidate data set is an empty set, all frequent sub-data sets are output to obtain a frequent data set.

[0036] In this embodiment, the terminal 100 scans all first candidate data sets. The following embodiments are all explained with the first candidate data set formed by the process parameter changes and defect parameter changes of the first group of trial mold processes. The terminal 100 scans the first candidate data set to generate a second candidate data set containing k elements. Since k=2, a second candidate data set containing 2 elements is generated. In other words, in the first candidate data set, the process parameter changes and defect parameter changes that are not zero in pairs can form a sub-data set. It can be seen that a single first candidate data set can be composed of multiple sub-data sets, and multiple sub-data sets can form a second candidate data set. Based on this, the terminal 100 calculates the support of each data in the second candidate data set, and removes the items in the second candidate data set whose support is less than the preset support threshold to obtain the third candidate data set. In the following, the second candidate data set includes three sub-data sets, one of which includes temperature parameter changes and injection parameter changes, another includes temperature parameter changes and holding pressure parameter changes, and another includes injection parameter changes and cooling parameter changes. According to the support calculation formula, the support of the temperature parameter change is 2 / 3, the support of the injection parameter change is 2 / 3, the support of the holding pressure parameter change is 1 / 3, and the support of the cooling parameter change is 1 / 3. Further, the terminal 100 compares the support of each parameter change with the preset support threshold, removes the parameter changes in the second candidate data set that are less than the preset support threshold through the comparison result, retains the parameter changes in the second candidate data set that are not less than the preset support threshold, and the retained parameter changes are used as the third candidate data set corresponding to the second candidate data set. When the third candidate data set is not an empty set, it means that there is at least one data in the second candidate data set whose support is not less than the preset support threshold. The third candidate data set can be used as a frequent sub-data set, and k=k+1 is assigned, and the above step 32 is returned to be executed; when the third candidate data set is an empty set, it means that there is no data in the second candidate data set whose support is not less than the preset support threshold, then all frequent sub-data sets are output to obtain a frequent data set.

[0037] In one embodiment of the present application, the calculation formula for the support of each data in the second candidate data set is: ; in, is the support of each data in the second candidate data set, is the number of sets containing the required items in the second candidate data set, is the total number of sets of the second candidate data set.

[0038] S40, filtering the frequent data set according to a preset confidence threshold to obtain a rule set; In this embodiment, the preset confidence threshold is a confidence threshold pre-stored in the terminal 100, and the preset confidence threshold can be set as needed. The terminal 100 screens the frequent data set according to the preset confidence threshold to obtain a rule set.

[0039] In one embodiment of the present application, refer to Figure 4 , Figure 4 This is a flow chart of another embodiment of the injection molding defect correction rule learning method of the present application, wherein the frequent data set is screened according to a preset confidence threshold to obtain a rule set including: S41, obtaining all frequent sub-datasets; S42, using every two frequent sub-data sets as rule antecedents and rule consequents, constructing a defect correction sub-rule, and calculating the confidence of each defect correction sub-rule; S43. Remove defect correction sub-rules whose confidence is less than a preset confidence, and obtain a rule set.

[0040] In this embodiment, the terminal 100 obtains all frequent sub-data sets in the frequent data, and uses every two frequent data sets as a rule antecedent and a rule consequent, and a rule antecedent and a rule consequent constitute a defect correction sub-rule, and calculates the confidence of each defect correction sub-rule. Further, the terminal 100 compares the confidence of each defect correction sub-rule with the preset confidence, removes the defect correction sub-rule with a confidence less than the preset confidence through the comparison result, retains the defect correction sub-rule with a confidence not less than the preset confidence, and forms a rule set.

[0041] In one embodiment of the present application, the calculation formula for the confidence of each defect correction sub-rule is: ; in, is the confidence of each defect correction sub-rule, Indicates the frequent subset of data, including and The support of the two items, Indicates that it contains support.

[0042] S50: Filter the rule set to obtain defect correction rules.

[0043] In this embodiment, the terminal 100 screens the rule set to obtain defect correction rules.

[0044] In one embodiment of the present application, refer to Figure 5 , Figure 5 This is a flow chart of another embodiment of the injection molding defect correction rule learning method of the present application, wherein the trial mold data includes process parameters and defect parameters; The step of screening the rule set to obtain the defect correction rule specifically includes: S51. Filter the rule set, retain the defect sub-rules whose rule antecedents include defect parameters and whose rule suffixes only include process parameters, and use them as defect correction rules.

[0045] In this embodiment, the terminal 100 screens the rule set, retains the defect correction sub-rule whose rule antecedent includes the defect parameter variation and whose rule consequent only includes the defect adjustment item, and uses it as the defect correction rule. That is, the rule antecedent of the defect correction rule can include not only the defect parameter but also the process parameter, but the corresponding rule consequent can only include the process parameter.

[0046] In one embodiment of the present application, refer to Figure 6 , Figure 6 This is a flow chart of another embodiment of the injection molding defect correction rule learning method of the present application. The step of screening the rule set to obtain the defect correction rule includes: S60, using a preset artificial neural network to match the defect parameters with the rule antecedents, and executing the rule consequences.

[0047] In this embodiment, for the defect correction rules finally obtained, the terminal 100 can use an artificial neural network to complete the data input to the conclusion output, thereby completing the matching of the rule antecedents and the calculation of the rule matching degree.

[0048] The technical solution of the present application proposes a method for learning injection molding defect correction rules, which includes: collecting multiple sets of trial mold data from trial mold processes; subtracting the data corresponding to each previous mold from the last mold data of each trial mold process to construct a training data set; filtering the training data set according to a preset support threshold to obtain a frequent data set; filtering the frequent data set according to a preset confidence threshold to obtain a rule set; filtering the rule set to obtain defect correction rules. The technical solution of the present application can collect data sets from multiple mold adjustment processes, automatically complete the learning of injection molding defect correction knowledge, without manual intervention, avoiding the individual differences, bias and locality of manually summarized knowledge, and can achieve better data compatibility.

[0049] The present application also proposes a terminal 100, see Figure 8The terminal 100 includes a memory 10, a processor 20, and an injection molding defect correction rule learning program stored in the memory 10 and executable on the processor 20, wherein the injection molding defect correction rule learning program is configured to implement the steps of the injection molding defect correction rule learning method.

[0050] The terminal 100 provided by the present application adopts the injection molding defect correction rule learning method in the above embodiment, and can collect data sets of multiple trial mold processes, automatically complete the learning of injection molding defect correction knowledge, and do not require manual intervention, thereby avoiding the individual differences, bias and locality of manual summary of knowledge, and can achieve better data compatibility. Compared with the prior art, the beneficial effects of the terminal 100 provided by the present application are the same as the beneficial effects of the injection molding defect correction rule learning method provided by the above embodiment, and the other technical features in the terminal 100 are the same as the features disclosed in the above embodiment method, which will not be repeated here.

[0051] The above description is only a preferred embodiment of the present application, and does not limit the patent scope of the present application. All equivalent structural changes made by using the contents of the present application specification and drawings under the inventive concept of the present application, or directly / indirectly applied in other related technical fields are included in the patent protection scope of the present application.

Claims

1. A method for learning injection molding defect correction rules, characterized in that: include: Collect multiple groups of mold trial data during the mold trial process; The last mold data of each mold trial process is subtracted from the data corresponding to each previous mold trial to construct a training data set; Filter the training data set according to the preset support threshold to obtain a frequent data set; Filter the frequent data set according to the preset confidence threshold to obtain the rule set; The rule set is filtered to obtain defect correction rules.

2. The injection molding defect correction rule learning method according to claim 1, characterized in that: The training dataset is defined as follows: ； ； ； ； in, is the training data set, Indicates The dataset of the mold trial process, Indicates During the mold trial Group data, is the product feature vector, is the material characteristic vector, is the device feature vector, Indicates the parameter change between two modules. The parameter change includes the process parameter change and the defect parameter change. Indicates the process parameters of the last mold in a single set of mold trials. Indicates During the mold trial Moulding process parameters.

3. The injection molding defect correction rule learning method according to claim 1, characterized in that: The step of constructing a training data set by subtracting the data corresponding to each previous mold from the last mold data of each mold trial process specifically includes: The last mold data of each mold trial process is subtracted from the data corresponding to each previous mold trial, and the parameters of the same field are normalized using the deviation standardization method; Using the equidistance dispersion method, each parameter is discretized within the normalized interval, and a training data set is constructed based on the discretized data.

4. The injection molding defect correction rule learning method according to claim 3, characterized in that: The normalized calculation formula is: ； in, is the minimum value of the same field parameter, The maximum value of the parameters in the same field.

5. The injection molding defect correction rule learning method according to claim 1, characterized in that: The step of filtering the training data set according to the preset support threshold to obtain the frequent data set specifically includes: Scan the training data set to generate the first candidate data set containing 1 element, set the variable k, and assign k=2; Scan the first candidate data set to generate a second candidate data set containing k elements, calculate the support of each data in the second candidate data set, and remove the items in the second candidate data set whose support is less than a preset support threshold to obtain a third candidate data set; If the third candidate data set is not an empty set, the third candidate data set is used as a frequent sub-data set, and k=k+1 is assigned, and the execution step is returned to scan the first candidate data set to generate a second candidate data set containing k elements, the support of each data in the second candidate data set is calculated, and the items whose support is less than a preset support threshold in the second candidate data set are removed to obtain the third candidate data set; If the third candidate data set is an empty set, all frequent sub-data sets are output to obtain a frequent data set.

6. The injection molding defect correction rule learning method according to claim 5, characterized in that: The calculation formula for the support of each data in the second candidate data set is: ； in, is the support of each data in the second candidate data set, is the number of sets containing the required items in the second candidate data set, is the total number of sets of the second candidate data set.

7. The injection molding defect correction rule learning method according to claim 1, characterized in that: The frequent data set is filtered according to the preset confidence threshold to obtain the rule set, which includes: Get all frequent sub-datasets; Take every two frequent sub-data sets as rule antecedents and rule consequents, construct a defect correction sub-rule, and calculate the confidence of each defect correction sub-rule; The defect correction sub-rules whose confidence is less than the preset confidence are removed to obtain a rule set.

8. The injection molding defect correction rule learning method according to claim 7, characterized in that: The calculation formula of the confidence of each defect correction sub-rule is: ； in, is the confidence of each defect correction sub-rule, Indicates the frequent sub-dataset, including and The support of the two items, Indicates that it contains support.

9. The injection molding defect correction rule learning method according to claim 7, characterized in that: The trial mold data includes process parameters and defect parameters; The step of screening the rule set to obtain the defect correction rule specifically includes: The rule set is screened, and defect sub-rules whose rule antecedents contain defect parameters and whose rule causality only contains process parameters are retained as defect correction rules.

10. The injection molding defect correction rule learning method according to claim 9, characterized in that: The step of screening the rule set to obtain the defect correction rule includes: The preset artificial neural network is used to match the defect parameters and the rule antecedents, and the rule consequences are executed.

11. A terminal, characterized in that: include: Memory; as well as A processor, an injection molding defect correction rule learning program stored on the memory and executed by the processor, wherein the injection molding defect correction rule learning program, when executed by the processor, implements the injection molding defect correction rule learning method as described in any one of claims 1 to 10.

Images