A fault early warning method and system of an in-mold trimming automatic machine

Abstract

The application discloses a kind of fault early warning method and system of burying injection molding up and down material automatic machine, it is related to equipment management field, wherein the method includes: obtaining the F-grade working condition information of first injection molding up and down material automatic machine;Based on multidimensional fault analysis index and F-grade working condition information, obtain the first injection machine fault record base;Based on sensitive convergence learning function, according to F-grade working condition information and the first injection machine fault record base carries out deep learning, generates first fault detection channel;Based on real-time operation data source, according to first fault detection channel carries out fault analysis, obtains fault detection coefficient;If fault detection coefficient meets fault detection early warning constraint, generates fault early warning signal.The technical problem that the fault detection accuracy of burying injection molding up and down material automatic machine in prior art is low, leading to the poor fault early warning effect of burying injection molding up and down material automatic machine.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of equipment management, in particular to a fault early warning method and system of an in-mold injection molding feeding and discharging robot. BACKGROUND

[0002] The in-mold injection molding feeding and discharging robot is an automatic device applied to plastic product production. Fault early warning is an important means to ensure the normal operation of the in-mold injection molding feeding and discharging robot. In the prior art, there is a technical problem that the fault detection accuracy of the in-mold injection molding feeding and discharging robot is low, resulting in poor fault early warning effect of the in-mold injection molding feeding and discharging robot. How to effectively perform fault early warning on the in-mold injection molding feeding and discharging robot has attracted widespread attention. SUMMARY

[0003] The present application provides a fault early warning method and system of an in-mold injection molding feeding and discharging robot. The technical problem of low fault detection accuracy of the in-mold injection molding feeding and discharging robot in the prior art, resulting in poor fault early warning effect of the in-mold injection molding feeding and discharging robot, is solved. The technical effect of improving the fault detection accuracy and comprehensiveness of the in-mold injection molding feeding and discharging robot and enhancing the fault early warning quality of the in-mold injection molding feeding and discharging robot is achieved.

[0004] In view of the above problems, the present application provides a fault early warning method and system of an in-mold injection molding feeding and discharging robot.

[0005] In a first aspect, the present application provides a fault early warning method of an in-mold injection molding feeding and discharging robot, wherein the method is applied to a fault early warning system of an in-mold injection molding feeding and discharging robot, and the method comprises: obtaining F-level working condition information of a first injection molding feeding and discharging robot, wherein F is a positive integer greater than 1; obtaining a multi-dimensional fault analysis index, wherein the multi-dimensional fault analysis index comprises a fault type, a fault probability and a fault influence; based on the multi-dimensional fault analysis index and the F-level working condition information, collecting fault records of the first injection molding feeding and discharging robot to obtain a first injection machine fault record library; based on a sensitive convergence learning function, performing deep learning according to the F-level working condition information and the first injection machine fault record library to generate a first fault detection channel; obtaining a real-time running data source of the first injection molding feeding and discharging robot, wherein the real-time running data source comprises a real-time running working condition identifier and real-time working condition state data corresponding to the real-time running working condition identifier; based on the real-time running data source, performing fault analysis according to the first fault detection channel to obtain a fault detection coefficient; and if the fault detection coefficient meets a fault detection early warning constraint, generating a fault early warning signal.

[0006] In a second aspect, the application also provides a fault early warning system of an in-mold injection molding up-down feeding automatic machine, wherein the system comprises: a working condition information obtaining module, configured to obtain F-level working condition information of a first injection molding up-down feeding automatic machine, wherein F is a positive integer greater than 1; a fault analysis index obtaining module, configured to obtain a multi-dimensional fault analysis index, wherein the multi-dimensional fault analysis index comprises a fault type, a fault probability and a fault influence; a fault record collecting module, configured to collect fault records of the first injection molding up-down feeding automatic machine based on the multi-dimensional fault analysis index and the F-level working condition information, and obtain a first injection machine fault record library; a deep learning module, configured to perform deep learning based on a sensitive convergence learning function, according to the F-level working condition information and the first injection machine fault record library, and generate a first fault detection channel; a running data source obtaining module, configured to obtain real-time running data sources of the first injection molding up-down feeding automatic machine, wherein the real-time running data sources comprise real-time running working condition identifiers and corresponding real-time working condition state data; a fault analysis module, configured to perform fault analysis based on the real-time running data sources, according to the first fault detection channel, and obtain a fault detection coefficient; and a fault early warning module, configured to generate a fault early warning signal if the fault detection coefficient meets a fault detection early warning constraint.

[0007] The one or more technical solutions provided in the application have at least the following technical effects or advantages:

[0008] According to the multi-dimensional fault analysis index and the F-level working condition information, the fault records of the first injection molding up-down feeding automatic machine are collected, and the first injection machine fault record library is obtained; based on the sensitive convergence learning function, deep learning is performed according to the F-level working condition information and the first injection machine fault record library, and the first fault detection channel is generated; the real-time running data sources of the first injection molding up-down feeding automatic machine are input into the first fault detection channel, and the fault detection coefficient is obtained; if the fault detection coefficient meets the fault detection early warning constraint, a fault early warning signal is generated. The technical effect of improving the fault detection accuracy and comprehensiveness of the in-mold injection molding up-down feeding automatic machine, and improving the fault early warning quality of the in-mold injection molding up-down feeding automatic machine is achieved.

[0009] The above description is only a summary of the technical solutions of the application. In order to enable one skilled in the art to better understand the technical means of the application, the application can be implemented according to the content of the description, and in order to make the above and other purposes, characteristics and advantages of the application more obvious and easy to understand, the following specific embodiments of the application are described. BRIEF DESCRIPTION OF DRAWINGS

[0010] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings of the embodiments of the present application will be briefly introduced as follows. Obviously, the drawings described in the following description only relate to some embodiments of the present application, and are not a limitation on the present application.

[0011] Figure 1 A flowchart of a fault early warning method of an in-mold injection molding up-down material automatic machine according to the present application;

[0012] Figure 2 A structural diagram of a fault early warning system of an in-mold injection molding up-down material automatic machine according to the present application. DETAILED DESCRIPTION

[0013] The present application provides a fault early warning method and system of an in-mold injection molding up-down material automatic machine. The technical problem of low fault detection accuracy of the in-mold injection molding up-down material automatic machine in the prior art, resulting in poor fault early warning effect of the in-mold injection molding up-down material automatic machine, is solved. The technical effect of improving the fault detection accuracy and comprehensiveness of the in-mold injection molding up-down material automatic machine, and improving the fault early warning quality of the in-mold injection molding up-down material automatic machine, is achieved.

[0014] Embodiment one

[0015] Please refer to the drawings Figure 1 The present application provides a fault early warning method of an in-mold injection molding up-down material automatic machine, wherein the method is applied to a fault early warning system of an in-mold injection molding up-down material automatic machine, and the method specifically includes the following steps:

[0016] Obtaining F-level working condition information of a first injection up-down material automatic machine, wherein F is a positive integer greater than 1;

[0017] The in-mold injection molding up-down material automatic machine is an automatic device applied to plastic product production. The in-mold injection molding up-down material automatic machine can complete multiple production nodes such as in-mold injection molding, mechanical hand up-down material, detection, packaging, etc. The first injection up-down material automatic machine can be any in-mold injection molding up-down material automatic machine that uses the fault early warning system of the in-mold injection molding up-down material automatic machine for intelligent fault early warning. The F-level working condition information includes F working condition parameters. The F working condition parameters include multiple production nodes (such as in-mold injection molding, mechanical hand up-down material, detection, packaging, etc.) of the first injection up-down material automatic machine. And F is a positive integer greater than 1.

[0018] Obtaining a multi-dimensional fault analysis index, wherein the multi-dimensional fault analysis index includes fault type, fault probability and fault influence;

[0019] Based on the multi-dimensional fault analysis index and the F-level working condition information, the fault records of the first injection molding up-down feeding automatic machine are collected to obtain a first injection molding machine fault record library;

[0020] Based on the multi-dimensional fault analysis index and the F-level working condition information, the fault records of the first injection molding up-down feeding automatic machine are collected to obtain a first injection molding machine fault record library, including:

[0021] A preset historical time zone constraint is obtained;

[0022] Based on the multi-dimensional fault analysis index and the preset historical time zone constraint, F search constraints are generated according to the F-level working condition information;

[0023] According to the F search constraints, the fault records of the first injection molding up-down feeding automatic machine are collected to obtain F working condition-fault record sources;

[0024] Based on the F working condition-fault record sources, principal component analysis is performed to generate the first injection molding machine fault record library.

[0025] Based on the multi-dimensional fault analysis index and the preset historical time zone constraint, F search constraints are generated according to the F-level working condition information. Each search constraint includes a multi-dimensional fault analysis index, a preset historical time zone constraint, and one working condition parameter in the F-level working condition information. The multi-dimensional fault analysis index includes fault type, fault probability, and fault impact. The preset historical time zone constraint includes a historical time range determined by the fault early warning system of the in-mold injection molding up-down feeding automatic machine.

[0026] Further, according to the F search constraints, the fault records of the first injection molding up-down feeding automatic machine are collected to obtain F working condition-fault record sources. Each working condition-fault record source includes a plurality of working condition-fault records of the first injection molding up-down feeding automatic machine corresponding to each working condition parameter within the preset historical time zone constraint and the F-level working condition information. Each working condition-fault record includes historical working condition state data of the first injection molding up-down feeding automatic machine under the working condition parameter, and a historical fault detection report corresponding to the historical working condition state data. The historical working condition state data includes historical equipment state data and historical equipment operation data of the first injection molding up-down feeding automatic machine. The historical equipment state data includes historical equipment temperature, historical equipment noise, historical equipment ambient temperature, etc. of the first injection molding up-down feeding automatic machine. The historical equipment operation data includes a plurality of historical equipment operation parameters of the first injection molding up-down feeding automatic machine. For example, when the working condition parameter is in-mold injection molding, the corresponding plurality of historical equipment operation parameters include historical injection volume, historical injection pressure, historical injection time, historical pressure holding time, historical cooling time, etc. The historical fault detection report includes historical fault type information, historical fault probability coefficient, and historical fault impact coefficient corresponding to the historical working condition state data.

[0027] Further, principal component analysis is performed on the F working condition-failure record sources to obtain a first injection molding machine failure record library. Principal component analysis is a commonly used linear dimension reduction method. The basic principle of principal component analysis is to map high-dimensional data (i.e., the F working condition-failure record sources) to a low-dimensional space through linear projection, and to expect that the information amount of the data in the projected dimension is maximum, so as to use fewer data dimensions while retaining more characteristics of the original data points, thereby realizing dimension reduction processing of the F working condition-failure record sources. The first injection molding machine failure record library includes the F working condition-failure record sources after dimension reduction processing.

[0028] Based on the sensitive convergence learning function, deep learning is performed according to the F-level working condition information and the first injection molding machine failure record library to generate a first failure detection channel.

[0029] Based on the sensitive convergence learning function, deep learning is performed according to the F-level working condition information and the first injection molding machine failure record library to generate a first failure detection channel, including:

[0030] The sensitive convergence learning function is constructed, wherein the sensitive convergence learning function is:

[0031] FSN = e fsc-fsw ;

[0032] Wherein, FSN represents the fault sensitive convergence degree, fsc represents the fault sensitive accuracy, and fsw represents the fault sensitive error degree.

[0033] Based on the first injection molding up-down automatic machine, a same-family injection molding machine group is matched.

[0034] According to the multi-dimensional failure analysis index, failure record data of the same-family injection molding machine group is retrieved to obtain a same-family failure record library.

[0035] Device basic information of the first injection molding up-down automatic machine is collected, and a same-family injection molding machine group is matched according to the device basic information. The device basic information includes device basic parameters such as the model specification and device structure composition of the first injection molding up-down automatic machine. The same-family injection molding machine group includes a plurality of same-family injection molding machines of the first injection molding up-down automatic machine. The same-family injection molding machine is a buried injection molding up-down automatic machine with the same device basic information as the first injection molding up-down automatic machine. Then, according to the multi-dimensional failure analysis index, failure record data of the same-family injection molding machine group is retrieved to obtain a same-family failure record library. The same-family failure record library includes a device failure record set corresponding to each same-family injection molding machine in the same-family injection molding machine group. The device failure record set includes a plurality of working condition-failure record information under each working condition parameter in the F-level working condition information corresponding to the same-family injection molding machine. Each working condition-failure record information includes historical working condition state data and historical failure detection reports of the same-family injection molding machine.

[0036] training F working condition-fault detection branches according to the same family fault record library and the first injection molding machine fault record library based on the F-level working condition information and the sensitive convergence learning function;

[0037] wherein, training F working condition-fault detection branches according to the same family fault record library and the first injection molding machine fault record library based on the F-level working condition information and the sensitive convergence learning function, comprises:

[0038] obtaining the fth working condition information according to the F-level working condition information, and f is a positive integer belonging to F;

[0039] performing feature recognition on the same family fault record library and the first injection molding machine fault record library based on the fth working condition information to obtain the fth working condition same family fault record source and the fth injection molding machine fault record data;

[0040] performing supervised training on the fth working condition same family fault record source to obtain the fth working condition fault analysis network;

[0041] performing test analysis on the fth working condition fault analysis network based on the fth injection molding machine fault record data and the sensitive convergence learning function to obtain the fth fault sensitive convergence degree;

[0042] judging whether the fth fault sensitive convergence degree meets a preset fault sensitive convergence constraint;

[0043] if the fth fault sensitive convergence degree meets the preset fault sensitive convergence constraint, obtaining the fth working condition-fault detection branch according to the fth working condition fault analysis network, and adding the fth working condition-fault detection branch to the F working condition-fault detection branches.

[0044] integrating the F working condition-fault detection branches to obtain the first fault detection channel.

[0045] extracting F working condition parameters in F-level working condition information in sequence to obtain the fth working condition information, and f is a positive integer belonging to F. The fth working condition information is sequentially each working condition parameter in the F working condition parameters. Then, according to the fth working condition information, feature recognition is performed on the same family fault record library and the first injection molding machine fault record library respectively, that is, a plurality of working condition-fault record information corresponding to the fth working condition information in the same family fault record library is recorded as the fth working condition same family fault record source. At the same time, a plurality of working condition-fault records corresponding to the fth working condition information in the first injection molding machine fault record library are set as the fth injection molding machine fault record data.

[0046] Preferably, the application adopts a BP neural network to perform supervised training on the same-family fault record source of the fth working condition, to obtain a fault analysis network for the fth working condition. The BP neural network is a kind of multi-layer feedforward neural network trained according to the error backpropagation algorithm. The BP neural network can learn and store a large number of input-output mode mapping relationships without needing to reveal the mathematical equations describing such mapping relationships in advance. The BP neural network continuously adjusts the weights and thresholds of the network through backpropagation, so as to minimize the sum of squares of errors of the network. Supervised training is a method of machine learning. The basic principle of supervised training is that the input sample data (i.e., the same-family fault record source of the fth working condition) is repeatedly adjusted and trained on the weights and biases of the BP neural network through backpropagation, so that the output of the network is as close as possible to the expected vector. The fault analysis network for the fth working condition includes an input layer, a hidden layer, and an output layer.

[0047] Further, the fth injection molding machine fault record data is input into the fault analysis network for the fth working condition, the fault analysis network for the fth working condition is tested according to the fth injection molding machine fault record data, and the fth fault sensitivity accuracy and the fth fault sensitivity inaccuracy are obtained. The fth fault sensitivity accuracy is the output accuracy rate of the fault analysis network for the fth working condition on the fth injection molding machine fault record data. The fth fault sensitivity inaccuracy is the difference between 1 and the fth fault sensitivity accuracy. Then, the fth fault sensitivity accuracy and the fth fault sensitivity inaccuracy are input into a sensitive convergence learning function, to obtain the fth fault sensitivity convergence degree. The sensitive convergence learning function is:

[0048] FSN = e fsc-fsw ;

[0049] wherein FSN represents the fault sensitivity convergence degree, i.e., the output fth fault sensitivity convergence degree; fsc represents the fault sensitivity accuracy, i.e., the input fth fault sensitivity accuracy; and fsw represents the fault sensitivity inaccuracy, i.e., the input fth fault sensitivity inaccuracy.

[0050] Further, the preset fault sensitivity convergence constraint includes a fault sensitivity convergence degree range determined in advance by the fault early warning system of the embedded injection molding upper and lower material automatic machine. It is judged whether the fth fault sensitivity convergence degree meets the preset fault sensitivity convergence constraint. If the fth fault sensitivity convergence degree meets the preset fault sensitivity convergence constraint, the fault analysis network for the fth working condition is set as the fth working condition-fault detection branch, and the fth working condition-fault detection branch is added to the F working condition-fault detection branches. The F working condition-fault detection branches are connected, to obtain a first fault detection channel. The first fault detection channel includes F working condition-fault detection branches corresponding to F working condition parameters in F-level working condition information. Each working condition-fault detection branch is constructed in the same way as the fth working condition-fault detection branch, and will not be described here again.

[0051] The F working condition-fault detection branches corresponding to the F working condition parameters in the F-level working condition information are constructed by using the sensitive convergence learning function, a comprehensive first fault detection channel is obtained, and the fault detection accuracy of the embedded injection molding upper and lower material automatic machine is improved.

[0052] The judgment of whether the fth fault sensitive convergence degree meets the preset fault sensitive convergence constraint further includes:

[0053] If the fth fault sensitive convergence degree does not meet the preset fault sensitive convergence constraint, a loss data set is obtained according to the fth injection molding machine fault record data;

[0054] An fth incremental working condition fault analysis network is obtained by performing incremental learning on the fth working condition fault analysis network according to the loss data set.

[0055] The fth working condition-fault detection branch is generated according to the fth incremental working condition fault analysis network.

[0056] When judging whether the fth fault sensitive convergence degree meets the preset fault sensitive convergence constraint, if the fth fault sensitive convergence degree does not meet the preset fault sensitive convergence constraint, the fth injection molding machine fault record data is set as the loss data set, then the fth incremental working condition fault analysis network is obtained by performing incremental learning on the fth working condition fault analysis network according to the loss data set, and the fth incremental working condition fault analysis network is output as the fth working condition-fault detection branch. Incremental learning is a machine learning method that allows a model (i.e., the fth working condition fault analysis network) to update the model by adding new data and adjusting existing parameters without retraining the entire model, thereby avoiding overfitting of the model. Incremental learning refers to a learning system that can continuously learn new knowledge from new samples and can preserve most of the previously learned knowledge. The fth working condition fault analysis network is a neural network composed of multiple neurons connected to each other. Through training of the loss data set, the fth incremental working condition fault analysis network retains the basic functions of the fth working condition fault analysis network and maintains the performance of the model, thereby improving the generalization ability and accuracy of the constructed fth working condition-fault detection branch.

[0057] The real-time running data source of the first injection molding upper and lower material automatic machine is obtained, wherein the real-time running data source includes a real-time running working condition identifier and real-time working condition state data corresponding to the real-time running working condition identifier;

[0058] Based on the real-time running data source, fault analysis is performed according to the first fault detection channel to obtain a fault detection coefficient;

[0059] Based on the real-time running data source, fault analysis is performed according to the first fault detection channel to obtain a fault detection coefficient, which includes:

[0060] characterizing the first fault detection channel based on the real-time running condition identifier, to obtain a matched condition-fault detection branch;

[0061] inputting the real-time condition state data into the matched condition-fault detection branch, to obtain a fault detection report, wherein the fault detection report comprises fault type information, a fault probability coefficient, and a fault influence coefficient;

[0062] obtaining a preset fault detection weighting operator, and performing weighting calculation on the fault probability coefficient and the fault influence coefficient according to the preset fault detection weighting operator, to generate the fault detection coefficient.

[0063] connecting the fault early warning system of the one embedded injection molding up-down feeding automatic machine, to retrieve real-time running data source of the first injection molding up-down feeding automatic machine. The real-time running data source comprises a real-time running condition identifier, and real-time condition state data corresponding to the real-time running condition identifier. The real-time running condition identifier comprises real-time condition parameters corresponding to the first injection molding up-down feeding automatic machine. The real-time condition state data comprises real-time equipment state data (real-time equipment temperature, real-time equipment noise, real-time equipment ambient temperature, etc.) and real-time equipment running data (a plurality of real-time equipment running parameters, for example, when the real-time running condition identifier is mechanical hand up-down feeding, the corresponding plurality of real-time equipment running parameters comprise real-time speed of the mechanical hand, real-time motion trajectory of the mechanical hand, real-time acceleration of the mechanical hand, etc.) of the first injection molding up-down feeding automatic machine corresponding to the real-time running condition identifier.

[0064] The first fault detection channel comprises F condition-fault detection branches corresponding to F condition parameters in F-level condition information. According to the real-time running condition identifier, the F condition-fault detection branches in the first fault detection channel are matched to obtain a matched condition-fault detection branch. The matched condition-fault detection branch comprises a condition-fault detection branch corresponding to the real-time running condition identifier in the first fault detection channel.

[0065] Further, the real-time condition state data is inputted into the matched condition-fault detection branch to obtain a fault detection report. The fault detection report comprises fault type information, a fault probability coefficient, and a fault influence coefficient. The fault type information comprises fault type parameters corresponding to the first injection molding up-down feeding automatic machine under the real-time condition state data. The fault probability coefficient is data information for representing the possibility of the first injection molding up-down feeding automatic machine under the real-time condition state data. The higher the possibility of the first injection molding up-down feeding automatic machine under the real-time condition state data, the greater the corresponding fault probability coefficient. The fault influence coefficient is data information for representing the fault influence degree of the first injection molding up-down feeding automatic machine under the real-time condition state data. The higher the fault influence degree of the first injection molding up-down feeding automatic machine under the real-time condition state data, the greater the corresponding fault influence coefficient.

[0066] Further, the preset fault detection weighting operator includes a fault probability weight and a fault influence weight determined by the fault early warning system of the embedded injection molding up-down material automatic machine. Then, the fault probability coefficient and the fault influence coefficient are calculated by weighting according to the preset fault detection weighting operator, that is, the product of the fault probability weight in the preset fault detection weighting operator and the fault probability coefficient is set as the weighted fault probability coefficient. Similarly, the product between the fault influence weight in the preset fault detection weighting operator and the fault influence coefficient is set as the weighted fault influence coefficient. The sum of the weighted fault probability coefficient and the weighted fault influence coefficient is set as the fault detection coefficient.

[0067] Through the first fault detection channel, accurate and efficient fault analysis is performed on the first injection molding up-down material automatic machine, and an accurate fault detection coefficient is obtained, thereby improving the fault early warning reliability of the embedded injection molding up-down material automatic machine.

[0068] If the fault detection coefficient meets the fault detection early warning constraint, a fault early warning signal is generated.

[0069] If the fault detection coefficient meets the fault detection early warning constraint, a fault early warning signal is generated.

[0070] It is judged whether the fault detection coefficient meets the fault detection early warning constraint.

[0071] If the fault detection coefficient meets the fault detection early warning constraint, a fault early warning signal level table is retrieved.

[0072] The fault detection coefficient is input into the fault early warning signal level table, and the fault early warning signal is generated.

[0073] The fault detection early warning constraint includes a fault detection coefficient early warning range determined by the fault early warning system of the embedded injection molding up-down material automatic machine. Then, it is judged whether the fault detection coefficient meets the fault detection early warning constraint. If the fault detection coefficient meets the fault detection early warning constraint, the fault early warning signal level table is retrieved. The fault detection coefficient is input into the fault early warning signal level table, and the fault early warning signal is generated, thereby realizing the hierarchical fault early warning of the embedded injection molding up-down material automatic machine and improving the fault early warning refinement degree of the embedded injection molding up-down material automatic machine. The fault early warning signal level table includes a plurality of preset fault detection coefficient ranges and a plurality of preset fault level early warning signals corresponding to the plurality of preset fault detection coefficient ranges, which are determined by the fault early warning system of the embedded injection molding up-down material automatic machine. The fault early warning signal includes a preset fault level early warning signal corresponding to the fault detection coefficient.

[0074] In summary, the fault early warning method of the in-mold injection molding upper and lower material automatic machine provided by the application has the following technical effects:

[0075] According to the multi-dimensional fault analysis index and the F-level working condition information, the fault record of the first injection molding upper and lower material automatic machine is collected to obtain a first injection machine fault record library; based on a sensitive convergence learning function, deep learning is performed according to the F-level working condition information and the first injection machine fault record library to generate a first fault detection channel; the real-time running data source of the first injection molding upper and lower material automatic machine is input into the first fault detection channel to obtain a fault detection coefficient; if the fault detection coefficient meets the fault detection early warning constraint, a fault early warning signal is generated. The technical effects of improving the fault detection accuracy and comprehensiveness of the in-mold injection molding upper and lower material automatic machine and improving the fault early warning quality of the in-mold injection molding upper and lower material automatic machine are achieved.

[0076] Embodiment two

[0077] Based on the fault early warning method of the in-mold injection molding upper and lower material automatic machine in the foregoing embodiment, the same inventive concept is provided, and the application also provides a fault early warning system of an in-mold injection molding upper and lower material automatic machine. Please refer to the accompanying Figure 2 , the system comprises:

[0078] A working condition information obtaining module is configured to obtain F-level working condition information of a first injection molding upper and lower material automatic machine, wherein F is a positive integer greater than 1;

[0079] A fault analysis index obtaining module is configured to obtain a multi-dimensional fault analysis index, wherein the multi-dimensional fault analysis index comprises a fault type, a fault probability and a fault influence;

[0080] A fault record collecting module is configured to collect fault records of the first injection molding upper and lower material automatic machine based on the multi-dimensional fault analysis index and the F-level working condition information to obtain a first injection machine fault record library;

[0081] A deep learning module is configured to perform deep learning according to the F-level working condition information and the first injection machine fault record library based on a sensitive convergence learning function to generate a first fault detection channel;

[0082] A running data source obtaining module is configured to obtain a real-time running data source of the first injection molding upper and lower material automatic machine, wherein the real-time running data source comprises a real-time running working condition identifier and real-time working condition state data corresponding to the real-time running working condition identifier;

[0083] a fault analysis module, configured to perform fault analysis according to the first fault detection channel based on the real-time operation data source, and obtain a fault detection coefficient;

[0084] a fault early warning module, configured to generate a fault early warning signal if the fault detection coefficient meets a fault detection early warning constraint.

[0085] Further, the fault record collection module is further configured to:

[0086] obtain a preset historical time zone constraint;

[0087] generate F search constraints according to the F-level working condition information based on the multi-dimensional fault analysis index and the preset historical time zone constraint;

[0088] perform fault record collection on the first injection molding up-down machine according to the F search constraints, and obtain F working condition-fault record sources;

[0089] perform principal component analysis based on the F working condition-fault record sources, and generate the first injection molding machine fault record library.

[0090] Further, the deep learning module is further configured to:

[0091] construct the sensitive convergence learning function, wherein the sensitive convergence learning function is:

[0092] FSN = e fsc-fsw ;

[0093] wherein FSN represents fault sensitive convergence degree, fsc represents fault sensitive accuracy, and fsw represents fault sensitive error degree;

[0094] match a same-family injection molding machine group based on the first injection molding up-down machine;

[0095] retrieve fault record data of the same-family injection molding machine group according to the multi-dimensional fault analysis index, and obtain a same-family fault record library;

[0096] train F working condition-fault detection branches according to the same-family fault record library and the first injection molding machine fault record library based on the F-level working condition information and the sensitive convergence learning function;

[0097] integrate the F working condition-fault detection branches, and obtain the first fault detection channel.

[0098] Further, the deep learning module is further configured to:

[0099] obtain fth working condition information according to the F-level working condition information, and f is a positive integer belonging to F;

[0100] based on the fth working condition information, feature recognition is performed on the same family fault record library and the first injection molding machine fault record library to obtain an fth working condition same family fault record source and an fth injection molding machine fault record data;

[0101] The fth working condition same family fault record source is supervised trained to obtain an fth working condition fault analysis network;

[0102] The fth working condition fault analysis network is tested and analyzed based on the fth injection molding machine fault record data and the sensitive convergence learning function to obtain an fth fault sensitive convergence degree;

[0103] It is judged whether the fth fault sensitive convergence degree meets a preset fault sensitive convergence constraint;

[0104] If the fth fault sensitive convergence degree meets the preset fault sensitive convergence constraint, an fth working condition-fault detection branch is obtained according to the fth working condition fault analysis network, and the fth working condition-fault detection branch is added to the F working condition-fault detection branches.

[0105] Further, the deep learning module is further used for:

[0106] If the fth fault sensitive convergence degree does not meet the preset fault sensitive convergence constraint, a loss data set is obtained according to the fth injection molding machine fault record data;

[0107] The fth working condition fault analysis network is incrementally learned according to the loss data set to obtain an fth incremental working condition fault analysis network;

[0108] The fth working condition-fault detection branch is generated according to the fth incremental working condition fault analysis network.

[0109] Further, the fault analysis module is further used for:

[0110] Based on the real-time running working condition identifier, feature recognition is performed on the first fault detection channel to obtain a matching working condition-fault detection branch;

[0111] The real-time working condition state data is input into the matching working condition-fault detection branch to obtain a fault detection report, wherein the fault detection report includes fault type information, a fault probability coefficient and a fault influence coefficient;

[0112] A preset fault detection weighting operator is obtained, and the fault probability coefficient and the fault influence coefficient are weighted calculated according to the preset fault detection weighting operator to generate the fault detection coefficient.

[0113] Further, the fault warning module is further used for:

[0114] determining whether the fault detection coefficient satisfies the fault detection early warning constraint;

[0115] If the fault detection coefficient satisfies the fault detection early warning constraint, a pre-constructed fault early warning signal level table is called.

[0116] The fault detection coefficient is input into the fault early warning signal level table to generate the fault early warning signal.

[0117] The fault early warning system of the injection molding upper and lower material automatic machine provided by the embodiment of the application can execute the fault early warning method of the injection molding upper and lower material automatic machine provided by any embodiment of the application, and has the corresponding function modules and beneficial effects of the execution method.

[0118] Each module included is only divided according to the function logic, but is not limited to the above division, as long as the corresponding function can be realized; in addition, the specific name of each function module is only for easy mutual differentiation, and is not used to limit the protection scope of the application.

[0119] The application provides a fault early warning method of an injection molding upper and lower material automatic machine, wherein the method is applied to a fault early warning system of an injection molding upper and lower material automatic machine, and the method comprises the following steps: collecting fault records of a first injection molding upper and lower material automatic machine according to multi-dimensional fault analysis indexes and F-grade working condition information, and obtaining a first injection machine fault record library; performing deep learning according to the F-grade working condition information and the first injection machine fault record library based on a sensitive convergence learning function, and generating a first fault detection channel; inputting real-time running data sources of the first injection molding upper and lower material automatic machine into the first fault detection channel, and obtaining a fault detection coefficient; and if the fault detection coefficient satisfies a fault detection early warning constraint, generating a fault early warning signal. The technical problem of low fault detection accuracy of the injection molding upper and lower material automatic machine in the prior art, which leads to poor fault early warning effect of the injection molding upper and lower material automatic machine, is solved. The technical effect of improving the fault detection accuracy and comprehensiveness of the fault detection of the injection molding upper and lower material automatic machine, and improving the fault early warning quality of the injection molding upper and lower material automatic machine is achieved.

[0120] Although the application is described in detail through the above embodiments, the application is not limited to the above embodiments only, and can include more other equivalent embodiments without departing from the concept of the application, and the scope of the application is determined by the appended claims.

Claims

1. A method for early warning of faults in an in-mold trimming automatic machine, characterized in that, The method includes: Obtain the F-level operating condition information of the first automatic injection molding machine, where F is a positive integer greater than 1; Obtain multidimensional fault analysis indicators, wherein the multidimensional fault analysis indicators include fault type, fault probability, and fault impact; Based on the multi-dimensional fault analysis indicators and the F-level working condition information, the fault records of the first automatic injection molding machine are collected to obtain the first injection molding machine fault record library. Based on the sensitive convergence learning function, deep learning is performed according to the F-level working condition information and the first injection molding machine fault record library to generate the first fault detection channel. Obtain the real-time operation data source of the first automatic injection molding machine, wherein the real-time operation data source includes a real-time operation condition identifier and real-time operation condition status data corresponding to the real-time operation condition identifier. Based on the real-time running data source, fault analysis is performed according to the first fault detection channel to obtain the fault detection coefficient; If the fault detection coefficients satisfy the fault detection and early warning constraints, a fault early warning signal is generated. Based on a sensitive convergence learning function, deep learning is performed using the F-level operating condition information and the first injection molding machine fault record library to generate a first fault detection channel, including: Construct the sensitive convergence learning function, wherein the sensitive convergence learning function is: FSN = e fsc-fsw ; Among them, FSN represents the fault-sensitive convergence, fsc represents the fault-sensitive accuracy, and fsw represents the fault-sensitive error. Based on the first automatic injection molding machine, a group of injection molding machines of the same family is matched; Based on the multidimensional fault analysis indicators, retrieve the fault record data of the same injection molding machine group to obtain the same family fault record library; Based on the F-level working condition information and the sensitive convergence learning function, F working condition-fault detection branches are trained according to the same family of fault record library and the first injection molding machine fault record library. By integrating the F operating condition-fault detection branches, the first fault detection channel is obtained; Based on the F-level operating condition information and the sensitive convergence learning function, and according to the same family of fault records and the first injection molding machine fault record library, F operating condition-fault detection branches are trained, including: Based on the F-level working condition information, the f-th working condition information is obtained, where f is a positive integer belonging to F; Based on the f-th working condition information, feature recognition is performed on the same family of fault record library and the first injection molding machine fault record library to obtain the same family of fault record source for the f-th working condition and the f-th injection molding machine fault record data. Supervised training is performed on the fault record sources of the same family under the f-th working condition to obtain the fault parsing network for the f-th working condition; Based on the fault record data of the f-th injection molding machine and the sensitive convergence learning function, the fault parsing network of the f-th working condition is tested and analyzed to obtain the sensitive convergence degree of the f-th fault. Determine whether the f-th fault-sensitive convergence satisfies the preset fault-sensitive convergence constraint; If the f-th fault sensitivity convergence satisfies the preset fault sensitivity convergence constraint, the f-th working condition fault detection branch is obtained according to the f-th working condition fault parsing network, and the f-th working condition fault detection branch is added to the F working condition fault detection branches.

2. The method of claim 1, wherein, Based on the multi-dimensional fault analysis index and the F-level working condition information, the fault record of the first injection molding up-down feeding automatic machine is collected, and a first injection molding machine fault record library is obtained, including: obtaining a preset historical time zone constraint; Based on the multi-dimensional fault analysis index and the preset historical time zone constraint, according to the F-level working condition information, F retrieval constraints are generated; According to the F retrieval constraints, the fault record of the first injection molding up-down feeding automatic machine is collected, and F working condition-fault record sources are obtained; Based on the F working condition-fault record sources, principal component analysis is performed to generate the first injection molding machine fault record library.

3. The method of claim 1, wherein, determine whether the fth fault sensitive convergence degree meets the preset fault sensitive convergence constraint, further comprising: if the fth fault sensitive convergence degree does not meet the preset fault sensitive convergence constraint, obtain a loss data set according to the fth injection molding machine fault record data; According to the loss data set, the fth incremental working condition fault analysis network is obtained by performing incremental learning on the fth working condition fault analysis network; According to the fth incremental working condition fault analysis network, the fth working condition-fault detection branch is generated.

4. The method of claim 1, wherein, Based on the real-time running data source, fault analysis is performed according to the first fault detection channel to obtain a fault detection coefficient, including: Based on the real-time running working condition identifier, the first fault detection channel is identified to obtain a matching working condition-fault detection branch; input the real-time working condition state data into the matching working condition-fault detection branch to obtain a fault detection report, wherein the fault detection report includes fault type information, fault probability coefficient and fault influence coefficient; obtain a preset fault detection weighting operator, and perform weighted calculation on the fault probability coefficient and the fault influence coefficient according to the preset fault detection weighting operator to generate the fault detection coefficient.

5. The method of claim 1, wherein, If the fault detection coefficient meets the fault detection early warning constraint, a fault early warning signal is generated, including: determine whether the fault detection coefficient meets the fault detection early warning constraint; if the fault detection coefficient meets the fault detection early warning constraint, the pre-constructed fault early warning signal level table is called; input the fault detection coefficient into the fault early warning signal level table to generate the fault early warning signal.

6. A fault early warning system for an in-mold trimming automatic machine, characterized by, The system is used to execute the method of any one of claims 1 to 5, and the system comprises: working condition information obtaining module, the working condition information obtaining module is used for obtaining F-level working condition information of first injection molding up-down feeding automatic machine, wherein F is a positive integer greater than 1; fault analysis index obtaining module, the fault analysis index obtaining module is used for obtaining multi-dimensional fault analysis index, wherein the multi-dimensional fault analysis index includes fault type, fault probability and fault influence; fault record collection module, the fault record collection module is used for collecting fault record of the first injection molding up-down feeding automatic machine based on the multi-dimensional fault analysis index and the F-level working condition information, and obtaining a first injection molding machine fault record library; The deep learning module is used for deep learning based on a sensitive convergence learning function according to the F-level working condition information and the first injection molding machine fault record library to generate a first fault detection channel; The running data source obtaining module is used for obtaining a real-time running data source of the first injection molding feeding and discharging automatic machine, wherein the real-time running data source includes a real-time running working condition identifier, and real-time working condition state data corresponding to the real-time running working condition identifier; The fault analysis module is used for fault analysis based on the real-time running data source according to the first fault detection channel to obtain a fault detection coefficient; The fault early warning module is used for generating a fault early warning signal if the fault detection coefficient meets a fault detection early warning constraint.

Images