Optimal control support apparatus for injection molding machine

Abstract

This invention relates to an optimal control auxiliary device and method for suppressing the generation of harmful gases in an injection molding machine. More specifically, by applying the Taguchi Method and using gas collection data as characteristic values, the signal-to-noise ratio (SNR) of the characteristic values ​​of the feature vectors based on control parameter data is obtained. A model is prepared to predict the SNR using the feature vectors. Among multiple control parameter data used in the process, the optimal control parameter data is recommended based on the SNR predicted by the model and applied to the process execution of the injection molding machine. This minimizes the variation in gas generation and reduces the amount of gas generated, thereby enabling stable purification of the generated gases before emission. This technology is the result of a regional intelligent innovation talent cultivation project developed with funding from the Korean government (Ministry of Science and ICT) and support from the Korea Information and Communication Planning and Evaluation Institute (RS-2024-00436765).

Description

[Technical Field]

[0001] This invention relates to an optimal control auxiliary device and method for suppressing the generation of harmful gases in an injection molding machine. [Previous Technology]

[0002] A plastic injection molding machine is a device that performs injection molding processes including mold closing, filling, packing, cooling, and mold opening. In the filling process, a barrel is used to melt solid plastic raw materials and compress them before injecting them into the mold cavity through a nozzle. In the packing process, a certain pressure is maintained. The molding conditions are determined according to the plastic raw materials, the shape and size of the molded product, the structure of the nozzle and the barrel, etc., and control parameters are set to obtain the molding conditions, so that the plastic injection molding machine operates based on the values ​​of the set control parameters.

[0003] However, harmful gases are generated and emitted during the process of melting plastic raw materials in the barrel. This not only deteriorates the working environment but also reduces the quality of the molded products. Therefore, a gas emitter is installed in the barrel or between the barrel and the nozzle, and a gas collector / purifier is operated at the same time so that the gas is drawn in and purified by the gas emitter.

[0004] However, depending on the raw materials and molding conditions, the type and amount of gas produced will vary. This may not only make it impossible to process the gas through the gas collector / purifier, but also affect the replacement cycle of the filter used for gas purification, making it difficult to replace it at the appropriate time.

[0005] Previously, based on empirical data, the temperature of the barrel that has the greatest impact on gas generation was set to the temperature with the lowest gas generation intensity. However, in addition to the barrel temperature, the type and amount of gas generated will vary depending on various control parameters. Therefore, there are limitations in suppressing gas generation.

[0006] In particular, even if the injection molding machine is controlled according to the same control parameter values, the actual molding conditions will vary due to various factors such as ambient temperature and humidity, varying flow rate and speed of molten resin, and the amount of gas generated will also vary according to these factors, making it impossible to provide a clear value for the optimal control parameters.

[0007] In addition, although there are techniques in the prior art that optimize molding conditions in order to minimize the defect rate of molded products, these techniques do not provide optimal control parameter values ​​for suppressing gas generation.

[0008] [Preferred Technical Documents] Patent Documents Patent Document 1: KR 10-2298636 B1 2021.08.31. Patent Document 2: KR 10-2311048 B1 2021.10.01. [Summary of the Invention]

[0009] [Problem to be solved by the invention] Therefore, the object of the present invention is to provide an optimal control auxiliary device and auxiliary method for injection molding machines, which can derive control parameters that can suppress gas generation, thereby applying them to the process of injection molding machines, and being able to cope with changes in the amount of gas generated in the actual process to operate in a state that suppresses gas generation.

[0010] [Method for Solving the Problem] To achieve the above objective, the present invention provides an optimal control auxiliary device for an injection molding machine, which is used in an injection molding machine 100 to suppress gas emissions. The injection molding machine melts plastic raw material and injects it into a mold 1, collects and detects the gas generated in the molten resin to be injected, purifies the gas, and then discharges it. The process is executed according to control parameter data set by a controller 160 to satisfy molding conditions. The optimal control auxiliary device for the injection molding machine includes: a data processing unit 12, which, upon receiving multiple control parameter data applicable to the process, embeds each control parameter data into a feature space to generate multiple feature vectors; an inference unit 14, which prepares a prediction model that uses the feature vectors to predict the signal-to-noise ratio (SNR) of the feature values ​​used for gas collection data in order to use the Taguchi Method, and derives the SNR based on each feature vector using the prediction model; and an application unit 15, which, according to the Taguchi Method... The method analyzes and predicts multiple signal-to-noise ratios, determines the control parameter data corresponding to any one of the signal-to-noise ratios as the optimal control parameter, and applies the optimal control parameter data to the process.

[0011] As an embodiment, the optimal control auxiliary device for injection molding machines according to the present invention further includes a training unit 13. When receiving multiple control parameter data applied to process execution and gas collection data measured by performing multiple processes using the various control parameter data, the training unit 13 uses multiple feature vectors generated by the data processing unit 12 on the multiple control parameter data and the gas collection data to generate training data, thereby training a prediction model that uses feature vectors to estimate the signal-to-noise ratio.

[0012] To achieve the above objective, the present invention provides an optimal control assistance method for an injection molding machine, which is used to suppress gas emissions in an injection molding machine 100. The injection molding machine 100 melts plastic raw material and injects it into a mold 1, collects and detects the gas generated in the molten resin to be injected, purifies the gas, and then discharges it. The process is executed according to control parameter data set by a controller 160 to meet molding conditions. This optimal control assistance method for an injection molding machine is implemented by an optimal control assistance device for an injection molding machine, including a data processing unit 12, an inference unit 14, and an application unit 15. The method includes: a data processing step S30, in which, when multiple control parameter data applicable to the process are received, the data processing unit 12 embeds each control parameter data into a feature space to generate multiple feature vectors; and an inference step S40, in which the feature vectors are used to predict the signal-to-noise ratio (SNR) of the feature values ​​used for gas collection data in order to use the Taguchi Method. The inference unit 14 of the prediction model of the Ratio (SNR) predicts the signal-to-noise ratio based on each feature vector by the prediction model; and the application step S50, in which the multiple signal-to-noise ratios predicted by the application unit 15 are analyzed according to the Taguchi Method, the control parameter data corresponding to any signal-to-noise ratio is determined as the optimal control parameter, and the optimal control parameter data is applied to the process.

[0013] As an embodiment, the optimal control assistance method for injection molding machines according to the present invention includes a training unit 13. When receiving multiple control parameter data applied to process execution and gas collection data measured by performing multiple processes using the various control parameter data, the training unit 13 receives multiple feature vectors generated by the data processing unit 12 from the multiple control parameter data. The method further includes: a training data generation step S10, in which the training unit 13 generates training data using the multiple feature vectors and the gas collection data; and a training step S20, in which a prediction model for estimating the signal-to-noise ratio using the feature vectors is trained using the training data.

[0014] [Effect of the Invention] The present invention applies the Taguchi Method to analyze characteristic values ​​based on control parameters, defines gas collection data as characteristic values, and uses a predictive model that predicts characteristic values ​​through the eigenvectors of control parameters. Thus, for various control parameters used in the injection molding machine, the optimal data that can effectively suppress gas generation is recommended and the process is executed, thereby minimizing the variation in gas generation and reducing the amount of gas generation to the greatest extent, so that the generated gas can be stably purified before being discharged.

Implementation Method

[0019] Hereinafter, embodiments of the present invention will be specifically described with reference to the drawings, so that those skilled in the art can easily implement the present invention. However, it is undeniable that various changes or modifications can be made to the embodiments of the present invention within the scope of the present invention, and therefore the present invention is not limited to the embodiments described below.

[0020] Known components, circuits, functions, methods, and typical related details can be added to and implemented in the embodiments of this invention by those skilled in the art; therefore, they will not be repeated here. Furthermore, the injection molding machine driven in connection with the embodiments of this invention will be briefly described to the extent that the embodiments of this invention can be understood.

[0021] Embodiments of the present invention can be implemented by a combination of software and hardware, which can be described as components, modules and units, and can be implemented by readable code in a computer presented on a recording medium.

[0022] The connection between components can be wired or wireless to transmit data, signals and messages, and can also include cases where they are indirectly connected through other components.

[0023] Regarding the inclusion of a certain component, unless otherwise specified, it should be understood as further including other components, rather than excluding other components.

[0024] Referring to FIG1, the injection molding machine optimal control auxiliary device 10 according to an embodiment of the present invention is configured to exchange data with the controller 160 of the injection molding machine 100 via wired or wireless means. As described in detail below, it receives control parameter data of the injection molding machine 100, derives optimal control parameters that can suppress gas generation using the Taguchi Method, and transmits them to the injection molding machine 100, thereby enabling management to generate only the appropriate level of gas that can be processed in the injection molding machine 100.

[0025] First, in order to facilitate understanding of the embodiments of the present invention, the hopper 110, barrel 120, nozzle 130, gas emitter 140, gas collector / purifier 150 and controller 160 constituting the injection molding machine 100 will be briefly described. The control parameters analyzed by statistical methods in the present invention and the molding conditions related to the control parameters will be illustrated by example, and the gas collection data will be explained.

[0026] Hopper 110 is typically used to feed plastic raw materials, which are manufactured into granular solid state, into barrel 120. Additionally, existing devices include a dryer for dehumidifying and drying the plastic raw materials. However, as described below, if the device can operate optimally to remove gases and moisture as in this invention, such a dryer is unnecessary. This not only reduces device costs but also lowers power consumption.

[0027] A screw 121 is provided inside the barrel 120. The screw 121 can move back and forth and rotate by a screw driver 122 provided at the rear end of the barrel 120. The position is adjusted by moving the screw 121 back and forth, and the rotation applies pressure to the plastic raw material supplied from the hopper 110 and conveys it to the front end head. A heater and a temperature sensor are provided on the outer periphery to control the temperature and heat it, so as to melt the plastic raw material and convey it to the front end head.

[0028] The screw 121 is configured to include a feeding zone, a compression zone, and a metering zone. Each zone is equipped with a heater and a temperature sensor to control the temperature of each zone. Therefore, the plastic raw material fed into the feeding zone is preheated and advanced at the same time. After being melted and compressed in the compression zone, it is transported to the metering zone. In the metering zone, the molten resin is supplied to the nozzle side for injection.

[0029] Nozzle 130 is used to inject molten resin supplied from barrel 120 into cavity 1a of mold 1. A heater and temperature sensor may also be provided in nozzle 130 to maintain the molten resin to be injected at a preset temperature.

[0030] The gas emitter 140 has a channel for emitting gases and moisture generated in the molten resin before it is injected through the nozzle 130. As shown, for example, as in the specific structure described in Korean Patent Registration No. 10-2298636, it can be disposed between the barrel 120 and the nozzle 130. In this case, a heater and a temperature sensor can be disposed. However, this is not a limitation on the gas emitter. For example, as disclosed in Korean Patent Registration No. 10-2379514, it can also be disposed on the metering zone side at the front end of the barrel 120.

[0031] Although not shown in the figure, it also includes a mechanism for moving the injection molding machine 100 to the mold.

[0032] The gas collector / purifier 150 draws in gas and moisture through the gas emitter 140, removes moisture using an oil-water filter, measures the oxygen concentration using a gas sensor, thereby measuring the amount of harmful gases, and emits purified gas that has been collected and had harmful gases and harmful substances such as tar removed by the gas filter. Here, the intake volume can be adjusted to an appropriate amount.

[0033] The controller 160 sets control parameter values ​​to obtain molding conditions determined based on plastic raw materials, the size and shape of the molded product, and the device characteristics of the injection molding machine, and controls the actions of the above-mentioned components so as to execute the injection molding process according to the control parameter values.

[0034] Here, molding conditions include, for example, resin temperature of each zone, injection pressure, injection speed, injection volume, injection capacity, holding pressure, metering distance, holding transition distance, injection time, holding time, and metering time. Additionally, control parameters include temperature of each zone, screw position, rotational speed and torque, and operating time.

[0035] When the injection molding process is controlled and processed using control parameter data set and used according to the injection molding process, and according to the control parameter data of each process, the injection molding machine optimal control auxiliary device according to an embodiment of the present invention receives gas collection data measured by the gas collector / purifier 150 as training data. Furthermore, the gas collection data is defined as characteristic values ​​of the Taguchi method, thereby obtaining a prediction model for the signal-to-noise ratio of the predicted characteristic values.

[0036] Here, the control parameter data refers to the data used when repeatedly performing the injection molding process using the same plastic raw material, and is the data used when the values ​​of each item are set to the same value and the process is performed at least twice, and when the values ​​of each item are changed and these multiple processes are repeated. If the injection molding machine 100 is used as an apparatus for producing products using different plastic raw materials, the control parameter data is the data applied according to the plastic raw material. Of course, considering that the gas generated varies depending on the size and shape of the molded product, the control parameter data may be the data obtained according to the molded product.

[0037] Gas collection data is data obtained by repeatedly executing the process using the same control parameter data, and can be the amount of gas generated as measured by a gas sensor.

[0038] Such control parameter data and gas collection data can be obtained by implementing process tests using an orthogonal permutation table based on the Taguchi Method, thereby obtaining the most information-rich data with the fewest number of process tests.

[0039] In addition, according to an embodiment of the present invention, when the injection molding machine optimal control auxiliary device receives multiple control parameter data for executing the actual process, it predicts the signal-to-noise ratio based on each control parameter by means of a prediction model, and then selects the optimal control parameter data by means of the signal-to-noise ratio according to the Taguchi method and applies it to the process execution of the injection molding machine 100.

[0040] Therefore, the optimal control auxiliary device 10 for injection molding machines according to an embodiment of the present invention will be described in detail.

[0041] Referring to the block diagram of FIG2, the injection molding machine optimal control auxiliary device 10 according to an embodiment of the present invention includes an interface unit 11, a data processing unit 12, a training unit 13, an inference unit 14, and an application unit 15.

[0042] The interface unit 11 is connected to the controller 160 of the injection molding machine 100 via wired or wireless means, thereby performing data communication and association operations, receiving data for training the prediction model, and transmitting the selected optimal control parameter data after receiving multiple control parameter data to be applied to the actual process execution.

[0043] Here, as described above, the data received for training is data obtained by conducting process experiments in advance, including multiple control parameter data applied to each process, and gas collection data of gases generated when performing the injection molding process using each control parameter data.

[0044] The multiple control parameter data used for actual process execution are multiple control parameter data that can be applied to the currently executed process in the controller 160. Among such multiple control parameter data, the optimal control parameter data selected by the method described later is transmitted to the controller 160.

[0045] Alternatively, after receiving the control parameter data and gas collection data used during the actual process execution, the prediction model can be trained as described later.

[0046] The data processing unit 12 is used to embed the control parameter data received by the interface unit 11 into the feature space to generate a feature vector.

[0047] The feature vector is generated by embedding the control parameter data into the high-dimensional feature space using a kernel function. As is well known, among various kernel functions, a kernel function suitable for the data characteristics can be used to generate the feature vector to estimate the regression coefficients of the prediction model used as a regression analysis model in the embodiments of the present invention.

[0048] When transmitting data for training, a set of control parameters consisting of multiple control parameter data is embedded into the feature space to generate multiple feature vectors, which are then transmitted to the training unit 13 for training the prediction model.

[0049] When transmitting multiple control parameter data for performing the actual process, the control parameter set consisting of multiple control parameter data can also be embedded into the feature space to generate multiple feature vectors, and transmitted to the inference unit 14 for deriving the signal-to-noise ratio by the prediction model.

[0050] In addition, when transmitting control parameter data and gas collection data used in the actual process execution, feature vectors can also be generated for training purposes and transmitted to training unit 13 to retrain the prediction model.

[0051] The control parameters include not only the contribution parameters that contribute significantly to the characteristic value (gas collection data), but also the noise parameters that contribute relatively little or nothing. Feature vectors reflecting these characteristics can be generated here.

[0052] Furthermore, when the data processing unit 12 receives control parameter data differentiated according to plastic raw materials, whether the control parameters are used for training or retraining or for executing the actual process, the control parameter data is differentiated according to the plastic raw materials, thereby generating multiple feature vectors according to the plastic raw materials. Furthermore, when it receives control parameter data for molded products produced by the injection molding machine 100, the control parameter data is differentiated according to the molded products, thereby generating multiple feature vectors according to the molded products.

[0053] After the training unit 13 generates multiple feature vectors using the control parameter set in the received training data and multiple gas collection data in the received training data, it trains a prediction model that estimates the signal-to-noise ratio (SNR) of the feature vectors based on the characteristic values ​​of the Taguchi Method (gas collection data in this invention).

[0054] The Taguchi method is a method for robust design using controllable factors to address uncontrollable noise factors. This method defines a signal-to-noise ratio (SNR) that reflects not only the target value of the quality characteristic value but also the dispersion (or deviation) of these characteristic values, thereby determining the degree of control factors with a large SNR.

[0055] In order to use the Taguchi method, the characteristic values ​​that represent quality in the Taguchi method are defined as gas collection data and used.

[0056] The characteristic values ​​used as gas collection data have the property that the lower the better, so the signal-to-noise ratio (SNR) can be expressed as the following mathematical formula 1.

[0057] [Mathematical Formula 1] Where i is the exponent used to distinguish control parameter data, j and n are the exponent and process number used to distinguish processes executed using the same control parameter data, respectively, is the characteristic value (gas collection data) measured when the process is executed using the i-th control parameter data, and is the signal-to-noise ratio of the characteristic value obtained after the n-th process is executed using the i-th control parameter data.

[0058] The signal-to-noise ratio is as follows: the smaller the dispersion or deviation of the characteristic value, the larger the value; and the smaller the average of the characteristic value, the larger the value. It also exhibits robustness to noise control parameters in the control parameters or noise factors other than noise control parameters.

[0059] The signal-to-noise ratio expressed in this mathematical formula 1 can be calculated using the characteristic values ​​(gas collection data) obtained according to the data of each control parameter, thereby generating an actual vector consisting of the signal-to-noise ratios of the actual measured characteristic values.

[0060] In this invention, instead of directly using the conventional Taguchi method to calculate the signal-to-noise ratio based on control parameter data from experimental applications to determine the optimal control parameter values, a model for predicting the signal-to-noise ratio is generated, and the signal-to-noise ratio based on control parameter data from actual applications is predicted, thereby determining the optimal control parameter data.

[0061] The prediction model for the signal-to-noise ratio used to predict this characteristic value can be a regression model, such as the Support Vector Regression (SVR) model, which represents the regression model.

[0062] Such regression models and training methods used for prediction models belong to known techniques, so only their main contents are introduced here.

[0063] The SVR model used to implement the present invention is for nonlinear regression, and as is well known, it can be represented by the regression function of mathematical formula 2.

[0064] [Mathematical Formula 2] where is the signal-to-noise ratio, is the vector of actual measurement used during training, is the weight vector, b is a constant representing the bias value, and is the kernel function.

[0065] Furthermore, the regression function is completed by using the feature vectors of the training data and the actual measurement vectors to derive the objective function shown in the following mathematical formula 3 and the weight vectors and bias values ​​b that satisfy the constraints.

[0066] [Mathematical Expression 3] satisfies the following constraints (subject to.)

[0067] Where is a precision parameter representing the radius of the regression function pipeline, and C is a normalization constant. The precision parameter and the constant C are preset. and represent the distance of the eigenvector from the regression function pipeline. represents the kernel function, corresponding to the one in mathematical formula 2. is the aforementioned eigenvector, is the value of the aforementioned actual measurement vector, and w is the slope variable used to represent the interval.

[0068] Next, multiple feature vectors from the training data are input into the regression function to obtain the predicted vector, and then the accuracy is calculated based on the difference between the predicted vector and the actual measured vector. Of course, multiple feature vectors for the testing purpose can also be used separately for training and testing purposes. Here, the accuracy can be the reciprocal of the difference between the predicted vector and the actual measured vector, and can be derived according to the following mathematical formula 4.

[0069] [Mathematical Expression 4]

[0070] Where Eacc represents accuracy, is the value of the actual measurement vector, is the value of the prediction vector, and i is the exponent of the training data.

[0071] If the accuracy derived therefrom is lower than the preset threshold, the training process of re-obtaining the weight vector and bias value using the training data is repeated. If the accuracy is above the threshold, the training is terminated and the regression function obtained at this time is selected as the prediction model.

[0072] In this invention, the training unit 13 is characterized in that it uses gas collection data as characteristic value, defines the signal-to-noise ratio based on the minimization characteristic of the characteristic value, and thereby obtains a prediction model for the signal-to-noise ratio.

[0073] Furthermore, when training data (multiple control parameter data and gas collection data) is received according to the plastic raw material, multiple feature vectors generated by the multiple control parameter data and the received gas collection data are used according to the plastic raw material to generate training data according to each plastic raw material and generate a prediction model according to the plastic raw material.

[0074] In addition, when training data (multiple control parameter data and gas collection data) is received according to the molded products produced by the injection molding machine 100, multiple feature vectors generated according to the molded products and the received gas collection data are used to generate training data according to the molded products and generate a prediction model for use according to the molded products.

[0075] In addition, when additional control parameter data and gas collection data received based on actual process processing are received, the generated feature vector and the received gas collection data are appended to the training data. When the training data increases by a predetermined amount, the prediction model is retrained to update the prediction model.

[0076] As described above, the prediction model generated in the training unit 13 is transmitted to the inference unit 14.

[0077] The inference unit 14 has completed the preparation of receiving and using the prediction model generated by the training unit 13. It can prepare prediction models that are distinguished by plastic raw materials, and it can also prepare prediction models that are distinguished by molded products.

[0078] In addition, when the data processing unit 12 transmits multiple feature vectors generated by embedding the received multiple control parameter data into the feature space based on the multiple control parameter data received by the interface unit 11 for performing the actual process, the inference unit 14 predicts the signal-to-noise ratio based on each feature vector by means of a prediction model.

[0079] Here, the multiple control parameter data are different control parameter data that can be applied to the process, and may include data that is different from the data used for training purposes. Furthermore, the prediction vector is the value of the predicted signal-to-noise ratio, which is derived by sequentially selecting multiple feature vectors and generating a prediction model.

[0080] If the plastic raw material used in the process is a specified feature vector, then the prediction vector is derived using the prediction model corresponding to the plastic raw material. If the molded product produced by the process is a specified feature vector, then the prediction vector is also derived using the prediction model corresponding to the molded product.

[0081] The prediction vector containing the predicted values ​​of each signal-to-noise ratio is transmitted to the application unit 15.

[0082] Furthermore, although the data flow is not shown in Figure 2, after receiving multiple control parameter data and gas collection data for training purposes, the inference unit 14 analyzes the influence of each control parameter on gas production, thereby inferring control parameters that are adjusted based on changes in the amount of harmful gas produced. For example, changes in gas collection data that occur when the values ​​of each control parameter change can be obtained through data analysis.

[0083] The application unit 15 compares and analyzes the multiple signal-to-noise ratios contained in the prediction vector, and determines the control parameter data corresponding to any given signal-to-noise ratio as the optimal control parameter data. At this time, the determined optimal control parameter data is transmitted to the injection molding machine 100, causing the injection molding machine 100 to execute the process using the optimal control parameters. Undoubtedly, an identification message for the determined optimal control parameter data can be transmitted so that the injection molding machine 100 can identify which of the multiple control parameter data has been transmitted.

[0084] Here, the control parameter data corresponding to the signal-to-noise ratio with the maximum value among multiple signal-to-noise ratios can be determined as the optimal control parameter data.

[0085] Furthermore, the application unit 15 can determine the gas intake amount based on the signal-to-noise ratio corresponding to the optimal control parameter data, and transmit it along with the optimal control parameter data to the injection molding machine 100 for use in controlling the gas intake amount. In other words, control is performed based on the gas intake amount determined here, and the gas intake amount can be adjusted according to the amount of gas generated. For example, during the process of controlling gas and moisture intake based on the gas intake amount determined here, the amount of harmful gas generated by the gas sensor can be monitored, and the intake amount can be adjusted according to the amount of harmful gas generated.

[0086] In addition, the application unit 15 receives control parameters adjusted based on changes in the amount of gas generated from the inference unit 14 and transmits them to the injection molding machine 100, thereby enabling the adjustment of the amount of gas and moisture inhaled by monitoring the amount of harmful gas generated by the gas sensor.

[0087] As described above, the application unit 15 is connected to the injection molding machine 100 for operation, thereby enabling the process to be executed using the optimal control parameters with small fluctuations in gas production and small gas production. Furthermore, the process can be controlled according to the gas production, thereby enabling the gas collector / purifier 150 to collect the entire amount of gas production and perform stable purification, thereby fundamentally preventing the emission of unpurified harmful gases.

[0088] Figure 3 is a structural diagram of an injection molding machine 100 that presents an optimal control auxiliary device 10 for injection molding machines according to a modified embodiment of the present invention in a controller 160.

[0089] Since the controller 160 of the injection molding machine 100 has a process control processor, the optimal control auxiliary device for this injection molding machine can be presented as a programmatic component in the processor. At this time, the interface unit 11 mentioned above can be omitted, and it can be presented as a program that receives data processed by the process control program and runs it.

[0090] Furthermore, the prediction model is not generated by the training unit 13, but is pre-generated and then transferred to the inference unit 14 when assembling the optimal control auxiliary device for the injection molding machine. At this time, it is also suitable to use the training unit 13 for retraining, and the training data used to generate the prediction model can be stored for use during retraining in the training unit 13.

[0091] Figure 4 is a flowchart of the optimal control auxiliary method for injection molding machine, which consists of an optimal control auxiliary device for injection molding machine.

[0092] The optimal control assistance method for injection molding machines is executed by the data processing unit 12, training unit 13, reasoning unit 14 and application unit 15 of the optimal control assistance device 10 for injection molding machines described in detail above. Therefore, repeated detailed descriptions are omitted, and the main contents of each step are explained.

[0093] The optimal control assistance method for injection molding machines may include the following steps: training data generation step S10 and training step S20 for training a prediction model; data processing step S30 for deriving and providing the optimal control parameter data executed in the actual process by means of the prediction model; inference step S40 and application step S50; and training data appending step S60 and retraining step S70 for retraining the prediction model.

[0094] The training data generation step S10 and the training step S20 are steps performed when data for training purposes is received from the injection molding machine 100, and are performed by the data processing unit 12 and the training unit 13. Here, as described above, the data for training purposes includes multiple control parameter data used in the test or actual process execution, and multiple gas collection data for each control parameter data.

[0095] In the training data generation step S10, the following steps are performed in sequence: the data processing unit 12 generates multiple feature vectors using multiple control parameter data and transmits them to the training unit 13; and the training unit 13 generates training data using multiple feature vectors and multiple gas collection data.

[0096] In this training step S20, a prediction model based on the signal-to-noise ratio estimated by the feature vector is trained using the training data, and then the prediction model is transferred to the inference step S40 for its use.

[0097] If the data used for training is distinguished according to plastic raw materials, the training data generation step S10 generates training data according to plastic raw materials based on multiple feature vectors generated according to plastic raw materials and multiple gas collection data, and the training step S20 generates a prediction model according to the plastic raw materials used.

[0098] If the data used for training is differentiated according to the molded product, then in the training data generation step S10, training data is generated according to the molded product based on multiple feature vectors generated according to the molded product differentiation and multiple gas collection data, and in the training step S20, a prediction model is generated according to the molded product.

[0099] When multiple control parameter data that can be applied to the process are received, the data processing unit 12, the inference unit 14 and the application unit 15 execute the data processing step S30, the inference step S40 and the application step S50 to recommend the optimal control parameter data.

[0100] In this data processing step S30, the data processing unit 12 embeds each control parameter data into the feature space to generate multiple feature vectors.

[0101] In this inference step S40, the inference unit 14 uses the prediction model prepared in the training step S20 to derive the signal-to-noise ratio based on each feature vector. Here, the prediction model is a model that predicts the signal-to-noise ratio of the characteristic value required when using the Taguchi Method using feature vectors. It is trained to use gas collection data as characteristic values ​​and derive the characteristic value, i.e., the signal-to-noise ratio, using the feature vectors of the control parameter data. Therefore, the signal-to-noise ratio can be predicted using feature vectors.

[0102] Here, if a prediction model is prepared according to the plastic raw material and the control parameter data is distinguished according to the plastic raw material, then in the data processing step S30, a feature vector is generated by distinguishing the control parameter data according to the plastic raw material, and in the inference step S40, the signal-to-noise ratio is predicted by using the feature vector with the prediction model corresponding to the plastic raw material used in the process.

[0103] If a prediction model is prepared according to the molded product and the control parameter data is distinguished according to the molded product, then in the data processing step S30, a feature vector is generated by distinguishing the control parameter data according to the molded product, and in the inference step S40, the signal-to-noise ratio is predicted by using the feature vector with the prediction model corresponding to the molded product produced by the process.

[0104] In this application step S50, the application unit 15 analyzes multiple signal-to-noise ratios (SNRs) according to the Taguchi Method, determines the control parameter data corresponding to any given SNR as the optimal control parameter, and applies the optimal control parameter data to the injection molding machine 100 process. As an example, the control parameter data corresponding to the SNR with the highest value among multiple SNRs can be determined as the optimal control parameter.

[0105] In addition, even if the injection molding machine 100 is controlled with optimal control parameters and the process is executed, the amount of gas generated may vary due to various factors. Therefore, it is preferable to process the process while dealing with the variation in the amount of gas generated.

[0106] To this end, the reasoning step S40 analyzes the multiple control parameter data and gas collection data used for training purposes, and selects from the multiple items that make up the control parameters the items that are adjusted according to the changes in the amount of harmful gas generated during process execution.

[0107] In addition, when performing the process, the control parameters that are adjusted to reduce the amount of gas generated based on the amount of harmful gas generated are the items selected above.

[0108] Furthermore, in this application step S50, a reference value for the gas intake amount is determined based on the signal-to-noise ratio corresponding to the optimal control parameter data, and this reference value, along with the optimal control parameter data, is transmitted to the injection molding machine 100 for use. Thus, the injection molding machine 100 can control the gas intake amount based on the reference value and adjust the gas intake amount according to the gas generation amount.

[0109] The training data addition step S60 and the retraining step S70 are steps performed when the control parameter data added based on the actual process processing and the gas collection data are received. As during training, they are performed by the data processing unit 12 and the training unit 13.

[0110] In the training data appending step S60, the training unit 13 generates training data by means of the data processing unit 12, which includes multiple feature vectors generated from multiple control parameter data and the received gas collection data.

[0111] When additional training data is generated in a predetermined amount, this retraining step S70 is performed, thereby retraining the prediction model using the training data. Here, the retrained prediction model is used in the inference step S40 performed later when recommending the optimal control parameter data.

[0112] Furthermore, the prediction model is not generated by the training unit 13, but is pre-generated and then transferred to the inference unit 14 when assembling the optimal control auxiliary device for the injection molding machine. At this time, the training data generation step S10 and the training step S20 are not executed, but the training unit 13 for retraining is preferably run, and the training data used when generating the prediction model is stored so that it can be used during retraining in the training data appending step S60 and the retraining step S70. [Simplified Explanation of the Diagram]

[0015] FIG1 is a structural diagram of an injection molding machine 100 connected to an injection molding machine optimal control auxiliary device 10 according to an embodiment of the present invention.

[0016] FIG2 is a block diagram of the optimal control auxiliary device 10 for injection molding machine according to an embodiment of the present invention.

[0017] FIG3 is a structural diagram of an injection molding machine 100 showing an injection molding machine optimal control auxiliary device 10 according to a modified embodiment of the present invention.

[0018] Figure 4 is a flowchart of an optimal control auxiliary method for an injection molding machine according to an embodiment of the present invention.

Claims

1. An optimal control auxiliary device for an injection molding machine, used in an injection molding machine (100) to suppress gas emissions, wherein the injection molding machine (100) melts plastic raw material and injects it into a mold (1), collects and detects the gas generated in the molten resin to be injected, purifies the gas before releasing it, and executes the process according to control parameter data set by a controller (160) in a manner to meet molding conditions, wherein, The optimal control auxiliary device for the injection molding machine includes: a data processing unit (12) that, when receiving multiple control parameter data applicable to the process, embeds each control parameter data into the feature space using a kernel function to generate multiple feature vectors input to the regression analysis model; an inference unit (14) that prepares a regression analysis model as a prediction model, which uses the feature vectors to predict the signal-to-noise ratio (SNR) of the characteristic values ​​used for gas collection data for the Taguchi Method, and derives the SNR based on each feature vector using the prediction model; and an application unit (15) that analyzes the predicted multiple SNRs according to the Taguchi Method, determines the control parameter data corresponding to any one SNR as the optimal control parameter, and applies the optimal control parameter data to the process.

2. The optimal control auxiliary device for the injection molding machine as described in claim 1, wherein, The application unit (15) determines the control parameter data corresponding to the largest signal-to-noise ratio among multiple signal-to-noise ratios as the optimal control parameter.

3. The optimal control auxiliary device for the injection molding machine as described in claim 1, wherein, The optimal control auxiliary device for the injection molding machine also includes a training unit (13). When receiving multiple control parameter data applied to process execution and gas collection data measured by performing multiple processes using the various control parameter data, the training unit (13) uses multiple feature vectors generated by the data processing unit (12) on the multiple control parameter data and the gas collection data to generate training data, thereby training a prediction model that uses feature vectors to estimate the signal-to-noise ratio.

4. The optimal control auxiliary device for the injection molding machine as described in claim 3, wherein, The reasoning unit (14) analyzes multiple control parameter data and gas collection data used in the training unit (13) to select control parameters that are adjusted according to the changes in the amount of harmful gas generated during process execution. The application unit (15) uses the selected control parameters as control parameters to reduce the amount of gas generated based on the amount of harmful gas generated during process processing.

5. The optimal control auxiliary device for the injection molding machine as described in claim 1, wherein, This signal-to-noise ratio uses a signal-to-noise ratio based on the lower-the-better characteristic.

6. The optimal control auxiliary device for the injection molding machine as described in claim 1, wherein, The application unit (15) determines the amount of gas inhalation based on the signal-to-noise ratio and includes it in the optimal control parameter data, thereby controlling the amount of gas inhalation.

Images