Injection molding method and computer-readable storage medium

Abstract

To provide a method for easily obtaining an appropriate molding condition that satisfies a quality required for a molded product, without depending on a technical level of a molding worker.SOLUTION: An injection molding method includes the steps of: constructing a prediction model based on an input parameter and an objective variable value including a quality value obtained by quantifying a required quality of a molded product for the input parameter; inferring a predictive distribution of the objective variable value with respect to the input parameter using the prediction model; and deriving the molding condition by a Bayesian optimization method using a regression model that uses a predictive distribution to find the input parameter that results in a highest quality value of an evaluation of the objective variable value compared to the initial quality value. The quality value includes a feature amount converted from an external image of the molded product, and a molding condition that satisfies a desired required quality is derived by using an optimization method that repeatedly evaluates the objective variable value under a derived molding condition and derives the molding condition that is most highly evaluated to achieve the desired quality value.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] This disclosure relates to an injection molding method and a computer-readable storage medium. [Background technology]

[0002] Injection molding is a widely used method for forming resin parts by injecting molten resin material into a mold. In injection molding, determining the appropriate molding conditions is essential to producing high-quality resin parts (hereinafter referred to as molded products) that meet the required quality standards. However, since the appropriate molding conditions differ depending on the shape of the molded product and the properties of the resin used, this process of determining the appropriate molding conditions is carried out by skilled workers with extensive knowledge and experience.

[0003] Furthermore, as a conventional technology, a method for optimizing molded product quality using a neural network has been proposed as a way to support molding by injection molding machines. To construct this neural network, molding conditions are used as input parameters, and quality values ​​obtained by measuring good molded products are used as output items (hereinafter referred to as the target variable) (see, for example, Patent Document 1 below). [Prior art documents] [Patent Documents]

[0004] [Patent Document 1] Japanese Patent Publication No. 2008-110486 [Overview of the Initiative] [Problems that the invention aims to solve]

[0005] The task of deriving appropriate molding conditions that meet the required quality of molded products is generally performed by skilled workers with extensive knowledge and experience. On the other hand, when workers with limited knowledge and experience derive molding conditions, they often have to go through a lot of trial and error, which can be very time-consuming.

[0006] Furthermore, when using neural networks as in conventional technologies, there is a problem in that hundreds to tens of thousands of training data are required to construct a predictive function for optimizing molding conditions.

[0007] Furthermore, while conventional technologies utilize measured quality values ​​(product weight, warpage, dimensions, etc.) to optimize molding conditions, there are problems such as the difficulty of preparing high-resolution measuring instruments (sink marks, flow marks, etc.) due to the high cost of such instruments, the need to cut out measurement samples, and the inability to easily perform the measurements.

[0008] This disclosure provides technology to solve the above-mentioned problems, and aims to provide an injection molding method that can easily obtain appropriate molding conditions that meet the required quality of the molded product, regardless of the skill level of the molding operator, and a computer-readable storage medium. [Means for solving the problem]

[0009] The injection molding method of this disclosure is A step of constructing a predictive model based on input parameters including molding conditions for a molded product, and a target variable value including a quality value that quantifies the required quality of the molded product for the input parameters, The steps include: inferring the predicted distribution of the target variable value for the input parameters using the prediction model; The method comprises the step of deriving molding conditions by a Bayesian optimization method that utilizes a regression model that uses the aforementioned predictive distribution to find the input parameters that result in the highest quality value compared to the initial quality value when evaluating the value of the objective variable, The aforementioned quality value includes feature quantities converted from the appearance image of the molded product. By using an optimization method that repeatedly evaluates the value of the objective variable in the derived molding conditions and derives the molding conditions that yield the highest evaluation resulting in the desired quality value, molding conditions that satisfy the desired required quality are derived. An injection molding method, The aforementioned exterior image is subjected to a black and white binarization process, and the proportion of the white area within the exterior image is used as the quality value of the flow mark. In addition, the injection molding method of the present disclosure includes a step of constructing a prediction model based on input parameters including molding conditions of a molded product and target variable values including quality values obtained by quantifying required quality of the molded product with respect to the input parameters; a step of inferring a prediction distribution of the target variable values with respect to the input parameters using the prediction model; and a step of deriving molding conditions by a Bayesian optimization method utilizing a regression model for obtaining the input parameters for which the evaluation of the target variable values becomes the highest quality value compared to the initial quality value using the prediction distribution, where the quality value is a sink mark feature amount calculated from measured values of a temperature sensor and a pressure sensor installed in a mold, By using an optimization method of repeating the evaluation of the target variable values under the derived molding conditions and the derivation of the molding conditions for which the evaluation that the quality value becomes a desired value is the highest, molding conditions satisfying desired required quality are derived An injection molding method, The aforementioned sink mark feature is the logarithm of the value obtained by dividing the time integral of the temperature sensor's measurement by the time integral of the pressure sensor's measurement. In addition, the computer-readable storage medium of the present disclosure is a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, includes the following steps: a step of constructing a prediction model based on input parameters including molding conditions of a molded product and target variable values including quality values obtained by quantifying required quality of the molded product with respect to the input parameters; a step of inferring a prediction distribution of the target variable values with respect to the input parameters using the prediction model; and a step of deriving molding conditions by a Bayesian optimization method utilizing a regression model for obtaining the input parameters for which the evaluation of the target variable values becomes the highest quality value compared to the initial quality value using the prediction distribution, where the quality value includes feature amounts converted from an appearance image of the molded product, By using an optimization method that repeats the evaluation of the target variable value under the derived molding conditions and the derivation of the molding conditions with the highest evaluation that the quality value becomes a desired value, the molding conditions that satisfy the desired required quality are derived and executed A computer-readable storage medium, The aforementioned exterior image is subjected to a black and white binarization process, and the proportion of the white area within the exterior image is used as the quality value of the flow mark. In addition, the computer-readable storage medium of the present disclosure is A computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the following steps are Constructing a prediction model based on input parameters including molding conditions of a molded product and target variable values including quality values that quantify the required quality of the molded product for the input parameters Inferring a prediction distribution of the target variable value for the input parameter using the prediction model Deriving molding conditions by a Bayesian optimization method that utilizes the regression model to obtain the input parameter for which the evaluation of the target variable value becomes the highest quality value compared to the initial quality value using the prediction distribution The quality value is a sink mark feature amount calculated from the measured value of a temperature sensor and the measured value of a pressure sensor installed in the mold By using an optimization method that repeats the evaluation of the target variable value under the derived molding conditions and the derivation of the molding conditions with the highest evaluation that the quality value becomes a desired value, the molding conditions that satisfy the desired required quality are derived and executed A computer-readable storage medium, The aforementioned sink mark feature is the logarithm of the value obtained by dividing the time integral of the temperature sensor's measurement by the time integral of the pressure sensor's measurement.

Advantages of the Invention

[0010] According to the injection molding method and the computer-readable storage medium of the present disclosure, a prediction function for optimizing molding conditions can be constructed even with a small number of data. Therefore, it is possible to easily derive molding conditions that satisfy the required quality without depending on the technical level of the molding operator

Brief Description of the Drawings

[0011] [Figure 1] This figure shows an example of the apparatus configuration necessary to realize the injection molding method according to Embodiment 1. [Figure 2] This figure shows an example of the apparatus configuration necessary to realize the injection molding method according to Embodiment 1. [Figure 3] This is a schematic diagram showing an example of setting range information for molding conditions. [Figure 4] This is a schematic diagram illustrating an example of the influence of molding conditions on various factors. [Figure 5] This is a schematic diagram showing an example of molded product information. [Figure 6] This characteristic diagram shows an example of the correlation between defined sink mark features and measured sink mark amounts. [Figure 7] Figures 7A, 7B, and 7C are explanatory diagrams showing an example of image processing applied to an image of a molded product. [Figure 8] This diagram schematically illustrates how to determine the next search conditions (molding conditions) using the EI value. [Figure 9] This diagram schematically illustrates a method for determining the next search condition (molding condition) using the EI value. [Figure 10] This flowchart shows an example of a series of processing steps for optimizing molding conditions in the injection molding method of this disclosure. [Figure 11] This flowchart shows an example of a series of processing steps for optimizing molding conditions in the injection molding method of this disclosure. [Figure 12] This flowchart shows an example of a series of processing steps for optimizing molding conditions in the injection molding method of this disclosure. [Figure 13] This flowchart shows an example of a series of processing steps for optimizing molding conditions in the injection molding method of this disclosure. [Figure 14] This is an explanatory diagram showing an example of the screen of the molding condition derivation program that operates in the molding condition derivation device. [Figure 15] This is an explanatory diagram showing an example of the screen of the molding condition derivation program after the initial number of data points have been molded and quality values ​​have been entered. [Figure 16] This is an explanatory diagram showing an example of the screen of the molding condition derivation program after the initial optimization. [Figure 17] This figure shows an example of the hardware configuration of the control processing unit of the present disclosure. [Figure 18] This is a block diagram showing the details of the control processing unit of the molding condition derivation device according to Embodiment 1. [Figure 19] This flowchart shows the steps performed in the control processing unit of the molding condition derivation device according to Embodiment 1. [Figure 20] This figure shows an example of the apparatus configuration necessary to implement the injection molding method according to Embodiment 2. [Figure 21] This figure shows an example of the apparatus configuration necessary to implement the injection molding method according to Embodiment 2. [Figure 22] This figure shows an example of the X-direction and Y-direction feature quantities of sensor values ​​acquired in the control processing unit of this disclosure. [Figure 23] This figure shows an example of a calculation result with low similarity to the reference sensor value in the control processing unit of this disclosure. [Figure 24] This figure shows an example of a calculation result that shows a high degree of similarity to the reference sensor value in the control processing unit of this disclosure. [Figure 25] This flowchart shows an example of a series of processing steps for performing preliminary preparations to optimize molding conditions in the injection molding method according to Embodiment 2. [Figure 26] This flowchart shows an example of a series of processing steps for optimizing molding conditions in the injection molding method according to Embodiment 2. [Figure 27] This flowchart shows an example of a series of processing steps for optimizing molding conditions in the injection molding method according to Embodiment 2. [Figure 28] This flowchart shows an example of a series of processing steps for optimizing molding conditions in the injection molding method according to Embodiment 2. [Figure 29] This flowchart shows an example of a series of processing steps for acquiring feature quantities of sensor values ​​in the injection molding method according to Embodiment 2. [Figure 30] This is an explanatory diagram showing an example of the screen for the molding condition derivation program in Embodiment 2. [Figure 31] This is an explanatory diagram showing an example of the screen of the molding condition derivation program after the molding of the initial number of data points in Embodiment 2 has been completed and the feature quantities of the sensor values ​​have been input. [Figure 32] This is an explanatory diagram showing an example of the screen of the molding condition derivation program after the initial optimization in Embodiment 2. [Figure 33] This is a block diagram showing the details of the control processing unit of the molding condition derivation device according to Embodiment 2. [Figure 34] This flowchart shows the steps performed in the control processing unit of the molding condition derivation device according to Embodiment 2. [Modes for carrying out the invention]

[0012] This disclosure relates to an injection molding method and a molding condition derivation apparatus that utilize a regression model capable of determining the posterior distribution of output with respect to input. The following will be described in detail based on embodiments.

[0013] Embodiment 1. [Configuration for realizing injection molding method] First, the configuration for performing the injection molding method of this disclosure will be described with reference to Figures 1 and 2. Figures 1 and 2 show an example of the apparatus configuration necessary to realize the injection molding method according to Embodiment 1. As shown in Figure 1, the injection molding machine 200 is equipped with a mold 210 for molding the molded product 211, and various sensors 212 are attached to the mold 210. The data measured by the sensors 212 is input to the control processing unit 120 of the molding condition derivation device 100, which will be described later, via a measurement amplifier 220. Meanwhile, the molded product 211 formed in the mold 210 is removed by the removal robot 300. The removed molded product 500 is placed on the conveyor belt 400, where it is measured by the shape measuring device 600 and its appearance is photographed by the camera 700. The measurement results from the shape measuring device 600 and the appearance photographs taken by the camera 700 are taken into the control processing unit 120 of the molding condition derivation device 100, which will be described later. The configuration shown in Figure 1 will be explained in detail later.

[0014] As shown in Figure 2, the molding condition extraction device 100 of this embodiment 1 includes a communication unit 110, a control processing unit 120, a display input unit 130, and a storage unit 140.

[0015] In this case, the molding condition extraction device 100 may be a single device, or it may consist of multiple devices or systems connected by a network such as a WAN (Wide Area Network) or LAN (Local Area Network). Furthermore, this molding condition extraction device 100 may be implemented using a distributed computing or cloud computing system, or by multiple computer devices.

[0016] The communication unit 110 includes, for example, a communication interface such as a NIC (Network Interface Card) and a DMA (Direct Memory Access) controller. This communication unit 110 can communicate with the injection molding machine 200 via a network such as a WAN or LAN.

[0017] The control processing unit 120 includes a molding condition output unit 121, a molding condition optimization unit 122, an indirect quality value processing unit 123, and a direct quality value processing unit 124. This control processing unit 120 is comprised of a processor 1000, such as a CPU (Central Processing Unit) or a GPU (Graphics Processing Unit), and a storage device 1010 (storage unit 140 described later), as shown in Figure 17 as an example of the hardware configuration. The processor 1000 executes a program stored in the storage device 1010 (storage unit 140 described later). Furthermore, the components of the control processing unit 120 may be implemented using hardware such as an FPGA (Field Programmable Gate Array), or they may be composed of both software and hardware.

[0018] The display input unit 130 includes a display device such as a liquid crystal display and can be used by a molding operator handling the molding condition extraction device 100 to understand the progress of optimizing the molding conditions and to set and operate the device through a GUI (Graphical User Interface).

[0019] The storage unit 140 includes, for example, an HDD (Hard Disk Drive), an SSD (Solid State Drive), a ROM (Read Only Memory), or a RAM (Random Access Memory). In addition to various programs for deriving molding conditions, such as firmware and application programs, this storage unit 140 stores molding condition setting range information 141, molding condition item influence 142, and molded product information 143, which will be described later.

[0020] As shown in Figure 1, the injection molding machine 200 is equipped with a mold 210 as a tool for molding the molded product 211 that is subject to optimization of molding conditions, and various sensors 212 are attached to this mold 210.

[0021] The sensor 212 mentioned above includes strain-type or piezoelectric pressure sensors for measuring resin pressure, strain gauges for measuring the amount of strain in the mold 210, thermocouple or infrared temperature sensors for measuring the temperature of the mold 210 and the resin temperature, and AE (Acoustic Emission) sensors for detecting sound emission inside the mold 210. Any type of sensor that can be attached to the mold 210 is acceptable. The various sensor data measured by the sensor 212 are taken into the control processing unit 120 via the measurement amplifier 220, and converted into quality values ​​for evaluating the quality of the molded product in the indirect quality value processing unit 123.

[0022] The molded product 211 formed in the mold 210 is removed by the removal robot 300. The removed molded product 500 is placed on the conveyor belt 400, and then shape data such as flatness and dimensions is measured by the shape measuring instrument 600. The shape measuring instrument 600 may be a measuring instrument such as a caliper or height gauge, or it may be a contact-type or non-contact-type 3D measuring machine.

[0023] Furthermore, the molded product 500 that has been removed is photographed by camera 700. Lighting, a blackout curtain, and a jig may be added as needed to photograph the appearance of the molded product 500 with camera 700. Also, there may be one camera 700 or multiple cameras. The photographed appearance of the molded product 500 is taken into the control processing unit 120 and converted into quality values ​​for evaluating the quality of the molded product 211 in the indirect quality value processing unit 123.

[0024] Figure 3 is a schematic diagram showing an example of molding condition setting width information 141 that is pre-stored in the memory unit 140. This molding condition setting range information 141 has pre-set upper and lower limits for each molding condition setting item (input parameter) for each type of molded product 211. The upper and lower limits may be set based on experience and knowledge, or based on the results of resin flow analysis. For example, for injection temperature, it may be set within the temperature range recommended by the resin material manufacturer for each resin material, or it may be set after confirming in advance whether resin can be filled into the mold 210 using resin flow analysis. Alternatively, it may be set after actually performing a preliminary molding and confirming that there are no problems during molding.

[0025] Now, let's explain each item in Figure 3. The mold temperature (movable) is a parameter used to control the temperature of the mold, and "movable" refers to the side of the mold that moves (opens) when removing the molded product. The mold has openings for circulating fluids such as hot water, cold water, or oil, and the mold temperature is controlled by controlling the temperature of the fluid flowing through these pipes with a temperature controller. The mold temperature (fixed) is a parameter used to control the temperature of the mold, as described above. "Fixed" refers to the side of the mold that does not move when removing the molded product. It is common practice to set a temperature difference between the movable and fixed sides of the mold (e.g., 50°C for the movable side and 30°C for the fixed side). The injection temperature parameters 1-5 control the temperature at which the resin pellets (resin particles) injected into the mold melt. The numbers "1-5" indicate the locations of the 4-6 thermocouples installed sequentially from the tip of the injection unit in the heating cylinder of the injection molding machine. The injection molding machine controls the heater in the heating cylinder so that the thermocouples reach the set temperature. The set value should be set within the range recommended by the resin material manufacturer, while checking the quality of the molded product or the production cycle time. Injection positions 1-4 are parameters that control the screw position of the injection molding machine, switching the speed at which molten resin is injected into the mold. The combination of injection speed and injection position controls the flow of the injected resin. The numbers "1-4" indicate how many positions there are where the injection speed can be switched. (Example: Injection speed 50 mm / s up to injection position (screw position) 100 mm. Injection speed 30 mm / s from injection position 100 mm to 40 mm.) • Speed ​​and pressure switching is the setting of the screw position that switches the control of the injection molding machine's screw from injection speed control to holding pressure control when injecting molten resin into the mold. The injection speed parameters 1-4 control the speed at which molten resin is injected into the mold. The combination of injection speed and injection position controls the flow of the injected resin. The numbers "1-4" represent how much the injection speed is changed relative to the injection position. • Holding pressure 1-3 are parameters that set the magnitude of the holding pressure control after injection speed control when injecting molten resin into the mold. Generally, in injection molding, after filling the mold with resin by injection speed control, shrinkage occurs due to the change in the state of the resin (liquid → solid). To compensate for this shrinkage, additional molten resin is filled into the mold by holding pressure control. The numbers "1-3" indicate how many stages there are and by what magnitude the holding pressure is changed when applying holding pressure in multiple stages. Also, each holding pressure control is time-controlled (e.g., holding pressure 1 is 50 MPa for 3 seconds, holding pressure 2 is 30 MPa for 4 seconds). The cooling time is the time it takes for the molten resin to cool and solidify after holding pressure has been applied to it within the mold. Generally, no external pressure is applied to the resin during the cooling time, and only heat exchange between the resin and the mold occurs until the molten resin solidifies. If the cooling time is too short, it can cause warping or mold release problems, while if it is too long, it can increase the production cycle time (increasing the cost of molded parts) or cause mold release problems.

[0026] Figure 4 is a schematic diagram showing the item influence 142 of the molding conditions that are pre-stored in the memory unit 140. The item influence 142 for the molding conditions sets the magnitude of the influence of each molding condition item (e.g., temperature, pressure, injection speed, etc.) on the quality values ​​(e.g., warpage, sink marks) of the target molded product 211. In this case, the influence level may be set based on experience and knowledge, or based on the results of resin flow analysis. Alternatively, it may be set based on the results of actual preliminary molding. For example, an orthogonal array can be created with the molding condition items as control factors, resin flow analysis can be performed for each condition, and characteristic values ​​can be calculated for the quality values ​​of the molded product. The item influence level for the molding conditions can then be set based on these characteristic values. Furthermore, even when performing preliminary molding, the influence level may be set using an orthogonal array.

[0027] Figure 5 is a schematic diagram showing molded product information 143 that is pre-stored in the memory unit 140. This molded product information 143 is individually set for each type of molded product 211, including an ID number for individual identification, the resin molding material used, and information on the required quality of the molded product 211 that needs to be optimized. Any number of quality information items can be set from the required quality items for the molded product 211. For example, in the case of molded product A in Figure 5, warpage, sink marks, and dimensions occurring in molded product A are set as quality information. In this case, for example, warpage (information 1) is set by the flatness of an arbitrary measurement point, sink marks (information 2) is set by the amount of indentation at an arbitrary measurement point, and dimensions (information 3) are set by the dimensional value and dimensional tolerance.

[0028] In realizing the configuration of this disclosure, the sensor 212, measuring amplifier 220, shape measuring instrument 600, and camera 700 are devices for quantifying the quality of the molded product 211, and any one of these can be used. By realizing the above configuration, the injection molding method of this disclosure can be implemented.

[0029] [Quantification of molded product quality] To find the optimal molding conditions that satisfy the required quality of the molded product 211 (hereinafter referred to as "optimization of molding condition parameters"), it is necessary to consider this as an optimization problem that optimizes the relationship between input and output. In this embodiment 1, the input is the value of the molding conditions, and the output is the required quality of the molded product. The input value of the molding conditions is a quantitative value, but the output value of the required quality of the molded product may be expressed by the molding operator's visual inspection or intuition, and may not even be defined as a quantitative value. Therefore, we will first explain how to obtain a quality value that quantifies the required quality of the molded product 211.

[0030] In this embodiment 1, easily measurable quality values ​​such as dimensions and warping are defined as direct quality values. On the other hand, difficult-to-measure quality values ​​such as sink marks and flow marks are quantified by replacing them with feature quantities extracted from the values ​​of a sensor 212 installed inside the mold 210, or by replacing them with feature quantities from images captured by a camera 700, and these feature quantities are defined as indirect quality values. To make sink marks and flow marks direct quality values ​​would require measurement with a high-resolution measuring instrument, but since such measuring instruments are expensive and difficult to prepare, and require the cutting of measurement samples, they are not easily measured, so they are treated as indirect quality values.

[0031] The direct method for determining quality values ​​involves directly measuring the dimensions, flatness, etc., of the molded product 211. For example, if the required quality is that an arbitrary dimension falls within a dimensional tolerance, then the dimensions are measured using a shape measuring instrument 600 (such as a caliper or a 3D measuring machine), and the measured value is used as the quality value for the required quality. In addition, if the required quality is warpage, the quality value can be determined by measuring the geometric tolerances such as the flatness and perpendicularity of any surface.

[0032] The measured quality values ​​are transmitted to the display input unit 130 either via network transmission or input by the molding operator via a GUI (Graphical User Interface). Subsequently, in the control processing unit 120, the direct quality value processing unit 124 performs preprocessing (such as combining data with other quality values ​​and converting it into array information) for Bayesian optimization, which will be described later.

[0033] On the other hand, the method for determining indirect quality values ​​involves converting sensor values ​​and images acquired by the sensor 212 and camera 700 for the molded product 211 into quality values. For example, if the required quality is minimization of sink marks, the sink mark feature (defined as the logarithm of the value obtained by dividing the time integral of the temperature sensor installed in the mold 210 by the time integral of the pressure sensor) becomes the quality value.

[0034] As shown in Figure 6, this sink mark feature has a direct correlation with the amount of sink mark (measured sink mark) measured with a high-resolution measuring instrument, meaning that a smaller sink mark feature corresponds to a smaller amount of sink mark (measured sink mark). In addition, if the required quality is to minimize flow marks, the image of the extracted molded product 500 captured by camera 700 is subjected to black and white binarization by applying various processes such as grayscale conversion, cropping, and image smoothing by passing the image through various low-pass filters to blur the image.

[0035] Figure 7 is an explanatory diagram showing an example of the results of image processing on an image of a molded product. Figure 7 shows the case where the flow marks are large (Figure 7A), small (Figure 7B), and absent (Figure 7C), arranged from left to right. Since the proportion of white area in the image changes depending on the size of the flow marks, this proportion of white area can be used as a quality value for the flow marks.

[0036] As described above, by using quality values ​​(direct quality values ​​and indirect quality values) that quantify the required quality of the molded product 211, the input can be treated as a value of the molding condition, and the output as an optimization problem of the required quality (quality value) of the molded product.

[0037] [Bayesian optimization method] Before explaining the series of processes for optimizing molding conditions, we will first outline the Bayesian optimization method used to derive molding conditions that satisfy the required quality of the molded product 211.

[0038] Bayesian optimization is a parameter optimization method that can be applied even when the function to be optimized is unknown. First, a predictive model is constructed based on the input parameters (molding conditions in this embodiment 1) and the value of the target variable for those input parameters (quality values ​​that quantify the required quality of the molded product 211 as described above in this embodiment 1). Then, using this predictive model, the predictive distribution of the target variable for the input parameters to be considered is inferred. Using this predictive distribution, the input parameters that yield the highest evaluation for the value of the target variable (the molding conditions to be implemented next) are presented. By repeating this process, the input parameters are optimized.

[0039] The prediction model used here is one that can determine the posterior distribution of the output for a given input (in this embodiment 1, a Gaussian process regression model), but other regression models, such as a random forest regression model, can also be used.

[0040] [Gaussian process regression model] Gaussian process regression is a nonparametric regression method that, compared to neural network regression, can construct a predictive function even with relatively small amounts of data.

[0041] Gaussian process regression is one of the models used to estimate an objective function y=f(x) from an input variable x to a real-valued target variable y. Specifically, Data D 1:t ={x 1:t ,y 1:t Given}, the objective function f(x) of the search target is t+1 The predicted distribution P(f(x)) t+1 )|D 1:t , x t+1 This is a model that calculates ) using the following formula (1).

[0042]

number

[0043] The right-hand side of equation (1) is the mean (expected value) μxt+1|x1:t , The distribution is a normal distribution (Gaussian distribution) with a standard deviation of σ 2 xt+1|x1:t . For example, in a Gaussian process regression model when standardized so that the mean of y is 0, if the covariance matrix K showing the relationship of the input parameter x is expressed by an arbitrary kernel function k(x, x'), the predicted mean and predicted variance can be obtained by the following mathematical formulas (2) and (3).

[0044]

Equation

[0045]

Equation

[0046] Here, k ** represents k(x * , x * ), k * represents the vector of (k(x * , x1), k(x * , x2), ···, k(x * , x n ), the matrix K represents an N×N covariance matrix with elements of k(x T , x' n , x' n ), and the vector y represents the vector of (y1, y2, ···, y n ). In this Embodiment 1, a Gaussian kernel of the following mathematical formula (4) (Θ1 and Θ2 are parameters determining the nature of the kernel) is used for the kernel function k(x, x'), but any of an exponential kernel, a periodic kernel, and a Matérn kernel may be used instead.

[0047]

Equation

[0048] [Parameter Search Method by Bayesian Optimization] In Bayesian optimization, an acquisition function is used to evaluate combinations of candidate input parameters in order to determine the next input parameter (shaping condition) to be tested. This acquisition function may be, for example, PI (Probability of Improvement), EI (Expected Improvement), UCB (Upper Confidence Bound), or MI (Mutual Information).

[0049] In this embodiment 1, as an example, we will describe a case in which an acquisition function called EI (Expected Improvement), which calculates the expected value of how much improvement can be made from the minimum (maximum) value of the target variable, is used in conjunction with multiple target variables.

[0050] In the case of minimizing the objective function y=f(x), Bayesian optimization using the EI value determines the search condition that maximizes the expected value of improvement shown in the following equation (5) as the search condition.

[0051]

number

[0052] One specific method for calculating the EI value is to use the following formula (6).

[0053]

number

[0054] f in equation (6) min σ(x) represents the minimum value of the objective function at the current number of searches, while μ(x) and σ(x) represent the predicted mean and predicted standard deviation output from the Gaussian process regression model. Furthermore, Φ represents the cumulative distribution function and φ represents the probability density function.

[0055] Figure 8 shows a schematic diagram of the method for determining the next search conditions using this EI value. The upper graph in Figure 8 shows the predicted distribution of the objective function f(x), where the black dots represent observed data points, μ(x) is the predicted mean, and CI is the CI. Upper and CI Lower These indicate the upper and lower limits of the confidence interval calculated from the predicted standard deviation. Furthermore, the graph in the lower part of Figure 8 shows the calculated EI values, and the black dots indicate that the EI value is small because it has already been observed. In this schematic diagram of Figure 8, x5, which maximizes the EI value, becomes the next search condition.

[0056] Subsequently, the results after trying the next search condition are shown in the upper graph of Figure 9, and a schematic diagram of the method for determining the next search condition is shown in the lower graph of Figure 9. The predictive distribution of x5, which was the next search condition, is revealed, and the EI value of x5 becomes smaller. As a result, the next search condition is determined to be one of the other search conditions with a high EI value. As shown in Figures 8 and 9 above, the optimal input parameters are found by repeatedly performing trials while determining the next search conditions based on the EI value.

[0057] In this embodiment 1, there may be multiple objective functions (for example, the amount of warpage, sink marks, and color difference in the appearance photograph of the molded product), so the acquisition function is unified by combining each evaluation value (EI value). As a method for unifying the evaluation values, a weighted linear sum method is used, but a simple sum or product method may also be used.

[0058] [Methods for optimizing molding conditions] Figures 10 to 13 are flowcharts illustrating an example of a series of processing steps for optimizing molding conditions in the injection molding method of this disclosure. In the following, we will specifically explain the case where the target variables for optimization are warpage and sink marks. In the figures, the symbol S represents a step.

[0059] To optimize the molding conditions, the initial data collection process is initiated (steps S100 and S101). This involves first starting the molding condition derivation device 100 and setting the molding condition setting range information 141, the molding condition item influence degree 142, and the molded product information 143 as described earlier. After that, the molding condition derivation program stored in the memory unit 140 is started.

[0060] Figure 14 shows, as an example, the startup screen of the molding condition derivation program for a case with five input parameters (mold temperature_movable, mold temperature_fixed, injection speed 4th stage, holding pressure 1st stage, holding pressure 2nd stage) and two objective variables (objective variable 1: warpage, objective variable 2: sink mark).

[0061] The molding condition derivation program stored in the memory unit 140 reads the optimized quality information of the molded product information 143 as the target variable when started, selects molding condition items to be input parameters from the molding condition item influence degree 142, and creates an initial molding condition table as shown in Figure 14 so that it fits within the range of the molding condition setting range information 141. Then, it displays the contents on the display input unit 130. Although Figure 14 shows the case where combinations of each molding condition are output randomly, it is also possible to output as a two-level orthogonal array with each molding condition as a control factor. Furthermore, the molding operator may change the molding condition items that become input parameters selected by the molding condition derivation program to other molding condition items.

[0062] The molding operator performs the molding operation according to the initial molding conditions table shown in Figure 14 on the display input unit 130 of the molding condition derivation device 100. At this time, the molding conditions may be entered manually by the molding operator, or they may be automatically entered via the communication unit 110 if the injection molding machine 200 is connected to a network. The molding operation is performed sequentially from the first row of the initial molding conditions table in Figure 14, and the molding stability is checked (step S102).

[0063] In injection molding, once molding conditions are set and subsequently modified, the mold temperature gradually rises or falls as the number of molding cycles increases. After a certain number of molding cycles, the temperature fluctuations gradually decrease, and molding can be performed at the same mold temperature. This state is used as an indicator to determine that the molding process has stabilized.

[0064] Specific verification methods include observing the time-dependent changes in the temperature sensor value attached to the mold, or observing the changes in the pressure value calculated from the load cell of the molding machine. In this embodiment 1, the condition for confirming molding stability was that the change in the temperature sensor value remained within ±1°C over three consecutive molding cycles.

[0065] Next, once molding stability has been confirmed, molding quality values ​​are acquired for the molded product 211 (steps S103, S501). During molding, time-series data of temperature and pressure acquired by the sensor 212 and the measurement amplifier 220 are transmitted to the molding condition derivation device 100 (step S505). The transmitted time-series data is sent to the indirect quality value processing unit 123 and converted into the aforementioned sensor features (sink marks) (steps S506, S507).

[0066] For the molded product 500 after molding, the flatness of a predetermined surface is measured using the shape measuring device 600 (steps S502, S503). The measured flatness is sent to the direct quality value processing unit 124 of the molding condition derivation device 100 by inputting it into the GUI (Graphical User Interface) on the display input unit 130, or by directly transmitting it if the shape measuring device 600 is on the same network, and is converted into a direct quality value (steps S503, S504).

[0067] In the case of quality requirements related to the design of the molded product, such as flow marks, the appearance of the molded product 500 is photographed by the camera 700 after molding to acquire the image (step S508), which is then transmitted to the molding condition derivation device 100. Subsequently, the indirect quality value processing unit 123 calculates the indirect quality value based on the aforementioned image processing (steps S509, S510). Once the acquisition of the required quality values ​​is complete, the molding quality value acquisition process is terminated (step S511).

[0068] Next, the injection molding machine 200 is driven to perform molding under each molding condition in the initial molding conditions table, and the process of obtaining the molding quality value, which is the target variable, is repeated for the number of initial data displayed in the initial molding conditions table (step S104).

[0069] Figure 15 is an explanatory diagram showing an example of the screen of the molding condition derivation program after the molding of the initial number of data points has been completed and quality values ​​have been entered. As shown in Figure 15, once the initial number of data has been molded (steps S105 and S200), the molding condition optimization unit 122 starts the molding condition optimization process (step S300).

[0070] When this molding condition optimization process is started (steps S300, S301), the molding conditions are optimized by performing repeated molding using the Bayesian optimization method described above. To this end, first, a Gaussian process regression model is created using the input parameters (molding condition values) and the target variable (quality value) collected in the initial data collection process (step S302). That is, a prediction model (prediction function) is created using the input parameters (molding condition values) and the target variable (quality value). The combination of molding conditions that have not yet been molded is input into the Gaussian process regression model thus created, and the predicted mean value and predicted standard deviation are calculated. Then, the evaluation value of the molding conditions input into the Gaussian process regression model is calculated using the acquisition function EI (Expected Improvement) (step S303).

[0071] The combination of molding conditions entered here, which has not yet been molded, is an arithmetic progression of all combinations created within the upper and lower limits of the molding condition setting range information 141 for the set input parameters. The molding condition with the highest evaluation value is displayed as the next molding condition to be tried in the GUI (Graphical User Interface) on the display input unit 130 (step S304).

[0072] Figure 16 is an explanatory diagram showing an example of the screen of the molding condition derivation program, which displays the molding conditions to be tried next. In Figure 16, the molding conditions displayed in the last row are the molding conditions to be tried next.

[0073] The process of displaying the next molding conditions described above can be executed by clicking the "Execute Bayesian Optimization" button on the molding condition derivation program screen shown in Figures 14 to 16.

[0074] Subsequently, similar to the initial data acquisition process, the injection molding machine 200 is driven and molding is performed under the displayed molding conditions (step S305). Similarly, molding stability is confirmed (step S306) and molding quality values ​​are obtained (step S307). If the obtained quality values ​​meet the required quality (step S308), the molding condition optimization process is completed (steps S310, S400). On the other hand, if the required quality is not met, a termination determination is made (step S309).

[0075] In this embodiment 1, the termination determination is set to perform 10 repeated molding cycles, but the number of cycles can be set arbitrarily. If the termination determination does not result in termination, the process returns to the molding condition optimization step (step S301) and the series of steps is repeated. On the other hand, if the termination determination results in termination, the molding condition optimization step is completed (steps S310, S400).

[0076] As described above, in this embodiment 1, molding conditions that satisfy the required quality of the molded product are derived using a Bayesian optimization method that utilizes a regression model capable of determining the posterior distribution of the output with respect to the input, particularly a Gaussian process regression model. Therefore, a predictive function for optimizing the molding conditions can be constructed even with a small amount of data. For this reason, appropriate molding conditions that satisfy the required quality of the molded product can be easily derived, regardless of the skill level of the molding operator.

[0077] In other words, the injection molding method according to Embodiment 1 has the following steps, and the computer-readable storage medium on which the computer program according to Embodiment 1 is stored performs the following steps when the computer program is executed by the processor. In other words, the steps to be performed include constructing a predictive model based on input parameters including molding conditions for the molded product, and a target variable value including a quality value that quantifies the required quality of the molded product for the input parameters, The steps include: inferring the predicted distribution of the target variable value for the input parameters using the prediction model; This step involves deriving molding conditions that satisfy the required quality of the molded product by using a Bayesian optimization method that utilizes a regression model that finds the input parameters that result in the evaluation of the objective variable value being the highest quality value compared to the initial quality value, based on the predictive distribution.

[0078] The above injection molding method and computer-readable storage medium are executed by the molding condition optimization unit 122 in the control processing unit 120 of the molding condition derivation device 100 shown in Figure 1, as shown in Figure 18. The molding condition optimization unit 122 shown in Figure 18 comprises: a prediction model construction unit 1200 that constructs a prediction model based on the input parameters, including the molding conditions of the molded product, and the objective variable value, which quantifies the required quality of the molded product for the input parameters; a prediction distribution inference unit 1210 that uses the prediction model to infer the prediction distribution of the objective variable value for the input parameters; and a molding condition derivation unit 1220 that uses a Bayesian optimization method that utilizes a regression model to find the input parameters for which the evaluation of the objective variable value results in the highest quality value compared to the initial quality value, thereby deriving molding conditions that satisfy the required quality of the molded product.

[0079] Furthermore, as shown in the flowchart of Figure 19, the injection molding method according to Embodiment 1 involves the following steps, and the computer-readable storage medium in which the computer program is stored performs the following steps when the computer program is executed by the processor. In other words, as shown in Figure 19, a predictive model is constructed based on input parameters including molding conditions for the molded product, and objective variable values ​​including quality values ​​that quantify the required quality of the molded product for the input parameters (step S1200). Using the prediction model, the predicted distribution of the target variable value for the input parameters is inferred (step S1210), The process involves using a Bayesian optimization method that utilizes a regression model to determine the input parameters that result in the highest quality value compared to the initial quality value, based on the predictive distribution, thereby deriving molding conditions that satisfy the required quality of the molded product (step S1220).

[0080] As mentioned above, the control processing unit 120 is comprised of a processor 1000, such as a CPU (Central Processing Unit) or a GPU (Graphics Processing Unit), and a storage device 1010 (storage unit 140), as shown in Figure 17 as an example of the hardware configuration. The processor 1000 executes a program stored in the storage device 1010 (storage unit 140). Therefore, the processes of the prediction model construction unit 1200, prediction distribution inference unit 1210, and molding condition derivation unit 1220 of the molding condition optimization unit 122 shown in Figure 18, as well as the steps (S1200, S1210, S1220) executed in the flowchart shown in Figure 19, are realized by the processor 1000 executing a program stored in the storage device 1010 (storage unit 140).

[0081] Embodiment 2. [Utilizing sensor value features] Embodiment 2 describes an injection molding method for maintaining and restoring good quality products by utilizing the characteristic features of sensor values. To find the optimal molding conditions that meet the required quality of the molded product (referred to as "optimization of molding condition parameters"), it is necessary to consider this as an optimization problem that optimizes the relationship between input and output, similar to Embodiment 1. In Embodiment 2, for example, by using the sensor value inside the mold when a good product meets the required quality (hereinafter referred to as the reference sensor value) as a reference, and changing the molding conditions based on Bayesian optimization utilizing the features extracted from the sensor value described later, it is possible to maintain a good product state or return from a defective state to a good state. The following explanation describes, as an example, how to restore a defective product to a working condition.

[0082] [Configuration for realizing injection molding method] The configuration for performing the injection molding method of Embodiment 2 will be explained using Figures 20 and 21, focusing on the differences from Embodiment 1. Figures 20 and 21 show an example of the apparatus configuration necessary to realize the injection molding method according to Embodiment 2. As shown in Figure 20, the injection molding machine 200 is equipped with a mold 210 for molding the molded product 211, and various sensors 212 are attached to the mold 210. The data measured by the sensors 212 is input to the control processing unit 120A of the molding condition derivation device 100A, which will be described later, via the measurement amplifier 220. Since the configuration in Figure 20 is the same as in Figure 1 of Embodiment 1, a detailed explanation will be omitted.

[0083] As shown in Figure 21, the molding condition extraction device 100A includes a communication unit 110, a control processing unit 120A, a display input unit 130, and a storage unit 140. The control processing unit 120A includes a molding condition output unit 121, a molding condition optimization unit 122A, an indirect quality value processing unit 123, a direct quality value processing unit 124, and a sensor value feature quantity processing unit 125. The storage unit 140 includes setting range information 141 for molding conditions, item influence information 142 for molding conditions, and molded product information 143. The molding condition derivation device 100A of Embodiment 2 differs from the molding condition derivation device 100 of Embodiment 1 in that it includes a sensor value feature quantity processing unit 125 within the control processing unit 120A. The sensor value feature processing unit 125 uses the sensor values ​​received by the control processing unit 120A via the measurement amplifier 220 to calculate the sensor value feature quantities, described later, which are used to optimize the molding conditions. By having the above configuration, the injection molding method according to Embodiment 2 can be implemented.

[0084] [Features extracted from sensor values] In Embodiment 2, the input is the value of the molding conditions, and the output is a feature quantity extracted from the measurement value of the sensor 212 inside the mold. In this example, there are three features extracted from the sensor values ​​inside the mold that will be the output. The first and second features extracted from the sensor values ​​are the X-direction features and Y-direction features of the sensor values, as shown in Figure 22 (specifically, the maximum value of the sensor value or the time it reached the maximum value, the sensor value at the time of completion of filling or the time it reached, etc.). In Figure 22, as an example, pressure sensor values ​​are used as the sensor values, with the X-direction feature of the sensor value being the time at which the maximum injection pressure is reached, and the Y-direction feature of the sensor value being the maximum injection pressure. Note that the combination of the X and Y directions of the sensor values ​​for the first and second feature quantities is not specified. Also, as shown in Figure 22, there may be two feature quantities in the X direction (x1, x2) or two feature quantities in the Y direction (y1, y2).

[0085] The third feature extracted from the sensor values ​​is the similarity between the reference sensor value and the sensor value obtained when the molding conditions are changed. Embodiment 2 shows an example of using Euclidean distance as the similarity measure. Euclidean distance is the shortest distance between two points of any dimension. Specifically, if formula (7) is the reference sensor value and formula (8) is the sensor value when the molding conditions are changed, the Euclidean distance d between these sensor values ​​is calculated using formula (9). As an example of the calculated Euclidean distance, Figure 23 shows the case with low similarity and Figure 24 shows the case with high similarity. In Figures 23 and 24, the solid line shows the waveform of the reference sensor value and the dotted line shows the waveform of the sensor value when the molding conditions are changed. In addition to Euclidean distance, Manhattan distance, cosine similarity, or the time integral of the sensor values ​​may also be used as the similarity measure between sensor values.

[0086]

number

[0087]

number

[0088]

number

[0089] [Parameter search method using Bayesian optimization] In Embodiment 2, similar to Embodiment 1, a Bayesian optimization method is used to determine the next input parameter (shaping condition) to be verified, and an acquisition function is used to evaluate combinations of candidate input parameters. This acquisition function may be, for example, PI (Probability of Improvement), EI (Expected Improvement), UCB (Upper Confidence Bound), or MI (Mutual Information).

[0090] In Embodiment 2, as an example, the acquisition function called EI (Expected Improvement), described in Embodiment 1, which calculates the expected value of how much improvement can be made from the minimum (or maximum) value of the target variable, is used in correspondence with the three target variables, which are features extracted from the aforementioned sensor values, to match the reference sensor values. As a specific method for doing so, we will explain the case where the acquisition function EI (Expected Improvement) is a minimization algorithm (equations (5) and (6) described above).

[0091] To match the reference sensor value, it is necessary to combine different optimization methods for the three feature quantities of the sensor value mentioned above, which serve as the target variable. The X and Y direction feature quantities of the sensor value are optimized to fall within a range of ±3% of the X and Y direction feature quantities of the reference sensor value, and optimization is performed with the aim of keeping them within that range. Using a minimization algorithm, optimization can be achieved by using the value μ' calculated by applying the predicted mean μ output from the Gaussian process regression model mentioned above to equation (10), thereby keeping the target range within the target range. In equation (10), RLupper represents the +3% value of the X and Y direction of the reference sensor value, and RLlower represents the -3% value of the X and Y direction of the reference sensor value. The ratio of the upper and lower limits to the reference sensor value can be arbitrarily set within a range of ±10%.

[0092]

number

[0093] Furthermore, the similarity to the reference sensor value is optimized to maximize it. Using the aforementioned minimization algorithm, in order to perform the processing that maximizes the similarity of the target variable, the target variable can be maximized by multiplying the predicted mean μ output from the Gaussian process regression model by -1 and reversing its sign.

[0094] In this second embodiment, the aforementioned processing is performed on the three target variables, and then the acquisition function is unified by combining the evaluation values ​​(EI values) obtained by EI (Expected Improvement) as described in the first embodiment. As a method for unifying the evaluation values, a weighted linear sum method is used, but a simple sum or product method may also be used.

[0095] In this example, we describe an example using pressure sensor values ​​as the sensor values, but you may also use temperature sensor values, AE (Acoustic Emission) sensor values, mold strain values ​​measured with strain gauges, etc., and perform the same processing as described above. Furthermore, in this example, we have used the X-direction and Y-direction features of a 2D coordinate system as sensor values ​​for explanation. However, we may also focus on the X-direction, Y-direction, and Z-direction features of a 3D coordinate system as sensor values, or we may focus on the x1-direction, x2-direction, ..., xN-direction features of an N-dimensional coordinate system (where N is an integer greater than or equal to 2) as sensor values ​​and perform the same processing as described above.

[0096] [Methods for optimizing molding conditions] Figures 25 to 28 are flowcharts illustrating an example of a series of processing steps for optimizing molding conditions in the injection molding method of Embodiment 2. In the following, we will specifically explain the process of restoring a molded product to a good condition when defects occur due to disturbances such as temperature changes or lot variations in resin material. In the figures, the symbol S represents a step.

[0097] Figure 25 is a flowchart showing an example of a series of processing steps for performing preliminary preparations to optimize molding conditions in the injection molding method according to Embodiment 2. First, before optimizing the molding conditions to restore the product to a good condition, the process of acquiring reference sensor values ​​is started as a preliminary step, following the flowchart in Figure 25 (step S601).

[0098] First, the molding stability is checked (step S602). In this embodiment 2, the condition for checking molding stability was that the change in the temperature sensor value remained within ±1°C over three consecutive molding cycles. After confirming molding stability, the molding quality value acquisition process described in Embodiment 1 (see Figure 13) is used to confirm that the molding quality value meets the required quality (steps S603, S604, S501 to S511).

[0099] If the molding quality value continues to fail to meet the required quality, the molding conditions for producing a good product are derived using the method of Embodiment 1. Then, the value from sensor 212 (pressure sensor in this example) when the required quality is met is saved from the measurement amplifier 220 to the storage unit 140 via the display input unit 130 (step S605). In this case, the sensor value is saved for a predetermined set number of times (step S606). In this example, for example, the set number of times the sensor value is saved is set to 30 repeated moldings, but the set number can be set arbitrarily.

[0100] After the acquisition of sensor values ​​is complete, reference sensor values ​​are created using the saved sensor values ​​(step S607). Specific methods for creating reference sensor values ​​include pre-processing such as removing or replacing missing values ​​and noise, followed by calculating the mean, median, and weighted mean values ​​of the sensor values. In this example, the mean value was used. At this time, the X-direction feature quantity (in this example, the time when the maximum injection pressure value is reached) and the Y-direction feature quantity (in this example, the maximum injection pressure value) of the sensor values ​​shown in Figure 22 are obtained from the created reference sensor values ​​and saved together in the storage unit 140. After the creation and saving of the reference sensor values ​​is complete, the reference sensor value acquisition process is terminated (step S608). This preliminary preparation can be performed not only when deriving the molding conditions for the mold, but also during mass production of the molded product.

[0101] Next, after creating reference sensor values ​​for the target molded product, the molding conditions are optimized to return it to a good product state. Specifically, the initial data acquisition process is started according to the flowcharts in Figures 26 to 28 (steps S100 and S101). This involves first starting the molding condition derivation device 100A and setting the molding condition setting range information 141, the molding condition item influence degree 142, and the molded product information 143 as described in Embodiment 1. After that, the molding condition derivation program stored in the storage unit 140 is started.

[0102] Figure 30 shows, as an example, the startup screen of the molding condition derivation program for a case with four input parameters (resin temperature 1st stage, injection speed 3rd stage, injection speed 4th stage, holding pressure 1st stage) and three objective variables (objective variable 1: X-direction feature of sensor value, objective variable 2: Y-direction feature of sensor value, objective variable 3: similarity of reference waveform to sensor value).

[0103] The molding condition derivation program stored in the memory unit 140 selects molding condition items that significantly affect the fluctuation of sensor values ​​from the molding condition item influence degree 142 at startup, and creates an initial molding condition table as shown in Figure 30 so that it falls within the range of the molding condition setting range information 141. Then, it displays the contents of this table on the display input unit 130. Figure 30 shows the case where combinations of molding conditions are output randomly. However, it is also possible to output the combinations as a two-level orthogonal array with each molding condition as a control factor, or to use the scalar value of [matrix of randomly selected molding conditions] × [transpose matrix of selected molding conditions] as the optimization criterion and select the initial conditions that maximize that optimization criterion.

[0104] The molding operator performs the molding operation according to the initial molding conditions table shown in Figure 30 on the display input unit 130 of the molding condition derivation device 100A. At this time, the molding conditions may be entered manually by the molding operator, or they may be automatically entered via the communication unit 110 if the injection molding machine 200 is connected to a network. The molding operation is performed sequentially from the first row of the initial molding conditions table in Figure 30, and the molding stability is checked (step S102).

[0105] In injection molding, once molding conditions are set and subsequently modified, the mold temperature gradually rises or falls as the number of molding cycles increases. After a certain number of molding cycles, the temperature fluctuations gradually decrease, and molding can be performed at the same mold temperature. This state is used as an indicator to determine that the molding process has stabilized. As a specific verification method, in this example, the condition for confirming molding stability was that the change in the temperature sensor value remained within ±1°C over three consecutive molding cycles, similar to the example in Embodiment 1.

[0106] Next, after confirming molding stability, the sensor value features are acquired for the molded product 211 according to the flowchart in Figure 29 (steps S1000, S801). During molding, time-series data of pressure acquired by the sensor 212 and the measurement amplifier 220 is transmitted to the molding condition derivation device 100A (step S802). The transmitted time-series data is sent to the sensor value feature processing unit 125. The sensor value feature processing unit 125 calculates the three sensor value features mentioned above: the similarity of the current sensor value to the reference sensor value (step S803), the X-direction feature of the sensor value (step S804), and the Y-direction feature of the sensor value (step S805). Once the acquisition of the necessary values ​​is complete, the sensor value feature acquisition process is terminated (step S806).

[0107] Next, the injection molding machine 200 is driven to perform molding under each molding condition in the initial molding conditions table, and the process of acquiring characteristic quantities of the sensor values ​​that will be the target variable is repeated for the number of initial data displayed in the initial molding conditions table (step S104). Figure 31 is an explanatory diagram showing an example of the screen of the molding condition derivation program after the initial number of data points have been molded and the sensor value features have been input. As shown in Figure 31, once the molding of the initial data is complete and the initial data collection process is finished (steps S105 and S200), the molding condition optimization unit 122A starts the molding condition optimization process (step S300).

[0108] The molding condition optimization process is initiated (steps S300 and S301), and the molding conditions are optimized by performing repeated molding using the Bayesian optimization method described above. To this end, a Gaussian process regression model is first created using the input parameters (molding condition values) and target variables (sensor value features) collected in the initial data collection process (step S302). That is, a prediction model (prediction function) is created using the input parameters (molding condition values) and target variables (sensor value features). The combination of molding conditions that have not yet been molded is input into the Gaussian process regression model thus created, and the predicted mean value and predicted standard deviation are calculated. Then, numerical processing is performed to apply the minimization algorithm described above to the predicted mean value, and the evaluation value of the molding conditions input into the Gaussian process regression model is calculated using the acquisition function EI (Expected Improvement) (step S303).

[0109] The combination of molding conditions entered here, which has not yet been molded, is an arithmetic progression of all combinations created within the upper and lower limits of the molding condition setting range information 141 for the set input parameters. The molding condition with the highest evaluation value is displayed as the next molding condition to be tried in the GUI (Graphical User Interface) on the display input unit 130 (step S304).

[0110] Figure 32 is an explanatory diagram showing an example of the screen of the molding condition derivation program, which displays the molding conditions to be tried next. In Figure 32, the molding conditions displayed in the last row are the molding conditions to be tried next. The process of displaying the next molding conditions described above can be executed by clicking the "Execute Bayesian Optimization" button on the molding condition derivation program screen shown in Figures 30 to 32.

[0111] Subsequently, similar to the initial data acquisition process, the injection molding machine 200 is driven and molding is performed under the displayed molding conditions (step S305). Furthermore, similar to the initial data acquisition process, molding stability is confirmed (step S306) and feature quantities of the sensor values ​​are acquired (steps S2000, S801). If the acquired feature quantities of the sensor values ​​meet the requirements (step S308), the molding condition optimization process is completed (steps S310, S400). In this example, the requirements were deemed met when the Euclidean distance used for similarity to the reference sensor value was 0.7 or higher. On the other hand, if the requirements are not met, a termination determination is made (step S309).

[0112] In this example, the termination determination (step S309) is set to perform 10 repeated molding cycles, but the number of cycles can be set arbitrarily. If the termination determination does not result in completion, the process returns to the molding condition optimization process (step S301) and the series of steps is repeated. On the other hand, if the termination determination results in completion, the molding condition optimization process is completed (steps S310, S400).

[0113] As described above, in this embodiment 2, molding conditions that satisfy the required quality of the molded product are derived using a Bayesian optimization method that utilizes a regression model capable of determining the posterior distribution of the output with respect to the input, particularly a Gaussian process regression model. Therefore, a predictive function for optimizing the molding conditions can be constructed even with a small amount of data. For this reason, regardless of the skill level of the molding operator, when defects occur in the molded product due to disturbances such as temperature changes or lot changes in the resin material, appropriate molding conditions to restore it to a good product state can be easily derived.

[0114] In other words, the injection molding method according to Embodiment 2 has the following steps, and the computer-readable storage medium storing the computer program according to Embodiment 2 performs the following steps when the computer program is executed by the processor. In other words, the steps to be performed include constructing a predictive model based on input parameters including molding conditions for a molded product, feature quantities of sensor values ​​from sensors placed on an injection molding machine for the input parameters, and a target variable value including the similarity of the sensor values ​​when the molding conditions for the molded product are changed to a reference sensor value which is the sensor value when the molded product meets the required quality. The steps include: inferring the predicted distribution of the target variable value for the input parameters using the prediction model; This step involves deriving molding conditions that satisfy the required quality of the molded product by using a Bayesian optimization method that utilizes a regression model that finds input parameters such that the evaluation of the target variable value is closer to the features of the reference sensor value than to the features of the initial sensor value, based on the predictive distribution.

[0115] The above injection molding method and computer-readable storage medium are executed by the molding condition optimization unit 122A in the control processing unit 120A of the molding condition derivation device 100A shown in Figure 21, as shown in Figure 33. The molding condition optimization unit 122A shown in Figure 33 comprises: a prediction model construction unit 1200A that constructs a prediction model based on the input parameters including the molding conditions of the molded product, the feature quantities of the sensor values ​​of a sensor placed on the injection molding machine for the input parameters, and a target variable value including the similarity of the sensor values ​​when the molding conditions of the molded product are changed to a reference sensor value which is the sensor value when the molded product satisfies the required quality; a prediction distribution inference unit 1210A that infers the prediction distribution of the target variable value for the input parameters using the prediction model; and a molding condition derivation unit 1220A that derives molding conditions that satisfy the required quality of the molded product by a Bayesian optimization method utilizing a regression model that finds the input parameters such that the evaluation of the target variable value is closer to the feature quantities of the reference sensor value than to the feature quantities of the initial sensor value, based on the prediction distribution.

[0116] Furthermore, as shown in the flowchart of Figure 34, the injection molding method according to Embodiment 1 involves the following steps, and the computer-readable storage medium storing the computer program executes the following steps when the computer program is executed by the processor. Specifically, as shown in Figure 34, a prediction model is constructed based on input parameters including molding conditions for the molded product, feature quantities of sensor values ​​from a sensor placed on the injection molding machine for the input parameters, and a target variable value including the similarity of the sensor values ​​when the molding conditions for the molded product are changed to a reference sensor value which is the sensor value when the molded product meets the required quality (step S1200A). Then, the prediction distribution of the target variable value for the input parameters is inferred using the prediction model (step S1210A). Finally, molding conditions that satisfy the required quality for the molded product are derived using a Bayesian optimization method that utilizes a regression model to find the input parameters for which the evaluation of the target variable value is closer to the feature quantities of the reference sensor value than to the feature quantities of the initial sensor value, based on the prediction distribution (step S1220A).

[0117] In Embodiment 2, as in Embodiment 1, the control processing unit 120A is comprised of a processor 1000, such as a CPU (Central Processing Unit) or a GPU (Graphics Processing Unit), and a storage device 1010 (storage unit 140), as shown in Figure 17 as an example of the hardware configuration. The processor 1000 executes a program stored in the storage device 1010 (storage unit 140). Therefore, the processes of the prediction model construction unit 1200A, the prediction distribution inference unit 1210A, and the molding condition derivation unit 1220A of the molding condition optimization unit 122A shown in Figure 33, as well as the steps (S1200A, S1210A, S1220A) executed in the flowchart shown in Figure 34, are realized by the processor 1000 executing a program stored in the storage device 1010 (storage unit 140).

[0118] While this disclosure describes various exemplary embodiments and examples, the various features, aspects, and functions described in one or more embodiments are not limited to the application of a particular embodiment, but are applicable individually or in various combinations to the embodiments. Accordingly, countless variations not illustrated are conceivable within the scope of the art disclosed in this specification. These include, for example, modifying, adding or omitting at least one component, or even extracting at least one component and combining it with components of other embodiments. [Explanation of symbols]

[0119] 100, 100A Molding condition extraction device, 110 Communication unit, 120, 120A Control processing unit, 121 Next condition output unit for molding conditions, 122, 122A Molding condition optimization unit, 123 Indirect quality value processing unit, 124 Direct quality value processing unit, 125 Sensor value feature quantity processing unit, 130 Display input unit, 140 Storage unit, 141 Setting width information for molding conditions, 142 Influence of molding conditions, 143 Molded product information, 200 Injection molding machine, 210 Mold, 211 Molded product (in the process of molding), 212 Sensor, 220 Measuring amplifier, 300 Picking robot, 400 Conveyor belt, 500 Molded product (after molding), 600 Shape measuring instruments, 700 Cameras, 1200, 1200A Predictive model building unit, 1210, 1210A Predictive distribution inference unit, 1220, 1220A Molding condition derivation unit.

Claims

1. A step of constructing a predictive model based on input parameters including molding conditions for a molded product, and a target variable value including a quality value that quantifies the required quality of the molded product for the input parameters, The steps include: inferring the predicted distribution of the target variable value for the input parameters using the prediction model; The method comprises the step of deriving molding conditions by a Bayesian optimization method that utilizes a regression model that uses the aforementioned predictive distribution to find the input parameters that result in the highest quality value compared to the initial quality value when evaluating the value of the objective variable, The aforementioned quality value includes feature quantities converted from the appearance image of the molded product. An injection molding method for deriving molding conditions that satisfy desired required quality by using an optimization method that repeatedly evaluates the value of the objective variable in the derived molding conditions and derives the molding conditions that yield the highest evaluation so that the quality value is the desired value, wherein the molding conditions satisfy desired required quality are derived, An injection molding method comprising performing a black and white binarization process on the aforementioned appearance image, and using the proportion of the white area within the appearance image as the quality value of the flow mark.

2. A step of constructing a predictive model based on input parameters including molding conditions for a molded product, and a target variable value including a quality value that quantifies the required quality of the molded product for the input parameters, The steps include: inferring the predicted distribution of the target variable value for the input parameters using the prediction model; The method comprises the step of deriving molding conditions by a Bayesian optimization method that utilizes a regression model that uses the aforementioned predictive distribution to find the input parameters that result in the highest quality value compared to the initial quality value when evaluating the value of the objective variable, The aforementioned quality value is a sink mark characteristic calculated from the measurement values ​​of a temperature sensor and a pressure sensor installed inside the mold. An injection molding method for deriving molding conditions that satisfy desired required quality by using an optimization method that repeatedly evaluates the value of the objective variable in the derived molding conditions and derives the molding conditions that yield the highest evaluation so that the quality value is the desired value, wherein the molding conditions satisfy desired required quality are derived, An injection molding method wherein the sink mark feature is the logarithm of the value obtained by dividing the time integral of the temperature sensor's measurement by the time integral of the pressure sensor's measurement.

3. A computer-readable storage medium on which a computer program is stored, wherein when the computer program is executed by a processor, the following steps are taken: A step of constructing a predictive model based on input parameters including molding conditions for a molded product, and a target variable value including a quality value that quantifies the required quality of the molded product for the input parameters, The steps include: inferring the predicted distribution of the target variable value for the input parameters using the prediction model; The method comprises the step of deriving molding conditions by a Bayesian optimization method that utilizes a regression model that uses the aforementioned predictive distribution to find the input parameters that result in the highest quality value compared to the initial quality value when evaluating the value of the objective variable, The aforementioned quality value includes feature quantities converted from the appearance image of the molded product. A computer-readable storage medium that performs the task of deriving molding conditions that satisfy a desired required quality by using an optimization method that repeatedly evaluates the value of the objective variable in the derived molding conditions and derives the molding conditions that yield the highest evaluation so that the quality value is the desired value, A computer-readable storage medium that performs a black and white binarization process on the aforementioned exterior image, and uses the proportion of white area in the exterior image as the quality value of the flow mark.

4. A computer-readable storage medium on which a computer program is stored, wherein when the computer program is executed by a processor, the following steps are taken: A step of constructing a predictive model based on input parameters including molding conditions for a molded product, and a target variable value including a quality value that quantifies the required quality of the molded product for the input parameters, The steps include: inferring the predicted distribution of the target variable value for the input parameters using the prediction model; The method comprises the step of deriving molding conditions by a Bayesian optimization method that utilizes a regression model that uses the aforementioned predictive distribution to find the input parameters that result in the highest quality value compared to the initial quality value when evaluating the value of the objective variable, The aforementioned quality value is a sink mark characteristic calculated from the measurement values ​​of a temperature sensor and a pressure sensor installed inside the mold. A computer-readable storage medium that performs the task of deriving molding conditions that satisfy a desired required quality by using an optimization method that repeatedly evaluates the value of the objective variable in the derived molding conditions and derives the molding conditions that yield the highest evaluation so that the quality value is the desired value, The sink mark feature is the logarithm of the value obtained by dividing the time integral of the temperature sensor's measurement by the time integral of the pressure sensor's measurement, and is a computer-readable storage medium.

5. The computer-readable storage medium according to claim 3 or 4, wherein the regression model is a Gaussian process regression model.

Images