Action condition inference device, method, and program for blow molding machine, and learning device

Abstract

A learning completion model (73) is prepared, taking the physical properties of the preform, the molding conditions of the bottle, the measured values ​​of the physical properties of the bottle, and the measured values ​​of the environmental information during the manufacturing of the bottle as input data, and the operating conditions of the blow molding machine as output data. By inputting the target values ​​(81a) of the physical properties that the molded bottle should meet, and the measured values ​​(81b) of the physical properties of the preform, the molding conditions of the bottle, and the environmental information, into the learning completion model (73), it is possible to infer the operating conditions (heater output, blow molding pressure, blow molding timing, bar speed, cooling water temperature) of the blow molding machine that should be set in the blow molding machine used to mold the bottle that meets the target values.

Description

Technical Field

[0001] This invention relates to a device for inferring the operating conditions of a roll forming machine, a roll forming machine, a method and procedure for inferring the operating conditions of a roll forming machine, and a learning device. Background Technology

[0002] Resin bottles, such as PET bottles, are mostly manufactured by pre-fabricating preforms with a shape similar to test tubes and then stretching and blow molding them. The preforms are stretched using a blow molding machine for stretch blow molding to form the bottles (see Patent Document 1 for details on blow molding machines).

[0003] Blow molding machines have a variety of settings, each with its own set of data (values). The machine then operates according to these settings. In recent years, the variety of bottle types has continuously increased, and consequently, with the demands for lightweighting and the use of reusable materials, the variety of preforms has also grown. If the type of bottle or the type of preform used to mold it differs, the settings of the blow molding machine will naturally need to be adjusted (changed). Setting up a blow molding machine largely relies on the intuition and experience of a skilled operator.

[0004] Existing technical documents

[0005] Patent documents

[0006] Patent Document 1: International Publication WO2019 / 048419 Summary of the Invention

[0007] The purpose of this invention is to assist in various settings of blow molding machines.

[0008] The operating condition estimation device of the blow molding machine according to at least several embodiments of the present invention heats the preform with a heater, moves the heated preform towards the mold, stretches the preform axially by inserting a stretching rod into the preform, and stretches the preform outward by supplying pressurized fluid into the preform, thereby forming a hollow bottle. It is characterized by comprising:

[0009] The learning completion model learns parameters by taking the physical properties or environmental information of the bottle as input data and the operating conditions of the blow molding machine as output data; the inference unit infers the operating conditions of the blow molding machine, which are inferred by inputting the target values ​​of the physical properties or environmental information that the molded bottle should meet into the learning completion model, and are the operating conditions that should be set in the blow molding machine in order to mold a bottle that meets the target values ​​or target quality.

[0010] According to the present invention, the operating conditions of a blow molding machine that should be set in order to form bottles that meet target values ​​or target quality are inferred. The inferred operating conditions are based on a learning-complete model that learns parameters by taking the physical properties of the bottle or environmental information as input data and the operating conditions of the blow molding machine as output data. The operating conditions of the blow molding machine for forming bottles with target physical properties or target quality can be inferred in a short time without relying on the operator's experience.

[0011] Preferably, the input data also includes at least one of the physical properties of the preform and the molding conditions of the bottle. This increases the possibility of more accurately and precisely inferring the operating conditions of the blow molding machine used to mold bottles with the target physical properties or target quality.

[0012] In one embodiment, the deduced operating conditions of the blow molding machine are at least one of the heater output conditions for heating the preform and the pressure conditions of the fluid supplied to the preform. This is because the heater output and fluid pressure (blow molding pressure) have a significant impact on the physical properties or quality of the bottle formed by the blow molding machine. By learning and completing the model, the heater output and / or fluid pressure of the blow molding machine should be set to meet the target physical property values ​​or quality in order to form a bottle. This improves the reliability of forming bottles that meet the specified physical property value targets or quality targets without relying on the operator's experience.

[0013] In addition to the heater output and fluid pressure (blow molding pressure) mentioned above, the inferred operating conditions of the blow molding machine may also include the timing conditions for supplying fluid to the preform and / or the speed conditions for inserting the tension rod into the preform.

[0014] Preferably, the input data includes information related to whether the bottle has any appearance abnormalities, and the inference unit infers the operating conditions of the blow molding machine that should be set in order to mold a bottle that meets the target value or target quality and has no appearance abnormalities. This allows for the inference of operating conditions for bottles with higher quality molded parts.

[0015] In one embodiment, the physical properties of the preform include data related to the preform's material and weight. The physical properties of the preform may include data distinguishing whether the preform is made of virgin resin or recycled resin. Furthermore, the physical properties of the preform may also include the preform's Intrinsic V iscosity value.

[0016] In other embodiments, the molding conditions of the bottle include data related to the bottle's capacity and shape.

[0017] In other embodiments, the physical properties of the bottle include at least one of the following: full capacity, buckling strength, total height, wall thickness, and section weight.

[0018] In other embodiments, the aforementioned environmental information includes at least one of the temperature and humidity around the blow molding machine.

[0019] Preferably, the blow molding machine operation condition inference device includes a setting unit that sets the blow molding machine operation conditions inferred by the inference unit. The blow molding machine can then be operated according to the inferred operation conditions.

[0020] The present invention also provides a learning device that generates a learning model suitable for use in a blow molding motion condition inference device. The learning device according to at least several embodiments of the present invention includes: a learning data acquisition unit that acquires learning data including the motion conditions of a blow molding machine and physical characteristics or environmental information of a bottle, wherein the motion conditions of the blow molding machine are: heating the preform using a heater, moving the heated preform towards a mold and inserting a stretching rod into the preform, thereby stretching the preform axially, and stretching the preform outward by supplying pressurized fluid internally, thereby forming a hollow bottle; and a model generation unit that uses the learning data to generate a learning completion model for inferring the motion conditions of the blow molding machine that should be set in order to form a bottle that meets the target values ​​or target quality of the physical characteristics that the formed bottle should satisfy, based on target values ​​or target quality of the physical characteristics that should be satisfied.

[0021] The present invention also provides a method for inferring the operating conditions of a blow molding machine and a program for executing the method by a computer. The program can be stored on a removable recording medium, such as a CD-ROM, a semiconductor memory, etc. Attached Figure Description

[0022] Figure 1 This is a rough top view of a blow molding machine.

[0023] Figure 2 This indicates the positional relationship between the preform and the heating zone.

[0024] Figure 3 It is a longitudinal cross-sectional view of a mold containing a preform.

[0025] Figure 4 This is a function block diagram of a blow molding machine.

[0026] Figure 5 This is a block diagram of a learning device.

[0027] Figure 6 This is a block diagram of the inference device.

[0028] Figure 7 This indicates the processing result of the inference device.

[0029] Figure 8 It is a graph showing the relationship between the target value and the measured value of the filled capacity.

[0030] Figure 9 This is another example of representing the processing result of the inference device.

[0031] Figure 10 This is another example of representing the processing result of the inference device. Detailed Implementation

[0032] Figure 1 It is a top view that roughly represents a blow molding machine.

[0033] Multiple preforms P made of resin, such as PET (polyethylene terephthalate), are prepared at preform supply station 1. The preforms P are supplied from preform supply station 1 to heating station (oven) 2 along a conveyor path. Heating station 2 has one or more heating zones 5 (in... Figure 1 The diagram shows six heating zones 5. The preform P passes through the heating zones 5 and is heated to a temperature suitable for blow molding. The number of heating zones 5 can be arbitrarily designed.

[0034] The preform P heated in heating station 2 enters molding station 3. Molding station 3 includes: a rotating wheel 6; and multiple molds 7 arranged along the rotating wheel 6, which rotate around molding station 3 at a constant speed as the rotating wheel 6 rotates. The preforms P supplied to molding station 3 are respectively placed in the molds 7. Alternatively, molding station 3 does not necessarily need to have a rotating wheel 6; the molds 7 can also be conveyed in a straight line. The number of molds 7 in molding station 3 is also arbitrary.

[0035] The mold 7 (the bottom mold 7c of the mold 7 described later) has a cooling water path (not shown) through which cooling water passes, and is cooled by cooling water at a specified temperature passing through the cooling water path. As will be explained below, if bottle B is adhered to the inner surface of the mold 7, bottle B is cooled and solidified.

[0036] A stretching rod is inserted into the preform P from the opening of the mold 7, thereby stretching the preform P axially (lengthwise). Fluid, typically compressed air, is blown into the preform P from its axially extending opening. Through this blowing, the preform P extends outwards (axially and radially). The blowing process is performed in two stages: first, the preform P expands by blowing in low-pressure air (hereinafter referred to as first air), and then the preform P adheres to the inner surface of the mold by blowing in high-pressure air (hereinafter referred to as second air), thus creating a bottle B shaped according to the mold 7. After removing the compressed air and stretching rod from bottle B, bottle B is removed from the mold 7. Bottle B exiting the forming station 3 proceeds to subsequent processes (illustrations omitted), such as sterilization, beverage filling, and cap installation, thus completing the bottle containing the beverage.

[0037] A thermometer 61 is provided to measure the ambient temperature (indoor temperature) of the blow molding machine, and a hygrometer 62 is provided to measure the ambient humidity of the blow molding machine. The temperature measured by the thermometer 61 and the humidity measured by the hygrometer 62 can be used to infer the operating conditions of the blow molding machine, as described later.

[0038] Figure 2 This is a diagram that roughly shows the positional relationship between the preform P and the heating zone 5 during transport at heating station 2, and also roughly shows the cross-section of heating zone 5. Additionally, in Figure 2 The diagram also illustrates the main shaft 8 used to hold and rotate the preform P. In the heating station 2, the preform P is conveyed while rotating along the main shaft 8 with its axis of rotation as the axis of rotation.

[0039] The heating station 2 of the blow molding machine, as described above, has a heating zone 5 where the preform P is softened, allowing it to be stretched by a stretching rod and air blowing. When multiple heating zones 5 are provided, these zones include types that heat the entire preform P, types that concentrate heating around the support ring of the preform P (focusing heater), types that blow air onto the entire preform P without a heater (surface cleaning), and types that blow air onto a portion of the preform P (air knife). Various types of heating zones can be used, thereby enabling precise control of the heating of the preform P. Figure 2 The heating zone 5 shown indicates the type of heating that applies to the entire preform P. Multiple heating zones 5 can be controlled independently.

[0040] Figure 2 The heater region 5 shown has a plurality of heaters spaced apart from each other along the axial direction of the preform P. Figure 2There are 7 heaters 5a (e.g., halogen lamps) in the heater zone 5. Each of the 7 heaters 5a in the heater zone 5 can be controlled independently. The number of heaters 5a in the heater zone 5 can also be arbitrary.

[0041] For example, if it is desired to locally thicken the wall of bottle B formed by a blow molding machine, it is possible to consider lowering the heating temperature of the preform P corresponding to that part. For example, using Figure 2 The heating zone 5 shown reduces the output of heater 5a (e.g., one of seven heaters 5a) that heats the portion of the preform P corresponding to the part where the wall thickness is to be increased during the molding of bottle B. This reduces the heating temperature of a portion of the preform P, enabling the molding of bottle B with locally thickened walls.

[0042] Figure 3 This is a diagram that roughly illustrates the blow molding process, showing a longitudinal section of the mold 7 containing the preform P. Figure 3 The diagram of the cooling water path of the bottom mold 7c of the mold 7 mentioned above is omitted.

[0043] The preform P, heated in heating station 2 and transferred to molding station 3, is surrounded by parting molds 7a and 7b, which constitute mold 7. The lower surfaces of parting molds 7a and 7b are closed by bottom mold 7c, and valve block 13 engages with the opening of preform P. A tension rod 11 is inserted into preform P through a hole formed in valve block 13. The tension rod 11 extends preform P axially.

[0044] Valve block 13 has two valves, V1 and V2, inside it. Compressed air (first air and second air) from air pump 12 is sent into bottle preform P through valve V1. The first air and second air are sequentially sent into bottle preform P through valve V1, thereby causing bottle preform P to expand outward and form bottle B.

[0045] After bottle B is formed, the compressed air inside bottle B is discharged outward through valve V2 in valve block 13. The tension rod 11 is pulled out of bottle B, and the bottom mold 7c and parting molds 7a and 7b are opened, thereby removing the completed bottle B from the blow molding machine.

[0046] In the forming station 3, the rod speed (travel speed) of the tension rod 11, the timing of starting to blow the first air, the pressure of the first air, the timing of switching from the first air to the second air, and the pressure of the second air can be controlled.

[0047] For example, if you want to thin the wall thickness near the bottom of a bottle B formed by a blow molding machine, consider advancing the timing of starting to blow the first air. By supplying the first air before the stretching rod is fully extended (before the preform P is fully extended longitudinally), the wall thickness at the bottom of bottle B is thinned, instead of thickening the wall thickness from the neck to the shoulder of bottle B.

[0048] Figure 4 This is a functional block diagram representing a blow molding machine.

[0049] The blow molding machine is generally divided into a heating function 20 and a blow molding function 30. The heating function 20 is achieved through the multiple heating zones 5 of the heating station 2. The blow molding function 30 is achieved through the tension rod 11, air pump 12, valve block 13, etc. of the molding station 3.

[0050] As described above, each of the multiple heating zones 5 in the heating station 2, and consequently each of the multiple heaters 5a in the heating zone 5, can be controlled individually. In the molding station 3, the pressure of the air supplied from the air pump 12, the air supply timing, and the rod speed (entry speed) of the tension rod 11 can also be controlled. The heating function 20 and the air supply function 30 are centrally managed by the control device 40.

[0051] A setting device 50 is connected to the control device 40. The setting device 50 has a display device (not shown), on which multiple setting items are displayed on the setting screen. According to the settings of the multiple setting items used on the setting screen, the control device 40 controls the heating function 20 and the air supply function 30.

[0052] The setting device 50 can set values ​​for multiple settings, and the heating function 20 and the blow molding function 30 of the blow molding machine can be precisely controlled according to these multiple settings. By changing the settings, as described above, for example, the wall thickness of a specific part of the final molded bottle B can be increased or decreased.

[0053] Specifically, the setting device 50 includes settings for the preform's physical properties (material and weight), the bottle forming conditions (capacity and shape of the formed bottle B), heater settings (output of heater 5a in heating zone 5), the speed (travel / insertion) of the tension rod 11, the timing for starting the first air, the pressure of the first air, the timing for switching from the first air to the second air, and the pressure of the second air, among other blow molding settings. The blow molding machine has numerous settings, and skilled operators can precisely adjust the values ​​of multiple settings on-site. This is because skilled operators have accumulated experience in how to adjust which of the multiple settings and how the result will appear in bottle B.

[0054] In order to mechanically achieve adjustments to minute settings performed by skilled operators, thereby reducing adjustment time and stabilizing the quality of molded bottle B, and further to adjust settings for improving the quality of bottle B that even skilled operators might not notice, machine learning is performed in this embodiment as described below. Target values ​​for the physical properties that the molded bottle B should meet are assigned, and the operating conditions set on the blow molding machine to mold bottle B that meets these target values ​​(the types of setting data to be set on the blow molding machine and their setting values) are inferred as the result of machine learning processing.

[0055] As explained below, the operating conditions of the blow molding machine inferred using machine learning are, in this embodiment, typically set as the heater output in the blow molding machine. Blow pressure can also be inferred based on this or alternatively. In addition, blow molding timing, lever speed, and the temperature of the cooling water passing through the cooling water path in the bottom mold 7C of the mold 7 can be added to the inferred operating conditions. In summary, the inferred operating conditions of the blow molding machine, such as heater output, blow molding pressure, blow molding timing, lever speed, and cooling water temperature, are related to the physical properties of the bottle B formed from the preform P by the blow molding machine.

[0056] Figure 5 This is a block diagram representing the learning device 70. The learning device 70 performs machine learning related to a blow molding machine and includes: a learning data acquisition unit 71 that acquires data used in the learning process, i.e., learning data; a model generation unit 72 that uses the learning data to generate a learning model for inferring heater output, blow molding output, blow molding timing, lever speed, cooling water temperature, etc., of the blow molding machine; and a generated completed learning model 73 (its storage unit). The learning device 70 is implemented, for example, by a computer device equipped with a processor, memory, storage device, communication device, etc., and operates according to a program that enables the computer device to perform the functions of the learning device 70. By installing the aforementioned program stored in a removable recording medium 75 onto the computer device, the computer device functions as the learning device 70.

[0057] The learning data acquisition unit 71 acquires (receives) data set or measured during the actual formation of bottle B from multiple bottle forming plants (bottled beverage manufacturing plants) equipped with blow molding machines via a network (e.g., the Internet) as learning data. The learning data includes the following data.

[0058] (1) Heater output

[0059] This refers to the output of the multiple heating zones 5 of the heating station 2 of the blow molding machine. The data representing the heater output can be the output ratio (combined output) relative to the total output of the heating station 2, each heating zone 5, or each heater 5a (e.g., 80%), or it can be the on / off data of the multiple heaters 5a of the heating zone 5.

[0060] (2) Blow molding pressure

[0061] These are the pressure values ​​of the first air blown into the preform P and the pressure values ​​of the second air blown into it.

[0062] (3) Blow molding timing

[0063] This refers to the start timing of the first air and the timing of switching from the first air to the second air (switching timing). Alternatively, a combination of the end timing of the first air and the start timing of the second air can be used instead of the switching timing. In the data representing blow molding timing, the rotation angle of the rotating wheel 6, with the position (0°) as the reference point for handing over the preform P to the molding station 3, can be used. For example, if the start timing of the first air is "33°", it means that the first air begins at the timing point when the preform P is handed over to the molding station 3 and the rotating wheel 6 has rotated 33°.

[0064] (4) Rod speed

[0065] It is the travel (insertion) speed of the tension rod 11 inserted into the preform P.

[0066] (5) Physical properties of the preform

[0067] This includes the material, weight, and dimensions of the preform P. Information related to the material of the preform P may also include data distinguishing whether it is made from virgin resin or recycled resin. Additionally, it may include the IV (Intrinsic Viscosity) value of the PET resin used in the preform P. Dimensions generally include the opening diameter, the diameter of the body portion, and the length from the support ring to the bottom (body portion length).

[0068] (6) Conditions during bottle forming

[0069] This refers to the capacity and shape of bottle B. Shapes are generally categorized as round or square.

[0070] (7) Environmental Information

[0071] These are the measured values ​​of temperature and humidity around the blow molding machine. The measurements are taken using a thermometer 61 and a hygrometer 62.

[0072] (8) Physical properties of the bottle

[0073] These are the measured values ​​of the filled capacity, buckling strength, overall height, wall thickness, and section counterweight (the weights of the individual sections formed by transversely cutting the bottle B at one or more specified locations (perpendicular to the axial direction)) of the molded bottle B. Regarding wall thickness and section counterweight, values ​​are typically used for multiple locations of the bottle B, such as the neck, shoulder, body, and bottom (or even more precisely).

[0074] (9) Bottle appearance information

[0075] Bottle B, formed from preform P, exhibits a whitening (cloudiness). The presence or absence of this whitening is included in the bottle's appearance information. This appearance information may also include whether there are structural defects in bottle B. Whitening can be categorized into two types: whitening caused by overstretching and whitening caused by crystallization. Overstretching-induced whitening is more likely to occur when preform P is underheated, while crystallization-induced whitening occurs when preform P is overheated, causing the material to crystallize.

[0076] (10) Cooling water temperature of the mold

[0077] The temperature of the cooling water passing through the cooling water path is for cooling the bottom mold 7c of the aforementioned mold 7.

[0078] The model generation unit 72 learns from the learning data acquisition unit 71 and generates a learning model for inferring the operating conditions that should be set on the blow molding machine in order to mold the bottle B that meets the target. Specifically, the heater output, blow molding pressure, blow timing, rod speed, cooling water temperature, etc.

[0079] The learning algorithm used by the model generation unit 72 can use known algorithms such as taught learning, untaught learning, and reinforcement learning.

[0080] For example, the LightGBM learning algorithm, which offers high prediction accuracy and short training time, can be used. The model learns the relationship between performance variables (specifically, the physical properties of the bottle, the physical properties of the preform, the bottle molding conditions, environmental information, etc.) and target variables (heater output, blow molding pressure, blow molding timing, lever speed, cooling water temperature, etc. of the blow molding machine), and learns the parameters used to calculate the values ​​of the target variables. External specification variables (hyperparameters) used to manage the training of the learning model (such as the number of branches in the decision tree) are optimized, for example, using a Bayesian optimization library.

[0081] When learning Model 73, the output data (target variable) can be prioritized to include data that significantly influences the physical properties of the bottle B formed by the blow molding machine. Here, we specifically illustrate an example where five data points—heater output, blow molding pressure, blow molding timing, lever speed, and cooling water temperature—are used as output data. However, more types of data can also be used as output data, or conversely, fewer types of data can be used, such as only heater output data, or only heater output and blow molding pressure (especially the pressure of the first air). In LightGBM, the importance (degree of influence) of each explanatory variable affecting the target variable is calculated. Therefore, when focusing on the physical properties of the bottle in question (e.g., full capacity) (one of the explanatory variables), the data (target variable) that significantly influences the physical properties of the bottle in question can be selected as the output data for learning Model 73.

[0082] The input data (description variables) for the learning completion model 73 can, of course, be limited to, for example, the physical characteristics of the bottle B formed by the blow molding machine (e.g., the filling capacity as one of them). Furthermore, for example, if the physical characteristics of the formed bottle B are fixed, the input data for the learning completion model 73 can also be limited to environmental information (either or both of temperature and humidity). However, if multiple input data are used, the possibility of inferring settings and values ​​that the operator may not have noticed increases.

[0083] Figure 6 It is a block diagram of a deduction device used to infer the operating conditions set for a blow molding machine.

[0084] The inference device 80 includes an inference data acquisition unit 82 and an inference unit 83. The inference device 80 may also be implemented, for example, by a computer device equipped with a processor, memory, storage device, communication device, etc., and operates according to a program that enables the computer device to perform the functions of the inference device 80. By installing the aforementioned program stored in a removable recording medium 85 onto the computer device, the computer device functions as the inference device 80.

[0085] The data acquisition unit 82 for inference assigns target values ​​(design values) (at least one of filling capacity, buckling strength, total height, wall thickness and section weight) 81a to the physical properties of the bottle B to be formed, preferably assigns physical properties (at least one of material and weight of the preform P) to the preform P used to form the bottle B, bottle forming conditions (at least one of bottle capacity and shape) and environmental information (at least one of temperature and humidity) 81b.

[0086] The inference unit 83 uses the learning completion model 73 to infer the operating conditions of the blow molding machine, including heater output, blow molding pressure, blow molding timing, lever speed, and cooling water temperature. In other words, the inference unit 83 inputs the inference data acquired in the inference data acquisition unit 82 into the learning completion model 73, and outputs the heater output, blow molding pressure, blow molding timing, lever speed, and cooling water temperature to indicate that the generated bottle possesses the target physical characteristics.

[0087] The target (value) of the physical properties of the bottle to be formed can be a specific numerical value or a numerical range with a specified width. Regarding wall thickness and section counterweight, values ​​are usually set for multiple parts of bottle B, such as the neck, shoulder, body, and bottom.

[0088] By using a learning completion model 73 that has learned from multiple learning data, the inference device 80 can output the heater output, blow molding pressure, blow timing, lever speed, and cooling water temperature that should be set in the setting device 50 of the blow molding machine to generate a bottle B with physical properties that meet or fall within the target range. The data on heater output, blow molding pressure, blow molding timing, lever speed, and cooling water temperature output from the inference device 80 are provided as setting data to the setting device 50 of the blow molding machine, and the blow molding machine operates according to the provided setting data. Under the operating conditions of the blow molding machine set by a skilled operator, and under operating conditions of the blow molding machine that even a skilled operator cannot detect, it is possible to manufacture a bottle B that meets the target physical properties.

[0089] As described above, since the learning data includes bottle appearance information, the inference device 80 can output the heater output, blow molding pressure, blow molding timing, bar speed, and cooling water temperature that should be set in the setting device 50 of the blow molding machine to generate a bottle B with the target physical properties and no appearance abnormalities (e.g., no whitening). This enables the molding of higher quality bottles B.

[0090] Alternatively, it can replace the target values ​​of the physical properties that the molded bottle B should meet, or, based on this, focus on the quality of the molded bottle B that it should meet (the target quality of bottle B). In this case, the quality of bottle B (distinguishing whether blow molding is possible, the presence or absence of eccentricity, the presence or absence of condensation, etc.) is learned in the learning completion model 73, and the inference device 80 infers the operating conditions that should be set on the blow molding machine in order to mold the bottle B that meets the target quality according to the learning completion model 73.

[0091] Figure 7This indicates the processing result of the inference device 80 using the learning completion model 73. The learning completion model 73 uses "bottle capacity (600ml)" and "bottle shape (round)" as conditions during bottle molding, "fill capacity" (ml) as the target value (explanation variable, input data) of the physical properties of the bottle to be molded, and "heater output" (%) (comprehensive output) and "blow molding pressure" (first air pressure) (bar), two operating conditions of the blow molding machine, as inferred values ​​(target variable, output data). Figure 7 The data also includes the heater output and blow molding pressure inferred when the target value for filling capacity is set in the blow molding machine, the measured values ​​of the filling capacity of the five bottles B actually formed from five preforms P of the same type, and the error between the target value and the measured value. The "error between target value and measured value" further includes "error (= measured value - target value)," "error rate (= error · measured value)," "absolute error rate (= absolute value of error rate)," and "average absolute error rate (= total absolute error rate ÷ number of data points)." Figure 8 It is a graph with the horizontal axis set to the target value of the full capacity and the vertical axis set to the measured value of the full capacity. The target value and measured value of each of the five data points are plotted by circular markers. Figure 8 The dashed line shown in the graph is an auxiliary line indicating the situation where the target value and the measured value are consistent.

[0092] The mean absolute error rate of the five data points was small at 0.124%, confirming that by using the heater output and blow molding pressure inferred from the learning-completed model 73 to form bottle B, it is possible to form bottle B with a full capacity that is quite close to the target value.

[0093] Figure 9 This indicates the inference result of the inference device 80 using the learning completion model 73. The learning completion model 73 takes the "weight", "difference between virgin resin and recycled resin" and IV (Intrinsic Viscosity) value (the inherent viscosity of the resin) as input data, and "heater output" (%) (comprehensive output), which is one of the operating conditions of the blow molding machine, as the inferred value (output data).

[0094] Figure 9The document also indicates the bottle quality (in this case, whether it can be molded and its eccentricity) of eight bottles B formed by setting the heater output on a blow molding machine with different input data (weight, difference between raw materials / recycled materials, and any one of the IV values) ("Reflecting the Inference Result"), and the bottle quality of eight bottles B formed by a blow molding machine with the heater output (overall output) fixed at 70% without controlling the heater output based on the inference value ("Before Inference"). "OK" for "Whether it can be molded" indicates that blow molding can be performed, and "NG" indicates that blow molding cannot be performed (forming a deformed bottle B). "Eccentricity" refers to the deviation (misalignment) between the center of the entire bottle B and the center of its bottom; "OK" indicates no eccentricity, and "NG" indicates eccentricity.

[0095] Comparing No.1 and No.3, these preforms P have the same weight and IV value, but differ in whether they are made from virgin resin (new resin) or recycled resin (used resin). By comparing No.1 and No.3, and using recycled resin preform P (No.3) as one of the input data, it was deduced that the heater output (68%) for No.3 is lower than that for No.1 preform P made from virgin resin (70%). As a result, for No.3, the eccentricity that occurs when the heater output is set to 70% can be eliminated. Compared to virgin resin preforms P, recycled resin preforms P soften more easily during heating, thus making them more prone to eccentricity. To prevent this, when using recycled resin preforms, the preform temperature after heating is lowered. This deduction is consistent with the measures taken based on experience to date.

[0096] Comparing No.1 and No.2, they share the same preform weight P and are both made of virgin resin, but differ in the intrinsic viscosity (limiting viscosity) of the PET resin, expressed by the IV value. Comparing No.1 and No.2, by assigning an IV value of 0.80 as one of the input data (No.2), it was deduced that the heater output (67%) for No.2 is lower than that for the preform P of No.1 made of PET resin with an IV value of 0.82 (70%). As a result, for No.2, the eccentricity that occurs when the heater output is set to 70% can be eliminated. This deduction is consistent with the following approach: preforms made of PET resin with a low IV value tend to become more flexible during heating, thus easily causing eccentricity; to prevent this, a treatment based on current experience is to reduce the preform temperature after heating when using preforms made of PET resin with a low IV value.

[0097] Comparing No.1 and No.5, both preforms P are made of virgin resin with the same IV value for PET resin, but their weights differ. By comparing No.1 and No.5, weight was assigned as one of the input data points. Therefore, for No.5, which is lighter than No.1, a larger heater output (80%) was inferred compared to the preform P for No.1 (70%). As a result, for No.5, bottles B can be formed normally from preform P. This inference is consistent with the empirically derived approach of increasing the preform heating temperature to eliminate molding abnormalities.

[0098] Figure 10 This indicates the inference result of the inference device 80 using the learning completion model 73. The learning completion model 73 takes the "temperature" and "humidity" around the blow molding machine as environmental information as input data, and takes the "heater output" (%) (comprehensive output) and "cooling water temperature of the mold" (°C) as inference values ​​(output data) as operating conditions of the blow molding machine.

[0099] exist Figure 10 The text represents the bottle quality (fill capacity and condensation) and preform temperature of six bottles B formed by setting the heater output and mold cooling water temperature on a blow molding machine, with different input data (either of the temperature and humidity around the blow molding machine). It also shows the bottle quality and preform temperature of the six bottles B formed by a blow molding machine with the heater output (overall output) fixed at 65% and the mold cooling water temperature fixed at 8°C.

[0100] Comparing No.1 and No.3, their humidity levels are similar (30% and 20% respectively), but their temperatures differ (No.1 is 25°C, No.3 is 33°C). Comparing No.1 and No.3, with temperature (33°C) assigned as one of the input data (No.3), a lower heater output (62%) than No.1 at 25°C (65%) was deduced. As a result, the deviation of bottle B's filling capacity from the target (525ml) was eliminated (528ml) when the heater output was set to 65%. This deduction is consistent with countermeasures derived from experience so far: if the ambient temperature of the blow molding machine is high, the temperature of the preform P after heating increases, and the filling capacity of bottle B increases. To prevent this, when the ambient temperature of the blow molding machine is high, the heater output is reduced to keep the temperature of the preform P constant.

[0101] Comparing No.1 and No.5, their humidity levels are similar (30% and 35% respectively), but their temperatures differ (No.1 is 25°C, No.5 is 20°C). Comparing No.1 and No.5, with temperature (20°C) assigned as one of the input data (No.5), a higher heater output (67%) than No.1 at 25°C (65%) was deduced. As a result, the deviation of bottle B's filling capacity from the target (525ml) was eliminated (521ml) when the heater output was set to 65%. This deduction is consistent with countermeasures derived from current experience: if the ambient temperature around the blow molding machine is low, the temperature of the preform P after heating decreases, resulting in a smaller filling capacity of bottle B. To prevent this, when the ambient temperature around the blow molding machine is low, the heater output is increased to maintain a constant temperature for the preform P.

[0102] Comparing No.1 and No.2, they have the same temperature but different humidity levels (No.1 has 30% humidity, and No.2 has 70% humidity). By comparing No.1 and No.2 and using humidity (70%) as one of the input data (No.2), a higher cooling water temperature (10°C) was deduced compared to the cooling water temperature of No.1 (8°C) at 30% humidity. As a result, condensation that would occur when the cooling water temperature of the mold is set to 8°C can be prevented. This deduction is consistent with countermeasures derived from experience to date, namely, that high humidity around the blow molding machine easily leads to condensation on the mold and adhesion of condensate to bottle B; to prevent this, the cooling water temperature of the mold is increased when the humidity around the blow molding machine is high, thus preventing condensation.

[0103] 1: Preform Supply Station

[0104] 2: Heating Station

[0105] 3: Molding Station

[0106] 5: Heater area

[0107] 5a: Heater

[0108] 7: Mold

[0109] 11: Tension bar

[0110] 12: Air pump

[0111] 13: Valve block

[0112] 20: Heating function

[0113] 30: Blow molding function

[0114] 40: Control device

[0115] 50: Setting device

[0116] 61: Thermometer

[0117] 62: Hygrometer

[0118] 70: Learning Device

[0119] 71: Learning Data Acquisition Department

[0120] 72: Model Generation Department

[0121] 73: Learning to complete the model

[0122] 75, 85: Recording media

[0123] 80: Inference device

[0124] 82: Inference Data Acquisition Department

[0125] 83: Inference Department

[0126] B: Bottle

[0127] P: Preform

Claims

1. A device for determining the operating conditions of a blow molding machine, wherein the blow molding machine uses a heater to heat a preform, moves the heated preform towards a mold, stretches the preform axially by inserting a tension rod, and stretches the preform outward by supplying pressurized fluid to the interior, thereby forming a hollow bottle, the device for determining the operating conditions of the blow molding machine being characterized by comprising: The learning model learns parameters by taking the physical properties of the bottle or environmental information as input data and the operating conditions of the blow molding machine as output data. The inference unit's inference should be set based on the operating conditions of the blow molding machine, which are inferred by inputting target values ​​or environmental information of the physical properties that the molded bottle should meet after being learned into the model, and used to mold bottles that meet the target values ​​or target quality.

2. The device for inferring the operating conditions of a blow molding machine according to claim 1, characterized in that, The input data also includes at least one of the physical properties of the preform and the molding conditions of the bottle.

3. The device for inferring the operating conditions of a blow molding machine according to claim 1, characterized in that, The operating conditions of the blow molding machine are at least one of the heater output conditions for heating the preform and the pressure conditions of the fluid supplied to the preform.

4. The device for inferring the operating conditions of a blow molding machine according to claim 3, characterized in that, The operating conditions of the blow molding machine also include timing conditions for supplying fluid to the preform or speed conditions for inserting the tension rod into the preform.

5. The device for inferring the operating conditions of a blow molding machine according to claim 1, characterized in that, The input data includes information related to whether the bottle has any abnormal appearance. The inference unit is a unit that infers the operating conditions of the blow molding machine to form a bottle that meets the target value or target quality and has no abnormal appearance.

6. The device for inferring the operating conditions of a blow molding machine according to claim 2, characterized in that, The physical properties of the preform include data related to the preform's material and weight.

7. The device for inferring the operating conditions of a blow molding machine according to claim 2, characterized in that, The physical properties of the preform include data that distinguish whether the preform is made of virgin resin or recycled resin.

8. The device for inferring the operating conditions of a blow molding machine according to claim 2, characterized in that, The physical properties of the preform include its IV value.

9. The device for inferring the operating conditions of a blow molding machine according to claim 2, characterized in that, The conditions for forming the bottle are data regarding the bottle's capacity and shape.

10. The device for inferring the operating conditions of a blow molding machine according to claim 1, characterized in that, The physical properties of the bottle include at least one of the following: full capacity, buckling strength, overall height, wall thickness, and segment weight.

11. The device for inferring the operating conditions of a blow molding machine according to claim 1, characterized in that, The environmental information is at least one of the temperature and humidity around the blow molding machine.

12. The device for inferring the operating conditions of a blow molding machine according to claim 1, characterized in that, The device includes a setting unit that sets the operating conditions of the blow molding machine as inferred by the inference unit.

13. A learning device, characterized in that, have: The learning data acquisition unit acquires learning data including the operating conditions of the blow molding machine and the physical properties or environmental information of the molded bottle. The operating conditions of the blow molding machine are: heating the preform with a heater, moving the heated preform towards the mold and inserting a stretching rod into the preform, thereby stretching the preform axially, and stretching the preform outward by supplying pressurized fluid to the inside, thereby forming the hollow bottle. The model generation unit uses the learning data to generate a learned model based on the target values ​​or target quality of the physical properties that the molded bottle should meet, to infer the operating conditions of the blow molding machine that should be set in order to mold the bottle that meets the target values ​​or target quality.

14. A method for inferring the operating conditions of a blow molding machine, wherein the blow molding machine heats a preform using a heater, moves the heated preform towards a mold, inserts a stretching rod into the preform, thereby stretching the preform axially, and stretches it outward by supplying pressurized fluid internally, thus forming a hollow bottle; the method for inferring the operating conditions of the blow molding machine is characterized in that... The learning process is complete. The learning process involves learning parameters by taking the physical properties of the bottle or environmental information as input data and the operating conditions of the blow molding machine as output data. The operating conditions of the blow molding machine are inferred by inputting target values ​​of physical properties or environmental information that the bottle after molding should meet from the learned model. These operating conditions are set in the blow molding machine in order to mold bottles that meet the target values ​​or target quality.

15. A program, characterized in that, Used to enable a computer to perform the method of claim 14.

Images