WO2024075852A1

WO2024075852A1 - Machine learning device, information processing device, reasoning device, machine learning method, control conditions prediction method, and reasoning method

Info

Publication number: WO2024075852A1
Application number: PCT/JP2023/036657
Authority: WO
Inventors: 洋一田所; 七勢大竹
Original assignee: 東洋製罐グループホールディングス株式会社
Priority date: 2022-10-07
Filing date: 2023-10-07
Publication date: 2024-04-11
Also published as: JP2024055692A

Abstract

[Problem] To provide a machine learning device which easily derives heating device control conditions, an information processing device, a reasoning device, a machine learning method, a control conditions prediction method and a reasoning method. [Solution] A machine learning device which generates a learning model 45 for predicting new control conditions 102b of a heating device 10 for heating a temperature control target 200. The machine learning device is equipped with: a learning data storage unit 43 for storing one or more sets of learning data 46; a machine learning control unit 41 for training the learning model 45 on the correlation between learning input parameters 101a and learning control conditions 101b, which are the control conditions for the heating device 10; and a machine learning model storage unit 44 for storing the trained learning model 45. The learning input parameters 101b comprise one or more of the following: material information, shape information and initial temperature information for the temperature control target 200, and target temperature distribution information and thermal environment information for the heating device 10.

Description

Machine learning device, information processing device, inference device, machine learning method, control condition prediction method, and inference method

This disclosure relates to a machine learning device, an information processing device, an inference device, a machine learning method, a control condition prediction method, and an inference method.

　Conventionally, a heating device equipped with multiple infrared heaters is used to heat the preforms being blown into hollow containers by a blow molding device while they are being transported (for example, Patent Document 1).

WO2014/208693A1

When using a heating device to heat a temperature-controlled object such as the above-mentioned preform, it is necessary to set detailed control conditions such as the heater output conditions in the heating device and the transport speed when the temperature-controlled object passes through the heating space. However, setting the control conditions for the heating device often relies on the experience of a skilled worker, and there is an issue in that it is not easy to determine them.

In view of the above problems, the present disclosure aims to provide a machine learning device, an information processing device, an inference device, a machine learning method, a control condition prediction method, and an inference method that can easily derive the control conditions of a heating device.

In order to achieve the above object, the machine learning device of the present disclosure includes:
A machine learning device that generates a learning model for predicting a control condition of a heating device that heats a temperature control target,
a learning data storage unit that stores one or more pieces of learning data;
a machine learning control unit that causes the learning model to learn based on the one or more pieces of learning data;
A machine learning model storage unit that stores the learned learning model;
Equipped with
Each of the learning data is composed of learning input parameters and learning control conditions, which are control conditions of the heating device corresponding to the learning input parameters,
the learning model learns a correlation between the learning input parameters in the learning data and the learning control conditions;
The learning input parameters are composed of at least one of material information of the temperature control object, shape information of the temperature control object, initial temperature information of the temperature control object, target temperature distribution information of the heating device, and thermal environment information of the heating device.

The machine learning device, information processing device, inference device, machine learning method, control condition prediction method, and inference method disclosed herein make it possible to easily determine control conditions.

　Problems, configurations and effects other than those mentioned above will be made clear in the detailed description of the invention described below.

1 is a schematic diagram showing a heating system according to a first embodiment. FIG. 2 is a side view showing the heating device of FIG. 1 . FIG. 2 is a configuration diagram showing the machine learning device of FIG. 1. FIG. 4 is a conceptual diagram showing elements of machine learning implemented by the machine learning device of FIG. 3 . FIG. 2 is a configuration diagram showing the information processing device of FIG. 1; FIG. 1 is a diagram illustrating the hardware configuration of a computer. 2 is a flowchart showing a machine learning method performed by the machine learning device of FIG. 1 . 4 is a flowchart showing a control condition prediction method by the information processing device of FIG. 1 . FIG. 11 is a schematic diagram showing a heating device and a transport device according to a second embodiment.

Below, an embodiment for carrying out the present invention will be described with reference to the drawings. Below, the scope necessary for the explanation to achieve the object of the present invention will be shown diagrammatically, and the scope necessary for the explanation of the relevant parts of this disclosure will be mainly explained, and the parts where explanation is omitted will be based on publicly known technology.

First embodiment
Fig. 1 is a schematic diagram showing a heating system 1 according to embodiment 1. Fig. 2 is a side view showing the heating device 10 of Fig. 1. The heating system 1 includes the heating device 10, a machine learning device 40, an information processing device 50, a control device 60, and a network 70 connecting the respective devices.

The heating system 1 is installed in a production system that produces containers such as PET bottles. The containers are produced by molding preforms 200 produced from the container raw material using a blow molding device (not shown).

When molding the preform 200 using a blow molding device, it is necessary to preheat the preform 200 in order to improve the moldability of the preform 200. The heating system 1 disclosed herein is a system that handles the pre-processing of the blow molding device, and is a system that heats the preform 200, which is an object of temperature control.

The preforms 200 are molded from the container raw material by an injection molding machine (not shown). The molded preforms 200 are successively carried into the heating system 1 and heated. The heated preforms 200 are successively fed to the blow molding device in the next process and molded into containers.

The heating device 10 has a heating device main body 11, a plurality of heaters 14 as heat sources arranged in the heating device main body 11, a heater power supply (not shown) that supplies power to each heater 14, a blower device 17 arranged in the heating device main body 11, and a conveying device 30.

The heating device main body 11 has a heating passage 12, which is a long, tunnel-shaped space, formed along the longitudinal direction. The heating passage 12 is large enough for the preform 200 to pass through. The preform 200 can pass through the heating passage 12 from the entrance side to the exit side. The transportation of the preform 200 within the heating passage 12 will be explained later.

Each heater 14 is a long near-infrared heater. Each heater 14 is arranged along the longitudinal direction of the heating passage 12. Note that, in addition to long near-infrared heaters, different types of heaters can be used for each heater 14.

Each heater 14 can heat the preform 200 in the heating passage 12 inside the heating device main body 11. Each heater 14 is electrically connected to a heater power supply, and can heat the inside of the heating device main body 11 with the power supplied from the heater power supply. Therefore, the heating space of the heating device 10 is defined by the heating device main body 11.

The heater power supply can adjust the amount of power supplied to each heater 14 according to commands from the control device 60. Each heater 14 radiates an amount of heat according to the amount of power supplied to it. By controlling the amount of power supplied to each heater 14, the amount of heat radiated from each heater 14 can be controlled.

The blower device 17 has a blower 18, which is an air source, a blower duct 19, and a blower power source (not shown). The blower 18 is disposed on the inlet side of the heating passage 12. The blower duct 19 is a pipe-shaped object.

The longitudinal direction of the air blower duct 19 is aligned with the longitudinal direction of the heating passage 12 and is disposed within the heating passage 12. One open end of the air blower duct 19 opens toward the outside of the heating device 10 at the inlet side of the heating passage 12, and the other open end of the air blower duct 19 opens toward the outside of the heating device 10 at the outlet side of the heating passage 12.

The air duct 19 has multiple through holes (not shown). The blower 18 is connected to the open end of the air duct 19 on the inlet side. The blower 18 can send air into the air duct 19 from the open end of the air duct 19 on the inlet side.

The air sent into the air duct 19 by the blower 18 flows through the air duct 19. The air flowing through the air duct 19 is further sent into the heating passage 12 through a number of through holes, and is also released into the atmosphere from the open end of the air duct 19 on the outlet side. That is, the air sent by the blower 18 is sent into the heating space via the air duct 19.

The blower power supply can control the amount of power supplied to the blower 18 according to commands from the control device 60. The control device 60 controls the amount of power supplied to the blower 18, thereby controlling the blower 18 and the flow rate of air sent into and flowing through the air duct 19.

The conveying device 30 has a conveying device main body 31 and a conveying control unit 32. The conveying device main body 31 is a conveyor. The conveying device main body 31 can support a plurality of preforms 200. The conveying device main body 31 is positioned so that the preforms 200 it supports pass through the heating passage 12.

The conveying device body 31 can rotatably support the multiple preforms 200 in an upright state. The conveying device body 31 has a rotation mechanism (not shown) and can impart a rotational motion to each of the multiple preforms 200 in an upright state. Therefore, the conveying device body 31 can convey the preforms 200 through the heating passage 12 while imparting a rotational motion to the upright preforms 200. In other words, each preform 200 supported by the conveying device body 31 can pass through the heating space while rotating in an upright state.

The conveying control unit 32 can control the conveying speed of the conveying device main body 31 based on commands from the control device 60. By controlling the conveying speed of the conveying device main body 31, the time that the preform 200 stays in the heating passage 12 can be controlled. Furthermore, the conveying control unit 32 can control the rotational speed of the rotational motion of the preform 200 by controlling the rotation mechanism of the conveying device main body 31 based on commands from the control device 60. That is, the conveying control unit 32 can control at least one of the conveying speed of the conveying device main body 31 and the rotational speed of the rotational motion of the preform 200 based on commands from the control device 60.

The machine learning device 40 is a device that operates as the main subject of the learning phase of machine learning. FIG. 3 is a configuration diagram showing the machine learning device 40 of FIG. 1. FIG. 4 is a conceptual diagram showing the elements of machine learning performed by the machine learning device 40 of FIG. 3. The machine learning device 40 has a machine learning control unit 41, a machine learning communication unit 42, a learning data storage unit 43, and a machine learning model storage unit 44.

The machine learning control unit 41 generates a learning model 45 that provides output information corresponding to input information. The machine learning control unit 41 uses one or more pieces of learning data 46 to generate the learning model 45. The learning data 46 is composed of learning input parameters 101a and learning control conditions 101b, which are control conditions for the heating device 10 that correspond to the learning input parameters 101a.

The learning data 46 is data used as training data, verification data, and test data, which are teacher data in supervised learning. The learning control conditions 101b are data used as correct answer labels in supervised learning. The learning input parameters 101a and the learning control conditions 101b, which are the control conditions of the heating device 10 corresponding to the learning input parameters 101a, represent data in a state in which the input parameters 100a prepared by the operator and the control conditions 100 of the heating device 10 corresponding to the input parameters 100a are stored in the learning data storage unit 43 as the learning data 46.

The input parameters 100a and control conditions 100b prepared by the operator, as well as the learning data storage unit 43 in which the learning data 46 is stored, will be explained later.

The machine learning control unit 41 can cause the learning model 45 to learn the correlation between the learning input parameters 101a and the learning control conditions 101b in one or more pieces of learning data 46. This allows the machine learning control unit 41 to generate a learned learning model 45.

The learning model 45 employs a neural network structure. The learning model 45 has an input layer 45a, an intermediate layer 45b, and an output layer 45c. The input layer 45a has a number of neurons corresponding to the number of parameters in the learning input parameters 101a. The output layer 45c has a number of neurons corresponding to the number of conditions in the learning control conditions 101b.

Between each layer, there are synapses (not shown) that connect each neuron. Weights can be assigned to each synapse.

The machine learning control unit 41 adjusts a group of weight parameters consisting of the weights of each synapse through machine learning. The group of weight parameters is reflected in the learning model 45.

In the machine learning control unit 41, each parameter of the learning input parameters 101a is input to each neuron of the input layer 45a, and the learning result control condition 101c of the heating device 10 corresponding to the learning input parameters 101a is output through the learning model 45. The machine learning control unit 41 examines the value of the learning result control condition 101c and adjusts the weight of each synapse based on the result of the examination.

When the learning model 45 is configured as a regression model, the learning control conditions 101b are output as numerical values normalized to a predetermined range (e.g., 0 to 1). When the learning model 45 is configured as a classification model, the learning control conditions 101b are output as scores (accuracies) for each class as numerical values normalized to a predetermined range (e.g., 0 to 1).

The learning data storage unit 43 can store multiple pieces of learning data 46 as a database. The specific configuration of the database that constitutes the learning data storage unit 43 can be designed as appropriate.

The learning data storage unit 43 receives input parameters 100a and control conditions 100b corresponding to the input parameters 100a from an operator using an operation terminal (not shown). The input parameters 100a and control conditions 100b input to the learning data storage unit 43 are stored as learning input parameters 101a and learning control conditions 101b, respectively. Therefore, the learning control conditions 101b correspond to the learning input parameters 101a. As a result, one piece of learning data 46 is stored.

Data from a known heating operation is used for the learning input parameters 101a and the learning control conditions 101b. That is, the input parameters 100a and control conditions 100b from a known heating operation are input as the input parameters 100a and control conditions 100b input to the learning data storage unit 43 as the basis for the learning input parameters 101a and the learning control conditions 101b.

The known heating operation is preferably a past heating operation using the heating device 10. The data actually used in the past heating operation using the heating device 10 is data with a track record of subsequent production of containers using a blow molding machine.

In other words, the data actually used in past heating operations is data that can produce containers with shape accuracy within an acceptable range. Therefore, since the correlation between the input parameters 100a and the control conditions 100b actually used in past heating operations is valid, it is desirable to use such data as learning data 46 for machine learning.

In addition, data on known heating operations may be data actually used in past heating operations, or data created by simulations or experienced workers, for example.

The input parameters 100a are parameters that are taken into consideration in order to calculate the control conditions 100b for controlling each device of the heating system 1. The input parameters 100a are parameters related to the preform 200, parameters related to the heating device 10, and parameters related to the relationship between the preform 200 and the heating device 10.

Specifically, the input parameters 100a are composed of at least one of the material information, shape information, initial temperature information, target temperature distribution information of the heating device 10, and thermal environment information of the heating device 10 of the preform 200. The material information of the preform 200 is at least one of the light transmittance, reflection, and refractive index of the preform 200, the L value which is color brightness information, the RGB value as color information, density information, specific heat information, and thermal conductivity information.

The material information can be obtained by measuring the preform 200 in advance. Any known method can be appropriately adopted as the method for measuring the material information.

The shape information of the preform 200 is at least one of design drawing information as shape information read from the design drawing of the preform 200, shape image information as shape information read from the shape image, dimensional information, and shape tracing measurement/scanning information. When reading shape information from the design drawing of the preform 200, the design drawing of the preform 200 is provided to a computer in advance, and the computer reads the shape information of the preform 200 from the design drawing, thereby obtaining the shape information of the preform 200.

The shape image is image information captured by an imaging device. When reading shape information from the shape image, the shape information of the preform 200 can be obtained from the image information. The dimensional information can be obtained by having the worker input the required information of the dimensions of the preform 200 in advance. The shape tracing measurement/scanning information is the shape information of the preform 200 that is grasped by scanning the shape of the preform 200 with a well-known scanning device.

The initial temperature information of the preform 200 can be obtained by measuring the temperature of the preform 200 before the preform 200 is provided to the heating system 1. The target temperature distribution information of the heating device 10 is information that indicates the ideal temperature distribution inside the heating passage 12 during the heating operation. The ideal temperature distribution inside the heating passage 12 can be calculated and obtained by prior analysis, etc.

The thermal environment information of the heating device 10 includes at least one of the following information: distance information from each heater 14 of the heating device 10 to the preform 200, position, irradiation direction, and irradiation angle information of each heater 14, dimensional information of the heating device 10, heat diffusion information outside the heating device 10, and ambient temperature and humidity information of the heating device 10. The distance in the distance information from each heater 14 of the heating device 10 to the preform 200 is the distance in the direction perpendicular to the transport direction of the preform 200.

Information on the distance between the preform 200 and each heater 14 can be obtained by an operator calculating and inputting the distance in advance, or by inputting the respective design drawings into a computer and analyzing the computer. Information on the position, irradiation direction, and irradiation angle of each heater 14 can be obtained by inputting the design drawings of the heating device 10 into a computer and analyzing the computer. Information on heat diffusion outside the heating device 10 can be obtained by computer simulation analysis, etc. Information on the ambient temperature and humidity of the heating device 10 can be obtained by measuring the temperature and humidity around the heating device 10.

The dimensional information of the heating device 10 can be obtained in advance by the worker inputting the required dimensional values of the heating device 10.

The control conditions 100b are the control conditions for the heating device 10, the blower device 17, and the conveying device 30. Specifically, they are the output command value for the blower 18, the output command value for the conveying device 30, and the output command value for each heater 14.

Each output command value is an arbitrary value between 0 and 100. Each output command value is the output value of each device when the maximum output of the corresponding device is set to 100. Note that the command to each heater 14 may not only be an output command value that is an arbitrary value between 0 and 100, but may also be an ON/OFF command for each heater 14. Furthermore, the command to each heater 14 may be the time of the ON command for each heater 14, the time of the OFF command, or the ratio of the time of the ON command to the time of the OFF command.

The machine learning control unit 41 can extract any one or more pieces of learning data 46 from the multiple pieces of learning data 46 stored in the learning data storage unit 43, and use them for machine learning.

The machine learning model storage unit 44 is a database that stores the trained learning model 45 generated by the machine learning control unit 41, i.e., the adjusted weight parameter group.

The machine learning communication unit 42 is a communication interface unit. The machine learning communication unit 42 is connected to an external device via the network 70, and is therefore capable of transmitting and receiving various types of data. The trained learning model 45 stored in the machine learning model storage unit 44 is provided to the information processing device 50 via the network 70, a storage medium, etc.

Note that in FIG. 3, the learning data storage unit 43 and the machine learning model storage unit 44 are shown as separate storage units, but they may be configured as a single storage unit.

FIG. 5 is a schematic diagram showing the information processing device 50 of FIG. 1. The information processing device 50 is a device that operates as the main body of the inference phase of machine learning. The information processing device 50 uses the learning model 45 generated by the machine learning device 40 to predict new control conditions 102b of the heating device 10 that correspond to newly input input parameters for prediction 102a.

The information processing device 50 has an information processing control unit 51, an information processing model storage unit 55, and an information processing communication unit 56. The information processing control unit 51 has an information acquisition unit 52, an information prediction unit 53, and an output processing unit 54.

The information acquisition unit 52 acquires prediction input parameters 102a input by the worker via an operation terminal (not shown). The prediction input parameters 102a are input parameters for a new heating operation, which is a heating operation in which a new heating device 10 is used for heating.

The control conditions for a new heating operation using the heating device 10 corresponding to the prediction input parameters 102a are referred to as new control conditions 102b.

The information prediction unit 53 predicts new control conditions 102b by inputting the prediction input parameters 102a acquired by the information acquisition unit 52 into the learning model 45. The output processing unit 54 outputs the new control conditions 102b predicted by the information prediction unit 53 to the information processing and communication unit 56.

The information prediction unit 53 can select and use one learning model 45 from among the multiple learning models 45 stored in the information processing model storage unit 55.

The information processing model storage unit 55 is a database that stores the trained learning models 45 used by the information prediction unit 53. The information processing model storage unit 55 can store multiple trained learning models 45 input from the machine learning device 40.

Each of the multiple learning models 45 is a multiple trained model that differs in, for example, machine learning method, type of data included in learning input parameters 101a, type of data included in learning control conditions 101b, etc.

The information processing model storage unit 55 may be replaced by a storage unit of an external computer such as a server-type computer or a cloud-type computer. In that case, the information prediction unit 53 can access the storage unit of the external computer to acquire the learning model 45.

The information processing and communication unit 56 is connected to be able to communicate with devices outside the heating system 1 via the network 70. The information processing and communication unit 56 is a communication interface unit that transmits and receives various types of data. The information processing and communication unit 56 can output the new control conditions 102b output by the output processing unit 54 to the information processing and communication unit 56.

Returning to FIG. 1, the explanation will be continued. The control device 60 can control each device of the heating system 1. The control device 60 can control each device of the heating system 1 based on the new control condition 102b transmitted from the information processing device 50.

The control device 60 outputs the output command value for the blower 18 among the new control conditions 102b to the blower power supply. The blower power supply controls the blower 18 based on the input output command value. This makes it possible to control the flow rate of air flowing through the blower duct 19.

The control device 60 outputs the output command value for the conveying device 30 among the new control conditions 102b to the conveying control unit 32. The conveying control unit 32 controls the conveying device 30 based on the input output command value. This makes it possible to control at least one of the conveying speed of the conveying device main body 31 and the rotational speed of the rotational motion of the preform 200.

The control device 60 outputs the output command values for each heater 14 in the new control condition 102b to the heater power supply. The heater power supply controls each heater 14 based on the input output command values. This makes it possible to control the output of each heater 14.

FIG. 6 is a hardware configuration diagram of the computer 900. The machine learning device 40 and the information processing device 50 of the heating system 1 are configured by a general-purpose or dedicated computer 900.

The computer 900 includes a bus 910, a processor 912, a memory 914, an input device 916, an output device 917, a display device 918, a storage device 920, a communication interface unit 922, an external device interface unit 924, an input/output device interface unit 926, and a media input/output unit 928. Note that the above components may be omitted as appropriate depending on the application for which the computer 900 is used.

The processor 912 is composed of one or more arithmetic processing devices (CPU (Central Processing Unit), MPU (Micro-processing unit), DSP (digital signal processor), GPU (Graphics Processing Unit), etc.) and operates as a control unit that oversees the entire computer 900.

Memory 914 stores various data and programs 930, and is composed of, for example, volatile memory (DRAM, SRAM, etc.) that functions as main memory, non-volatile memory (ROM), flash memory, etc.

The input device 916 is, for example, a keyboard, a mouse, a numeric keypad, an electronic pen, etc., and functions as an input unit. The output device 917 is, for example, a sound output device including voice, a vibration device, etc., and functions as an output unit. The display device 918 is, for example, a liquid crystal display, an organic EL display, electronic paper, a projector, etc., and functions as an output unit.

The input device 916 and the display device 918 may be configured as one unit, such as a touch panel display. The storage device 920 is configured, for example, as an HDD, SSD (Solid State Drive), etc., and functions as a memory unit. The storage device 920 stores various data necessary for the execution of the operating system and the program 930.

The communication interface unit 922 is connected to a network 940 such as the Internet or an intranet by wire or wirelessly, and functions as a communication unit that transmits and receives data to and from other computers in accordance with a specific communication standard. The network 940 may be the same as the network 70.

The external device interface unit 924 is connected to an external device 950 such as a camera, printer, scanner, or reader/writer via a wired or wireless connection, and functions as a communication unit that transmits and receives data to and from the external device 950 in accordance with a specified communication standard.

The input/output device interface unit 926 is connected to input/output devices 960 such as various sensors and actuators, and functions as a communication unit that transmits and receives various signals and data between the input/output devices 960, such as detection signals from sensors and control signals to actuators.

The media input/output unit 928 is composed of a drive device such as a DVD drive or a CD drive, and reads and writes data from and to the media 970, which is a storage medium such as a DVD or a CD.

In the computer 900 having the above configuration, the processor 912 calls up the program 930 stored in the storage device 920 into the memory 914, executes it, and controls each part of the computer 900 via the bus 910. Note that the program 930 may be stored in the memory 914 instead of the storage device 920.

The program 930 may be recorded on the medium 970 in an installable file format or an executable file format, and provided to the computer 900 via the media input/output unit 928. The program 930 may be provided to the computer 900 by downloading it over the network 940 via the communication interface unit 922.

In addition, the computer 900 may realize the various functions realized by the processor 912 executing the program 930 using hardware such as an FPGA or ASIC.

The computer 900 may be, for example, a stationary computer or a portable computer, and may be any type of electronic device. The computer 900 may be a client-type computer, a server-type computer, or a cloud-type computer. The computer 900 may also be applied to devices other than the machine learning device 40 and the information processing device 50 of the heating system 1.

Next, the machine learning method will be explained. Figure 7 is a flowchart showing the machine learning method by the machine learning device 40 of Figure 1.

First, the operator inputs one or more pieces of learning data 46 in advance and stores them in the learning data storage unit 43. The number of pieces of learning data 46 to be stored is set taking into consideration the inference accuracy required for the learning model 45 that is ultimately obtained.

In the machine learning method, a learning model preparation process is carried out as step S100. The machine learning control unit 41 prepares a learning model 45 before learning. In the prepared learning model 45, the weights of each synapse are set to initial values.

Next, a machine learning process is performed in step S110. In the machine learning process, first, a learning data acquisition process is performed in step S111. The machine learning control unit 41 randomly acquires one piece of learning data 46 from the multiple pieces of learning data 46 stored in the learning data storage unit 43.

Next, an inference result output process is carried out as step S112. The machine learning control unit 41 inputs the learning input parameters 101a contained in one acquired piece of learning data 46 to the input layer 45a of the prepared learning model 45. As a result, the learning result control condition 101c is output as an inference result from the output layer 45c of the learning model 45.

The learning result control condition 101c output as the inference result is generated by the learning model 45 before or during learning. Therefore, the learning result control condition 101c is different from the learning control condition 101b, which is the correct label included in the learning data 46.

Next, a weight adjustment process is performed in step S113. The machine learning control unit 41 compares the learning control conditions 101b in the learning data 46 acquired in step S111, which is the correct label, with the learning result control conditions 101c output as the inference result in step S112. Based on this comparison, the machine learning control unit 41 performs backprovisioning, which is a process of adjusting the weights of each synapse, and performs machine learning.

As a result, the machine learning control unit 41 causes the learning model 45 to learn the correlation between the learning input parameters 101a and the learning control conditions 101b.

Next, a machine learning end determination process is performed as step S114. The machine learning control unit 41 determines whether a predetermined learning end condition has been met. This determination is performed, for example, based on the evaluation value of an error function based on the learning control condition 101b, which is the correct label, and the learning result control condition 101c, and the remaining amount of unlearned learning data 46 stored in the learning data storage unit 43, etc.

If, in step S114, the machine learning control unit 41 determines that the learning termination condition has not been met and that machine learning is to continue, i.e., if step S114 returns No, the process returns to step S111. In this way, the processes of steps S111 to S114 are performed multiple times on the unlearned learning data 46 for the learning model 45 being trained.

On the other hand, in step S114, if the machine learning control unit 41 determines that the learning end condition is met and machine learning is to be ended, i.e., if step S114 returns Yes, the process proceeds to step S120.

Then, in step S120, a trained model storage process is performed. The machine learning control unit 41 stores the trained learning model 45 in which the weights of each synapse have been adjusted, i.e., the learning model 45 reflecting the adjusted weight parameter group, in the machine learning model storage unit 44. This ends the machine learning method.

Next, a method for predicting new control conditions 102b for the heating system 1 using the information processing device 50 will be described. Figure 8 is a flowchart showing a method for predicting learning control conditions using the information processing device 50 of Figure 1.

First, the prediction input parameter acquisition process is carried out as step S200. The operator inputs the prediction input parameters 102a to the information processing device 50, and the information acquisition unit 52 acquires the prediction input parameters 102a.

Next, a prediction process is carried out as step S210. The information prediction unit 53 inputs the prediction input parameters 102a acquired in step S200 to the learning model 45. As a result, the information prediction unit 53 predicts new control conditions 102b corresponding to the prediction input parameters 102a.

Next, the output processing step is carried out as step S220. As the output processing, the output processing unit 54 transmits the new control conditions 102b generated in step S210 to the control device 60. This completes the prediction and output of the new control conditions 102b.

The control device 60 can control each device of the heating system 1 based on the new control conditions 102b input from the output processing unit 54.

The present disclosure can also be provided in the form of a machine learning program that is a program that causes the computer 900 to function as each unit of the machine learning device 40, or a machine learning program that is a program that causes the computer 900 to execute each step of the machine learning method.

The present disclosure can also be provided in the form of a program for causing the computer 900 to function as each component of the heating system 1, or a learning control condition prediction program that is a program for causing the computer 900 to execute each step of the learning control condition prediction method according to the above embodiment.

The machine learning device 40 according to embodiment 1 can generate a learning model 45 for predicting new control conditions 102b of the heating device 10 that heats the preform 200. The machine learning device 40 also includes a learning data storage unit 43 that stores one or more pieces of learning data 46, a machine learning control unit 41 that causes the learning model 45 to learn based on the one or more pieces of learning data 46, and a machine learning model storage unit 44 that stores the learned learning model 45. Each piece of learning data 46 is composed of learning input parameters 101a and learning control conditions 101b of the heating device 10 that correspond to the learning input parameters 101a. The learning model 45 also learns the correlation between the learning input parameters 101a and the learning control conditions 101b in the learning data 46. In addition, the learning input parameters 101a are composed of at least one of the material information of the preform 200, the shape information of the preform 200, the initial temperature information of the preform 200, the target temperature distribution information of the heating device 10, and the thermal environment information of the heating device 10. As a result, the machine learning control unit 41 can acquire a learning model 45 that has learned the correlation between the learning input parameters 101a and the learning control conditions 101b. Therefore, using the acquired learning model 45, the new control conditions 102b of the heating device 10 can be easily derived. In addition, by controlling the heating device under the new control conditions 102b, it is possible to prevent poor heating of the preform 200, and by molding the preform 200 in a good heating state with a blow molding device, it is possible to produce a product having appropriate finished dimensions.

According to the machine learning device 40 according to the first embodiment, the material information of the preform 200 is at least one of the light transmission/reflection/refractive index information of the preform 200, the color brightness information of the preform 200, the color information of the preform 200, the density information of the preform 200, the specific heat information of the preform 200, and the thermal conductivity information of the preform 200. As a result, new control conditions 102b for the heating device 10 can be easily derived using a learning model 45 that has learned the correlation between the learning input parameters 101a, including at least one of the light transmission/reflection/refractive index information of the preform 200, the color brightness information of the preform 200, the color information of the preform 200, the density information of the preform 200, the specific heat information of the preform 200, and the thermal conductivity information of the preform 200, and the learning control conditions 101b.

According to the machine learning device 40 according to the first embodiment, the shape information of the preform 200 is at least one of the design drawing information of the preform 200, the shape image information of the preform 200, the dimensional information of the preform 200, and the shape tracing measurement/scanning information of the preform 200. As a result, the learning input parameters 101a include at least one of the design drawing information of the preform 200, the shape image information of the preform 200, the dimensional information of the preform 200, and the shape tracing measurement/scanning information of the preform 200. Therefore, new control conditions 102b of the heating device 10 can be easily derived using the learning model 45 that has learned the correlation between the learning input parameters 101a and the learning control conditions 101b.

According to the machine learning device 40 according to the first embodiment, the thermal environment information of the heating device 10 is at least one of the following: distance information from the heaters 14 of the heating device 10 to the preform 200; information on the position, irradiation direction, and irradiation angle of the heaters 14 of the heating device 10; dimensional information of the heating device 10; information on thermal diffusion to the outside of the heating device 10; and information on the ambient temperature and humidity of the heating device 10. As a result, the learning input parameters 101a include at least one of the following: distance information from the heaters 14 of the heating device 10 to the preform 200; information on the position, irradiation direction, and irradiation angle of the heaters 14 of the heating device 10; dimensional information of the heating device 10; information on thermal diffusion to the outside of the heating device 10; and information on the ambient temperature and humidity of the heating device 10. Therefore, the new control conditions 102b of the heating device 10 can be easily derived using the learning model 45 that has learned the correlation between the learning input parameters 101a and the learning control conditions 101b.

According to the machine learning device 40 according to the first embodiment, the heating device 10 has a heating device main body 11 having a heating passage 12, and an air blower 17, and the air blower 17 has a blower 18 and an air duct 19 for sending air from the blower 18 into the heating passage 12. The learning control condition 101b is the flow rate of air flowing through the air duct 19. This makes it possible to obtain a new control condition 102b, which is the flow rate of air flowing through the air duct 19, by using the learning model 45 that has learned the correlation between the learning input parameter 101a and the learning control condition 101b. Therefore, the new control condition 102b of the heating device 10 can be easily derived.

According to the machine learning device 40 according to the first embodiment, the heating device 10 has a heating device main body 11 having a heating passage 12, and a conveying device 30 that conveys the preform 200 while rotating the preform 200 in the heating space. The learning control condition 101b is at least one of the conveying speed of the conveying device 30 and the rotational speed of the rotational motion of the preform 200. As a result, a new control condition 102b, which is at least one of the conveying speed of the preform 200 conveyed by the conveying device main body 31 and the rotational speed of the rotational motion of the preform 200, can be obtained using a learning model 45 that has learned the correlation between the learning input parameters 101a and the learning control conditions 101b. Therefore, the new control condition 102b of the heating device 10 can be easily derived.

According to the machine learning device 40 according to the first embodiment, the learning control conditions 101b are output command values for each heater 14 of the heating device 10. This makes it possible to obtain new control conditions 102b, which are output command values for each heater 14, using the learning model 45 that has learned the correlation between the learning input parameters 101a and the learning control conditions 101b. Therefore, the new control conditions 102b for the heating device 10 can be easily derived.

According to the information processing device 50 of the first embodiment, the new control conditions 102b of the heating device 10 can be predicted based on the prediction input parameters 102a consisting of at least one of the material information of the preform 200, the shape information of the preform 200, the initial temperature information of the preform 200, the target temperature distribution information of the heating device 10, and the thermal environment information of the heating device 10. The information processing device 50 also includes an information acquisition unit 52 that acquires the prediction input parameters 102a, and an information prediction unit 53 that predicts the new control conditions 102b. The information prediction unit 53 also predicts the new control conditions 102b by inputting the prediction input parameters 102a to a learning model 45 that has learned the correlation between the learning input parameters 101a and the learning control conditions 101b of the heating device 10 corresponding to the learning input parameters 101a by machine learning. In addition, the learning input parameters 101a are composed of at least one of the material information of the preform 200, the shape information of the preform 200, the initial temperature information of the preform 200, the target temperature distribution information of the heating device 10, and the thermal environment information of the heating device 10. This makes it possible to predict the new control conditions 102b using the learning model 45 that has learned the correlation between the learning input parameters 101a and the learning control conditions 101b. Therefore, the new control conditions 102b of the heating device 10 can be easily derived using the learning model 45. In addition, by controlling the heating device under the new control conditions 102b, it is possible to prevent poor heating of the preform 200 and produce a product with appropriate finished dimensions.

According to the machine learning method of the first embodiment, a learning model 45 for predicting new control conditions 102b of the heating device 10 for heating the preform 200 can be generated. The machine learning method includes a learning data storage process for storing one or more pieces of learning data 46, a machine learning process for training the learning model 45 based on the one or more pieces of learning data 46, and a trained model storage process for storing the trained learning model 45. Each piece of learning data 46 is composed of learning input parameters 101a and learning control conditions 101b corresponding to the learning input parameters 101a. The learning model 45 learns the correlation between the learning input parameters 101a and the learning control conditions 101b in the learning data 46. The learning input parameters 101a are composed of at least one of material information of the preform 200, shape information of the preform 200, initial temperature information of the preform 200, target temperature distribution information of the heating device 10, and thermal environment information of the heating device 10. This makes it possible to obtain a learning model 45 that has learned the correlation between the learning input parameters 101a and the learning control conditions 101b. Therefore, using the obtained learning model 45, new control conditions 102b for the heating device 10 can be easily derived.

According to the control condition prediction method of the first embodiment, the new control conditions 102b of the heating device 10 can be predicted based on the prediction input parameters 102a, which are composed of at least one of the material information of the preform 200, the shape information of the preform 200, the initial temperature information of the preform 200, the target temperature distribution information of the heating device 10, and the thermal environment information of the heating device 10. The control condition prediction method also includes a prediction input parameter acquisition step for acquiring the prediction input parameters 102a, and a prediction step for predicting the new control conditions 102b corresponding to the prediction input parameters 102a acquired by the prediction input parameter acquisition step. In addition, in the prediction step, the new control conditions 102b are predicted by inputting the prediction input parameters 102a into a learning model 45 that has learned the correlation between the learning input parameters 101a and the learning control conditions 101b of the heating device 10 corresponding to the learning input parameters 101a by machine learning. In addition, the learning input parameters 101a are composed of at least one of the material information of the preform 200, the shape information of the preform 200, the initial temperature information of the preform 200, the target temperature distribution information of the heating device 10, and the thermal environment information of the heating device 10. This makes it possible to predict the new control conditions 102b using the learning model 45 that has learned the correlation between the learning input parameters 101a and the learning control conditions 101b. Therefore, the new control conditions 102b of the heating device 10 can be easily derived using the learning model 45.

Embodiment 2.
9 is a schematic diagram showing a heating device and a transport device according to embodiment 2. In embodiment 2, the learning control conditions 101b and the new control conditions 102b are different from embodiment 1 in that the learning control conditions 101b and the new control conditions 102b include the time of the annealing treatment performed on the preform 200. That is, the learning control conditions 101b and the new control conditions 102b in embodiment 2 are the learning control conditions 101b and the new control conditions 102b in embodiment 1 with the time of the annealing treatment added. In addition, the heating device 10 is different from embodiment 1 in that it has two heating device bodies 11 and five transport device bodies 31. The other configurations are the same as those in embodiment 1, so the description will be omitted.

The heating device 10 has a first heating device body 11a arranged on the upstream side and a second heating device body 11b arranged on the downstream side of the first heating device body 11a. The specific configurations of the first heating device body 11a and the second heating device body 11b are similar to the specific configuration of the heating device body 11 in embodiment 1, so a description thereof will be omitted.

The conveying device 30 has an upstream conveying device body 31a, a first conveying device body 31b, an intermediate conveying device body 31c, a second conveying device body 31d, and a downstream conveying device body 31e. The specific configurations of the upstream conveying device body 31a, the first conveying device body 31b, the intermediate conveying device body 31c, the second conveying device body 31d, and the downstream conveying device body 31e are similar to the specific configuration of the conveying device body 31 in embodiment 1, so a description thereof will be omitted.

The upstream conveying device body 31a, the first conveying device body 31b, the intermediate conveying device body 31c, the second conveying device body 31d, and the downstream conveying device body 31e are arranged so as to be connected in sequence. The preform 200 conveyed by the upstream conveying device body 31a is conveyed in sequence from the upstream conveying device body 31a to the first conveying device body 31b, the intermediate conveying device body 31c, the second conveying device body 31d, and the downstream conveying device body 31e.

The first conveying device body 31b is disposed at a position where the preform 200 supported by the first conveying device body 31b passes through the heating passage 12 of the first heating device body 11a. The second conveying device body 31d is disposed at a position where the preform 200 supported by the second conveying device body 31d passes through the heating passage 12 of the second heating device body 11b.

In this embodiment, in order to perform an annealing process on the preform 200, the preform 200 is removed from the heating passage 12 and left as it is. Specifically, with the preform 200 placed on the intermediate conveying device body 31c, control is performed such as slowing down the conveying speed of the intermediate conveying device body 31c or temporarily stopping the conveying of the intermediate conveying device body 31c.

Therefore, when the time for annealing the preform 200, i.e., the time for removing the preform 200 from the heating space, is output as the new control condition 102b, the control device 60 performs control such as slowing down the conveying speed of the intermediate conveying device body 31c or temporarily stopping the conveying of the intermediate conveying device body 31c according to that time. The control device 60 causes the preform 200 on the intermediate conveying device body 31c to remain on the intermediate conveying device body 31c for a time equivalent to the time for annealing.

In other words, the preform 200 is placed outside the heating space of the heating device 10, and the annealing process is performed on the preform 200. This provides time for the annealing process to be performed on the preform 200.

According to the machine learning device 40 according to the second embodiment, the learning control conditions 101b are the time of the annealing treatment performed on the preform 200. This allows the appropriate annealing treatment time to be obtained using the learning model 45 that has learned the correlation between the learning input parameters 101a and the learning control conditions 101b. Therefore, new control conditions 102b for the heating device 10 can be easily derived.

The present disclosure can be provided not only in the form of the information processing device 50, the learning control condition prediction method, or the learning control condition prediction program according to the first and second embodiments, but also in the form of an inference device, an inference method, or an inference program used to infer the learning control condition 101b.

In this case, the inference device, inference method, or inference program may include a memory and a processor, and the processor may execute a series of processes. The series of processes includes an information acquisition process that acquires prediction input parameters 102a, and an inference process that infers new control conditions 102b based on the acquired learning input parameters 101a.

By providing the inference device, inference method, or inference program, it is possible to more easily apply the inference device, inference method, or inference program to various devices compared to implementing the heating system 1. It will be obvious to those skilled in the art that when the inference device, inference method, or inference program infers the new control condition 102b, the prediction method implemented by the information prediction unit 53 may be applied using the trained learning model 45 generated by the machine learning method in the information processing device 50 in embodiment 1.

The inference device according to the first and second embodiments includes a memory and a processor, and can infer new control conditions 102b of the heating device 10 that heats the preform 200. The processor also executes an information acquisition process to acquire prediction input parameters 102a consisting of at least one of material information of the preform 200, shape information of the preform 200, initial temperature information of the preform 200, target temperature distribution information of the heating device 10, and thermal environment information of the heating device 10, and an inference process to infer new control conditions 102b of the heating device 10 based on the prediction input parameters 102a acquired in the information acquisition process. This makes it possible to predict new control conditions 102b using a learning model 45 that has learned the correlation between the learning input parameters 101a and the learning control conditions 101b. Therefore, the new control conditions 102b of the heating device 10 can be easily derived using the learning model 45.

According to the inference method of the first and second embodiments, the inference method is executed by an inference device having a memory and a processor to infer new control conditions 102b of the heating device 10 that heats the preform 200. The processor also executes an information acquisition step of acquiring prediction input parameters 102a consisting of at least one of material information of the preform 200, shape information of the preform 200, initial temperature information of the preform 200, target temperature distribution information of the heating device 10, and thermal environment information of the heating device 10, and a prediction step of predicting new control conditions 102b corresponding to the prediction input parameters 102a acquired by the information acquisition step. In addition, in the prediction step, the new control conditions are predicted by inputting the prediction input parameters 102a into a learning model 45 that has learned the correlation between the learning input parameters 101a and the learning control conditions 101b of the heating device 10 corresponding to the learning input parameters 101a by machine learning. In addition, the learning input parameters 101a are composed of at least one of the material information of the preform 200, the shape information of the preform 200, the initial temperature information of the preform 200, the target temperature distribution information of the heating device 10, and the thermal environment information of the heating device 10. This makes it possible to predict the new control conditions 102b using the learning model 45 that has learned the correlation between the learning input parameters 101a and the learning control conditions 101b. Therefore, the new control conditions 102b of the heating device 10 can be easily derived using the learning model 45.

The heating system 1 in the first and second embodiments heats the preform 200. However, this is not limited to this. The heating system 1 can be a system adapted to various temperature control targets. In this case, the input parameters 100a and the control conditions 100b can be suitably selected based on the heating device and the temperature control target to be applied.

Furthermore, in the heating system 1 of the first and second embodiments, the learning control conditions 101b and the new control conditions 102b are the flow rate of air flowing through the air blower duct 19, the conveying speed of the conveying device 30, the rotational speed of the rotational motion of the preform 200, the time of the annealing treatment performed on the preform 200, and the output command values of the multiple heaters 14. However, they are not limited to this. In the heating system 1, appropriate control conditions can be set.

Furthermore, in the first and second embodiments, the machine learning device 40, the information processing device 50, and the control device 60 are configured as separate devices. However, this is not limited to this. For example, the machine learning device 40, the information processing device 50, and the control device 60 may be configured as a single device. Furthermore, any two of the three devices may be configured as a single device.

Furthermore, in the heating system 1 of the first and second embodiments, the learning model 45 generated by the machine learning device 40 is transmitted to the information processing device 50 via the network 70, and the new control condition 102b predicted by the information processing device 50 is transmitted to the control device 60 via the network 70. However, this is not limited to this. The machine learning device 40 and the information processing device 50 do not have to be connected to the heating system 1 via the network 70. For example, the machine learning device 40 and the information processing device 50 may be arranged independently. In this case, the learning model 45 generated by the machine learning device 40 may be transmitted to the information processing device 50 manually. Furthermore, the new control condition 102b predicted by the information processing device 50 may be transmitted to the control device 60 manually.

Furthermore, in the first and second embodiments, a neural network is adopted as the learning model 45 that realizes machine learning by the machine learning control unit 41. However, this is not limited to this. Other machine learning models may be adopted as the learning model 45. Examples of other machine learning models include tree types such as decision trees and regression trees, ensemble learning such as bagging and boosting, recurrent neural networks, convolutional neural networks, neural net types (including deep learning) such as LSTM, clustering types such as hierarchical clustering, non-hierarchical clustering, k-nearest neighbors, and k-means, multivariate analyses such as principal component analysis, factor analysis, and logistic regression, and support vector machines.

In addition, in the first and second embodiments, the air blowing device 17 has a blower 18 as an air blowing source. However, this is not limited to this. The air blowing source may be, for example, a pressure supply source that stores high-pressure air. In this case, the flow rate of air sent from the pressure supply source to the air blower duct 19 can be controlled by controlling a flow valve or the like provided in the pressure supply source.

Furthermore, in the second embodiment, the annealing process is performed on the preform 200 on the intermediate conveying device body 31c. However, this is not limited to this. For example, the annealing process may be performed on the preform 200 on the downstream conveying device body 31e. In this case, the annealing process time can be given to the preform 200 on the downstream conveying device body 31e by controlling the downstream conveying device body 31e.

Furthermore, in the second embodiment, there are five conveying device bodies 31 and two heating device bodies 11. However, this is not limited to this. The number of conveying device bodies 31 and heating device bodies 11 can be set appropriately. Furthermore, the annealing process may be performed on the preforms 200 on any of the conveying device bodies 31.

This disclosure is not limited to the above-described embodiment, and various modifications can be made without departing from the spirit of the present invention. All such modifications are within the scope of the technical concept of the present invention.

The various aspects of this disclosure are summarized below as appendices.

(Appendix 1)
A machine learning device that generates a learning model for predicting a control condition of a heating device that heats a temperature control target,
a learning data storage unit that stores one or more pieces of learning data;
a machine learning control unit that causes the learning model to learn based on the one or more pieces of learning data;
A machine learning model storage unit that stores the learned learning model;
Equipped with
Each of the learning data is composed of learning input parameters and learning control conditions, which are control conditions of the heating device corresponding to the learning input parameters,
the learning model learns a correlation between the learning input parameters in the learning data and the learning control conditions;
The learning input parameters are composed of at least one of material information of a temperature control object, shape information of the temperature control object, initial temperature information of the temperature control object, target temperature distribution information of the heating device, and thermal environment information of the heating device.
Machine learning device.
(Appendix 2)
The material information is at least one of light transmission/reflection/refractive index information of the temperature control object, color brightness information of the temperature control object, color information of the temperature control object, density information of the temperature control object, specific heat information of the temperature control object, and thermal conductivity information of the temperature control object.
2. The machine learning device of claim 1.
(Appendix 3)
The shape information is at least one of design drawing information of the temperature control object, shape image information of the temperature control object, dimensional information of the temperature control object, and shape tracing measurement/scanning information of the temperature control object.
3. The machine learning device according to claim 1 or 2.
(Appendix 4)
The thermal environment information is at least one of distance information from a heat source of the heating device to the temperature control target, information on the arrangement position, irradiation direction, and irradiation angle of the heat source of the heating device, dimensional information of the heating device, information on thermal diffusion outside the heating device, and ambient temperature and humidity information of the heating device.
4. The machine learning device according to claim 1 .
(Appendix 5)
The heating device has a heating device body that defines a heating space and a blower device,
The air blowing device has an air blowing source and an air blowing tube for blowing air from the air blowing source into the heating space,
The learning control condition is a flow rate of the air flowing through the air duct.
5. The machine learning device according to claim 1 .
(Appendix 6)
The heating device includes a heating device body that defines a heating space, and a conveying device that conveys the temperature-control target while rotating the temperature-control target within the heating space,
The learning control condition is at least one of a transport speed of the transport device and a rotation speed of the rotational motion of the temperature control target.
5. The machine learning device according to claim 1 .
(Appendix 7)
The learning control condition is a time of an annealing treatment performed on the temperature control target.
5. The machine learning device according to claim 1 .
(Appendix 8)
The learning control condition is an output command value of a heat source of the heating device.
5. The machine learning device according to claim 1 .
(Appendix 9)
An information processing device that predicts a control condition of a heating device based on prediction input parameters constituted by at least one of material information of a temperature control object, shape information of the temperature control object, initial temperature information of the temperature control object, target temperature distribution information of a heating device, and thermal environment information of the heating device,
an information acquisition unit that acquires the prediction input parameters;
An information prediction unit that predicts the control condition;
Equipped with
the information prediction unit predicts the control condition by inputting the prediction input parameters into a learning model that has learned, by machine learning, a correlation between a learning input parameter and a learning control condition, which is a control condition of the heating device corresponding to the learning input parameter;
The learning input parameters are composed of at least one of material information of a temperature control object, shape information of the temperature control object, initial temperature information of the temperature control object, target temperature distribution information of the heating device, and thermal environment information of the heating device.
Information processing device.
(Appendix 10)
An inference device that includes a memory and a processor and infers a control condition of a heating device that heats a temperature control target,
The processor,
an information acquisition process for acquiring prediction input parameters including at least one of material information of the temperature control object, shape information of the temperature control object, initial temperature information of the temperature control object, target temperature distribution information of the heating device, and thermal environment information of the heating device;
an inference process for inferring the control condition of the heating device based on the input parameters for prediction acquired in the information acquisition process;
Inference device.
(Appendix 11)
A machine learning method for generating a learning model for predicting a control condition of a heating device that heats a temperature control target, comprising:
a learning data storage step of storing one or more learning data;
a machine learning process for training the learning model based on the one or more pieces of learning data;
A trained model storage step of storing the trained model;
Equipped with
Each of the learning data is composed of learning input parameters and learning control conditions, which are control conditions of the heating device corresponding to the learning input parameters,
the learning model learns a correlation between the learning input parameters in the learning data and the learning control conditions;
The learning input parameters are composed of at least one of material information of a temperature control object, shape information of the temperature control object, initial temperature information of the temperature control object, target temperature distribution information of the heating device, and thermal environment information of the heating device.
Machine learning methods.
(Appendix 12)
A control condition prediction method for predicting a control condition of a heating device based on prediction input parameters including at least one of material information of a temperature control object, shape information of the temperature control object, initial temperature information of the temperature control object, target temperature distribution information of a heating device, and thermal environment information of the heating device,
a prediction input parameter acquisition step of acquiring the prediction input parameters;
a prediction step of predicting the control condition corresponding to the input parameter for prediction acquired in the input parameter for prediction acquisition step,
In the prediction step, the control condition is predicted by inputting the prediction input parameters into a learning model that has learned, by machine learning, a correlation between a learning input parameter and a learning control condition, which is a control condition of the heating device corresponding to the learning input parameter;
The learning input parameters are composed of at least one of material information of a temperature control object, shape information of the temperature control object, initial temperature information of the temperature control object, target temperature distribution information of the heating device, and thermal environment information of the heating device.
Control condition prediction methods.
(Appendix 13)
An inference method for inferring a control condition of a heating device that heats a temperature-controlled object, the method being executed by an inference device having a memory and a processor, the method comprising:
The processor,
an information acquisition step of acquiring prediction input parameters including at least one of material information of the temperature control object, shape information of the temperature control object, initial temperature information of the temperature control object, target temperature distribution information of the heating device, and thermal environment information of the heating device;
a prediction step of predicting the control condition corresponding to the input parameter for prediction acquired by the information acquisition step;
In the prediction step, the control condition is predicted by inputting the prediction input parameters into a learning model that has learned, by machine learning, a correlation between a learning input parameter and a learning control condition, which is a control condition of the heating device corresponding to the learning input parameter;
The learning input parameters are composed of at least one of material information of a temperature control object, shape information of the temperature control object, initial temperature information of the temperature control object, target temperature distribution information of the heating device, and thermal environment information of the heating device.
Inference methods.

1 heating system, 10 heating device, 11 heating device main body, 11a first heating device main body, 11b second heating device main body, 12 heating passage (heating space), 14 heater (heat source), 17 blower, 18 blower (air source), 19 air duct, 30 conveying device, 31 conveying device main body, 31a upstream conveying device main body, 31b first conveying device main body, 31c intermediate conveying device main body, 31d second conveying device main body , 31e downstream conveying device main body, 32 conveying control unit, 40 machine learning device, 41 machine learning control unit, 42 machine learning communication unit, 43 learning data storage unit, 44 machine learning model storage unit, 45 learning model, 45a input layer, 45b intermediate layer, 45c output layer, 46 learning data, 50 information processing device, 51 information processing control unit, 52 information acquisition unit, 53 information prediction unit, 54 output processing unit, 55 Information processing model storage unit, 56 information processing communication unit, 60 control device, 70 network, 100a input parameters, 100b control conditions, 101a learning input parameters, 101b learning control conditions, 101c learning result control conditions, 102a prediction input parameters, 102b new control conditions, 200 preform (temperature control target), 900 computer, 910 bus, 912 processor, 914 memory, 916 input device, 917 output device, 918 display device, 920 storage device, 922 communication interface unit, 924 external device interface unit, 926 input/output device interface unit, 928 media input/output unit, 930 program, 940 network, 950 external device, 960 input/output device, 970 media.

Claims

A machine learning device that generates a learning model for predicting a control condition of a heating device that heats a temperature control target,
a learning data storage unit that stores one or more pieces of learning data;
a machine learning control unit that causes the learning model to learn based on the one or more pieces of learning data;
A machine learning model storage unit that stores the learned learning model;
Equipped with
Each of the learning data is composed of learning input parameters and learning control conditions, which are control conditions of the heating device corresponding to the learning input parameters,
the learning model learns a correlation between the learning input parameters in the learning data and the learning control conditions;
The learning input parameters are composed of at least one of material information of a temperature control object, shape information of the temperature control object, initial temperature information of the temperature control object, target temperature distribution information of the heating device, and thermal environment information of the heating device.
Machine learning device.
The material information is at least one of light transmission/reflection/refractive index information of the temperature control object, color brightness information of the temperature control object, color information of the temperature control object, density information of the temperature control object, specific heat information of the temperature control object, and thermal conductivity information of the temperature control object.
The machine learning device according to claim 1 .
The shape information is at least one of design drawing information of the temperature control object, shape image information of the temperature control object, dimensional information of the temperature control object, and shape tracing measurement/scanning information of the object.
The machine learning device according to claim 1 .
The thermal environment information is at least one of distance information from a heat source of the heating device to the temperature control target, information on the arrangement position, irradiation direction, and irradiation angle of the heat source of the heating device, dimensional information of the heating device, information on thermal diffusion outside the heating device, and ambient temperature and humidity information of the heating device.
The machine learning device according to claim 1 .
The heating device has a heating device body that defines a heating space and a blower device,
The air blowing device has an air blowing source and an air blowing tube for blowing air from the air blowing source into the heating space,
The learning control condition is a flow rate of the air flowing through the air duct.
The machine learning device according to claim 1 .
The heating device includes a heating device body that defines a heating space, and a conveying device that conveys the temperature-control target while rotating the temperature-control target within the heating space,
The learning control condition is at least one of a transport speed of the transport device and a rotation speed of the rotational motion of the temperature control target.
The machine learning device according to claim 1 .
The learning control condition is a time of an annealing treatment performed on the temperature control target.
The machine learning device according to claim 1 .
The learning control condition is an output command value of a heat source of the heating device.
The machine learning device according to claim 1 .
An information processing device that predicts a control condition of a heating device based on prediction input parameters constituted by at least one of material information of a temperature control object, shape information of the temperature control object, initial temperature information of the temperature control object, target temperature distribution information of a heating device, and thermal environment information of the heating device,
an information acquisition unit that acquires the prediction input parameters;
An information prediction unit that predicts the control condition;
Equipped with
the information prediction unit predicts the control condition by inputting the prediction input parameters into a learning model that has learned, by machine learning, a correlation between a learning input parameter and a learning control condition, which is a control condition of the heating device corresponding to the learning input parameter;
The learning input parameters are composed of at least one of material information of a temperature control object, shape information of the temperature control object, initial temperature information of the temperature control object, target temperature distribution information of the heating device, and thermal environment information of the heating device.
Information processing device.
An inference device that includes a memory and a processor and infers a control condition of a heating device that heats a temperature control target,
The processor,
an information acquisition process for acquiring prediction input parameters including at least one of material information of the temperature control object, shape information of the temperature control object, initial temperature information of the temperature control object, target temperature distribution information of the heating device, and thermal environment information of the heating device;
an inference process for inferring the control condition of the heating device based on the input parameters for prediction acquired in the information acquisition process;
Inference device.
A machine learning method for generating a learning model for predicting a control condition of a heating device that heats a temperature control target, comprising:
a learning data storage step of storing one or more learning data;
a machine learning process for training the learning model based on the one or more pieces of learning data;
A trained model storage step of storing the trained model;
Equipped with
Each of the learning data is composed of learning input parameters and learning control conditions, which are control conditions of the heating device corresponding to the learning input parameters,
the learning model learns a correlation between the learning input parameters in the learning data and the learning control conditions;
The learning input parameters are composed of at least one of material information of a temperature control object, shape information of the temperature control object, initial temperature information of the temperature control object, target temperature distribution information of the heating device, and thermal environment information of the heating device.
Machine learning methods.
A control condition prediction method for predicting a control condition of a heating device based on prediction input parameters including at least one of material information of a temperature control object, shape information of the temperature control object, initial temperature information of the temperature control object, target temperature distribution information of a heating device, and thermal environment information of the heating device, comprising:
a prediction input parameter acquisition step of acquiring the prediction input parameters;
a prediction step of predicting the control condition corresponding to the input parameter for prediction acquired in the input parameter for prediction acquisition step,
In the prediction step, the control condition is predicted by inputting the prediction input parameters into a learning model that has learned, by machine learning, a correlation between a learning input parameter and a learning control condition, which is a control condition of the heating device corresponding to the learning input parameter;
The learning input parameters are composed of at least one of material information of a temperature control object, shape information of the temperature control object, initial temperature information of the temperature control object, target temperature distribution information of the heating device, and thermal environment information of the heating device.
Control condition prediction methods.
An inference method for inferring a control condition of a heating device that heats a temperature-controlled object, the method being executed by an inference device having a memory and a processor, the method comprising:
The processor,
an information acquisition step of acquiring prediction input parameters including at least one of material information of the temperature control object, shape information of the temperature control object, initial temperature information of the temperature control object, target temperature distribution information of the heating device, and thermal environment information of the heating device;
a prediction step of predicting the control condition corresponding to the input parameter for prediction acquired by the information acquisition step;
In the prediction step, the control condition is predicted by inputting the prediction input parameters into a learning model that has learned, by machine learning, a correlation between a learning input parameter and a learning control condition, which is a control condition of the heating device corresponding to the learning input parameter;
The learning input parameters are composed of at least one of material information of a temperature control object, shape information of the temperature control object, initial temperature information of the temperature control object, target temperature distribution information of the heating device, and thermal environment information of the heating device.
Inference methods.