An artificial intelligence-based injection molding system in which molding conditions can be changed to manufacture fair-quality products when defective products are manufactured because of disturbances during the injection molding including an injection molding machine which performs injection molding by injecting a molding material into a mold; an injection state data acquisition unit for acquiring, during injection molding, current injection state data that includes the viscosity profile of the molding material injected into the mold and/or the injection pressure value thereof; a determination unit, which inputs the current injection state data into a molding quality maintenance model trained with predetermined target injection state data, so as to determine whether to maintain molding quality; and a molding condition setting unit for changing a preset molding condition so that the current injection state data follows the target injection state data, when the determination unit determines not to maintain the molding quality.

Provided is a machine learning method of a learning model that outputs a variable parameter that is configured to reduce the degree of defect of a molded article obtained by actual molding and relates to molding conditions of a molding machine in a case where observation data obtained by observing a physical quantity relating to actual molding using the molding machine is input. The machine learning method includes: a step of simulating a molding process by setting a variable parameter and a fixed parameter to a fluid analysis device; a step of acquiring a defect-related parameter that is obtained by simulation and relates to the degree of defect of the molded article; a step of calculating the degree of defect of the molded article on the basis of the acquired defect-related parameter; and a step of causing the learning model to perform machine learning by using the variable parameter set to the fluid analysis device and reward corresponding to the calculated degree of defect.

A method for simulating a fiber orientation in an injection-molded part made of a fiber-reinforced plastic. An orientation of the fibers in the injection-molded part to be manufactured that is present after the injection molding is determined via a macroscopic simulation of the injection molding. The macroscopic simulation of the injection molding takes place using macroscopic physical parameters of the fiber-reinforced plastic. In the macroscopic simulation, a temporal development of the fiber orientation tensor is determined via a combination of two macroscopic models. A first temporal development of the fiber orientation tensor is determined via a first macroscopic model based on shear flows. A second temporal development of the fiber orientation tensor is determined via a second macroscopic model based on elongation flows. The method is applied in a method for designing an injection-molded part made of a fiber-reinforced plastic.

An artificial intelligence-based injection molding system comprising a standard data extraction unit for extracting target standard data of a product produced by a mold from mold information about the mold to which a molding material is supplied; a molding condition output unit inputing the extracted target standard data into a pre-learned molding condition generation model to output a molding condition; an injection molding device, supplying the molding material to the mold according to the molding condition to produce the product; and a determination unit, comparing production standard data of the produced product and the target standard data to determine whether the molding condition is appropriate, wherein, if the determination unit determines that the molding condition is inappropriate, the molding condition output unit generates the production standard data and the molding condition as one set of feedback data, and trains the molding condition generation model with the set of feedback data.

An injection molding system includes a measuring unit to measure a movement of a line of sight of a worker observing a molded article, a line-of-sight data storage unit to store line-of-sight information representing movements of the worker's line of sight and a measurement time, an identifying unit to identify a focus area of the molded article, a focus area storage unit to store an image of the identified focus area, a molding defect type input unit to input or select a molding defect type, and a machine learning device to machine learn the molding defect type from the image of the focus area. The machine learning device inputs a type of a molding defect that has occurred in the molded article and carries out machine learning to learn and automatically recognize a feature quantity of the molding defect from the image of the focus area.

An injection molding system includes a measuring unit to measure a movement of a line of sight of a worker observing a molded article, a line-of-sight data storage unit to store line-of-sight information representing movements of the worker's line of sight and a measurement time, an identifying unit to identify a focus area of the molded article, a focus area storage unit to store an image of the identified focus area, a molding defect type input unit to input or select a molding defect type, and a machine learning device to machine learn the molding defect type from the image of the focus area. The machine learning device inputs a type of a molding defect that has occurred in the molded article and carries out machine learning to learn and automatically recognize a feature quantity of the molding defect from the image of the focus area.

An injection molding condition generation system improves the quality of injection molding. The injection molding condition generation system acquires a material property value of a resin material and an injection molding condition using the first resin material based on a target value of a quality parameter related to a quality of a molded article, the first material property value, and a predetermined relational expression. The predetermined relational expression indicates a relation among the material property value of the resin material, a plurality of injection molding conditions input to an injection molding machine, and a quality parameter related to the quality of the molded article molded by the injection molding machine based on the material property value and the injection molding conditions, and is generated based on data accumulated in the memory in association with the material property value of the resin material, the injection molding conditions, and the quality parameter.

A reinforcement learning method of a learning machine including a first agent adjusting a manufacture condition of a manufacturing device based on observation data obtained by observing a state of the manufacturing device and a second agent having a functional model or a functional approximator representing a relationship between the observation data and the manufacture condition in a different way from the first agent, comprises: adjusting the manufacture condition searched by the first agent that is performing reinforcement learning, using the observation data and the functional model or the functional approximator of the second agent; calculating reward data in accordance with a state of a product manufactured by the manufacturing device under the manufacture condition adjusted; and performing reinforcement learning on the first agent and the second agent based on the observation data and the reward data calculated.

A reinforcement learning method of a learning machine including a first agent adjusting a manufacture condition of a manufacturing device based on observation data obtained by observing a state of the manufacturing device and a second agent having a functional model or a functional approximator representing a relationship between the observation data and the manufacture condition in a different way from the first agent, comprises: adjusting the manufacture condition searched by the first agent that is performing reinforcement learning, using the observation data and the functional model or the functional approximator of the second agent; calculating reward data in accordance with a state of a product manufactured by the manufacturing device under the manufacture condition adjusted; and performing reinforcement learning on the first agent and the second agent based on the observation data and the reward data calculated.

An operating condition correction method of an industrial machine, comprises: acquiring measurement data obtained by measuring a state of the industrial machine and inspection data obtained by inspecting a state of a product manufactured by the industrial machine; calculating a correction quantity of the operating condition based on the acquired measurement data and inspection data using a plurality of learners trained with an association between the measurement data as well as the inspection data and a correction quantity of the operating condition; determining appropriateness of a correction direction of each of a plurality of the correction quantities calculated using the plurality of learners; and correcting the operating condition based on one or more of the correction quantities determined as having the correction direction being appropriate.