JP2017132260A - System capable of calculating optimum operating conditions in injection molding - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an injection molding system capable of adjusting operation conditions including molding conditions in a short time and capable of enabling molding under conditions of lower power consumption.SOLUTION: In a system in which a plurality of injection molding systems according to the present invention are interconnected via communication means, each of the injection molding system comprises: a state observing section for observing a physical quantity relating to injection molding; a physical quantity data storing section for storing the physical quantity data; a reward calculating section for calculating a reward on the basis of the physical quantity data and the reward condition; an operation condition adjustment learning section for machine-learning operation condition adjustments; a learning result storing section for storing a result of machine-learning by the operation condition adjustment learning section; and an operation condition adjustment amount outputting section for determining and outputting a target for operation condition adjustment and an amount of adjustment on the basis of the machine-learning performed by the operation condition adjustment learning section. The individual injection molding systems send/receive and share the data stored by the physical quantity data storing section and/or the learning result storing section.SELECTED DRAWING: Figure 2

Description

  The present invention relates to an injection molding system, and more particularly to an injection molding system capable of calculating an optimum operation condition without adjustment by an operator.

  When a mold for a new molded product is produced, it is necessary to calculate an optimum value for the operating conditions including the molding conditions before starting mass production of a molded product based on the mold. In the operation condition calculation work to calculate the optimum operation condition of the injection molding machine, the operator sets process conditions that are approximate standards based on experience and performs the molding work, while the operator measures the process monitoring data and the molded product weight It is necessary to adjust various operating conditions while checking the molding state by visually checking the molded product, etc., so that the optimum operating conditions are obtained. For this reason, it is necessary for the operator to calculate the optimum operating conditions over time while adjusting various operating conditions and comparing molded products molded under each operating condition.

  On the other hand, as a conventional technology that supports the operator to set molding conditions, a technique for storing molding conditions and molding data in a non-volatile memory for comparison display, and reading past molding conditions in response to operator requests And the like are disclosed (for example, Patent Documents 1 and 2).

Japanese Patent Laid-Open No. 06-039889 JP 11-333899 A

  Depending on the skill of the operator performing the operation, it may take time to calculate the optimum operation condition, or the operator may have a difference in the level (quality) of the optimum operation condition. There is a problem that it is difficult to calculate the operation condition based on the same standard every time.

  In the operation condition determination work, it is also an important viewpoint to derive an operation condition capable of suppressing power consumption during molding from the viewpoint of manufacturing cost of a molded product during mass production. However, there is a problem that it is difficult even for a skilled operator to derive operation conditions for molding while suppressing power consumption while maintaining high quality of a molded product.

  And such a subject cannot be solved only by memorize | storing and using the log | history of shaping | molding conditions, as disclosed by patent document 1,2.

  Accordingly, an object of the present invention is to provide an injection molding system that can adjust operation conditions including molding conditions in a short time and that can perform molding under conditions with less power consumption. It is.

  The invention according to claim 1 of the present application is a system that includes at least one injection molding machine, and is configured by connecting a plurality of injection molding systems having artificial intelligence that perform machine learning to each other via communication means. Each of the injection molding systems stores a physical quantity data observed by the state observing unit and a state observing unit for observing a physical quantity related to the injection molding being performed when the injection molding by the injection molding machine is executed. Based on the physical quantity data storage unit, the reward condition setting unit for setting the reward condition in the machine learning, the physical quantity data observed by the state observation unit and the reward condition set in the reward condition setting unit A remuneration calculation unit to calculate, the remuneration calculated by the remuneration calculation unit, operation conditions set in the injection molding system, and the physical quantity data An operation condition adjustment learning unit that performs machine learning of operation condition adjustment based on a learning result storage unit that stores a learning result obtained by machine learning by the operation condition adjustment learning unit, and a learning result performed by the operation condition adjustment learning unit. An operation condition adjustment amount output unit for determining and outputting an operation condition adjustment target and an adjustment amount based on the physical quantity data stored in the physical quantity data storage unit included in each injection molding system and a learning result storage unit It is a system that sends and receives stored learning results and shares them.

  The invention according to claim 2 of the present application is the system according to claim 1, wherein the learning result stored in the learning result storage unit is used for learning of the operation condition adjustment learning unit.

  In the invention according to claim 3 of the present application, each of the injection molding systems further includes a measurement unit, and the physical quantity data observed by the state observation unit is the weight, size, and molding of the molded product measured by the measurement unit. Including at least one of the appearance, length, angle, area, volume, optical inspection result of the optical molded product, and measurement result of the molded product strength calculated from the image data of the product, and the physical quantity data storage unit The system according to claim 1 or 2, wherein the physical quantity data is stored as physical quantity data of one molded product together with the physical quantity data.

  The invention according to claim 4 of the present application is capable of inputting reward conditions from the display device provided in the injection molding machine to the reward condition setting unit. It is a system as described in one.

  The invention according to claim 5 of the present application is that, when the reward calculation unit contributes to at least one of stabilization of physical quantity data, reduction of cycle time, and energy saving, a positive reward is given according to the degree. The system according to claim 1, wherein the system is characterized in that

  In the invention according to claim 6 of the present application, the reward calculation unit gives a negative reward depending on the degree when at least one of the physical quantity data occurs, the cycle time is extended, or the energy consumption is increased. The system according to any one of claims 1 to 5.

  The invention according to claim 7 of the present application is characterized in that an allowable value is set in advance in the physical quantity data, and the reward calculation unit gives a positive reward when the physical quantity data falls within the allowable value. It is a system as described in any one of Claims 1-6.

  In the invention according to claim 8 of the present application, an allowable value is set in advance in the physical quantity data, and when the physical quantity data deviates from the allowable value, the reward calculation unit gives a negative reward based on the deviation amount. The system according to claim 1, wherein:

  In the invention according to claim 9 of the present application, a target value is set in advance in the physical quantity data, and when the physical quantity data approaches the target value, the reward calculation unit is based on a deviation amount between the target value and the physical quantity. The system according to claim 1, wherein a positive reward is given.

  In the invention according to claim 10 of the present application, a target value is set in advance in the physical quantity data, and when the physical quantity data is separated from the target value, the reward calculation unit is based on a deviation amount between the target value and the physical quantity. The system according to claim 1, wherein a negative reward is given.

  The invention according to claim 11 of the present application is characterized in that the reward calculation unit gives a negative reward depending on the degree of occurrence of a state indicating a molding defect, according to any one of claims 1 to 10. The described system.

  In the invention according to claim 12 of the present application, the molding defects are burrs, sink marks, warps, bubbles, shorts, flow marks, welds, silver streaks, color unevenness, discoloration, carbonization, mixing of impurities, and the optical axis of the lens molded product. The system according to claim 11, wherein the system is at least one of a deviation outside an allowable value and a molded product thickness defect.

  According to the thirteenth aspect of the present invention, the operation conditions that the operation condition adjustment learning unit performs machine learning include mold clamping conditions, ejection conditions, injection pressure holding conditions, measurement conditions, temperature conditions, nozzle touch conditions, resin supply conditions, molds The system according to claim 1, wherein the system is at least one of a thickness condition, a molded product extraction condition, and a hot runner condition.

  A fourteenth aspect of the present invention is the system according to the thirteenth aspect, further comprising a robot as a molded product extraction means for setting the molded product extraction condition.

  The invention according to claim 15 of the present application is characterized in that at least one of the operation conditions is varied within a predetermined range and is learned by the operation condition adjustment learning unit. It is a system as described in one.

  In the present invention, by incorporating machine learning into the calculation of the optimum operating conditions, various operating conditions can be adjusted in a short time and more stable molding can be performed. Further, it becomes possible to perform molding under conditions with less power consumption. Furthermore, by sharing the molding data and learning data that are being adjusted in a plurality of injection molding systems and using them for machine learning, it is possible to realize machine learning that can obtain better results for each injection molding system. .

It is a figure explaining the basic concept of a reinforcement learning algorithm. It is a schematic block diagram of the injection molding system in embodiment of this invention. This is an example in which injection holding pressure data for one shot is displayed as a pressure waveform.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.
In the present invention, a machine learning device that becomes an artificial intelligence is introduced into the injection molding system, and machine learning related to operation conditions related to injection molding is performed to adjust various conditions in injection molding. In this way, an injection molding system capable of calculating the optimum operation condition in a short period of time and stabilizing the molding or contributing to energy saving is proposed.

<1. Machine learning>
In general, machine learning is classified into various algorithms depending on the purpose and conditions such as supervised learning and unsupervised learning. The purpose of the present invention is to learn the operation condition setting work for the mold. In the injection molding system, there are parameters that cannot be directly measured in the injection environment, and what kind of condition is given to the state of the molded product as a result of injection. Considering that it is difficult to explicitly indicate whether it is correct to perform an action (adjustment of operation conditions), the machine learner automatically learns the action to reach the goal simply by giving a reward Reinforcement learning algorithm is adopted.

FIG. 1 is a diagram for explaining the basic concept of the reinforcement learning algorithm. In reinforcement learning, agent learning and actions are performed by interaction between an agent (machine learning device) as a subject to learn and an environment (control target system) as a control target. More specifically, (1) The agent observes the state s _t environment in some point, (2) Observation and executing an action a _t Select they take actions a _t on the basis of past learning and, (3) action a _t the state s _t environment by runs is changed to the next state s _{t + 1,} the agent based on the change of state as a result of (4) action a _t receive a reward r _{t + 1,} (5) the agent state s _t, act a _t, based on the reward r _{t + 1} and the results of past learning advancing learning, exchanges such is performed between the agent and the environment .

The learning in the above (5), d - stringent as serving as a reference information for determining the amount of compensation that can be acquired in the future, acquired observed state s _t, act a _t, the mapping reward r _{t + 1} To do. For example, the number of possible states at each time m, the number of actions that can be taken is when the n, the m × n for storing a reward r _{t + 1} for the set of states s _t and action a _t by repeating the action A two-dimensional array is obtained.
Based on the mapping obtained above, the value function (evaluation function), which is a function indicating how good the current state or action is, is updated while the action is repeated. To learn the best behavior for the situation.

State value function is a value function that indicates whether it is how much good state a state s _t is. The state value function is expressed as a function with the state as an argument, and is based on the reward obtained for the action in a certain state in learning while repeating the action, the value of the future state that is shifted by the action, etc. Updated. The state value function update equation is defined according to the reinforcement learning algorithm. For example, in TD learning, which is one of the reinforcement learning algorithms, the state value function is defined by the following equation (1). In Equation 1, α is called a learning coefficient, and γ is called a discount rate, and is defined in the range of 0 <α ≦ 1 and 0 <γ ≦ 1.

In addition, action-value function is a value function that indicates whether it is how much good behavior action a _t is in a certain state s _t. The action value function is expressed as a function with the state and action as arguments, and in learning while repeating the action, the reward obtained for the action in a certain state and the action in the future state that is shifted by the action Updated based on value etc. The update formula of the action value function is defined according to the reinforcement learning algorithm. For example, in Q learning which is one of the typical reinforcement learning algorithms, the action value function is defined by the following equation (2). . In Equation 2, α is called a learning coefficient, and γ is called a discount rate, and is defined in the range of 0 <α ≦ 1 and 0 <γ ≦ 1.

As a method for storing the value function (evaluation function), in addition to a method using an approximate function and a method using an array, for example, when the state s takes many states, the state _st , worth a _t as input (evaluation) may be used supervised learning device such as SVM or a neural network of multi-valued output for outputting.

Then, in the selection of the behavior in the above (2), reward future in the current state s _t with the value created by the previous learning function (cost _{function) (r t + 1 + r} t + 2 + If the ...) is using action a _t (state value function becomes maximum, most actions to move to a higher-value state, most valuable high action in the condition in case of using the action value function ) Is selected. During the learning of the agent, a random action may be selected with a certain probability in the action selection in (2) for the purpose of learning progress (ε-greedy method).

  Thus, learning is advanced by repeating (1) to (5). After learning is completed in a certain environment, learning can be advanced so as to adapt to the environment by performing additional learning even in a new environment. Therefore, by applying to the operation condition setting work as in the present invention, even when a new mold for a molded product is produced, a new mold for the new molded product is added to the learning of the past operation condition calculation work. It is possible to adjust various conditions in a short time by learning the additional operation condition setting work in a simple environment.

  In reinforcement learning, a system in which a plurality of agents are connected via a network or the like, and information such as state s, action a, and reward r is shared between the agents and used for each learning. Efficient learning can be performed by performing distributed reinforcement learning in which an agent learns considering the environment of other agents. Also in the present invention, by performing distributed machine learning in a state where a plurality of agents (machine learning devices) that control a plurality of environments (injection molding machines to be controlled) are connected via a network or the like, This makes it possible to efficiently learn the operation condition setting work.

Various methods such as Q learning, SARSA method, TD learning, and AC method are well known as reinforcement learning algorithms, but any reinforcement learning algorithm may be adopted as a method applied to the present invention. . In addition, since each reinforcement learning algorithm is well-known, detailed description of each algorithm in this specification is abbreviate | omitted.
Below, the injection molding system of this invention which introduced the machine learning device is demonstrated based on specific embodiment.

<2. Embodiment>
FIG. 2 is a diagram showing a schematic configuration of an injection molding system according to an embodiment of the present invention. The injection molding system 1 according to the present embodiment includes a control object such as an injection molding machine 2, a mold 3, and other peripheral devices, and a machine learning device 20 serving as an artificial intelligence that performs machine learning. When the configuration shown in FIG. 2 is compared with the elements in the reinforcement learning shown in FIG. 1, the machine learning device 20 corresponds to the agent and includes control objects such as the injection molding machine 2, the mold 3, and other peripheral devices. The whole corresponds to the environment.

The injection molding system 1 is composed of various devices, and each may be equipped with a control device, a sensor, and the like.
As the control device provided in the injection molding machine 2, the mold clamping force control device 7, mold opening / closing, mold thickness adjustment, eject, screw rotation, screw forward / backward movement, injection unit forward / backward movement, rotation of an appropriate amount feeder feeder, etc. are driven. There are a nozzle / cylinder, a hopper lower temperature control device, a pressure control device, a nozzle touch force control device and the like in addition to the control device 9 for the above.

  In relation to the mold 3, a mold temperature control device, a hot runner temperature / nozzle opening / closing control device, a movable member control device for injection compression, a core set / core pull control device, a control device for slide table / rotary table such as multicolor molding, There are multicolor molding material flow path switching devices, mold vibration devices, and the like.

Peripheral devices include a molded product take-out device (robot) 5 controlled by the control device 8, an insert product insertion device, a insert insertion device, an in-mold molding foil feeding device, a hoop molding hoop feeding device, and a gas assist. Gas injection device for molding, gas injection device for foam molding using supercritical fluid, long fiber injection device, two-material mixing device for LIM molding, deburring device for molded product, runner cutting device, molded product weight meter, molding There are a product strength testing machine, a molded product optical inspection device, a molded product photographing device and an image processing device, a molded product transport robot, and the like.
Among these devices, there is also a control device that includes a sensor and performs feedback control and feedforward control by a closed loop. Some devices output only data.

  Further, the machine learning device 20 that performs machine learning includes a state observation unit 21, a physical quantity data storage unit 22, a reward condition setting unit 23, a reward calculation unit 24, an operation condition adjustment learning unit 25, a learning result storage unit 26, and an operation condition adjustment. A quantity output unit 27 is provided. The machine learning device 20 may be provided in the injection molding machine 2 or may be provided in a personal computer or the like outside the injection molding machine 2.

  The state observation unit 21 is a functional unit that observes physical quantity data related to injection molding output from each device of the injection molding system 1 and acquires the physical quantity data in the machine learning device 20. Physical quantity data includes temperature, position, speed, acceleration, current, voltage, pressure, time, image data, image analysis data, torque, force, strain, power consumption, mold opening, backflow, tie bar deformation, heater Heating rate, weight of molded product, strength of molded product, dimensions of molded product, appearance calculated from image data of molded product, length, angle, area, volume of each part of molded product, optical inspection result of optical molded product, There are a molded product strength measurement result, an optical axis shift amount of the transparent molded product, and a calculated value calculated by calculating each physical quantity. An environment state s used for machine learning is defined by a set of these physical quantity data.

  The physical quantity data storage unit 22 is a functional unit that inputs and stores physical quantity data and outputs the stored physical quantity data to the reward calculation unit 24 and the operation condition adjustment learning unit 25. The physical quantity data storage unit 22 stores the physical quantity data observed by the state observation unit 21 at the time of injection molding as physical quantity data of one molded product molded by the injection molding. The input physical quantity data may be data acquired in the latest molding operation or data acquired in the past molding operation. Further, physical quantity data stored in another injection molding system 1 or the centralized management system 30 can be input and stored or output.

The reward condition setting unit 23 is a functional means for setting conditions for giving reward in machine learning. There are positive and negative rewards, which can be set as appropriate. Furthermore, the input to the reward condition setting unit 23 may be from a personal computer or a tablet terminal used in the centralized management system, but by enabling the input via the display 6 provided in the injection molding machine 2, It becomes possible to set more simply.
The reward calculation unit 24 analyzes the physical quantity data input from the state observation unit 21 or the physical quantity data storage unit 22 based on the conditions set by the reward condition setting unit 23, and calculates the calculated rewards to the operation condition adjustment learning unit 25. Output to. The reward output by the reward calculation unit 24 corresponds to the reward r used for machine learning.

Below, the example of the reward conditions set with the reward condition setting part 23 in this embodiment is shown.
● [Reward 1: A case where a positive reward is given if it contributes to at least one of stabilization of physical quantity data, reduction of cycle time, and energy saving]
The determination of the stabilization of the physical quantity data may be made so as to give a positive reward depending on the degree when the physical quantity data contributes to the reduction of variation as a result of statistical processing. A standard deviation is generally used as an index of variation.
In the determination of the cycle time shortening, when the cycle time is shortened, a positive reward is given according to the degree.
Energy conservation is based on the power consumption of the injection molding machine alone, the power consumption of the entire injection molding system, and the total power consumption of multiple injection molding systems. give.
Conversely, if the physical quantity data becomes unstable, the cycle time is extended, or the energy consumption is increased, a negative reward is given according to the degree.

● [Reward 2: A case where a permissible value is set in advance for physical quantity data and the reward calculation unit gives a positive reward when the physical quantity data falls within the permissible value]
When the maximum injection pressure exceeds 200 MPa, burrs are generated in the molded product, and when it is less than 190 MPa, it is known that a short circuit occurs in the molded product. If the physical quantity data falls within the allowable value, a positive reward is given. Further, when the deviation is out of the allowable value, a larger negative reward may be given as the deviation amount increases according to the deviation amount.
In addition, as shown in FIG. 3, the injection molding machine screen has a function to display injection holding pressure data for one shot in a pressure waveform with a horizontal axis (time or screw position) and a vertical axis (pressure). There is known a function of setting upper and lower limits in a plurality of sections of the pressure waveform and displaying an alarm message when the waveform deviates from these upper and lower limits, and determining whether the waveform is good or bad. It is also possible to give a reward to the pressure waveform by using the upper and lower limits set in the pressure waveform as an allowable value set by the reward condition setting unit 23.

● [Reward 3: A case where a target value is set in advance in the physical quantity data and the reward calculation unit gives a positive reward based on the amount of deviation between the target value and the physical quantity when the physical quantity data approaches the target value]
A target value may be set for the molded product weight based on the mold design and resin selection, and a larger positive reward may be given as the target value is approached.
Conversely, when the physical quantity data is separated from the target value, a negative reward may be given based on the amount of deviation between the target value and the physical quantity. In addition, when the rate of change in the amount of deviation increases, if a negative reward is given based on the rate of change, it becomes possible to give a larger negative reward when the amount of deviation increases at an accelerated rate. .

  Further, the target value may be set near the upper limit within the allowable value by combining the setting of the allowable value of the reward 2 and the setting of the target value of the reward 3 described above. Even if the resin used in injection molding is the same grade, it has a factor that the molecular weight distribution and viscosity at the time of melting vary depending on the lot, so even if molding is performed within the allowable value, it is short near the lower limit of the allowable value. There are cases where the probability of occurrence increases. In order to avoid such a risk, the target value is set below the upper limit of the allowable value and close to the upper limit. As a result, the molding becomes stable within the allowable value range and near the upper limit of the allowable value, and the probability of occurrence of a short circuit can be reduced. In this way, a plurality of reward conditions can be used in combination.

● [Reward 4: A case where a negative reward is given if a condition indicating molding failure occurs]
Images obtained by photographing molded products or image analysis data obtained by analyzing images, burrs, sink marks, sleds, bubbles, shorts, flow marks, welds, silver streaks, color irregularities, discoloration, detected by optical inspection devices, etc. If molding defects such as carbonization, contamination of impurities, deviation of the optical axis of the lens molded product from an allowable value, and defective thickness of the molded product occur, a negative reward is given.
Further, the magnitude of the negative reward may be changed according to the degree of these defects. For example, when discoloration occurs, the degree of discoloration is converted into a numerical value based on a color difference meter or image analysis of a captured image, and the amount of negative reward is changed according to the degree of discoloration.

Returning to FIG. 2, the operation condition adjustment learning unit 25 is based on the physical quantity data, the adjustment of the operation condition including the molding condition of the injection molding system performed by itself, and the reward calculated by the reward calculation unit 24. Perform machine learning (reinforcement learning). At this time, machine learning (reinforcement learning) may be performed using learning results stored in a learning result storage unit 26 described later.
Then, the operation condition adjustment amount output unit 27 described later is based on the learning result of the operation condition adjustment learning unit 25, and the mold clamping condition, the eject condition, the injection pressure holding condition, the measurement condition, the temperature condition, the nozzle touch condition, and the resin supply. The adjustment amount of the operation conditions of the injection molding system such as the conditions, the mold thickness conditions, the molded product take-out conditions, the hot runner conditions, and the setting conditions of the control device of each peripheral device is output. The adjustment of the operation condition here corresponds to the action a used for machine learning.

Here, in the machine learning performed by the operating condition adjusting learning unit 25, the state s _t by a combination of physical quantity data is defined at a certain time t, the operating conditions of the injection molding system to be made to defined states s _t adjusting is possible to output the result of adjustment by the operation condition adjustment amount output unit 27 to be described later act a _t next, then the reward based on the data obtained as a result of injection molding was performed based on the adjustment result The value calculated by the calculation unit 24 is the reward r _{t + 1} . The value function used for learning is determined according to the learning algorithm to be applied. For example, when Q learning is used, learning may be advanced by updating the action value function Q (s _t , a _t ) according to the above-described equation (2).

In learning, each operation condition may be set to an initial value in advance, and at least one operation condition may be varied within a predetermined range. For example, each physical quantity is captured when molding is performed by automatically increasing the screw rotational speed during measurement from the initial value of 100 rpm by 10 rpm, or by automatically increasing the back pressure from the initial value of 5 MPa to 1 MPa. Thus, it is possible to learn a combination of the screw rotation speed and the back pressure that consumes the least amount of power within a range in which a good product can be stably molded without causing molding defects.
Further, the above-described ε-greedy method may be adopted, and learning may be progressed by selecting a random action with a predetermined probability.

  Note that the molded product take-out device 5 in which the molded product take-out conditions are set may be an articulated robot. If the molded product (green body) is very fragile, such as metal powder molding, the molded product cannot be taken out using a general take-out machine. Good. However, since the speed of taking out the molded product greatly affects the cycle time, it is preferable to learn to take out the molded product at as high a speed as possible without breaking the molded product. The presence or absence of damage to the molded product can be output and determined as physical quantity data by measuring the weight of the molded product or analyzing the image.

The learning result storage unit 26 stores the result learned by the operation condition adjustment learning unit 25. When the operation condition adjustment learning unit 25 reuses the learning result, the stored learning result is output to the operation condition adjustment learning unit 25. As described above, the learning function is stored with an approximate function, an array, or a supervised learning device such as an SVM or a neural network having a multi-value output, as described above. You can do it.
In addition, the learning result stored in the other injection molding system 1 or the centralized management system 30 is input to the learning result storage unit 26 and stored, or the learning result stored in the learning result storage unit 26 is stored in another learning result. It is also possible to output to the injection molding system 1 or the centralized management system 30.

  The operation condition adjustment amount output unit 27 determines and outputs an operation condition adjustment target and adjustment amount based on the learning result of the operation condition learned by the operation condition adjustment learning unit 25. By changing the operating condition of the injection molding system 1 based on the adjustment amount output from the operation condition adjustment amount output unit 27 and repeating learning using the physical quantity data input to the state observation unit 21 again, Excellent learning results can be obtained.

  In addition, machine learning can be performed using an evaluation function expressing physical quantity data and an operation with arguments so as to maximize the reward. The machine learning may be performed while acquiring the latest physical quantity data of molding, or may be executed with the acquired physical quantity data stored in the physical quantity data storage unit.

  When performing machine learning, it is possible to cause the operation condition adjustment learning unit 25 to learn by varying at least one of the operation conditions within a predetermined range. By learning with intentional variation, it is possible to efficiently learn the influence on the variation.

Further, when each of the plurality of injection molding systems 1 further includes a communication means with the outside, the physical quantity data stored in each physical quantity data storage unit 22 and the learning result stored in the learning result storage unit 26 are transmitted and received and shared. And machine learning can be performed more efficiently. For example, when learning by varying operating conditions within a predetermined range, different operating conditions are varied within a predetermined range in a plurality of injection molding systems 1 while molding between the respective injection molding systems 1. Thus, it is possible to efficiently learn by exchanging physical quantity data and learning data to advance learning in parallel.
In addition, when exchanging between a plurality of injection molding systems 1, communication may be performed directly via the host computer such as the centralized management system 30 or between the injection molding systems 1, and a cloud is used. However, since a large amount of data may be handled, a communication means with a communication speed as fast as possible is preferable.

  Although the embodiments of the present invention have been described above, the present invention is not limited to the above-described embodiments, and can be implemented in various modes by making appropriate changes.

DESCRIPTION OF SYMBOLS 1 Injection molding system 2 Injection molding machine 3 Mold 5 Molded product taking-out apparatus 6 Indicator 7 Clamping force control apparatus 8 Control apparatus 9 Control apparatus 20 Machine learning device 21 State observation part 22 Physical quantity data storage part 23 Compensation condition setting part 24 Reward calculation unit 25 Operation condition adjustment learning unit 26 Learning result storage unit 27 Operation condition adjustment amount output unit 30 Centralized management system

Claims (15)

A system comprising a plurality of injection molding systems each having at least one injection molding machine and having artificial intelligence for machine learning, connected to each other via communication means,
Each of the injection molding systems
When performing injection molding by the injection molding machine, a state observing unit that observes physical quantities related to the injection molding being performed;
A physical quantity data storage unit for storing physical quantity data observed by the state observation unit;
A reward condition setting unit for setting a reward condition in the machine learning;
A reward calculation unit that calculates a reward based on the physical quantity data observed by the state observation unit and the reward condition set in the reward condition setting unit;
An operation condition adjustment learning unit that performs machine learning of operation condition adjustment based on the reward calculated by the reward calculation unit, the operation condition set in the injection molding system, and the physical quantity data;
A learning result storage unit for storing a learning result obtained by machine learning by the operation condition adjustment learning unit;
An operation condition adjustment amount output unit that determines and outputs an operation condition adjustment target and an adjustment amount based on a learning result performed by the operation condition adjustment learning unit;
With
The physical quantity data stored in the physical quantity data storage unit provided in each injection molding system and the learning result stored in the learning result storage unit are transmitted and received and shared,
system.   The system according to claim 1, wherein the learning result stored in the learning result storage unit is used for learning of the operation condition adjustment learning unit. Each of the injection molding systems further comprises measuring means,
The physical quantity data observed by the state observing unit is the weight, dimension, appearance, length, angle, area, volume, and optical molded product calculated from the image data of the molded product measured by the measuring means. Including at least one of optical inspection results and molded product strength measurement results,
The physical quantity data storage unit stores physical quantity data of one molded product together with other physical quantity data.
The system according to claim 1 or 2, characterized by the above-mentioned. Reward conditions can be input from the display provided in the injection molding machine to the reward condition setting unit.
The system according to any one of claims 1 to 3, wherein: If the reward calculation unit contributes to at least one of stabilization of physical quantity data, cycle time reduction, and energy saving, a positive reward is given according to the degree,
The system according to any one of claims 1 to 4, characterized in that: The reward calculation unit gives a negative reward depending on the degree of occurrence of at least one of physical quantity data instability, cycle time extension, and energy consumption increase,
The system according to any one of claims 1 to 5, characterized in that: An allowable value is set in advance in the physical quantity data,
The reward calculation unit gives a positive reward when the physical quantity data is within the allowable value,
The system according to any one of claims 1 to 6, characterized in that: An allowable value is set in advance in the physical quantity data,
When the physical quantity data deviates from the allowable value, the reward calculation unit gives a negative reward based on the deviation amount,
A system according to any one of claims 1 to 7, characterized in that A target value is set in advance in the physical quantity data,
When the physical quantity data approaches the target value, the reward calculation unit gives a positive reward based on the amount of deviation between the target value and the physical quantity,
The system according to any one of claims 1 to 8, characterized in that: A target value is set in advance in the physical quantity data,
The reward calculation unit gives a negative reward based on the amount of deviation between the target value and the physical quantity when the physical quantity data is separated from the target value,
10. A system according to any one of claims 1 to 9, characterized in that The reward calculation unit gives a negative reward according to the degree of occurrence of a state indicating defective molding,
The system according to any one of claims 1 to 10, wherein: The molding defects are burrs, sink marks, warps, bubbles, shorts, flow marks, welds, silver streaks, uneven color, discoloration, carbonization, contamination of impurities, deviation of the optical axis of the lens molded product from the allowable value, molded product. At least one of the thickness defects,
The system according to claim 11. The operation conditions that the operation condition adjustment learning unit learns are mold clamping conditions, ejection conditions, injection pressure holding conditions, measurement conditions, temperature conditions, nozzle touch conditions, resin supply conditions, mold thickness conditions, molded product removal conditions, hot At least one of the runner conditions,
A system according to any one of claims 1 to 12, characterized in that A robot as a molded product take-out means in which the molded product take-out conditions are set;
The system of claim 13. Fluctuating at least one of the operation conditions within a predetermined range and learning the operation condition adjustment learning unit;
15. A system according to any one of the preceding claims.