Temperature control system for processing machines

Abstract

To enable a processing machine to perform a process with high accuracy regardless of a difference in outside air temperature.SOLUTION: A transition of outside air temperatures at days when a processing machine 10 is operated is estimated on the basis of data about weather forecast automatically obtained. A learning tool 85 is memorized in a memorizing part 82 of an electronic control device 80. The learning tool 85 is made to learn in a mode in which control target values (a target oil temperature, a target coolant temperature, a starting time of an oil supply part 40, a starting time of a coolant supply part 50, a target inside air temperature, and a starting time of an air conditioner 60) are outputted, with data on estimated transition of the outside air temperatures as input data. A reinforcement learning part 89 performs reinforcement learning of the learning tool 85 in a mode in which a higher reward is paid when variations in temperatures of parts of the processing machine 10 detected by temperature sensors 90-96 and in the inside air temperatures are small than when the variations are large. A control part 88 controls operation of the oil supply part 40, the coolant supply part 50 and the air conditioner 60, on the basis of control target values outputted from the learning tool 85.SELECTED DRAWING: Figure 3

Description

[Technical Field]

[0001] The present invention relates to a temperature adjustment system for a processing machine that processes a workpiece with a processing tool. [Background technology]

[0002] BACKGROUND ART Conventionally, processing machines that use tools to process the surface of an object to be processed (a so-called workpiece) fixed to a table have been widely used (see, for example, Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Laid-Open No. 2002-361552 Summary of the Invention [Problem to be solved by the invention]

[0004] Because processing machines deform slightly due to temperature changes, variations in the degree of temperature change in each part of the processing machine will also result in variations in the degree of deformation in each part of the processing machine. In this case, the straightness of the table may decrease, causing a misalignment between the tool and the workpiece, which may lead to a decrease in the processing accuracy of the workpiece. To prevent such a decrease in processing accuracy, it is preferable to manage the temperature of the processing machine and the room in which the processing machine is installed. Such temperature management can be achieved by installing an air conditioning device such as an air conditioner in the room in which the processing machine is installed.

[0005] However, due to the effects of climate change in recent years, the difference between the highest temperature in summer and the lowest temperature in winter has become greater. This increases the temperature difference between the inside and outside of the room where the processing machine is installed, which may make it difficult to properly control the temperature using an air conditioning system. This may result in a decrease in the processing accuracy of the workpieces produced by the processing machine. [Means for solving the problem]

[0006] A temperature adjustment system for a processing machine for solving the above problem includes a processing machine that processes a workpiece with a processing tool, a processing machine temperature adjustment device that adjusts the temperature of the processing machine to a first set temperature, an air conditioning device that adjusts the temperature of a room in which the processing machine is installed to a second set temperature, a processing machine temperature detection unit that detects temperatures at multiple locations on the processing machine, an internal temperature detection unit that detects the temperature inside the room, a data acquisition unit that automatically acquires weather forecast data via a communication network, a temperature estimation unit that estimates a transition in outside air temperature on an operating day of the processing machine based on the weather forecast data acquired by the data acquisition unit, and a temperature adjustment unit that uses the transition in outside air temperature estimated by the temperature estimation unit as input data to calculate a constant first control target value for the first set temperature and a temperature adjustment value for the processing machine. The system includes a learning value memory unit that stores a learning device that has been trained to output a first start-up timing of the temperature adjustment device, a constant second control target value for the second set temperature, and a second start-up timing of the air conditioning device; a reinforcement learning unit that performs reinforcement learning on the learning device in a manner that awards a higher reward when the variation in the temperatures at the multiple locations detected by the processing machine temperature detection unit and the internal temperature detected by the internal temperature detection unit is small than when the variation in the temperatures is large; and a control unit that controls the operation of the processing machine temperature adjustment device based on the first control target value and the first start-up timing output from the learning device, and controls the operation of the air conditioning device based on the second control target value and the second start-up timing output from the learning device.

[0007] In the above configuration, to prevent a decrease in machining accuracy due to differences in outside air temperature, it is preferable to operate the processing machine, processing machine temperature regulator, and air conditioner as follows: First, start the processing machine temperature regulator and air conditioner. Then, after the temperature of the room in which the processing machine is installed (inside air temperature) and the temperature of each part of the processing machine (processing machine temperature) have risen to a set temperature, temperature regulation by the processing machine temperature regulator and temperature regulation by the air conditioner are performed in such a manner that the set temperature is maintained constant. Then, processing by the processing machine is performed during the period in which the inside air temperature and the temperature of each part of the processing machine are maintained constant. This makes it possible to process a workpiece by the processing machine while minimizing variations in the amount of thermal deformation in each part of the processing machine.

[0008] According to the above configuration, the learning device can be trained to achieve the above-described desirable situation, i.e., a situation in which the processing machine performs processing while the inside air temperature and the processing machine temperature are maintained at constant set temperatures. The trained learning device can then be used to set a first constant control target value for the first set temperature (processing machine temperature), a first start timing for the processing machine temperature regulator, a second constant control target value for the second set temperature (inside air temperature), and a second start timing for the air conditioner, all of which correspond to the change in outside air temperature predicted from weather forecast data on the day of operation. Therefore, the processing machine temperature regulator and the air conditioner can be started in accordance with the change in outside air temperature predicted from weather forecast data on the day of operation, and the operation control of the processing machine temperature regulator and the air conditioner can be executed. This realizes the above-described desirable situation, thereby suppressing a decrease in processing accuracy due to differences in outside air temperature. [Effects of the Invention]

[0009] According to the temperature adjustment system for a processing machine of the present invention, it is possible to suppress a decrease in processing accuracy caused by differences in outside air temperature. [Brief explanation of the drawings]

[0010] [Figure 1]FIG. 2 is a schematic diagram of a temperature adjustment system for a processing machine. [Figure 2] FIG. [Figure 3] FIG. 2 is a block diagram showing the electrical configuration of the temperature adjustment system. [Figure 4] 4 is a time chart showing an example of the transition of the temperature of each part in a preferable situation. [Figure 5] FIG. 2 is a block diagram showing an example of the configuration of a control unit. [Figure 6] FIG. 2 is a block diagram illustrating an example of the configuration of a long-term short-term memory layer. DETAILED DESCRIPTION OF THE INVENTION

[0011] An embodiment of a temperature adjustment system for a processing machine will be described below. (processing machine) As shown in Figures 1 and 2, the processing machine 10 of this embodiment is a gate-type surface grinding machine. The processing machine 10 has a bed 11. A pair of rails 12 extending in the X direction is provided on the upper surface of the bed 11. A work table 13 is provided above the bed 11. The work table 13 is capable of reciprocating in the X direction relative to the bed 11 via the pair of rails 12. When a workpiece (hereinafter referred to as workpiece W) is processed by the processing machine 10, the workpiece W is supported on the work table 13.

[0012] The processing machine 10 has a gate-shaped column 14. The column 14 has a pair of upright pillars 15 extending vertically on both sides of the bed 11, and a beam 16 installed to connect the upper ends of the upright pillars 15. The column 14 is installed so as to straddle the running surface of the work table 13 on the bed 11.

[0013] A crossbar 17 is provided on the column 14. The crossbar 17 is installed across the pair of upright pillars 15 so as to connect the vertically middle portions of the upright pillars 15. A pair of rails 18 extending in the Y direction are provided on one side surface (the left side in FIG. 1) of the crossbar 17. A support plate 19 is also provided on the side of the crossbar 17. The support plate 19 is capable of reciprocating movement in the Y direction relative to the crossbar 17 via the pair of rails 18.

[0014] A pair of rails 20 extending in the Z direction are provided on one side surface (the left side in FIG. 1) of the support plate 19. A processing head 21 is also provided on one side of the support plate 19. The processing head 21 is capable of moving back and forth in the Z direction relative to the support plate 19 via the pair of rails 20. A rotary grindstone 22 serving as a processing tool is attached to the processing head 21. The rotary grindstone 22 is rotated while in contact with the top surface of the workpiece W, and the work table 13 is moved back and forth, thereby grinding the processing surface of the workpiece W.

[0015] As shown in FIGS. 1 to 3, the processing machine 10 has an X-direction moving unit 31 (FIG. 3) that moves the work table 13 in the X direction relative to the bed 11, and a Y-direction moving unit 32 that moves the support plate 19 in the Y direction relative to the crossbar 17. The processing machine 10 also has a Z-direction moving unit 33 that moves the processing head 21 in the Z direction relative to the support plate 19. Each of the X-direction moving unit 31, Y-direction moving unit 32, and Z-direction moving unit 33 is configured with a moving mechanism that enables relative movement of the two members and an actuator that operates the moving mechanism. In addition, the processing machine 10 has a grindstone driving unit 34 that rotates and drives the rotary grindstone 22.

[0016] During processing by the processing machine 10, first, the processing head 21 is moved in the Y direction and the Z direction through operation control of the Y-direction moving unit 32 and operation control of the Z-direction moving unit 33. This puts the processing machine 10 in a state where it can bring the rotary grinding wheel 22 into contact with a desired position on the processing surface of the workpiece W. Then, in this state, the rotary grinding wheel 22 is rotationally driven through operation control of the grinding wheel driving unit 34, and the work table 13 and the workpiece W are reciprocated in the X direction through operation control of the X-direction moving unit 31. This causes the processing surface of the workpiece W to be ground.

[0017] (oil supply section) 1 and 3, the processing machine 10 is equipped with an oil supply unit 40 that supplies oil to lubrication parts. The oil supply unit 40 has an oil tank 41 that stores oil and an oil pump 42 that pumps the oil. The oil supply unit 40 also has a temperature adjustment unit 43 that adjusts the temperature of the oil to a desired temperature. In this embodiment, based on a control command signal input to the oil supply unit 40 from an external device (an electronic control device 80, described later), it is possible to start oil supply by the oil supply unit 40 and set a control target value for the oil temperature TO (target oil temperature TTO).

[0018] In this embodiment, a constant value is set as the target oil temperature TTO. This is for the following reason: If the target oil temperature TTO is changed while the processing machine 10 is in operation, the temperature of the processing machine 10 will gradually change over time due to the heat capacity of the processing machine 10. As the temperature of the processing machine 10 changes, temperature variations and, ultimately, thermal deformation variations will occur in various parts of the processing machine 10, which may result in a decrease in processing accuracy. For this reason, in order to maintain the processing accuracy of the processing machine 10, it is preferable to maintain the target oil temperature TTO constant while the processing machine 10 is in operation.

[0019] (Coolant supply section) The processing machine 10 is equipped with a coolant supply unit 50 that supplies coolant to a processing location where the workpiece W is processed by the rotary grinding wheel 22. The coolant supply unit 50 has a coolant tank 51 that stores coolant and a coolant pump 52 that pumps the coolant. The coolant supply unit 50 also has a temperature adjustment unit 53 that adjusts the temperature of the coolant to a desired temperature. Based on a control command signal input to the coolant supply unit 50 from an electronic control device 80 (described later), it is possible to start the supply of coolant by the coolant supply unit 50 and to set a control target value for the coolant temperature TC (target coolant temperature TTC).

[0020] In this embodiment, a constant value is set as the target coolant temperature TTC. This is for the following reason: If the target coolant temperature TTC is changed while the processing machine 10 is in operation, the temperature of the processing machine 10 will gradually change over time due to the heat capacity of the processing machine 10. As the temperature of the processing machine 10 changes, temperature variations and, ultimately, thermal deformation variations will occur in various parts of the processing machine 10, which may result in a decrease in processing accuracy. For this reason, in order to maintain the processing accuracy of the processing machine 10, it is preferable to maintain the target coolant temperature TTC constant while the processing machine 10 is in operation.

[0021] In this embodiment, the oil supply unit 40 and the coolant supply unit 50 correspond to a processing machine temperature control device that adjusts the temperature of the processing machine to a first set temperature, and the target oil temperature TTO and the target coolant temperature TTC correspond to a constant first control target value for the first set temperature.

[0022] (machine control device) 3, the processing machine 10 has an operation panel 23 operated by an operator, and a processing machine control device 24 that controls the operation of the processing machine 10. The processing machine control device 24 has an arithmetic processing unit consisting of one or more processors, and a memory unit that stores various data related to the operation control of the processing machine 10. The processing machine control device 24 executes various controls related to the operation of the processing machine 10 based on various information related to the operation of the processing machine 10 input through operation of the operation panel 23. The various controls include operation control of the X-direction moving unit 31, operation control of the Y-direction moving unit 32, operation control of the Z-direction moving unit 33, operation control of the grinding wheel driving unit 34, operation control of the oil supply unit 40, and operation control of the coolant supply unit 50.

[0023] (Air conditioner) The system of this embodiment has an air conditioner 60 for adjusting the temperature (inside air temperature TI) inside the room R in which the processing machine 10 is installed. In this embodiment, it is possible to start the operation of the air conditioner 60 and set a control target value for the inside air temperature TI (target inside air temperature TTI) based on a control command signal input to the air conditioner 60 from an external device (an electronic control device 80, described later).

[0024] In this embodiment, a constant value is set as the target indoor air temperature TTI. This is for the following reason. If the target indoor air temperature TTI is changed while the processing machine 10 is operating, the temperature of the processing machine 10 will gradually change over time due to the heat capacity of the air in the room R and the heat capacity of the processing machine 10. As the temperature of the processing machine 10 changes, temperature variations and, ultimately, variations in the amount of thermal deformation will occur in various parts of the processing machine 10, which may result in a decrease in processing accuracy. For this reason, in order to maintain the processing accuracy of the processing machine 10, it is preferable to maintain the target indoor air temperature TTI constant while the processing machine 10 is operating. Note that in this embodiment, the target indoor air temperature TTI corresponds to a constant second control target value for the second set temperature.

[0025] Here, in order to prevent a decrease in the machining accuracy of the processing machine 10 due to differences in outside air temperature, it is preferable to operate the processing machine 10 and the air conditioning device 60 as follows. That is, first, as shown in Fig. 4, before the processing machine 10 starts machining the workpiece W, the oil supply unit 40, the coolant supply unit 50, and the air conditioning device 60 of the processing machine 10 are started (time t1). Then, temperature control by the temperature adjustment units 43, 53 and temperature control by the air conditioning device 60 are performed in such a manner that the temperature of the room R in which the processing machine 10 is installed (inside air temperature TI) and the temperatures of each part of the processing machine 10 (processing machine temperature) are raised to a set temperature and then maintained at the set temperature. Then, processing by the processing machine 10 is performed during a period TS (times t2 to t3) in which the inside air temperature TI and the processing machine temperature are maintained constant. By operating the processing machine 10 and the air conditioning unit 60 in this manner, it becomes possible to process the workpiece W using the processing machine 10 while suppressing temperature variations in each part of the processing machine 10, and ultimately variations in the amount of thermal deformation.

[0026] The temperature adjustment system of this embodiment achieves the above-described preferable conditions in the following manner. First, weather forecast data is acquired from the external device 70, and from this weather forecast data, the change in the outside air temperature on the operating day of the processing machine 10 (the day on which the processing machine 10 processes the workpiece W) is predicted. Then, based on this predicted data on the outside air temperature, control target values ​​([A] to [F] below) for operation control of the oil supply unit 40, coolant supply unit 50, and air conditioner 60 of the processing machine 10 are calculated.

[0027] [A] The control target value (target oil temperature TTO) of the oil temperature TO related to the operation control of the oil supply unit 40 (including the temperature adjustment unit 43). [B] Control target value for the start timing of the oil supply unit 40 (start time STO).

[0028] [C] The control target value (target coolant temperature TTC) of the coolant temperature TC related to the operation control of the coolant supply unit 50 (including the temperature adjustment unit 53). [D] Control target value for the start timing of the coolant supply unit 50 (start time STC).

[0029] [E] A control target value of the inside air temperature TI (target inside air temperature TTI) for controlling the operation of the air conditioner 60. [F] Control target value for the start timing of the air conditioner 60 (start time STI).

[0030] In the temperature adjustment system of this embodiment, values ​​that satisfy all of the following (Condition 1) to (Condition 4) are calculated as these control target values. (Condition 1) The oil supply unit 40, the coolant supply unit 50, and the air conditioning device 60 of the processing machine 10 are started up before the opening time (for example, 8:00 AM) of the factory where the processing machine 10 is installed.

[0031] (Condition 2) Between the start of the processing machine 10 and the air conditioner 60 and the start of work in the factory, the temperatures of the various parts of the processing machine 10 and the inside air temperature TI rise to a predetermined temperature. (Condition 3) During the operating hours of the factory (the time from the start of work to the end of work), the temperatures of the various parts of the processing machine 10 and the inside air temperature TI are kept substantially constant at predetermined temperatures.

[0032] (Condition 4) The indoor air temperature TI during factory operating hours is a temperature that workers do not find uncomfortable. Then, based on these control target values, the oil supply unit 40, the coolant supply unit 50, and the air conditioning device 60 of the processing machine 10 are started and operated.

[0033] (Temperature control system) A specific configuration for realizing the above-mentioned preferable situation will be described below. The temperature adjustment system of this embodiment has an electronic control device 80 that controls the operation of the oil supply unit 40, coolant supply unit 50, and air conditioning unit 60 of the processing machine 10 to adjust the temperature of each part of the processing machine 10 and the inside air temperature TI.

[0034] The electronic control device 80 has an arithmetic processing unit 81 consisting of one or more processors, a memory unit 82 that stores various data related to the operation control of the processing machine 10 and the air conditioning device 60, and a communication unit 83 that communicates with external equipment 70.

[0035] The memory unit 82 stores a program 84 for causing the arithmetic processing unit 81 to execute various processes related to the operation control of the oil supply unit 40, the operation control of the coolant supply unit 50, and the operation control of the air conditioning unit 60. The arithmetic processing unit 81 executes this program 84 to execute various processes related to the operation control of the oil supply unit 40, the operation control of the coolant supply unit 50, and the operation control of the air conditioning unit 60. The memory unit 82 stores a machine learning model (hereinafter, referred to as a learning device 85). This learning device 85 is used to calculate control target values ​​related to the operation control of the processing machine 10 and the control target values ​​related to the operation control of the air conditioning unit 60. Details of the learning device 85 will be described later. In this embodiment, the memory unit 82 corresponds to a learned value storage unit.

[0036] The communication unit 83 is a communication device that transmits and receives data to and from an external device 70 (more specifically, a server at an information center that provides weather forecast data) via a predetermined communication network 71 such as the Internet. In this embodiment, data related to a weather forecast (hereinafter, weather forecast data) is imported from the external device 70 to the electronic control unit 80 via the communication unit 83, and the weather forecast data is stored in the memory unit 82.

[0037] (Temperature sensor) The temperature adjustment system of this embodiment has temperature sensors 90 to 96 for detecting the temperature of each part. The output signals of the temperature sensors 90 to 96 are taken into the electronic control unit 80 and stored in the memory unit .

[0038] 1 and 2, six temperature sensors 90 for detecting the temperature of the upper portion are provided at intervals on the work table 13 of the processing machine 10, and six temperature sensors 91 for detecting the temperature of the lower portion are provided at intervals on the work table 13. Eight temperature sensors 92 for detecting the temperature of the upper portion are provided at intervals on the bed 11 of the processing machine 10, and eight temperature sensors 93 for detecting the temperature of the lower portion are provided at intervals on the bed 11. Note that Fig. 1 shows only three each of the temperature sensors 90 and 91, and only four each of the temperature sensors 92 and 93.

[0039] The oil supply unit 40 is provided with a temperature sensor 94 for detecting the oil temperature TO. The coolant supply unit 50 is provided with a temperature sensor 95 for detecting the coolant temperature TC.

[0040] A temperature sensor 96 is provided on the column 14 of the processing machine 10 to detect the internal temperature (internal air temperature TI) of the room R in which the processing machine 10 is installed. In this embodiment, the temperature sensors 90 to 95 correspond to a processing machine temperature detection unit that detects the temperature at a plurality of locations in the processing machine, and the temperature sensor 96 corresponds to an internal temperature detection unit that detects the temperature inside the room R.

[0041] The electronic control device 80 has a data acquisition unit 86, a temperature estimation unit 87, a control unit 88, and a reinforcement learning unit 89 as functional units for maintaining the temperatures of each part of the processing machine 10 and the inside air temperature TI constant.

[0042] (Data acquisition section) The data acquisition unit 86 acquires weather forecast data from the external device 70 via the communication network 71 and the communication unit 83. In this embodiment, the weather forecast data is provided by an external information providing center. The data acquisition unit 86 automatically acquires the weather forecast data from the external device 70 at a predetermined time (for example, 2:00 AM on each operating day of the processing machine 10).

[0043] The weather forecast data acquired by the data acquisition unit 86 is data for a predetermined area including the facility (factory) where the processing machine 10 is installed. This weather forecast data is data that includes at least temperature information (maximum temperature, minimum temperature, etc.). In addition to temperature information, the weather forecast data may also include weather information (sunny, cloudy, rainy) and humidity information.

[0044] (Temperature estimation part) The temperature estimation unit 87 estimates the transition of the outside air temperature on the day the processing machine 10 operates, based on the weather forecast data acquired by the data acquisition unit 86. In this embodiment, a relationship (e.g., a calculation model) that enables accurate calculation of the transition of the actual outside air temperature on the day of the weather forecast data based on the weather forecast data has been determined in advance from the results of various experiments and simulations conducted by the inventors, etc. This relationship is then stored in advance in the memory unit 82 of the electronic control device 80. The temperature estimation unit 87 calculates time-series data (hereinafter, estimated data) that indicates the transition of the estimated value of the outside air temperature from the relationship stored in the memory unit 82, based on the weather forecast data acquired by the data acquisition unit 86.

[0045] (Control unit) The control unit 88 calculates each control target value (the above [A] to [F]) based on the estimated data and outputs these control target values ​​as control command signals to the processing machine control device 24 and the air conditioner 60. As a result, the operation of the processing machine 10 on an operating day is controlled, such as starting the oil supply unit 40 at the start time STO and operating the oil supply unit 40 (including the temperature adjustment unit 43) based on the target oil temperature TTO. The operation of the processing machine 10 on an operating day is also controlled, such as starting the coolant supply unit 50 at the start time STC and operating the coolant supply unit 50 (including the temperature adjustment unit 43) based on the target coolant temperature TTC. Furthermore, the operation of the air conditioner 60 on an operating day is controlled, such as starting the air conditioner 60 at the start time STI and operating based on the target inside air temperature TTI. In this embodiment, the start times STO and STC correspond to the first start timing of the processing machine temperature adjustment device, and the start time STI corresponds to the second start timing of the air conditioner.

[0046] (Learning device) As shown in FIG. 5, the control unit 88 is mainly configured with the learning device 85. In this embodiment, a recurrent neural network (RNN) is used as the learning device 85. The learning device 85 has an input layer 100 to which estimated data is input as input data, an output layer 101 that outputs each control target value, and multiple hidden layers 102 located between the input layer 100 and the output layer 101. Hereinafter, the hidden layer located at the bottom is referred to as hidden layer 102-1, and the hidden layer located at the top is referred to as hidden layer 102-n (n is an integer equal to or greater than 1). The input layer 100 performs various calculations based on the estimated data and outputs the calculation results to the hidden layer 102-1. Each hidden layer 102 uses the calculation result at time t-1 as a return value for calculation at time t. The output layer 101 calculates each control target value using the calculation results of the hidden layer 102-n.

[0047] Normalized data of the estimated data is input to the input layer 100. The hidden layers 102 calculate one or more feature quantities as the final calculation results. Each hidden layer 102 is composed of a long short-term memory (LSTM) layer.

[0048] As shown in FIG. 6, the long short-term memory layer (hereinafter referred to as LSTM 110) calculates an output h(t) at time t based on an input x(t) at time t. The output h(t) is input to the upper layer as input x(t). The LSTM 110 also calculates a cell state C(t) in the process of calculating the output h(t), and treats the calculated cell state C(t) and the output state h(t) indicating the output h(t) as the return value for the next cycle of calculation. The LSTM 110 calculates the cell state C(t) and output state h(t) at time (t) based on the input x(t) from the lower layer at time (t), as well as the cell state C(t-1) and output state h(t-1), which are the calculation results at the previous time (t-1). The LSTM 110 has a forget gate layer 111 that controls the cell state C(t-1) at time t-1, an input gate layer 112, a tanh layer 113, and an output gate layer 114. In the learning device 85 of this embodiment, the weight vectors and bias vectors determined in the calculations in each of the layers 111 to 114 of the LSTM 110 are updated by learning.

[0049] The learning device 85 extracts feature amounts based on the estimated data, and learns in a manner that calculates and outputs each control target value related to the operation control of the oil supply unit 40, the operation control of the coolant supply unit 50, and the operation control of the air conditioner 60 based on these feature amounts. The learning device 85 learns in a manner that realizes the above-mentioned preferable situation (see FIG. 4). In detail, the learning device 85 learns in a manner that outputs the following values ​​as each control target value:

[0050] As the start time STO, a time (e.g., 4:00 AM) preceding the start time (e.g., 8:00 AM) by the start time TB (times t1 to t2 in FIG. 4) from the start of the oil supply unit 40 under the above-mentioned favorable conditions until the oil temperature TO reaches a substantially constant stable state is output. As the target oil temperature TTO, a value equivalent to the oil temperature TO when the oil temperature TO reaches a substantially constant stable state under the above-mentioned favorable conditions is output. As the start time STC, a time preceding the start time of the factory by the start time TB (times t1 to t2 in FIG. 4) from the start of the coolant supply unit 50 under the above-mentioned favorable conditions until the coolant temperature TC reaches a substantially constant stable state is output. As the target coolant temperature TTC, a value equivalent to the coolant temperature TC when the coolant temperature TC reaches a substantially constant stable state under the above-mentioned favorable conditions is output. As the start time STI, a time preceding the start time of the factory by the start-up time TB (times t1 to t2 in FIG. 4) from the start of the air conditioner 60 under the above-mentioned favorable circumstances until the inside air temperature TI reaches a substantially constant stable state is output. As the target inside air temperature TTI, a value corresponding to the inside air temperature TI when the inside air temperature TI reaches a substantially constant stable state under the above-mentioned favorable circumstances is output.

[0051] (Reinforcement Learning Department) The reinforcement learning unit 89 performs reinforcement learning to train the learner 85 using a reward. The reinforcement learning unit 89 sets a reward based on the temperatures of each part detected by the temperature sensors 90-96 while the processing machine 10 and the air conditioner 60 are in operation. Specifically, the reinforcement learning unit 89 calculates a higher reward the smaller the variation in temperatures at multiple locations, including the temperatures of each part of the processing machine 10 and the inside air temperature TI (specifically, the temperatures detected by the temperature sensors 90-96). In this embodiment, the smaller the value indicating the difference between the temperatures at multiple locations at any given time (for example, the difference between the maximum and minimum values) (hereinafter referred to as temperature difference ΔT), the higher the reward set by the reinforcement learning unit 89.

[0052] The reinforcement learning unit 89 calculates the temperature difference ΔT and sets a learning coefficient based on the temperature difference ΔT. For example, the learning coefficient is set to "1.0" when no reward is given because the temperature difference ΔT is large, and a value greater than "1.0" is calculated when a reward is given because the temperature difference ΔT is small. Reinforcement learning is then performed using this learning coefficient as a reward. In reinforcement learning, the weight vectors and bias vectors determined in the calculations in each of the layers 111 to 114 of the LSTM 110 are updated.

[0053] (action) The operation of the temperature adjustment system of this embodiment will be described below. First, at a predetermined time, the electronic control device 80 retrieves weather forecast data from the external device 70 via the communication network 71 and stores it in the memory unit 82. Then, based on this weather forecast data, the electronic control device 80 calculates time-series data (estimated data) that indicate the transition of estimated values ​​of the outside air temperature on the day the processing machine 10 is in operation.

[0054] Thereafter, based on this estimated data, the electronic control device 80 calculates each control target value (TTO, STO, TTC, STC, TI, STI) for controlling the operation of the oil supply unit 40, the coolant supply unit 50, and the air conditioning unit 60 of the processing machine 10. These control target values ​​are calculated using a trained learning device 85 stored in the memory unit 82.

[0055] Then, the electronic control device 80 outputs each control target value to the corresponding device (machine control device 24 or air conditioner 60) as a control command signal. When the control command signal is input, the processing machine control device 24 controls the operation of the oil supply unit 40 of the processing machine 10 on an operating day based on the start time STO and the target oil temperature TTO. Specifically, the oil supply unit 40 is started at the start time STO (time t1) before the start time of the factory (e.g., time t2 in FIG. 4). After the oil supply unit 40 is started (after time t1), the operation of the oil supply unit 40 is controlled so that the actual oil temperature TTO becomes the target oil temperature TTO. In this embodiment, through this operation control of the oil supply unit 40, the actual oil temperature TTO reaches the target oil temperature TTO before the start time of the factory (time t2). Moreover, during the factory operating hours (times t2 to t3), the actual oil temperature TTO is maintained substantially constant at the target oil temperature TTO.

[0056] On the other hand, when the control command signal is input, the processing machine control device 24 controls the operation of the coolant supply unit 50 of the processing machine 10 on an operating day based on the start time STC and the target coolant temperature TTC. Specifically, the coolant supply unit 50 is started at the start time STC (time t1) before the start time of the factory (e.g., time t2 in FIG. 4). After the start of the coolant supply unit 50 (after time t1), the operation of the coolant supply unit 50 is controlled so that the actual coolant temperature TC becomes the target coolant temperature TTC. In this embodiment, through this control of the operation of the coolant supply unit 50, the actual coolant temperature TC reaches the target coolant temperature TTC before the start time (time t2). Moreover, during the factory operating hours (times t2 to t3), the actual coolant temperature TC is maintained substantially constant at the target coolant temperature TTC.

[0057] On the other hand, when a control command signal is input, the air conditioner 60 controls its operation on an operating day based on the start time STI and the target inside air temperature TTI. Specifically, the air conditioner 60 is started at a start time STI (time t1) before the start of work at the factory (for example, time t2 in FIG. 4). After the air conditioner 60 is started (after time t1), the operation of the air conditioner 60 is controlled so that the actual inside air temperature TI becomes the target inside air temperature TTI. In this embodiment, through this operation control of the air conditioner 60, the actual inside air temperature TI reaches the target inside air temperature TTI before the start of work (time t2). Moreover, during the factory operating hours (times t2 to t3), the actual inside air temperature TI is maintained substantially constant at the target inside air temperature TTI.

[0058] As described above, in this embodiment, the oil supply unit 40, coolant supply unit 50, and air conditioning unit 60 of the processing machine 10 are automatically started and operated before the start of business in the factory. Through the preliminary operation of the oil supply unit 40, coolant supply unit 50, and air conditioning unit 60, the temperatures of the various parts of the processing machine 10 and the inside air temperature TI are maintained at approximately constant levels at the start of business in the factory. Furthermore, the temperatures of the various parts of the processing machine 10 and the inside air temperature TI are maintained at approximately constant levels throughout the subsequent operating hours of the factory.

[0059] As a result, during factory operation, the processing machine 10 processes the workpiece W in a state where there is little temperature variation in each part of the processing machine 10. Therefore, variation in the amount of thermal deformation in each part of the processing machine 10 when the processing machine 10 processes the workpiece W is reduced. This reduces a decrease in the straightness of the work table 13 of the processing machine 10 caused by differences in outside air temperature, and ultimately reduces a decrease in the straightness of the workpiece W. Therefore, according to this embodiment, it is possible to achieve high-precision processing of the workpiece W by the processing machine 10, regardless of differences in outside air temperature.

[0060] According to this embodiment, the following effects can be obtained. (1) According to this embodiment, the learning device 85 can be trained to achieve the above-described preferable situation, i.e., a situation in which the processing machine 10 performs processing while the inside air temperature TI and the temperatures of each part of the processing machine 10 are maintained at constant set temperatures. Then, using the trained learning device 85, values ​​appropriate for the change in outside air temperature on the day of operation predicted from weather forecast data (specifically, estimated data) can be set as various control target values ​​for the operation control of the processing machine 10 and the operation control of the air conditioner 60.

[0061] Therefore, the oil supply unit 40, the coolant supply unit 50, and the air conditioning unit 60 can be started in accordance with the change in the outside air temperature on the working day predicted from weather forecast data. Moreover, the oil temperature TO can be adjusted by the oil supply unit 40, the coolant temperature TC can be adjusted by the coolant supply unit 50, and the inside air temperature TI can be adjusted by the air conditioning unit 60 in accordance with the change in the outside air temperature on the working day. This makes it possible to realize the above-mentioned preferable situation. Therefore, the workpiece W can be machined with high precision by the processing machine 10 regardless of differences in outside air temperature, such as differences in outside air temperature on the working day (the difference between the maximum and minimum temperatures of the day) or differences in outside air temperature between seasons (the difference between the maximum temperature in summer and the minimum temperature in winter).

[0062] (Example of change) The above embodiment can be modified as follows: The above embodiment and the following modifications can be combined with each other within the scope of technical compatibility.

[0063] The manner in which weather forecast data is acquired from the external device 70 can be changed as desired. The timing for acquiring the weather forecast data is not limited to a predetermined time before the start of work on the day the processing machine 10 is to be operated, but can also be a predetermined time (e.g., 11:00 p.m.) after the end of work on the previous day (e.g., 5:00 p.m.). The weather forecast data acquired by the data acquisition unit 86 may be weather forecast data for one day or weather forecast data for one week. When acquiring weather forecast data for one week from the external device 70, a predetermined time on a specific day of the week (e.g., late Sunday night, early Monday morning, etc.) may be set as the timing for acquiring the weather forecast data.

[0064] When the estimated data is calculated before the start of work on a working day, the actual temperature (outside air temperature) outside the room R in which the processing machine 10 is installed may be used as a calculation parameter in addition to weather forecast data. In this case, a temperature sensor for detecting the outside air temperature may be provided outside the room R (specifically, outdoors). According to the above configuration, the actual outside air temperature at the time of calculation (or before the calculation timing on the working day) can be taken into account when calculating the estimated data, so that the estimated data can be calculated with high accuracy.

[0065] The reinforcement learning of the learning device 85 may be configured to take into account the actual temperature (outside air temperature) outside the room R in which the processing machine 10 is installed. With this configuration, the reinforcement learning of the learning device 85 can be performed in a manner that corresponds to the error between the estimated value of the outside air temperature indicated by the estimation data and the actual value of the outside air temperature. Therefore, the reinforcement learning of the learning device 85 can be performed accurately in a manner that is in line with the actual situation. In this case, the reinforcement learning of the learning device 85 can be performed in a manner that, for example, gives a higher reward when the difference between the estimated value and the actual value of the outside air temperature is small than when the difference is large.

[0066] For the learning of the learner 85, supervised learning based on training data may be performed in addition to the reinforcement learning. The locations where temperature sensors for detecting the temperatures of various parts of the processing machine 10 are installed can be changed as desired. The key is to determine in advance from the results of various experiments and simulations which detection locations will enable the processing accuracy of the workpiece W by the processing machine 10 to be improved, and then install the temperature sensors at those detection locations.

[0067] The location where the temperature sensor for detecting the inside air temperature TI is installed can be changed as desired. Furthermore, temperature sensors for detecting the inside air temperature TI may be installed in multiple locations. The key point is to determine in advance from the results of various experiments and simulations which detection location will enable the machining accuracy of the workpiece W by the machining machine 10 to be improved, and then install the temperature sensor at that detection location.

[0068] As a control to prevent a decrease in the machining accuracy of the machining machine 10 due to differences in outside air temperature, one of the operation control of the oil supply unit 40 and the operation control of the coolant supply unit 50 may be executed, while the other may not be executed.

[0069] The parameters used to set the reward for reinforcement learning can be any value, such as the difference between the maximum and minimum values ​​of the temperatures detected by the temperature sensors 90 to 96, or the standard deviation in the distribution (normal distribution) of the temperatures detected by the temperature sensors 90 to 96. Any value that indicates the degree of variation in temperatures at multiple locations, including the temperatures of each part of the processing machine 10 and the inside air temperature TI, can be used as the parameter.

[0070] The learner 85 may be a convolutional neural network or other neural network other than a recurrent neural network. The temperature adjustment system according to the above embodiment is not limited to a temperature adjustment system for a grinding machine that grinds the surface of a workpiece with a grinding wheel, but can also be applied to a temperature adjustment system for a gear cutting machine, etc. The temperature adjustment system according to the above embodiment can be applied to a temperature adjustment system for any type of cutting machine that processes a workpiece with a processing tool. [Explanation of symbols]

[0071] 10…Processing machine 22...Rotary grinding wheel 24...Processing machine control device 40...Oil supply section 43...Temperature control unit 50...Coolant supply section 53...Temperature control unit 60...Air conditioner 71...Communication Network 80...Electronic control device 82...Storage section 85...Learning device 86...Data acquisition section 87…Temperature estimation section 88...Control unit 89…Reinforcement Learning Department 90~96...Temperature sensor

Claims

[Claim 1] A temperature adjustment system for a processing machine that adjusts the temperature of the processing machine prior to the start of processing a workpiece by the processing machine, the processing machine that processes the workpiece with a processing tool; a processing machine temperature adjusting device that adjusts the temperature of the processing machine to a first set temperature; an air conditioner that adjusts the temperature of a room in which the processing machine is installed to a second set temperature; a processing machine temperature detection unit that detects temperatures at a plurality of locations in the processing machine; an internal temperature detection unit that detects the temperature inside the room; a data acquisition unit that automatically acquires weather forecast data through a communication network; a temperature estimation unit that estimates a change in outside air temperature on an operating day of the processing machine based on the weather forecast data acquired by the data acquisition unit; a learning value storage unit that stores a learning unit that has been trained in a manner that uses the transition of the outside air temperature estimated by the temperature estimation unit as input data and outputs a constant first control target value for the first set temperature, a first start timing of the processing machine temperature adjustment device, a constant second control target value for the second set temperature, and a second start timing of the air conditioning device; a reinforcement learning unit that performs reinforcement learning on the learning device in a manner that a higher reward is given when the variations in the temperatures at the plurality of locations detected by the processing machine temperature detection unit and the internal temperature detected by the internal temperature detection unit are small than when the variations in the temperatures are large; A temperature adjustment system for a processing machine comprising: a control unit that controls the operation of the processing machine temperature adjustment device based on the first control target value and the first start-up timing output from the learning device, and controls the operation of the air conditioning device based on the second control target value and the second start-up timing output from the learning device.

Images