Working machinery

Abstract

To provide a work machine with high control robustness which can maintain a constant relationship between an operation input and a control amount of a control target element of a hydraulic drive system, regardless of fluctuations in an operating state of a hydraulic drive system, including changes in an operating environment, and performance variations and deterioration of hydraulic components.SOLUTION: A control device 114 includes: a learning data storage unit 114i that stores target control amounts of control target elements 113a to 113e and an operating state of a hydraulic drive system 902 as learning data; a learning unit 114k that machine-learns control amounts of the control target elements 113a to 113e with respect to the target control amounts based on the learning data stored in the learning data storage unit 114i; a prediction unit 114l that calculates a predicted control amount which is a predicted value of the control amount for the target control amount, using learning result of the machine-learning in the learning unit 114k; and a control signal correction unit 114d that corrects a control signal so that the predicted control amount approaches the target control amount.SELECTED DRAWING: Figure 4

Description

Technical Field

[0001] The present invention relates to a working machine such as a hydraulic excavator.

Background Art

[0002] Generally, the hydraulic system of a working machine powered by hydraulic pressure is composed of a plurality of hydraulic pumps, a plurality of hydraulic actuators, and a plurality of flow control valves for controlling the hydraulic oil supplied from the plurality of hydraulic pumps to the plurality of hydraulic actuators. It is common for these hydraulic components to be mechanically directly controlled. For example, the movement of an operation lever mechanically acts on the operation of a hydraulic valve, and the operating amount of the hydraulic components such as the pump and valve is determined by the hydraulic command generated thereby. In recent years, an electronically controlled hydraulic system that controls hydraulic components using an electric signal has been developed instead of the mechanical control of hydraulic components.

[0003] In an electronically controlled hydraulic system, an input command can be converted into a current or a signal for driving the operation of a hydraulic component, and by controlling the operation of the hydraulic component by a controller, it is easy to perform operations in various operation modes and implement various functions. On the other hand, in an electronically controlled hydraulic system, there are many components intervening between the operation instruction input and driving the hydraulic component, and due to the accumulation of errors caused by variations in performance and non-linear characteristics included in these components, the deviation between the desired operation and the actual operation of the working machine or hydraulic component tends to increase, which is an issue.

[0004] As a document disclosing a prior art for solving this problem, for example, there is Patent Document 1. According to Patent Document 1, by using a machine learning system based on a training dataset, a combination of an input command indicating the target speed of a hydraulic component and machine state data is mapped to a predicted displacement map of a hydraulic valve that causes the movement of the hydraulic component in response to the input command, and by determining the valve current using this predicted displacement map, the hydraulic component can be moved at the target speed.

Prior Art Documents

[0005] [Patent Document 1] Special table 2022-532740 publication [Overview of the project] [Problems that the invention aims to solve]

[0006] However, in the machine control system described in Patent Document 1, training data is constructed based on parameters measured during the execution of a specific machine operation. Therefore, it is extremely difficult to cover all external disturbances, such as changes in the operating environment, and internal disturbances, such as performance variations of hydraulic components, that may occur in the machine control system with the training dataset. Consequently, the amount of control of the controlled element of the hydraulic drive system in response to the operation input may fluctuate depending on whether or not external or internal disturbances that are not covered in the training dataset occur.

[0007] The present invention has been made in view of the above problems, and its objective is to provide a highly controllable work machine that can maintain a constant relationship between the control amount of the controlled element of the hydraulic drive system and the operation input, regardless of fluctuations in the operating state of the hydraulic drive system, including changes in the operating environment and variations and degradation of the performance of hydraulic components. [Means for solving the problem]

[0008] To achieve the above objective, the present invention provides a work machine comprising a vehicle body, a work device attached to the vehicle body, a hydraulic actuator for driving the work device, a hydraulic drive system for driving the hydraulic actuator, an operating device for instructing the operation of the hydraulic actuator, and a control device for outputting a control signal to the hydraulic drive system according to the amount of operation of the operating device, wherein the control device comprises: a target control amount calculation unit that calculates a target control amount of a controlled element of the hydraulic drive system based on the amount of operation of the operating device; an operating state detection unit that detects the operating state of the hydraulic drive system; a learning data storage unit that stores the target control amount and the detection result of the operating state detection unit as learning data; a learning unit that performs machine learning on the control amount of the controlled element with respect to the target control amount based on the learning data stored in the learning data storage unit; a prediction unit that calculates a predicted control amount, which is a predicted value of the control amount with respect to the target control amount, using the learning result of the machine learning in the learning unit; and a control signal correction unit that corrects the control signal so that the predicted control amount approaches the target control amount. [Effects of the Invention]

[0009] According to the present invention, regardless of fluctuations in the operating state of the hydraulic drive system, including changes in the operating environment and variations and degradation of the performance of hydraulic components, the relationship between the control amount of the controlled element of the hydraulic drive system and the operation input is kept constant, thereby improving the control robustness of the work machine. [Brief explanation of the drawing]

[0010] [Figure 1] Side view of a hydraulic excavator according to an embodiment of the present invention [Figure 2A] Circuit diagram (1 / 2) of the hydraulic drive system in the first embodiment of the present invention [Figure 2B] Circuit diagram of the hydraulic drive system in the first embodiment of the present invention (2 / 2) [Figure 3] A diagram showing the operating characteristics of the solenoid valve in the first embodiment of the present invention. [Figure 4]Functional block diagram of the controller in the first embodiment of the present invention [Figure 5] This figure shows a method for correcting the solenoid valve target command pressure in the first embodiment of the present invention. [Figure 6] Flowchart showing the computation process related to the controller learning process in the first embodiment of the present invention [Figure 7] A flowchart showing the calculation process related to the control of the hydraulic drive system of the controller in the first embodiment of the present invention. [Figure 8] Functional block diagram of the controller in the second embodiment of the present invention [Figure 9] A flowchart showing the computation process related to the controller learning process in the second embodiment of the present invention. [Figure 10] A flowchart showing the calculation process related to the control of the hydraulic drive system of the controller in the second embodiment of the present invention. [Figure 11] Circuit diagram showing a part of the hydraulic drive system in the third embodiment of the present invention. [Figure 12] Functional block diagram of the controller in the third embodiment of the present invention [Modes for carrying out the invention]

[0011] Hereinafter, a hydraulic excavator will be used as an example of a working machine according to the embodiment of the present invention, and will be described with reference to the drawings. In each figure, equivalent components will be denoted by the same reference numeral, and redundant explanations will be omitted as appropriate.

[0012] Figure 1 is a side view of a hydraulic excavator according to this embodiment. The hydraulic excavator 901 comprises a traveling body 201, a slewing body 202 that is rotatably positioned on the traveling body 201 and constitutes the vehicle body, and a front working device 203 (working device) that is rotatably attached to the slewing body 202 in the vertical direction and performs excavation work of earth and sand, etc. The slewing body 202 is driven by a slewing motor 211, which is a hydraulic actuator.

[0013] The working device 203 includes a boom 204 rotatably attached to the revolving body 202 in the vertical direction, an arm 205 rotatably attached to the tip of the boom 204 in the vertical direction, a bucket 206 rotatably attached to the tip of the arm 205 in the vertical direction, a boom cylinder 204a which is a hydraulic actuator for driving the boom 204, an arm cylinder 205a which is a hydraulic actuator for driving the arm 205, and a bucket cylinder 206a which is a hydraulic actuator for driving the bucket 206. In the working device 203, operation detection devices 212, 213, 214 for detecting the operations (such as posture and speed) of the boom 204, the arm 205, and the bucket 206 are installed. On the revolving body 202, operation detection devices 215, 216 for detecting the posture and rotation speed of the revolving body 202 are installed. As the operation detection devices 212 to 216, various sensors such as an inertial measurement unit (IMU), an inclination sensor, a rotation angle sensor, a stroke sensor, and an acceleration sensor can be used.

[0014] A cab 207 is provided at the front position on the revolving body 202, and a counterweight 209 for ensuring weight balance with the working device 203 is attached at the rear position. A machine room 208 is provided between the cab 207 and the counterweight 209. The machine room 208 houses an engine (not shown), hydraulic pumps 1, 2, 3 (shown in FIG. 2A), a swing motor 211, a control valve 210, etc. The control valve 210 controls the flow of hydraulic oil from the hydraulic pumps to each hydraulic actuator.

[0015] The hydraulic excavator 901 according to the present embodiment is equipped with a hydraulic drive system described in the following examples.

Examples

[0016] Figures 2A and 2B are circuit diagrams of a hydraulic drive system in a first embodiment of the present invention. The hydraulic drive system 902 comprises three main hydraulic pumps (for example, a first hydraulic pump 1, a second hydraulic pump 2, and a third hydraulic pump 3, each consisting of a variable displacement hydraulic pump), a pilot pump 111, and a hydraulic oil tank 4 that supplies oil to the hydraulic pumps 1-3 and the pilot pump 111. The hydraulic pumps 1-3 and the pilot pump 111 are driven by an engine (not shown).

[0017] The tilt angle of the first hydraulic pump 1 is controlled by a regulator attached to the first hydraulic pump 1. The regulator of the first hydraulic pump 1 has a flow control command pressure port 1a and is driven by the command pressure acting on the flow control command pressure port 1a. The tilt angle of the second hydraulic pump 2 is controlled by a regulator attached to the second hydraulic pump 2. The regulator of the second hydraulic pump 2 has a flow control command pressure port 2a and is driven by the command pressure acting on the flow control command pressure port 2a. The tilt angle of the third hydraulic pump 3 is controlled by a regulator attached to the third hydraulic pump 3. The regulator of the third hydraulic pump 3 has a flow control command pressure port 3a and is driven by the command pressure acting on the flow control command pressure port 3a.

[0018] The discharge line 41 of the first hydraulic pump 1 is connected to the hydraulic fluid tank 4 via the center bypass line 42. The center bypass line 42 has, in order from upstream, a right travel directional control valve 6, a bucket directional control valve 7, a second arm directional control valve 8, and a first boom directional control valve 9. The right travel directional control valve 6 controls the flow of pressurized oil supplied from the first hydraulic pump 1 to the right travel motor (not shown) of a pair of travel motors that drive the travel body 201. The bucket directional control valve 7 controls the flow of pressurized oil supplied from the first hydraulic pump 1 to the bucket cylinder 206a. The second arm directional control valve 8 controls the flow of pressurized oil supplied from the first hydraulic pump 1 to the arm cylinder 205a. The first boom directional control valve 9 controls the flow of pressurized oil supplied from the first hydraulic pump 1 to the boom cylinder 204a.

[0019] The supply ports for the bucket directional control valve 7, the second arm directional control valve 8, and the first boom directional control valve 9 are connected in parallel to a parallel line 43 that branches off from the section of the center bypass line 42 connecting the right-travel directional control valve 6 and the bucket directional control valve 7, via oil passages 44, 45, 46, 47, and 48, 49, respectively. A bleed-off valve 34 is located at the downstream end of the center bypass line 42 to control the flow of pressurized oil discharged from the center bypass line 42 to the hydraulic oil tank 4. The discharge line 41 is connected to the hydraulic oil tank 4 via an oil passage 50. A relief valve 31 is provided in the oil passage 50 to protect the circuit from excessive pressure rise.

[0020] The discharge line 51 of the second hydraulic pump 2 is connected to the hydraulic oil tank 4 via the center bypass line 52. The center bypass line 52 has, in order from upstream, a second boom directional control valve 10, a first arm directional control valve 11, a first attachment directional control valve 12, and a left travel directional control valve 13. The second boom directional control valve 10 controls the flow of pressurized oil supplied from the second hydraulic pump 2 to the boom cylinder 204a. The first arm directional control valve 11 controls the flow of pressurized oil supplied from the second hydraulic pump 2 to the arm cylinder 205a. The first attachment directional control valve 12 controls the flow of pressurized oil supplied from the second hydraulic pump 2 to a first actuator (not shown) that drives a first special attachment, such as a crusher installed in place of a bucket 206. The left travel directional control valve 13 controls the flow of pressurized oil supplied from the second hydraulic pump 2 to a left travel motor (not shown) of a pair of travel motors that drive the travel body 201.

[0021] The supply ports for the second boom directional control valve 10, the first arm directional control valve 11, the first attachment directional control valve 12, and the left-travel directional control valve 13 are connected in parallel to the parallel line 53 branching from the discharge line 51 via oil passages 54, 55, 56, 57, 58, 59, and 60, respectively. A bleed-off valve 35 is located at the downstream end of the center bypass line 52 to control the flow of pressurized oil discharged from the center bypass line 52 to the hydraulic oil tank 4. The parallel line 53 and the discharge line 41 of the first hydraulic pump are connected via an oil passage 69, and a merging valve 37 is installed in the oil passage 69. Downstream of the merging valve 37 in the oil passage 69, a check valve 38 is grounded to prevent backflow from the parallel line 53 to the discharge line 41. A check valve 39 is installed at the downstream end of the parallel line 53 to prevent backflow from the discharge line 41 to the parallel line 53. The parallel line 53 is connected to the hydraulic fluid tank 4 via an oil passage 61. The oil passage 61 is equipped with a relief valve 32 to protect the circuit from excessive pressure rise.

[0022] The discharge line 62 of the third hydraulic pump 3 is connected to the hydraulic fluid tank 4 via the center bypass line 63. The center bypass line 63 has, in order from upstream, a slewing directional control valve 14, a third boom directional control valve 15, and a second attachment directional control valve 16. The slewing directional control valve 14 controls the flow of pressurized oil supplied from the third hydraulic pump 3 to the slewing motor 211. The third boom directional control valve 15 controls the flow of pressurized oil supplied from the third hydraulic pump 3 to the boom cylinder 204a. The second attachment directional control valve 16 controls the flow of pressurized oil supplied from the third hydraulic pump 3 to the second actuator when a second special attachment equipped with a second actuator in addition to the first special attachment is installed, or when a second special attachment equipped with two actuators, a first actuator and a second actuator, is installed in place of the first special actuator.

[0023] The supply ports for the slewing directional control valve 14, the third boom directional control valve 15, and the second attachment directional control valve 16 are connected in parallel to a parallel line 64 branching from the discharge line 62 via oil passages 65, 66, 67, 68, and 69, 70, respectively. A bleed-off valve 36 is located at the downstream end of the center bypass line 63 to control the flow of pressurized oil discharged from the center bypass line 63 to the hydraulic oil tank 4. The parallel line 64 is connected to the hydraulic oil tank 4 via an oil passage 71. A relief valve 33 is provided in the oil passage 71 to protect the circuit from excessive pressure rise.

[0024] Flow control valves 21, 22, and 23 are provided in the oil passages 44 and 45 connected to the supply port of the bucket directional control valve 7, the oil passages 46 and 47 connected to the supply port of the second arm directional control valve 8, and the oil passages 48 and 49 connected to the supply port of the first boom directional control valve 9, respectively, to control the flow rate of pressurized oil supplied from the first hydraulic pump 1 to each directional control valve during combined operation. Flow control valves 24, 25, and 26 are provided in the oil passages 54 and 55 connected to the supply port of the second boom directional control valve 10, the oil passages 56 and 57 connected to the supply port of the first arm directional control valve 11, and the oil passages 58 and 59 connected to the supply port of the first attachment directional control valve 12, respectively, to control the flow rate of pressurized oil supplied from the second hydraulic pump 2 to each directional control valve during combined operation. Flow control valves 27, 28, and 29 are provided in the oil passages 65 and 66 connected to the supply port of the slewing directional control valve 14, the oil passages 67 and 68 connected to the supply port of the third boom directional control valve 15, and the oil passages 69 and 70 connected to the supply port of the second attachment directional control valve 16, respectively, to control the flow rate of pressurized oil supplied from the third hydraulic pump 3 to each directional control valve during combined operation.

[0025] The flow control valve 25, which controls the flow rate of pressurized oil supplied to the first arm directional control valve 11, has a seat-type main valve 25a that forms an auxiliary variable throttle, a control variable throttle 25b provided on the main valve 25a that changes its opening area according to the amount of movement of the main valve 25a, and a pilot variable throttle 25i. The housing in which the main valve 25a is housed has a first pressure chamber 25c formed at the connection between the main valve 25a and the oil passage 56, a second pressure chamber 25d formed at the connection between the main valve 25a and the oil passage 57, and a third pressure chamber 25e that communicates with the first pressure chamber 25c via an oil passage 25g and a control variable throttle 25b provided inside the main valve 25a. A check valve 25f for preventing backflow is provided in the oil passage 25g. The pilot variable throttle 25i is located in the oil passage 25h that connects the third pressure chamber 31e and the oil passage 57. For the sake of simplicity, some diagrams have been omitted, but the flow control valves 21-29, as well as the surrounding equipment, piping, and wiring, all have the same configuration.

[0026] The hydraulic oil tank 4 is equipped with a temperature sensor 91 for detecting the temperature of the hydraulic oil inside the tank 4. The discharge line 41 of the first hydraulic pump 1 is equipped with a pressure sensor 84 for detecting the discharge pressure of the first hydraulic pump 1. The discharge line 51 is equipped with a pressure sensor 85 for detecting the discharge pressure of the second hydraulic pump 2. The discharge line 62 of the third hydraulic pump 3 is equipped with a pressure sensor 86 for detecting the discharge pressure of the third hydraulic pump 3.

[0027] An actuator line 72a connecting the bottom side of the boom cylinder 204a to the boom directional control valves 9, 10, and 15 is provided with a pressure sensor 87a for detecting the actuator pressure on the bottom side of the boom cylinder 204a. An actuator line 72b connecting the rod side of the boom cylinder 204a to the boom directional control valves 9, 10, and 15 is provided with a pressure sensor 87b for detecting the actuator pressure on the rod side of the boom cylinder 204a. An actuator line 73a connecting the bottom side of the arm cylinder 205a to the arm directional control valves 8 and 11 is provided with a pressure sensor 88a for detecting the actuator pressure on the bottom side of the arm cylinder 205a. An actuator line 73b connecting the rod side of the arm cylinder 205a to the arm directional control valves 8 and 11 is provided with a pressure sensor 88b for detecting the actuator pressure on the rod side of the arm cylinder 205a. An actuator line 74a connecting the bottom side of the bucket cylinder 206a to the bucket directional control valve 7 is provided with a pressure sensor 89a for detecting the actuator pressure on the bottom side of the bucket cylinder 206a. An actuator line 74b connecting the rod side of the bucket cylinder 206a to the bucket directional control valve 7 is provided with a pressure sensor 89b for detecting the actuator pressure on the rod side of the bucket cylinder 206a. An actuator line 75a, 75b connecting the swing motor 211 to the swing directional control valve 14 is provided with pressure sensors 90a, 90b for detecting the actuator pressure of the swing motor 211. For the sake of simplicity, the left travel motor (not shown), the right travel motor (not shown), and the pressure sensors for detecting the actuator pressure of the actuators driving the attachments (not shown) are not shown.

[0028] In Figure 2B, the discharge port of the pilot pump 111 is connected to the hydraulic oil tank 4 via a pilot relief valve 112 for generating pilot primary pressure, and is also connected via a pilot line 121 to one input port of solenoid valves 113a to 113e built into the solenoid valve unit 113. The other input port of solenoid valves 113a to 113e is connected to the hydraulic oil tank 4 via a tank line 122. Each of the solenoid valves 113a to 113e reduces the pilot primary pressure according to the control current from the controller 114, which will be described later, and outputs it as a command pressure.

[0029] The output port of solenoid valve 113a is connected to the flow control command pressure port 2a of the regulator of the second hydraulic pump 2 via pilot line 123. The output ports of solenoid valves 113b and 113c are connected to the command pressure ports 11a and 11b of the directional control valve 11 for the first arm via pilot lines 124 and 125. The output port of solenoid valve 113d is connected to the command pressure port 25j of the flow control valve 25 via pilot line 126. The output port of solenoid valve 113e is connected to the command pressure port 35a of the bleed-off valve 35 via pilot line 127. For the sake of simplicity, the solenoid valves for hydraulic pumps 1 and 3, directional control valves 6-10 and 12-16, flow control valves 21-24 and 26-29, bleed-off valves 34 and 36, and the junction valve 37 are not shown in the diagram.

[0030] A temperature sensor 92 is provided on pilot line 121 to detect the temperature of the hydraulic fluid flowing through pilot line 121. A pressure sensor 133 is provided on pilot line 123 to detect the flow control command pressure of the second hydraulic pump 2 output from solenoid valve 113a. Pressure sensors 134 and 135 are provided on pilot lines 124 and 125 to detect the command pressure of the first arm directional control valve 11 output from solenoid valves 113b and 113c. A pressure sensor 136 is provided on pilot line 126 to detect the command pressure of the flow control valve 25 output from solenoid valve 113d. A pressure sensor 137 is provided on pilot line 127 to detect the command pressure of the bleed-off valve 35 output from solenoid valve 113e. Pressure sensors are also provided on pilot lines connected to the output ports of solenoid valves, which are not shown in the diagram. In this embodiment, the pressure sensors 133 to 137 constitute a control quantity measuring device that measures the control quantity (command pressure) of the solenoid valves 113a to 113e, which are controlled elements of the hydraulic drive system 902.

[0031] Current sensors 143-147 are provided on the signal line connecting the controller 114 and the solenoid valves 113a-113e as electrical signal detection devices to detect the electrical control signal (control current) output from the controller 114 to the solenoid valves 113a-113e. Similarly, current sensors are also provided on the signal line connecting the solenoid valves (not shown) and the controller 114. Furthermore, if the electrical control signal output from the controller 114 to the solenoid valves 113a-113e is a control voltage, voltage sensors are provided instead of current sensors 143-147.

[0032] The hydraulic drive system 902 includes a boom operating lever 115a (operating device) capable of switching the first boom directional control valve 9, the second boom directional control valve 10, and the third boom directional control valve 15, and an arm operating lever 115b (operating device) capable of switching the first arm directional control valve 11 and the second arm directional control valve 8. For the sake of simplicity, the right travel operating lever for switching the right travel directional control valve 6, the bucket operating lever for switching the bucket directional control valve 7, the first attachment operating lever for switching the first attachment directional control valve 12, the left travel operating lever for switching the left travel directional control valve 13, the slewing operating lever for switching the slewing directional control valve 14, and the second attachment operating lever for switching the second attachment directional control valve 16 are not shown in the illustration.

[0033] The hydraulic drive system 902 includes a controller 114 that controls solenoid valves 113a to 113e, a target surface setting device 116 that sets the target surface when the bucket 206 excavates, and an additional learning instruction device 117 that instructs the controller 114 to perform additional learning (described later). The controller 114 outputs control currents for the solenoid valves 113a to 113e in accordance with operation signals from operating devices 115a and 115b, target surface data from the target surface setting device 116, additional learning instructions from the additional learning instruction device 117, output values ​​from pressure sensors 84 to 86, 87a, 87b, 88a, 88b, 89a, 89b, 90a, 90b, and 133 to 137, output values ​​from temperature sensors 91 and 92, output values ​​from current sensors 143 to 147, and attitude data from motion detection devices 212 to 216.

[0034] Figure 3 shows the operating characteristics of solenoid valves 113a to 113e. Solenoid valves 113a to 113e reduce the pilot primary pressure in accordance with the solenoid valve control current from the controller 114 and output it as the solenoid valve command pressure. The solenoid valve control current is determined according to the solenoid valve target command pressure. At this time, a difference occurs between the solenoid valve target command pressure and the solenoid valve command pressure that is actually generated. This difference is not dependent on the passage of time and is therefore an error in the static characteristics of solenoid valves 113a to 113e. Furthermore, when the solenoid valve target command pressure is changed, there is a delay between the change in the target command pressure and the change in the actually generated solenoid valve command pressure. This delay is dependent on the passage of time and is therefore an error in the dynamic characteristics of solenoid valves 113a to 113e. These errors are caused by various factors such as individual variations in solenoid valves 113a to 113e, changes in the operating environment, and deterioration of solenoid valves 113a to 113e, or by a combination of these factors.

[0035] Figure 4 is a functional block diagram of the controller 114 in the first embodiment. The controller 114 includes an actuator target speed calculation unit 114a, an actuator target flow rate calculation unit 114b, a solenoid valve target command pressure calculation unit 114c, a solenoid valve target command pressure correction unit 114d, a solenoid valve control current output unit 114e, a target operation detection unit 114f, an operating environment detection unit 114g, an operating state detection unit 114h, a learning data storage unit 114i, a deterioration state detection unit 114j, a learning unit 114k, a prediction unit 114l, and a prediction accuracy evaluation unit 114m.

[0036] The actuator target speed calculation unit 114a calculates the actuator target speed required for the operation of the front work machine 203 based on the operation signals from the operating devices 115a and 115b, the attitude data from the motion detection devices 212 to 216, and the target surface data from the target surface setting device 116. The actuator target flow rate calculation unit 114b calculates the actuator target flow rate based on the actuator target speed calculated by the actuator target speed calculation unit 114a.

[0037] The solenoid valve target command pressure calculation unit 114c calculates the solenoid valve target command pressure based on the actuator target flow rate from the actuator target flow rate calculation unit 114b. In this embodiment, the solenoid valve target command pressure calculation unit 114c is a target control amount calculation unit that calculates the target control amount (solenoid valve target command pressure) of the controlled elements (solenoid valves 113a to 113e) of the hydraulic drive system 902 based on the operating amounts of the operating devices 115a and 115b. The solenoid valve target command pressure correction unit 114d corrects the solenoid valve target command pressure from the solenoid valve target command pressure calculation unit 114c based on the predicted solenoid valve command pressure from the prediction unit 114l. The solenoid valve target command pressure correction unit 114d also switches whether or not to correct the solenoid valve target command pressure in response to an additional learning instruction from the additional learning instruction device 117.

[0038] Figure 5 shows a method for correcting the target command pressure of a solenoid valve. In the hydraulic drive system 902, when an electromagnetic source control current corresponding to the target command pressure shown by the dashed line in the figure is output, it is expected that the actual command pressure will rise without significant lag behind the target command pressure, as shown by the solid line in the figure, and will match the target command pressure after stabilization. However, depending on the operating state of the hydraulic drive system 902, it is predicted that the actual command pressure will rise with a significant lag behind the target command pressure, as shown by the dashed line in the figure, and will be lower than the target command pressure after stabilization. Therefore, the target command pressure is corrected based on the predicted command pressure, as shown by the dotted line in the figure, so that the actual command pressure approaches the target command pressure (i.e., the rise of the actual command pressure approaches the rise of the target command pressure, and the actual command pressure after stabilization approaches the target command pressure before correction). This makes it possible to bring the actual command pressure closer to the target command pressure before correction, regardless of fluctuations in the operating state of the hydraulic drive system 902 (individual variations in solenoid valves 113a to 113e, operating environment disturbances, performance degradation, etc.).

[0039] Returning to Figure 4, the solenoid valve control current output unit 114e generates and outputs a solenoid valve control current based on the solenoid valve target command pressure from the solenoid valve target command pressure correction unit 114d. In this embodiment, the solenoid valve target command pressure correction unit 114d constitutes a control signal correction unit that corrects the control signal to the hydraulic drive system 902 (control current to solenoid valves 113a to 113e).

[0040] The target operation detection unit 114f generates the solenoid valve target command pressure from the solenoid valve target command pressure calculation unit 114c as target operation data for solenoid valves 113a to 113e. The operating environment detection unit 114g generates operating environment data based on the output values ​​of temperature sensors 91 and 92. In this embodiment, the operating environment data is generated based on the temperature of the hydraulic fluid, but the operating environment data may also be generated based on the temperature of the cooling water circulating in the hydraulic drive system 902 or the ambient temperature. The operating state detection unit 114h generates operating state data for solenoid valves 113a to 113e based on the output values ​​of current sensors 143 to 147 and pressure sensors 133 to 137. In this embodiment, the target operation detection unit 114f, the operating environment detection unit 114g, the operating state detection unit 114h, and the deterioration state detection unit 114j constitute an operating state detection unit that detects the operating state of the hydraulic drive system 902.

[0041] The learning data storage unit 114i receives degradation state data from the target operation data (i.e., solenoid valve target command pressure) from the target operation detection unit 114f, the operating environment data from the operating environment detection unit 114g, and the operating state data from the operating state detection unit 114h, and stores it as learning data. The degradation state detection unit 114j analyzes the target operation data, operating environment data, and operating state data stored in the learning data storage unit 114i and generates data (degradation state data) indicating the degradation state of the solenoid valves 113a to 113e. The learning data storage unit 114i also stores the degradation state data from the degradation state detection unit 114j as learning data. The learning unit 114k receives evaluation results from the prediction accuracy evaluation unit 114m (described later), performs machine learning processing based on the learning data in the learning data storage unit 114i, and generates an estimator that predicts the behavior of the solenoid valve command pressure in relation to the solenoid valve target command pressure that would be generated by the solenoid valves 113a to 113e when the solenoid valve control current corresponding to the solenoid valve target command pressure is output from the solenoid valve target command pressure calculation unit 114c.

[0042] The prediction unit 114l uses an estimator generated by the learning unit 114k to calculate a predicted value of the solenoid valve command pressure (predicted command pressure) based on the solenoid valve target command pressure calculation unit 114c. The prediction accuracy evaluation unit 114m evaluates the prediction accuracy of the solenoid valve command pressure in the prediction unit 114l by comparing the predicted command pressure from the prediction unit 114l with the command pressure measured by the pressure sensors 133-137, for example, when an additional learning instruction is input from the additional learning instruction device 117.

[0043] Figure 6 is a flowchart showing the computation process related to the learning process of the controller 114. Each step will be explained in order below.

[0044] First, the controller 114 determines whether a specific timing (for example, immediately after the start of operation of the hydraulic excavator 901) has occurred, or whether a certain amount of time has elapsed (step S101). In parallel with step S101, the controller 114 also determines whether there is an additional learning instruction from the additional learning instruction device 117 (step S102). If the result of the determinations in steps S101 and S102 is NO, the flow is terminated.

[0045] If the result of step S101 or S102 is YES, the controller 114 outputs the solenoid valve control current without correcting the solenoid valve target command pressure using the solenoid valve target command pressure correction unit 114d (step S103).

[0046] Following step S103, the controller 114 calculates the degree of discrepancy (prediction error) between the predicted solenoid valve command pressure from the prediction unit 114l and the solenoid valve command pressure measured by the pressure sensors 133-137 in the prediction accuracy evaluation unit 114m (step S406), and determines whether the prediction error exceeds a preset tolerance (step S105). If the result of the determination in step S105 is NO, the flow is terminated.

[0047] If the result of step S105 is YES, the controller 114 generates the solenoid valve target command pressure from the solenoid valve target command pressure calculation unit 114c as target operation data in the target operation detection unit 114f, generates operating environment data based on the output value of the temperature sensor 91 in the operating environment detection unit 114g, generates operating state data based on the output values ​​of the current sensors 143~147 and pressure sensors 84~86, 87a, 87b, 88a, 88b, 89a, 89b, 90a, 90b, 133~137 in the operating state detection unit 114h, and generates deterioration state data in the deterioration state detection unit 114j (step S106).

[0048] Following step S106, the controller 114 stores target motion data from the target motion detection unit 114f, operating environment data from the operating environment detection unit 114g, operating state data from the operating state detection unit 114h, and degradation state data from the degradation state detection unit 114j in the learning data storage unit 114i (step S107).

[0049] Following step S107, the controller 114 performs machine learning processing in the learning unit 114k based on the learning data in the learning data storage unit 114i, generates an estimator that predicts the behavior of the solenoid valve command pressure based on the results of the machine learning processing (step S108), and terminates the flow.

[0050] Figure 7 is a flowchart showing the calculation process related to the control of the hydraulic drive system 902 by the controller 114. Each step will be explained in order below.

[0051] First, the controller 114 determines whether or not there is input from the operating devices 115a and 115b (step S201). If the result of the determination in step S201 is NO, the flow is terminated.

[0052] If the result of step S201 is YES, the controller 114 calculates the actuator target speed using the actuator target speed calculation unit 114a based on the operation signal, attitude data, target surface data, etc. (step S202).

[0053] Following step S202, the controller 114 calculates the target actuator flow rate based on the target actuator speed using the actuator target flow rate calculation unit 114b (step S203).

[0054] Following step S203, the controller 114 calculates the solenoid valve target command pressure based on the actuator target flow rate using the solenoid valve target command pressure calculation unit 114c (step S204).

[0055] Following step S204, the controller 114 calculates the predicted command pressure for the solenoid valve relative to the target command pressure using the estimator generated by the learning unit 114k in the prediction unit 114l (S205).

[0056] Following step S205, the controller 114 corrects the solenoid valve target command pressure in the solenoid valve target command pressure correction unit 114d so that the predicted solenoid valve command pressure from the prediction unit 114l approaches the solenoid valve target command pressure calculated by the solenoid valve target command pressure calculation unit 114c (step S206).

[0057] Following step S206, the controller 114 outputs a solenoid valve control current corresponding to the corrected solenoid valve target command pressure at the solenoid valve control current output unit 114e (step S207), and the flow ends.

[0058] (operation) The operation of the hydraulic drive system 902 is described below. In response to operator input, control signals are input to the controller 114 from the control devices 115a and 115b. The controller 114 calculates the solenoid valve target command pressure based on the input control signals, etc., and outputs a solenoid valve control current corresponding to the solenoid valve target command pressure. As a result, a command pressure corresponding to the solenoid valve control current is generated in the solenoid valve, and the actuator is driven.

[0059] While the actuator is driven, a predicted value of the solenoid valve command pressure (predicted solenoid valve command pressure) is calculated, which is the solenoid valve control current that would be generated if the solenoid valve control current corresponding to the solenoid valve target command pressure were output. The solenoid valve target command pressure is then corrected so that the difference between the predicted solenoid valve command pressure and the target solenoid valve command pressure based on the operation signal, etc., is minimized. This allows the command pressure generated by the solenoid valve to be brought closer to the solenoid valve target command pressure before correction.

[0060] Furthermore, at specific timings, after a certain period of time has elapsed, or when an additional learning instruction is input, the solenoid valve is driven without applying any correction to the solenoid valve target command pressure. The solenoid valve command pressure obtained at that time is compared with the solenoid valve command pressure predicted by the prediction unit 114l (solenoid valve predicted command pressure), and the prediction accuracy of the solenoid valve command pressure in the prediction unit 114l is evaluated. If it is determined that the prediction accuracy has decreased, target operation data, operating environment data, operating state data, and deterioration state data (learning data) are stored in the learning data storage unit 114i, and additional learning is performed in the learning unit based on the stored learning data. This makes it possible to maintain the prediction accuracy of the solenoid valve command pressure in the prediction unit 114l.

[0061] (summary) In the first embodiment, a work machine 901 comprises a vehicle body 202, a work device 203 attached to the vehicle body 202, hydraulic actuators 204a, 205a, 206a, 211 that drive the work device 203, a hydraulic drive system 902 that drives the hydraulic actuators 204a, 205a, 206a, 211, operating devices 115a, 115b that instruct the operation of the hydraulic actuators 204a, 205a, 206a, 211, and a control device 114 that outputs a control signal to the hydraulic drive system 902 according to the amount of operation of the operating devices 115a, 115b. The control device 114 includes a target control amount calculation unit that calculates the target control amount of the controlled elements 113a to 113e of the hydraulic drive system 902 based on the amount of operation of the operating devices 115a, 115b. The system includes 114c, operating state detection units 114f, 114g, 114h, and 114j for detecting the operating state of the hydraulic drive system 902, a learning data storage unit 114i for storing the target control amount and the detection results of the operating state detection units 114f, 114g, 114h, and 114j as learning data, a learning unit 114k for machine learning the control amounts of the controlled elements 113a to 113e with respect to the target control amount based on the learning data stored in the learning data storage unit 114i, a prediction unit 114l for calculating a predicted control amount, which is a predicted value of the control amount with respect to the target control amount, using the learning results of the machine learning in the learning unit 114k, and a control signal correction unit 114d for correcting the control signal so that the predicted control amount approaches the target control amount.

[0062] According to the first embodiment configured as described above, a predicted control amount for the target control amount of the controlled elements 113a to 113e is calculated using the learning results of machine learning, which uses the target control amount of the controlled elements 113a to 113e and the operating state of the hydraulic drive system 902 as learning data. The control signal to the hydraulic drive system 902 is then corrected so that the predicted control amount approaches the target control amount. As a result, regardless of fluctuations in the operating state of the hydraulic drive system 902 (individual variations in the controlled elements 113a to 113e, operating environment disturbances, performance degradation, etc.), the relationship between the control amount of the controlled elements 113a to 113e and the operation input is kept constant, thereby improving the control robustness of the work machine 901.

[0063] Furthermore, the operating state detection units 114f, 114g, 114h, and 114j in the first embodiment include an operating state detection unit 114h for detecting the operating state of the hydraulic drive system 902, an operating environment detection unit 114g for detecting the operating environment of the hydraulic drive system 902, and a deterioration state detection unit 114j for detecting the deterioration state of the hydraulic drive system 902. This makes it possible to detect the operating state, operating environment, and deterioration state of the hydraulic drive system 902 as the operating state of the hydraulic drive system 902.

[0064] Furthermore, the work machine 901 in the first embodiment is equipped with motion detection devices 212 to 216 that detect the operation of the vehicle body 202, the work device 203, or the hydraulic actuators 204a, 205a, 206a, and 211, and the operation state detection unit 114h detects the operation state of the hydraulic drive system 902 based on the detection results of the motion detection devices 212 to 216. This makes it possible to detect the operation state of the hydraulic drive system 902 based on the operation of the hydraulic actuators 204a, 205a, 206a, and 211.

[0065] Furthermore, the work machine 901 in the first embodiment is equipped with pressure sensors 84-86, 87a, 87b, 88a, 88b, 89a, 89b, 90a, 90b, and 133-137 for detecting the pressure of the hydraulic fluid circulating within the hydraulic drive system 902. The operating state detection unit 114h detects the operating state of the hydraulic drive system 902 based on the detection results of the pressure sensors 84-86, 87a, 87b, 88a, 88b, 89a, 89b, 90a, 90b, and 133-137. This makes it possible to detect the operating state of the hydraulic drive system 902 based on the pressure of the hydraulic fluid circulating within the hydraulic drive system 902.

[0066] Furthermore, the work machine 901 in the first embodiment is equipped with electrical signal detection devices 143 to 147 that detect the current or voltage of the electrical control signal output from the control device 114 to the hydraulic drive system 902, and the operating state detection unit 114h detects the operating state of the hydraulic drive system 902 based on the detection results of the electrical signal detection devices 143 to 147. This makes it possible to detect the operating state of the hydraulic drive system 902 based on the electrical control signal output from the control device 114 to the hydraulic drive system 902.

[0067] Furthermore, the working machine 901 in the first embodiment is equipped with temperature sensors 91 and 92 that detect the temperature of the hydraulic fluid or coolant circulating within the hydraulic drive system 902, or the ambient temperature. The operating environment detection unit 114g detects the operating environment of the hydraulic drive system 902 based on the detection results of the temperature sensors 91 and 92. This makes it possible to detect the operating state of the hydraulic drive system 902 based on the temperature of the hydraulic fluid or coolant circulating within the hydraulic drive system 902, or the ambient temperature.

[0068] Furthermore, the deterioration state detection unit 114j in the first embodiment detects the deterioration state of the hydraulic drive system 902 based on the detection results of the target control amount, the operating state detection unit 114h, and the operating environment detection unit 114g. This makes it possible to detect the deterioration state of the hydraulic drive system 902 based on the target control amount, operating state, and operating environment of the controlled elements 113a to 113e of the hydraulic drive system 902.

[0069] Furthermore, the work machine 901 in the first embodiment includes control amount measuring devices 133 to 137 for measuring the control amounts of the controlled elements 113a to 113e, and the control device 114 includes a prediction accuracy evaluation unit 114m that determines whether the prediction error, which is the degree of deviation between the measured value of the control amount and the predicted control amount when the control signal correction unit 114d does not correct the control signal, exceeds a predetermined tolerance error, and the learning unit 114k executes the machine learning when the prediction accuracy evaluation unit 114m determines that the prediction error exceeds the tolerance error. This makes it possible to execute machine learning when the prediction accuracy of the solenoid valve command pressure in the prediction unit 114l decreases.

[0070] Furthermore, the work machine 901 in the first embodiment is equipped with an instruction device 117 that instructs the execution of the machine learning, and the prediction accuracy evaluation unit 114m receives instructions from the instruction device 117 and determines whether the prediction error exceeds a predetermined tolerance. This makes it possible to evaluate the prediction accuracy of the solenoid valve command pressure in the prediction unit 114l in accordance with the operator's instructions.

[0071] Furthermore, the working machine 901 in the first embodiment includes hydraulic pumps 1 to 3 and control valves 6 to 16, 21 to 29 that control the flow of pressurized oil supplied from hydraulic pumps 1 to 3 to hydraulic actuators 204a, 205a, 206a, and 211. The controlled elements 113a to 113e are solenoid valves 113a to 113e that generate the command pressure for control valves 6 to 16, 21 to 29, and the target control amount is the target command pressure for solenoid valves 113a to 113e. As a result, the relationship between the command pressure of solenoid valves 113a to 113e and the operation input is kept constant regardless of fluctuations in the operating state of the hydraulic drive system 902. [Examples]

[0072] A second embodiment of the present invention will be described, focusing on the differences from the first embodiment.

[0073] Figure 8 is a functional block diagram of the controller 114A in the second embodiment. The controller 114A includes an actuator target speed calculation unit 114Aa, an actuator target speed correction unit 114An, an actuator target flow rate calculation unit 114Ab, a solenoid valve target command pressure calculation unit 114Ac, a solenoid valve control current output unit 114Ae, a target operation detection unit 114Af, an operating environment detection unit 114Ag, an operating state detection unit 114Ah, a learning data storage unit 114Ai, a deterioration state detection unit 114Aj, a learning unit 114Ak, a prediction unit 114Al, and a prediction accuracy evaluation unit 114Am.

[0074] The actuator target speed calculation unit 114Aa calculates the actuator target speed required for the operation of the front work implement 203 based on the operation signals from the operating devices 115a and 115b, the attitude data from the motion detection devices 212 to 216, and the target surface data from the target surface setting device 116. In this embodiment, the actuator target speed calculation unit 114Aa constitutes a target control amount calculation unit that calculates the target control amount (actuator target speed) of the controlled element (hydraulic actuator) of the hydraulic drive system 902 based on the operation amounts of the operating devices 115a and 115b. Furthermore, in this embodiment, the motion detection devices 212 to 216 constitute a control amount measuring device that measures the control amount (actuator speed) of the hydraulic actuators 204a, 205a, 206a, and 211.

[0075] The actuator target speed correction unit 114An corrects the actuator target speed calculated by the actuator target speed calculation unit 114Aa based on the actuator predicted speed from the prediction unit 114Al. The actuator target speed correction unit 114An also switches whether or not to correct the actuator target speed in response to an additional learning instruction from the additional learning instruction device 117. In this embodiment, the actuator target speed correction unit 114An constitutes a control signal correction unit that corrects the control signal to the hydraulic drive system 902 (control current to solenoid valves 113a to 113e).

[0076] The actuator target flow rate calculation unit 114Ab calculates the actuator target flow rate based on the actuator target speed from the actuator target speed correction unit 114An. The solenoid valve target command pressure calculation unit 114Ac calculates the solenoid valve target command based on the actuator target flow rate from the actuator target flow rate calculation unit 114Ab. The solenoid valve control current output unit 114Ae generates a control current corresponding to the target control command from the solenoid valve target command pressure calculation unit 114Ac and outputs it to the solenoid valves 113a to 113e.

[0077] The target motion detection unit 114Af generates the actuator target speed from the actuator target speed calculation unit 114Aa as the actuator target motion data. The operating environment detection unit 114Ag generates operating environment data based on the output value of the temperature sensor 91. The operating state detection unit 114Ah generates actuator operating state data based on the output values ​​of the current sensors 143~147 and the pressure sensors 84~86, 87a, 87b, 88a, 88b, 89a, 89b, 90a, 90b, 133~137, as well as attitude data.

[0078] The learning data storage unit 114Ai receives target motion data (i.e., actuator target speed) from the target motion detection unit 114Af, operating environment data from the operating environment detection unit 114Ag, and operating state data from the operating state detection unit 114Ah, and stores them as learning data. The deterioration state detection unit 114Aj analyzes the target motion data, operating environment data, and operating state data stored in the learning data storage unit 114Ai and generates data indicating the deterioration state of the hydraulic drive system (deterioration state data). The learning data storage unit 114Ai also stores the deterioration state data from the deterioration state detection unit 114Aj as learning data. The learning unit 114Ak receives the evaluation results (described later) from the prediction accuracy evaluation unit 114Am, performs machine learning processing based on the learning data in the learning data storage unit 114Ai, and generates an estimator that predicts the behavior of the actuator speed that would be obtained if the solenoid valve control current corresponding to the actuator target speed from the actuator target speed calculation unit 114Aa were output, based on the results of the machine learning processing.

[0079] The prediction unit 114Al uses an estimator generated by the learning unit 114Ak to calculate a predicted value of the actuator speed (actuator predicted speed) based on the actuator target speed calculated from the actuator target speed calculation unit 114Aa. The prediction accuracy evaluation unit 114Am evaluates the prediction accuracy of the actuator speed in the prediction unit 114Al by comparing the actuator predicted speed from the prediction unit 114Al with the actuator speed calculated from the attitude data, for example, when an additional learning instruction is input from the additional learning instruction device 117. In this embodiment, the system is configured to predict the behavior of the actuator speed, but it may also be configured to predict the angular velocity or rotational speed of the actuator.

[0080] Figure 9 is a flowchart showing the computation process related to the learning process of the controller 114A in the second embodiment. Each step will be explained in order below.

[0081] First, the controller 114A determines whether a specific timing (for example, immediately after the start of operation of the hydraulic excavator 901) has occurred, or whether a certain amount of time has elapsed (step S101A). In parallel with step S101A, the controller 114A also determines whether there is an additional learning instruction from the additional learning instruction device 117 (step S102A). If the determination results in steps S101A and S102A are NO, the flow is terminated.

[0082] If the result of step S101A or S102A is YES, the controller 114A outputs the solenoid valve control current without correcting the actuator target speed at the actuator target speed correction unit 114An (step S103A).

[0083] Following step S103A, the controller 114A calculates the degree of discrepancy (prediction error) between the actuator prediction speed from the prediction unit 114Al and the actuator speed calculated from the attitude data using the prediction accuracy evaluation unit 114Am (step S104A), and determines whether the prediction error exceeds a preset tolerance (step S105A). If the result of the determination in step S105A is NO, the flow is terminated.

[0084] If the result of step S105A is YES, the controller 114A generates the target actuator speed from the actuator target speed calculation unit 114Aa as target motion data in the target motion detection unit 114Af, generates operating environment data based on the output values ​​of temperature sensors 91 and 92 in the operating environment detection unit 114Ag, generates operating state data based on the output values ​​of current sensors 143 to 147 and pressure sensors 84 to 86, 87a, 87b, 88a, 88b, 89a, 89b, 90a, 90b, 133 to 137 and attitude data in the operating state detection unit 114Ah, and generates deterioration state data in the deterioration state detection unit 114Aj (step S106A).

[0085] Following step S106A, the controller 114A stores target motion data from the target motion detection unit 114Af, operating environment data from the operating environment detection unit 114Ag, operating state data from the operating state detection unit 114Ah, and degradation state data from the degradation state detection unit 114Aj in the learning data storage unit 114i (step S107A).

[0086] Following step S107A, the controller 114A performs machine learning processing in the learning unit 114Ak based on the learning data in the learning data storage unit 114Ai, generates an estimator that predicts the behavior of the actuator speed based on the results of the machine learning processing (step S108A), and terminates the flow.

[0087] Figure 10 is a flowchart showing the calculation process related to the control of the hydraulic drive system 902 of the controller 114A in the second embodiment. Each step will be described in order below.

[0088] First, the controller 114A determines whether or not there is input from the operating devices 115a and 115b (step S201A). If the result of the determination in step S201A is NO, the flow is terminated.

[0089] If the result of step S201A is YES, the controller 114A calculates the actuator target speed in the actuator target speed calculation unit 114Aa based on the operation signal, attitude data, target plane data, etc. (step S202A).

[0090] Following step S202A, the controller 114A calculates the predicted actuator speed relative to the target actuator speed using the estimator generated by the learning unit 114Ak in the prediction unit 114Al (S203A).

[0091] Following step S203A, the controller 114A corrects the actuator target speed in the actuator target speed correction unit 114An so that the actuator predicted speed from the prediction unit 114Al approaches the actuator target speed calculated by the actuator target speed calculation unit 114Aa (step S204A).

[0092] Following step S204A, the controller 114A calculates the actuator target flow rate based on the corrected actuator target speed using the actuator target flow rate calculation unit 114b (step S205A).

[0093] Following step S205A, the controller 114A calculates the solenoid valve target command pressure based on the actuator target flow rate using the solenoid valve target command pressure calculation unit 114c (step S206A).

[0094] Following step S206A, the controller 114A outputs a solenoid valve control current corresponding to the solenoid valve target command pressure at the solenoid valve control current output unit 114Ae (step S207A), and the flow ends.

[0095] (operation) The operation of the hydraulic drive system 902 is described below. In response to operator input, control signals are input to the controller 114 from the control devices 115a and 115b. The controller 114 calculates the target actuator speed based on the input control signals and outputs a solenoid valve control current corresponding to the target actuator speed. As a result, a command pressure corresponding to the solenoid valve control current is generated in the solenoid valve, and the actuator is driven.

[0096] While the actuator is driven, a predicted value of the actuator speed (actuator predicted speed) is calculated, which is the value of the actuator speed that would be obtained if the solenoid valve control current corresponding to the actuator target speed were output. The actuator target speed is then corrected so that the difference between the actuator target speed (based on the operation signal, etc.) and the actuator predicted speed is minimized. This allows the actuator speed to be brought closer to the actuator target speed before correction.

[0097] Furthermore, at specific timings, after a certain period of time has elapsed, or when an additional learning instruction is input, the solenoid valve is driven without applying any correction to the actuator target speed. The actuator speed obtained at that time is compared with the actuator speed predicted by the prediction unit 114Al (actuator predicted speed), and the prediction accuracy of the actuator speed in the prediction unit 114Al is evaluated. If it is determined that the prediction accuracy has decreased, target operation data, operating environment data, operating state data, and degradation state data (learning data) are stored in the learning data storage unit 114Ai, and additional learning is performed in the learning unit based on the stored learning data. This makes it possible to maintain the prediction accuracy of the actuator speed in the prediction unit 114Al.

[0098] (summary) In the second embodiment, the controlled elements 204a, 205a, 206a, and 211 of the hydraulic drive system 902 are the hydraulic actuators 204a, 205a, 206a, and 211, and the target control quantities of the controlled elements 204a, 205a, 206a, and 211 are the target speed, target angular velocity, or target rotational speed of the hydraulic actuators 204a, 205a, 206a, and 211.

[0099] According to the second embodiment configured as described above, regardless of fluctuations in the operating state of the hydraulic drive system 902 (individual variations in hydraulic actuators 204a, 205a, 206a, 211, operating environment disturbances, performance degradation, etc.), the relationship between the speed, angular velocity, or rotational speed of the hydraulic actuators 204a, 205a, 206a, 211 and the operation input is kept constant, thereby improving the control robustness of the work machine 901. [Examples]

[0100] A third embodiment of the present invention will be described, focusing on the differences from the first embodiment.

[0101] Figure 11 is a circuit diagram showing a part of the hydraulic drive system 902B in the third embodiment. The hydraulic drive system 902B includes a degradation-related information input device 118 that inputs information related to the degradation of the hydraulic drive system 902B (degradation-related information) to the controller 114B, instead of the temperature sensor 92 (shown in Figure 2B) provided on the pilot line 121. Degradation-related information includes, for example, information on oxidation of the hydraulic fluid, precipitates, and degradation of components, and can be obtained from an external system with more advanced analytical capabilities than the controller 114B.

[0102] Figure 12 is a functional block diagram of the controller 114B in the third embodiment. Degradation-related information from the degradation-related information input device 118 is input to the degradation state detection unit 114Bj. The degradation state detection unit 114Bj analyzes the target operation data, operating environment data, operation state data, and degradation-related information stored in the learning data storage unit 114i and generates degradation state data.

[0103] (summary) In the third embodiment, the work machine 901 is equipped with a deterioration-related information input device 118 that inputs deterioration-related information, which is information related to the deterioration of the hydraulic drive system 902B, to the controller 114B. The deterioration state detection unit 114Bj detects the deterioration state of the hydraulic drive system 902B based on the target control amount of the controlled elements 204a, 205a, 206a, and 211, the detection results of the operating state detection unit 114h and the operating environment detection unit 114g, and the deterioration-related information.

[0104] According to the third embodiment configured as described above, it becomes unnecessary for the controller 114B, which controls the hydraulic drive system 902B, to perform tasks such as storing large amounts of long-term operational data necessary for determining the deterioration state of the hydraulic drive system 902B, and to perform computationally intensive advanced calculations for diagnosing the deterioration state. As a result, the computational load on the controller 114B can be reduced while improving the accuracy of determining the deterioration state.

[0105] Although embodiments of the present invention have been described in detail above, the present invention is not limited to the embodiments described above and includes various modifications. For example, the embodiments described above are described in detail in order to explain the present invention in an easy-to-understand manner and are not necessarily limited to those having all the configurations described. Furthermore, it is possible to add parts of the configuration of one embodiment to the configuration of another embodiment, and it is also possible to delete parts of the configuration of one embodiment or replace parts of parts of another embodiment. [Explanation of symbols]

[0106] 1…First hydraulic pump, 1a…Flow control command pressure port, 2…Second hydraulic pump, 2a…Flow control command pressure port, 3…Third hydraulic pump, 3a…Flow control command pressure port, 4…Hydraulic oil tank, 6…Right travel directional control valve, 7…Bucket directional control valve, 8…Second arm directional control valve, 9…First boom directional control valve, 10…Second boom directional control valve, 11…First arm directional control valve, 11a,11b…Command pressure ports, 12…First attachment directional control valve, 13…Left travel directional control valve, 14…Slewing directional control valve, 15…Third boom directional control valve, 16…Second attachment Directional control valve for treadmills, 21-25...flow control valve, 25a...main valve, 25c...first pressure chamber, 25d...second pressure chamber, 25e...third pressure chamber, 25f...check valve, 25g,25h...oil passage, 25j...command pressure port, 26-29...flow control valve, 31...relief valve, 31e...third pressure chamber, 32...relief valve, 33...relief valve, 34,35...bleed-off valve, 35a...command pressure port, 36...bleed-off valve, 37...junction valve, 38,39...check valve, 41...discharge line, 42...center bypass line, 43...parallel line, 44-50...oil passage, 51...discharge line, 52...center bypass line Center bypass line, 53…Parallel line, 54~61…Oil passage, 62…Discharge line, 63…Center bypass line, 64…Parallel line, 65~71…Oil passage, 72a,72b,73a,73b,74a,74b,75a,75b…Actuator line, 84~86,87a,87b,88a,88b,89a,89b,90a,90b…Pressure sensor, 91,92…Temperature sensor, 111…Pilot pump, 112…Pilot relief valve, 113…Solenoid valve unit, 113a~113e…Solenoid valve (controlled element), 114,114A,114B …Controller (control device), 114a…Actuator target speed calculation unit, 114b,114Ab…Actuator target flow rate calculation unit, 114c…Solenoid valve target command pressure calculation unit (target control amount calculation unit), 114d…Solenoid valve target command pressure correction unit (control signal correction unit), 114e,114Ae…Solenoid valve control current output unit, 114f,114Af…Target operation detection unit (operating state detection unit), 114g,114Ag…Operating environment detection unit (operating state detection unit), 114h,114Ah…Operating state detection unit (operating state detection unit), 114i,114Ai…Learning data storage unit, 114j,114Aj,114Bj...Deterioration state detection unit (operation state detection unit), 114k,114Ak...Learning unit, 114l,114Al...Prediction unit, 114m,114Am...Prediction accuracy evaluation unit, 114Aa...Actuator target speed calculation unit (target control amount calculation unit), 114Ac...Solenoid valve target command pressure calculation unit, 114An...Actuator target speed correction unit (control signal correction unit), 115a...Boom operation lever (operating device), 115b...Arm operation lever (operating device), 116...Target surface setting device, 117...Additional learning instruction device, 118...Deterioration-related information input device, 121...Pilot line, 122...Tank line, 123~127...Pilot line, 133~137...Pressure sensor (control amount) Measuring device), 143-147...Current sensor (electrical signal detection device), 201...Travel body, 202...Slewing body (vehicle body), 203...Front work implement (working device), 204...Boom, 204a...Boom cylinder (hydraulic actuator), 205...Arm, 205a...Arm cylinder (hydraulic actuator), 206...Bucket, 206a...Bucket cylinder (hydraulic actuator), 207...Operator's cab, 208...Machine room, 209...Counterweight, 210...Control valve, 211...Slewing motor (hydraulic actuator), 212-216...Motion detection device (control amount measurement device), 901...Hydraulic excavator (working machine), 902, 902B...Hydraulic drive system.

Claims

1. The car body and, A work device attached to the vehicle body, A hydraulic actuator that drives the aforementioned work device, A hydraulic drive system for driving the aforementioned hydraulic actuator, An operating device for instructing the operation of the hydraulic actuator, A work machine comprising a control device that outputs a control signal to the hydraulic drive system according to the amount of operation of the operating device, The control device is A target control amount calculation unit that calculates a target control amount for the controlled element of the hydraulic drive system based on the operating amount of the operating device, An operating state detection unit for detecting the operating state of the hydraulic drive system, A learning data storage unit stores the target control amount and the detection result of the operating state detection unit as learning data, A learning unit that uses the learning data stored in the learning data storage unit to machine-learn the control amount of the controlled element with respect to the target control amount, A prediction unit calculates a predicted control amount, which is a predicted value of the control amount relative to the target control amount, using the learning results of the machine learning in the learning unit. It includes a control signal correction unit that corrects the control signal so that the predicted control amount approaches the target control amount. A work machine characterized by the following features.

2. In the work machine described in claim 1, The aforementioned operating state detection unit is An operating state detection unit for detecting the operating state of the hydraulic drive system, An operating environment detection unit for detecting the operating environment of the hydraulic drive system, It includes a deterioration state detection unit for detecting the deterioration state of the hydraulic drive system. A work machine characterized by the following features.

3. In the work machine described in claim 2, The vehicle body, the work device, or the hydraulic actuator is equipped with an operation detection device for detecting the operation of the hydraulic actuator. The operating state detection unit detects the operating state of the hydraulic drive system based on the detection results of the operating state detection device. A work machine characterized by the following features.

4. In the work machine described in claim 2, The hydraulic drive system includes a pressure sensor that detects the pressure of the hydraulic fluid circulating within the system. The operating state detection unit detects the operating state of the hydraulic drive system based on the detection result of the pressure sensor. A work machine characterized by the following features.

5. In the work machine described in claim 2, The system includes an electrical signal detection device that detects the current or voltage of the electrical control signal output from the control device to the hydraulic drive system, The operating state detection unit detects the operating state of the hydraulic drive system based on the detection result of the electrical signal detection device. A work machine characterized by the following features.

6. In the work machine described in claim 2, The hydraulic drive system is equipped with a temperature sensor that detects the temperature of the hydraulic fluid or cooling water circulating within the system, or the ambient temperature. The operating environment detection unit detects the operating environment of the hydraulic drive system based on the detection result of the temperature sensor. A work machine characterized by the following features.

7. In the work machine described in claim 2, The deterioration state detection unit detects the deterioration state of the hydraulic drive system based on the target control amount, the operating state detection unit, and the operating environment detection unit's detection results. A work machine characterized by the following features.

8. In the work machine described in claim 2, The system includes a deterioration-related information input device that inputs deterioration-related information, which is information related to the deterioration of the hydraulic drive system, to the control device. The deterioration state detection unit detects the deterioration state of the hydraulic drive system based on the target control amount, the detection results of the operating state detection unit and the operating environment detection unit, and the deterioration-related information. A work machine characterized by the following features.

9. In the work machine described in claim 1, The system includes a control quantity measuring device for measuring the control quantity of the controlled element, The control device includes a prediction accuracy evaluation unit that determines whether the prediction error, which is the degree of deviation between the measured value of the control amount and the predicted control amount when the control signal correction unit does not correct the control signal, exceeds a predetermined tolerance error. The learning unit executes the machine learning when the prediction accuracy evaluation unit determines that the prediction error exceeds the tolerance error. A work machine characterized by the following features.

10. In the work machine described in claim 9, The system includes an instruction device that instructs the execution of the aforementioned machine learning, The prediction accuracy evaluation unit, upon receiving an instruction from the instruction device, determines whether the prediction error exceeds a predetermined tolerance. A work machine characterized by the following features.

11. In the work machine described in claim 1, Hydraulic pump and The system includes a control valve that controls the flow of pressurized oil supplied from the hydraulic pump to the hydraulic actuator, The controlled element is a solenoid valve that generates the command pressure for the control valve. The target control amount is the target command pressure of the solenoid valve. A work machine characterized by the following features.

12. In the work machine described in claim 1, The controlled element is the hydraulic actuator, The target control amount is the target speed, target angular velocity, or target rotational speed of the hydraulic actuator. A work machine characterized by the following features.

Images