Self-adaptive frozen soil excavation device based on iterative learning admittance control algorithm

Abstract

The invention discloses a self-adaptive frozen soil excavation device based on an iterative learning admittance control algorithm, the device is composed of a high-frequency breaking hammer, a transport bucket and an excavator mechanical arm, the high-frequency breaking hammer is fixedly connected to bucket teeth of the transport bucket, the transport bucket is connected with the excavator mechanical arm, and the excavator mechanical arm is connected with the transport bucket. The self-adaptive frozen soil excavation device carries out excavation control through an admittance control method based on iterative learning. According to the excavation device, high-frequency crushing and transfer cleaning of frozen soil are combined, excavation and transfer of the frozen soil can be completed at a time, excavation work of the frozen soil can be completed more efficiently, the excavation device is controlled based on an iterative learning admittance control algorithm, and the excavation efficiency is improved. The device can ensure the stable adaptability of the excavation device to the changing environment through learning and self-adaption, so that the requirement for rapid excavation of frozen soil is met, and the automation and intelligence level of frozen soil excavation is effectively improved.

Description

Technical Field

[0001] The invention relates to a frozen soil excavation device, in particular to an adaptive frozen soil excavation device based on an iterative learning admittance control algorithm. Background Art

[0002] As more projects are being built in cold regions, including housing, water conservancy, electricity, roadbeds, and oil and gas pipelines, these projects are inseparable from the excavation of frozen soil. During the excavation of frozen soil, the crushing speed is slow due to the high strength of frozen soil. In addition, changes in temperature, moisture content, and freeze-thaw cycles make the mechanical properties of frozen soil more complex, thereby increasing the difficulty of crushing. These factors significantly affect the efficiency of frozen soil excavation and increase the cost of engineering construction.

[0003] At present, the commonly used methods of crushing frozen soil mainly include melting method, blasting method and mechanical method. The melting method is to melt the frozen soil by heating, thereby reducing the strength and facilitating excavation, but it is mainly suitable for small-scale and shallow excavation projects; the blasting method is to use the instantaneous energy of explosives in the explosion hole to break the frozen soil, which is suitable for thicker and harder frozen soil layers and has higher efficiency, but it has greater safety hazards, and the relevant procedures are difficult to handle and there are more restrictions on its use; the mechanical method mainly includes mechanical cutting and impact cutting, which has the characteristics of good stability, high work efficiency and wide application range. However, the traditional excavator loading mechanism needs to cooperate with other excavation equipment (such as high-frequency breaker) to complete excavation and transportation, and the construction efficiency is low. Therefore, there is a lack of an integrated equipment for frozen soil excavation and transportation. In the field of automatic excavation control, a large number of AI-based methods use artificial neural networks that require a large amount of data to train. Without training, they may show unpredictable behavior or poor performance. Therefore, there is a lack of adaptive control algorithms for improving the excavator bucket loading capacity under a small amount of sample data.

[0004] In summary, developing an adaptive frozen soil excavation device based on iterative learning admittance control algorithm is an urgent problem to be solved in the industry. Summary of the invention

[0005] In view of the above problems existing in the background technology, the present invention provides an adaptive frozen soil excavation device based on an iterative learning admittance control algorithm. The excavation device combines high-frequency crushing and transportation cleaning of frozen soil, can complete the excavation and transportation of frozen soil at one time, can complete the frozen soil excavation operation more efficiently, and uses an iterative learning admittance control algorithm to control the excavation device, so that the device can ensure the stable adaptability of the excavation device to the changing environment through learning and adaptation, thereby meeting the needs of rapid excavation of frozen soil, and effectively improving the automation and intelligence level of frozen soil excavation.

[0006] The objective of the present invention is achieved through the following technical solutions:

[0007] An adaptive frozen soil excavation device based on iterative learning admittance control algorithm is composed of a high-frequency breaker, a transport bucket and an excavator mechanical arm, wherein:

[0008] The high-frequency breaker is fixedly connected to the bucket teeth of the transport bucket;

[0009] The transport bucket is connected to the excavator mechanical arm;

[0010] The excavator mechanical arm is composed of a bucket connecting rod, an arm connecting rod, a bucket cylinder, an arm, an arm cylinder, a boom and an arm cylinder, on which a control, sensor and drive device are installed;

[0011] The control, sensing and driving device includes a pressure sensor, a position sensor, a control system, a load weighing system and an electro-hydraulic servo valve;

[0012] The pressure sensor is installed on the head and rod side of the boom cylinder and the movable arm cylinder to measure the hydraulic pressure;

[0013] The position sensor is installed in the bucket cylinder to measure its extension length;

[0014] The control system is installed in the excavator operator's cabin and is used to read sensors, send actuator commands and record data;

[0015] The load weighing system is installed at the boom joint and uses static load analysis to calculate the bucket payload weight;

[0016] The electro-hydraulic servo valve is used to control the fluid flow to the bucket cylinder, the boom cylinder and the boom cylinder so that the valve command corresponds to the near-linear response of the cylinder;

[0017] The adaptive frozen soil excavation device performs excavation control by an admittance control method based on iterative learning;

[0018] The admittance control method based on iterative learning includes autonomous admittance control and iterative learning control;

[0019] The autonomous admittance control measures the interaction force between the robot and the environment and quickly adjusts the reference speed of the robot to cope with unknown and changing environments;

[0020] The iterative learning control automatically updates the control parameters according to the error between the actual bucket weight and the target bucket weight during each excavation process, so that the bucket has a higher load each time.

[0021] Compared with the prior art, the present invention has the following advantages:

[0022] 1. The excavation device of the present invention can simultaneously complete the high-frequency crushing and transportation and cleaning of frozen soil, so that it can be completed separately on one device, thereby improving construction efficiency and reducing costs, and the device can also be widely used in other similar projects.

[0023] 2. The present invention uses an admittance control algorithm based on iterative learning to perform excavation control. This control method can utilize the error information of the previously executed task and automatically update the control parameters during the subsequent execution process, so that it is suitable for excavation materials with different characteristics. Compared with the AI-based method, this method can reach the target loading weight of the bucket within 2 to 3 excavation cycles, without the need for a large number of training samples and complex computing hardware. It is more practical and feasible, thereby improving the efficiency of frozen soil transportation.

[0024] 3. The iterative learning-based admittance control algorithm of the present invention can also be used in other human-machine collaborations. The adaptive control strategy adopted by the algorithm can quickly respond to environmental interaction forces and has potential application value in human-machine collaborative tasks. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 This is a schematic diagram of a frozen soil excavation device;

[0026] Figure 2 It is a schematic diagram of a high-frequency breaker device;

[0027] Figure 3 It is a schematic diagram of the transport bucket device;

[0028] Figure 4 Provide schematic diagrams for system and hardware layout;

[0029] Figure 5 Strategy diagram for autonomous mining admittance control algorithm. DETAILED DESCRIPTION

[0030] The technical solution of the present invention is further described below in conjunction with the accompanying drawings, but is not limited thereto. Any modification or equivalent replacement of the technical solution of the present invention without departing from the spirit and scope of the technical solution of the present invention should be included in the protection scope of the present invention.

[0031] The present invention provides an adaptive frozen soil excavation device based on an iterative learning admittance control algorithm, such as Figure 1 As shown, the device is composed of a high-frequency breaker 1, a transport bucket 2 and an excavator mechanical arm 3, wherein:

[0032] The high-frequency breaker 1 needs to be connected to the bucket teeth of the transport bucket 2 for fixing. The connection is achieved by bolts and can be fixed and removed as needed.

[0033] The transport bucket 2 needs to be connected to the excavator mechanical arm 3, so as to be installed on the excavator to excavate the frozen soil;

[0034] The device can realize the vibration excavation of frozen soil and the transfer of thawed soil at one time. The excavation function is mainly completed by the high-frequency breaker 1. After the vibration crushing is completed, the high-frequency breaker 1 assists the transport bucket 2 to transport the frozen soil. During the transportation, the control, sensing and driving devices installed on the excavator mechanical arm 3 will realize adaptive adjustment of the power of the transportation process.

[0035] Furthermore, if Figure 2 As shown, the high-frequency breaker 1 is mainly composed of a frame 11, an upper pull rod 12, a lower pull rod 13, a vibrating tooth 14, a vibrating knife row 15, a vibrating box 16, a hydraulic motor 17 and a connecting piece 18. The vibrating box 16 is arranged on the upper part of the vibrating knife row 15, the vibrating tooth 14 is installed at the head end of the vibrating knife row 15, and upper pull rods 12 and lower pull rods 13 are arranged on both sides of the vibrating box 16. One end of the upper pull rod 12 is hinged to the upper part of the vibrating box 16, and the other end of the upper pull rod 12 is hinged to the inner wall of the frame 11, one end of the lower pull rod 13 is hinged to the lower part of the vibrating box 16, and the other end of the lower pull rod 13 is hinged to the inner wall of the frame 11. The inner wall of the frame 11 is hinged, and the hydraulic motor 17 is connected to the eccentric gear assembly inside the vibration box 16. A connecting piece 18 connected to the excavator mechanical arm 3 is arranged above the frame 11. When starting to work, the hydraulic motor 17 will use the high-pressure oil of the host hydraulic system as the power source, so that the eccentric gear assembly inside the vibration box 16 rotates at a high speed, thereby generating an eccentric force, the processing component of which is a periodically changing interference force, which causes the shaft to generate a high-frequency exciting force of radial forced vibration, and then the exciting box 16 drives the vibrating cutter row 15 and the vibrating teeth 14 to vibrate, thereby impacting and crushing the frozen soil structure.

[0036] Furthermore, if Figure 3 As shown, the transport bucket 2 is mainly composed of a box body 21, a wear-resistant plate 22, bucket teeth 23, side teeth 24, a connecting hole 25 and a rear groove 26. A wear-resistant plate 22 is provided at the bottom of the box body 21, and the box body 21 and the wear-resistant plate 22 are connected together by welding. The side teeth 24 are fixedly connected to the left and right sides of the box body 21 for auxiliary excavation. The bucket teeth 23 are fixedly connected to the front end of the box body 21 to facilitate the bucket to pick up the excavated frozen soil. A connecting hole 25 is provided at the front end of the box body 21. The high-frequency breaker 1 is fixedly connected to the middle of the front end of the box body 21 through the connecting hole 25 and the connecting piece 18 to ensure that sufficient pressure is applied to the high-frequency breaker 1 when it is working. A rear groove 26 for placing a hydraulic pipe is provided on the box body 21. The hydraulic pipe connects the main engine hydraulic system with the hydraulic motor 17 of the high-frequency breaker 1 to ensure the delivery of high-pressure oil and thus provide sufficient crushing force.

[0037] like Figure 4As shown, the excavator mechanical arm 3 is mainly composed of a bucket connecting rod 31, an arm connecting rod 32, a bucket cylinder 33, an arm 34, an arm cylinder 35, a boom 36 and an arm cylinder 37, on which necessary control, sensing and driving devices are installed. The control, sensing and driving devices are installed at different positions of the excavator mechanical arm, and mainly include pressure sensors, position sensors, control systems, load weighing systems and electro-hydraulic servo valves. The pressure sensor is installed at the head and rod side of the arm cylinder and the boom cylinder to measure the hydraulic pressure. The pressure measurement value P and the cross-sectional area A of the cylinder head and the connecting rod are used to estimate the interaction force. Where P h Refers to the pressure measurement value of the cylinder head, A h Refers to the cross-sectional area of ​​the cylinder head, P r Refers to the pressure measurement value of the cylinder rod, A r Refers to the cross-sectional area of ​​the cylinder rod. The position sensor refers to a wire potentiometer, which is installed in the bucket cylinder to measure its extension length. The control system is installed in the excavator operator's cab and runs in real time at a frequency of 20Hz to read sensors, send actuator commands and record data. The load weighing system uses static load analysis to calculate the bucket payload weight when the excavator's mechanical arm is lifted to a specific angle, and uses an absolute encoder to measure it at the boom joint; the electro-hydraulic servo valve uses a servo valve to control the fluid flow to the bucket cylinder, boom cylinder and boom cylinder, so that the valve command corresponds to the near-linear response of the cylinder.

[0038] Furthermore, after the deployment of relevant systems and hardware, it is necessary to complete the iterative learning-based admittance control method to realize excavation control. The iterative learning-based admittance control method includes autonomous admittance control and iterative learning control. The autonomous admittance control measures the interaction force between the manipulator and the environment and quickly adjusts the reference speed of the manipulator to cope with unknown and changing environments. The iterative learning control (ILC) is based on the idea of ​​improving system performance by learning the error information in the previous execution processes. Specifically, the algorithm automatically updates the control parameters according to the error between the actual bucket weight and the target bucket weight during each excavation process, so that the bucket has a higher load each time.

[0039] like Figure 5 As shown in Figure 1, autonomous admittance control is implemented in the actuator space of the excavator, which consists of the throttle, bucket cylinder, boom cylinder, and boom cylinder. During the excavation process, the throttle actuator provides a constant input. Control the traction force of the bucket cylinder, where u t Refers to the displacement applied to the throttle, It means that the displacement applied to the throttle is a constant amount, while the boom cylinder and the boom cylinder are not started, and there is ul =0, that is, the arm cylinder and the boom cylinder do not produce displacement, that is, the pressure sensor installed at the arm cylinder and the boom cylinder can reflect the interaction force between the bucket and the frozen soil This value can be used as feedback for the bucket cylinder admittance controller. The admittance transfer function of the bucket cylinder can be expressed as: H(s) = v f (s) / f e (s), where f e The force error is expressed by To calculate, that is, use the measured interaction force Subtract the set target force f d Then the speed change v is calculated using the admittance controller H(s) f , and compare it with the reference speed v d Combined to generate a new speed command v c =v d +v f , and then control the bucket cylinder to run at the new speed command. The target force f d This can be done by specifying a constant admittance transfer function H(s) = K A >0, constant reference speed And the measured force signal In one embodiment, the bucket cylinder is driven at a constant nominal rate. Extension, when When the rate v f Increase, which will make the bucket take soil faster. When the rate v f This will make the bucket pick up soil more slowly. Filtering refers to the use of a windowed median filter to filter the measured force signal. Filter to obtain the target force f d , which is achieved by giving the discrete time of the control system. In the time step k, the target force filter is defined as:

[0040]

[0041] Here, N>0 refers to the size of the window.

[0042] In this control method, there are four parameters that need to be adjusted: K A 、N、 and v d , where K A refers to the constant admittance transfer function value, N refers to the window size, The displacement applied to the throttle is a constant, v d Refers to the reference speed, where K AThe higher the value, the higher the speed command to the bucket cylinder, so it is necessary to adjust K A To minimize the saturation of the actuator, the window size N needs to be adjusted so that f d Follow Avoid higher N resulting in target force f d These two parameters have no significant effect on the filling of the bucket, only on the level of response to the changing forces. and the reference speed v of the bucket cylinder d The largest influence is on the filling of the bucket. Therefore, these two parameters must be learned and adapted to the excavation conditions, which is mainly achieved through iterative learning control (ILC).

[0043] To apply iterative learning control to autonomous mining, we first define the task error e for each mining channel j: j and system input u j To this end, the autonomous excavation task is considered as a tracking problem, where the system aims to track the target bucket filling weight W during each excavation process. d , so the error of the task is defined as the actual bucket filling weight W j and target bucket filling weight W d :

[0044] e j =W j -W d

[0045] Under this execution tracking task system, the system can be modeled in step j as:

[0046] y j =Gu j +d

[0047] Among them, y j The subject that performs the tracking task; j refers to the input condition; d refers to the unmodeled disturbance.

[0048] In order to reduce the task error e j , ILC passes the error information of one pass through e j Merge into the subsequent control input u j+1 The simplest ILC formula is the proportional learning function:

[0049] u j+1 =u j +Ke j

[0050] Where K is the learning gain function.

[0051] Since the reference speed input of the bucket cylinder and throttle input The biggest influence on the filling of the bucket. Therefore, based on equation u j+1 =u j +Ke j Assuming two proportional learning functions, two input parameters are iteratively updated according to the bucket filling weight error:

[0052]

[0053] This formula introduces two learning gain coefficients α and β, which must be properly designed to achieve stable learning behavior. It is possible to make e→0 and j→∞, in which case the dynamic iteration can be described as:

[0054] y j+1 -y j =G(u j+1 -u j )

[0055] Therefore, we can get:

[0056] y j+1 =y j +G(Ke j )

[0057] Next, subtract the target trajectory y from both sides of the equation d , and the dynamic error of the system is obtained:

[0058] e j+1 =(1+GK)e j

[0059] Therefore, for iteration stability:

[0060] |1+GK|<1

[0061] If the system model G is known, K can be designed to make the error asymptotically stable. If the system model is unknown, K must be adjusted by trial and error on the actual system to achieve a stable learning process.

[0062] Therefore, this control method can utilize the error information of the previously executed tasks to continuously improve the performance in the subsequent execution process, and can achieve the target loading weight of the bucket within 2 to 3 excavation cycles, showing good stability and convergence.

Claims

1. An adaptive frozen soil excavation device based on iterative learning admittance control algorithm, characterized in that The device is composed of a high-frequency breaker, a transport bucket and an excavator mechanical arm, wherein: The high-frequency breaker is fixedly connected to the bucket teeth of the transport bucket; The transport bucket is connected to the excavator mechanical arm; The excavator mechanical arm is composed of a bucket connecting rod, an arm connecting rod, a bucket cylinder, an arm, an arm cylinder, a boom and an arm cylinder, on which a control, sensor and drive device are installed; The control, sensing and driving device includes a pressure sensor, a position sensor, a control system, a load weighing system and an electro-hydraulic servo valve; The pressure sensor is installed on the head and rod side of the boom cylinder and the movable arm cylinder to measure the hydraulic pressure; The position sensor is installed in the bucket cylinder to measure its extension length; The control system is installed in the excavator operator's cabin and is used to read sensors, send actuator commands and record data; The load weighing system is installed at the boom joint and uses static load analysis to calculate the bucket payload weight; The electro-hydraulic servo valve is used to control the fluid flow to the bucket cylinder, the boom cylinder and the arm cylinder so that the valve command corresponds to the near-linear response of the cylinder.

2. The adaptive frozen soil excavation device based on iterative learning admittance control algorithm according to claim 1 is characterized in that The high-frequency breaker consists of a frame, an upper pull rod, a lower pull rod, a vibrating tooth, a vibrating cutter row, a vibrating box, a hydraulic motor and connecting parts. The vibrating box is arranged on the upper part of the vibrating cutter row, and the vibrating tooth is installed at the head end of the vibrating cutter row. Upper pull rods and lower pull rods are arranged on both sides of the vibrating box. One end of the upper pull rod is hinged to the top of the vibrating box, and the other end of the upper pull rod is hinged to the inner wall of the frame. One end of the lower pull rod is hinged to the bottom of the vibrating box, and the other end of the lower pull rod is hinged to the inner wall of the frame. The hydraulic motor is connected to the eccentric gear assembly inside the vibrating box, and a connecting part connected to the excavator mechanical arm is arranged on the top of the frame.

3. The adaptive frozen soil excavation device based on iterative learning admittance control algorithm according to claim 1 is characterized in that The transport bucket consists of a box body, a wear-resistant plate, bucket teeth, side teeth, connecting holes and rear grooves. A wear-resistant plate is arranged at the bottom of the box body. The side teeth are fixedly connected to the left and right sides of the box body. The bucket teeth are fixedly connected to the front end of the box body. A connecting hole is arranged at the front end of the box body. The high-frequency breaker is fixedly connected to the middle of the front end of the box body through the connecting hole and the connecting piece. A rear groove for placing a hydraulic pipe is arranged on the box body. The hydraulic pipe connects the main engine hydraulic system with the hydraulic motor of the high-frequency breaker.

4. The adaptive frozen soil excavation device based on iterative learning admittance control algorithm according to claim 1 is characterized in that The adaptive frozen soil excavation device performs excavation control through an admittance control method based on iterative learning, and the admittance control method based on iterative learning includes autonomous admittance control and iterative learning control; the autonomous admittance control quickly adjusts the reference speed of the robotic arm by measuring the interaction force between the robotic arm and the environment to cope with unknown and changing environments; the iterative learning control automatically updates the control parameters according to the error between the actual bucket weight and the target bucket weight during each excavation process, so that the bucket has a higher load each time.

5. The adaptive frozen soil excavation device based on iterative learning admittance control algorithm according to claim 4 is characterized in that The autonomous admittance control is implemented in the actuator space of the excavator, which consists of a throttle, bucket cylinder, boom cylinder and boom cylinder. During the excavation process, a constant input is provided by the throttle actuator. Control the traction force of the bucket cylinder, where u t Refers to the displacement applied to the throttle, The displacement applied to the throttle is a constant amount, the arm cylinder and the boom cylinder are not started, and there is u l = 0, that is, the arm cylinder and the boom cylinder do not produce displacement, that is, the pressure sensor installed at the arm cylinder and the boom cylinder reflects the interaction force between the bucket and the frozen soil This value is used as the feedback of the bucket cylinder admittance controller. The admittance transfer function of the bucket cylinder is expressed as: H(s) = v f (s) / f e (s), where f e Expressing the force error, through the formula To calculate, that is, use the measured interaction force Subtract the set target force f d Then use the admittance controller H(s) to calculate the speed change v f , and compare it with the reference speed v d Combined to generate a new speed command v c =v d +v f , and then control the bucket cylinder to run at the new speed instruction.

6. The adaptive frozen soil excavation device based on iterative learning admittance control algorithm according to claim 5 is characterized in that The target force f d By specifying a constant admittance transfer function H(s) = K A >0, constant reference speed And the measured force signal Obtained by filtering.

7. The adaptive frozen soil excavation device based on iterative learning admittance control algorithm according to claim 6 is characterized in that The force signal measured Filtering refers to the use of a windowed median filter to filter the measured force signal. Filter to obtain the target force f d , which is achieved by giving the discrete time of the control system. In the time step k, the target force filter is defined as: Here, N>0 refers to the size of the window.

8. The adaptive frozen soil excavation device based on iterative learning admittance control algorithm according to claim 4 is characterized in that The iterative learning control is used for autonomous mining. First, the task error e of each mining channel j is defined. j and system input u j , the autonomous excavation task is considered as a tracking problem, where the system aims to track the target bucket filling weight W during each excavation process d , the error of the task is defined as the actual bucket filling weight W j and target bucket filling weight W d : have been j =W j -W d Under the execution tracking task system, in step j, this system is modeled as: the j =To j +d Among them, y j The subject that performs the tracking task; j refers to the input condition; d refers to the unmodeled disturbance; In order to reduce the task error e j , ILC passes the error information of one pass through e j Merge into the subsequent control input u j+1 middle: you j+1 =you j +I j Where K is the learning gain function; Based on the equation u j+1 =u j +Ke j Assuming two proportional learning functions, two input parameters are iteratively updated according to the bucket filling weight error: Where α and β are learning gain coefficients; in order to achieve stable learning behavior, e→0, j→∞, in this case, the dynamic iteration is described as: y j+1 -y j =G(u j+1 -u j ) Therefore, we can get: yes j+1 =y j +G(I j ) Next, subtract the target trajectory y from both sides of the equation d , and the dynamic error of the system is obtained: e j+1 =(1+GK)e j Therefore, for iterative stability: |1+GK|<1 If the system model G is known, K is designed to make the error asymptotically stable. If the system model is unknown, K is adjusted through repeated trials on the actual system to achieve a stable learning process.

Images