An ultrasonic vibration inner cooling and lubricating grinding system based on deep reinforcement learning

Abstract

The application discloses an ultrasonic vibration inner cooling and lubrication grinding system based on deep reinforcement learning, which comprises a PC terminal deep reinforcement learning system, a power distribution acquisition control system, an ultrasonic vibration platform system, a sensor measurement system, an inner cooling and lubrication grinding wheel system and a cooling and lubricating liquid supply system. The system realizes effective cooperation of vibration, cooling and lubrication in the ultrasonic vibration platform and the inner cooling and lubrication grinding wheel during processing by combining the ultrasonic vibration platform with the embedded micro heat pipe array and the piezoelectric ceramic gradient polarization array and the inner cooling and lubrication asymmetric fractal flow channel grinding wheel, and realizes efficient, high-quality and low-consumption processing of difficult-to-process materials by means of auxiliary multi-physical field sensing and AI deep reinforcement learning.

Description

Technical Field

[0001] This invention relates to the field of precision grinding technology, specifically to an ultrasonic vibration internal cooling and lubrication grinding system based on deep reinforcement learning. Background Technology

[0002] Ultrasonic vibration-assisted grinding can effectively reduce grinding force and improve the surface quality of ground materials. Currently, it is widely used in aerospace and other difficult-to-machine materials. The implementation methods of ultrasonic vibration-assisted grinding are mainly divided into workpiece ultrasonic vibration and tool ultrasonic vibration. However, existing ultrasonic grinding methods mostly rely on small-diameter grinding wheel vibration, which not only limits the application of large-size grinding wheels but also restricts the improvement of grinding speed and processing efficiency, failing to meet the basic requirements of ultra-high-speed grinding technology for complex irregular surface components. Compared with tool ultrasonic vibration grinding, workpiece ultrasonic vibration grinding has the advantages of high energy utilization and simple structure, making it very suitable for situations with large grinding forces and high material removal rates, thus having broad application prospects and good development potential. However, workpiece ultrasonic vibration grinding also has problems such as uneven vibration of the ultrasonic vibration platform, platform deformation due to heat generation during processing, and deviations in processing accuracy.

[0003] In the grinding process of difficult-to-machine materials, a large amount of frictional heat and grinding force are generated, thus ensuring effective cooling and lubrication. Traditional cast-type cooling and lubrication cannot effectively penetrate the grinding air barrier layer, resulting in poor cooling and lubrication in the grinding zone and a significant waste of cooling and lubricating fluid, increasing processing costs. Internally cooled and lubricated grinding wheels, on the other hand, can effectively cool and lubricate the internal grinding zone, significantly improving the surface finish. However, most currently available internally cooled and lubricated grinding wheels use a single vertical flow channel or symmetrical vertical flow channels. When the grinding wheel speed is too high, the cooling and lubricating fluid is thrown out at high speed under centrifugal force, leading to uneven cooling and lubrication in the internal grinding zone, and also resulting in significant waste of cooling and lubricating fluid.

[0004] The adjustment and coordination of process parameters are crucial in grinding. Currently, most grinding systems rely on experience to adjust process parameters, which is time-consuming, labor-intensive, and lacks precision. Some systems use feedback control to automatically adjust processing parameters, but feedback algorithms cannot meet the demands of complex multi-objective optimization control, resulting in limited accuracy. Existing ultrasonic vibration grinding systems operate independently, with vibration, cooling, and lubrication functions without coordination. Furthermore, they often use ordinary grinding wheels for ultrasonic vibration grinding without a coordinating internally cooled and lubricated grinding wheel, thus failing to further improve processing efficiency and quality.

[0005] In summary, a grinding system based on deep reinforcement learning, combining an ultrasonic vibration platform with an internally cooled and lubricated grinding wheel, is proposed. The existing problems of ultrasonic vibration platforms and internally cooled and lubricated grinding wheels are improved. At the same time, multi-physics perception and AI deep reinforcement learning are used to achieve effective synergy of vibration, cooling and lubrication during the processing of ultrasonic vibration platforms and internally cooled and lubricated grinding wheels, so as to achieve efficient, high-quality and low-consumption processing of difficult-to-machine materials. Summary of the Invention

[0006] To address the aforementioned issues, this invention proposes a grinding system combining an ultrasonic vibration platform with an internally cooled and lubricated grinding wheel based on deep reinforcement learning. The system utilizes an ultrasonic vibration platform equipped with an embedded micro heat pipe array and a piezoelectric ceramic gradient polarization array, along with an internally cooled and lubricated asymmetric fractal flow channel grinding wheel, for collaborative processing. Simultaneously, multi-physics sensing and AI deep reinforcement learning are employed to achieve effective synergy between vibration, cooling, and lubrication during the processing of the ultrasonic vibration platform and the internally cooled and lubricated grinding wheel. This enables efficient, high-quality, and low-consumption processing of difficult-to-machine materials.

[0007] The technical solution of the present invention is: an ultrasonic vibration internal cooling and lubrication grinding system based on deep reinforcement learning, characterized in that: it includes a PC-based deep reinforcement learning system, a power distribution acquisition and control system, an ultrasonic vibration platform system, a sensor measurement system, an internal cooling and lubrication grinding wheel system, and a cooling and lubrication fluid supply system.

[0008] The power distribution acquisition and control system has a built-in power supply unit, control unit, data acquisition unit, and ultrasonic generation unit. It is responsible for simultaneously supplying power and transmitting signals to the PC-based deep reinforcement learning system, ultrasonic vibration platform system, sensor measurement system, internal cooling and lubrication grinding wheel system, and cooling and lubrication fluid supply system.

[0009] The ultrasonic vibration platform system comprises an ultrasonic vibration platform and a workpiece. The ultrasonic vibration platform includes a transducer, amplitude transformer, vibration conversion module, support base, vibration platform, embedded micro heat pipe array, and piezoelectric ceramic gradient polarization array. The support base, transducer, amplitude transformer, vibration conversion module, and vibration platform are sequentially connected to each other. The workpiece is fixed on the vibration platform. The embedded micro heat pipe array is parallel to and embedded within the vibration platform. A specific amount of special low-melting-point alloy is filled inside the micro heat pipes to absorb heat through melting and prevent platform deformation. The piezoelectric ceramic gradient polarization array is installed at the bottom of the vibration platform in a rectangular array. The polarization intensity of the piezoelectric ceramic gradient polarization array in the edge region of the vibration platform is 20%-35% higher than that in the center region, solving the problem of insufficient vibration at the edges. The transducer is connected to a power distribution and acquisition control system. The power supply unit and the ultrasonic wave generation unit of the power distribution and acquisition control system respectively supply power to the ultrasonic vibration platform system and transmit ultrasonic waves.

[0010] The sensor measurement system includes a force gauge, a CCD camera, and a semiconductor laser. The force gauge measures the grinding force during the grinding process, and the signal collected by the force gauge is transmitted to the data acquisition unit of the power distribution acquisition and control system. The CCD camera and semiconductor laser are used in combination to measure the surface roughness during the grinding process, and the measured signal is also transmitted to the data acquisition unit of the power distribution acquisition and control system.

[0011] The internally cooled and lubricated grinding wheel system includes an internally cooled and lubricated grinding wheel and a machine tool spindle, with the internally cooled and lubricated grinding wheel mounted on the machine tool spindle.

[0012] The internally cooled and lubricated grinding wheel includes a grinding wheel holder, an electro-hydraulic slip ring, a grinding wheel base, and MEMS thin-film thermocouples. The electro-hydraulic slip ring is mounted on the grinding wheel holder, which is connected to the grinding wheel base. The MEMS thin-film thermocouples are evenly distributed inside the grinding wheel base near the grinding surface, and their terminals are connected to the electro-hydraulic slip ring. The grinding wheel base has a built-in asymmetric tree-shaped fractal flow channel. This fractal flow channel is a biomimetic tree structure with at least three levels of branches, and the branch angles are asymmetric to utilize centrifugal force to transport fluid. The branch angle range is (θ = 15°-75°), which can effectively and actively lubricate the grinding zone by utilizing the centrifugal transport of fluid during the high-speed rotation of the grinding wheel, while controlling and conserving the cooling lubricant. The inlet end of the electro-hydraulic slip ring is connected to the cooling lubricant supply system through a gas-liquid delivery pipe. The cooling lubricant supply system pumps the cooling lubricant through the internal flow channel of the electro-hydraulic slip ring, through the tree-shaped flow channel of the grinding wheel base, and then to the grinding zone. The power distribution acquisition and control system is connected to the input terminal of the electro-hydraulic slip ring to supply power to the electro-hydraulic slip ring. The current is transmitted to the MEMS thin-film thermocouple, which measures the temperature of the grinding surface. The measured temperature signal is then transmitted to the data acquisition unit of the power distribution acquisition and control system.

[0013] The coolant and lubricant supply system comprises two parts: coolant supply and lubricant supply. These two parts can be controlled independently or in tandem, and each is equipped with its own flow control valve. The coolant and lubricant supply are powered by the built-in power supply unit of the power distribution and data acquisition control system, enabling the pumping of coolant and lubricant. The flow control valves are powered by the built-in power supply unit of the power distribution and data acquisition control system, and their flow is controlled by the built-in control unit of the system.

[0014] The PC-based deep reinforcement learning system has a built-in DRL decision-maker. Real-time grinding force, surface roughness, and temperature measured by various sensors are used as target signals and transmitted to the DRL decision-maker via the data acquisition unit of the power distribution acquisition and control system. The internal reward function is: R=α·(1-ΔT / T_max)+β·Ra_min / Ra+γ·F_c / F_c0 (where ΔT is the temperature rise, Ra is the surface roughness, F_c is the cutting force, and α+β+γ=1). The system performs self-deep reinforcement learning by setting target weights to find the optimal input strategy and then makes a decision. The decision information is transmitted to the built-in control unit of the power distribution acquisition and control system. The control unit then adjusts the input process parameters such as spindle speed, coolant flow rate, lubricant flow rate, and ultrasonic vibration frequency to make the machining target result close to the optimal solution.

[0015] The operation process of this invention is as follows: The power distribution and acquisition control system is activated, simultaneously supplying power and transmitting signals to the PC-based deep reinforcement learning system, ultrasonic vibration platform system, sensor measurement system, internal cooling and lubrication grinding wheel system, and cooling and lubrication fluid supply system. The cooling and lubrication fluid supply system pumps cooling and lubrication fluid to the internal cooling and lubrication grinding wheel system. The internal cooling and lubrication grinding wheel system and the ultrasonic vibration platform system simultaneously begin processing. The sensor measurement system and the internal cooling and lubrication grinding wheel system respectively measure the grinding force, workpiece surface roughness, and grinding temperature in real time. The measured signals are transmitted to the built-in control unit of the power distribution and acquisition control system, and then transmitted via the control unit to the built-in DRL decision-maker of the PC-based deep reinforcement learning system. The DRL decision-maker performs self-deep reinforcement learning through a set reward function and target weights to find the optimal input strategy, and then makes a decision. The decision information is transmitted to the built-in control unit of the power distribution and acquisition control system. The control unit then adjusts the input process parameters such as spindle speed, coolant flow rate, lubricant flow rate, and ultrasonic vibration frequency to make the processing target (grinding force, workpiece surface roughness, and grinding temperature) approach the optimal solution.

[0016] Compared with the prior art, the technical effect of the present invention is: to propose a grinding system based on deep reinforcement learning, which combines an ultrasonic vibration platform with an internally cooled and lubricated grinding wheel. The system achieves efficient, high-quality, and low-consumption processing of difficult-to-machine materials by coordinating the ultrasonic vibration platform with an embedded micro heat pipe array and a piezoelectric ceramic gradient polarization array with an internally cooled and lubricated asymmetric fractal flow channel grinding wheel. At the same time, it assists in multi-physics field perception and AI deep reinforcement learning to achieve effective coordination of vibration, cooling, and lubrication during the processing of the ultrasonic vibration platform and the internally cooled and lubricated grinding wheel. Attached Figure Description

[0017] Figure 1 This is a structural schematic diagram according to an embodiment of the present invention;

[0018] Figure 2This is a system hardware composition diagram according to an embodiment of the present invention;

[0019] Figure 3 This is a structural diagram of an ultrasonic vibration platform according to an embodiment of the present invention;

[0020] Figure 4 This is a cross-sectional view of an internally cooled and lubricated grinding wheel according to an embodiment of the present invention;

[0021] Figure 5 This is a flowchart of the Deep Reinforcement Learning (DRL) decision-maker according to an embodiment of the present invention; Detailed Implementation

[0022] To provide a clearer understanding of the technical features, objectives, and effects of the present invention, the present invention will be further described below in conjunction with the accompanying drawings and embodiments, but the scope of the present invention is not limited thereto.

[0023] like Figure 1 and 2 As shown, an ultrasonic vibration internal cooling and lubrication grinding system based on deep reinforcement learning is characterized by including a PC-based deep reinforcement learning system (1), a power distribution acquisition and control system (2), an ultrasonic vibration platform system (3), a sensor measurement system (4), an internal cooling and lubrication grinding wheel system (5), and a cooling and lubrication fluid supply system (6).

[0024] The power distribution acquisition and control system (2) has a built-in power supply unit, control unit, data acquisition unit, and ultrasonic generation unit, which is responsible for simultaneously supplying power and transmitting signals to the PC-side deep reinforcement learning system (1), ultrasonic vibration platform system (3), sensor measurement system (4), internal cooling and lubrication grinding wheel system (5), and cooling and lubrication fluid supply system (6).

[0025] like Figure 2 and 3As shown, the ultrasonic vibration platform system (3) includes an ultrasonic vibration platform (31) and a workpiece (32). The ultrasonic vibration platform (31) includes a support base (310), a transducer (311), an amplitude transformer (312), a vibration conversion module (313), a vibration platform (314), an embedded micro heat pipe array (315), and a piezoelectric ceramic gradient polarization array (316). The support base (310), transducer (311), amplitude transformer (312), vibration conversion module (313), and vibration platform (314) are connected to each other in sequence. The workpiece (32) is fixed on the vibration platform (314). The embedded micro heat pipe array (315) is installed parallel to the vibration platform (314). A certain amount of special low-melting-point alloy is filled inside the micro heat pipes to absorb heat by melting and prevent the vibration platform (314) from deforming. A piezoelectric ceramic gradient polarization array (316) is installed at the bottom of the vibration platform (314) in a rectangular array. The polarization intensity of the piezoelectric ceramic gradient polarization array (316) in the edge region of the vibration platform (314) is 20%-35% higher than that in the center region, thus solving the problem of insufficient vibration at the edge. The transducer (311) is connected to the power distribution acquisition and control system (2). The power supply unit and the ultrasonic wave generation unit of the power distribution acquisition and control system (2) provide power and ultrasonic wave transmission to the ultrasonic vibration platform system (3), respectively.

[0026] like Figure 2 As shown, the sensor measurement system (4) includes a force gauge (41), a CCD camera (42), and a semiconductor laser (43). The force gauge (41) measures the grinding force during the grinding process. The signal collected by the force gauge (41) is transmitted to the data acquisition unit of the power distribution acquisition control system (2). The CCD camera (42) and the semiconductor laser (43) are used in combination to measure the surface roughness during the grinding process. The measured signal is also transmitted to the data acquisition unit of the power distribution acquisition control system (2).

[0027] like Figure 2 and 4 As shown, the internal cooling and lubrication grinding wheel system (5) includes an internal cooling and lubrication grinding wheel (51) and a machine tool spindle (52), with the internal cooling and lubrication grinding wheel (51) mounted on the machine tool spindle (52);

[0028] The internally cooled and lubricated grinding wheel (51) includes a grinding wheel holder (510), an electro-hydraulic slip ring (511), a grinding wheel base (512), and MEMS thin-film thermocouples (513). The electro-hydraulic slip ring (511) is mounted on the grinding wheel holder (510), which is connected to the grinding wheel base (512). The MEMS thin-film thermocouples (513) are evenly distributed inside the grinding wheel base (512) near the grinding surface. 13) The terminal is connected to the electro-hydraulic slip ring (511). The grinding wheel base (512) has an asymmetric tree-shaped fractal flow channel (512-1) built in. The fractal flow channel is a biomimetic tree structure with at least three levels of branches. The branch angle is designed asymmetrically to use centrifugal force to transport fluid. The branch angle range is (θ=15°-75°). It can effectively and actively lubricate the grinding zone by using centrifugal fluid transport during the high-speed rotation of the grinding wheel, while controlling and saving coolant. The grinding wheel base (512) is coated with an abrasive layer (512-2). The liquid inlet of the electro-hydraulic slip ring (511) is connected to the coolant supply system through a gas-liquid delivery pipe. The coolant supply system (6) pumps the coolant through the internal flow channel of the electro-hydraulic slip ring (511) to the asymmetric tree-shaped fractal flow channel (512-1) of the grinding wheel base (512), and then to the grinding zone. The power distribution acquisition and control system (2) is connected to the input terminal of the electro-hydraulic slip ring (511) to supply power to the electro-hydraulic slip ring (511). The current is transmitted to the MEMS thin film thermocouple (513), which measures the temperature of the grinding surface. The measured temperature signal is then transmitted to the data acquisition unit of the power distribution acquisition and control system (2).

[0029] like Figure 2 As shown, the coolant and lubricant supply system (6) includes two parts: coolant supply and lubricant supply. These two parts can be controlled independently or in tandem, and each is equipped with a separate flow control valve. The coolant and lubricant supply are powered by the built-in power supply unit of the power distribution and acquisition control system (2), enabling the pumping of coolant and lubricant. The flow control valve is powered by the built-in power supply unit of the power distribution and acquisition control system (2), and its flow is controlled by the built-in control unit of the power distribution and acquisition control system (2).

[0030] like Figure 2 and 5As shown, the PC-side deep reinforcement learning system (1) has a built-in DRL decision-maker. The real-time grinding force, surface roughness, and temperature measured by each sensor are used as target signals and are uniformly transmitted to the DRL decision-maker through the data acquisition unit of the power distribution acquisition control system (2). The internal reward function is: R=α·(1-ΔT / T_max)+β·Ra_min / Ra+γ·F_c / F_c0 (where ΔT is the temperature rise, Ra is the surface roughness, F_c is the cutting force, α+β+γ=1). The system performs self-deep reinforcement learning by setting target weights, finds the optimal input strategy, and then makes decision information. The decision information is transmitted to the built-in control unit of the power distribution acquisition control system (2). The control unit then adjusts the input process parameters such as spindle speed, coolant flow rate, lubricant flow rate, and ultrasonic vibration frequency to make the machining target result close to the optimal solution.

[0031] The operation process of this invention is as follows: start the power distribution acquisition and control system (2) to simultaneously supply power and transmit signals to the PC-side deep reinforcement learning system (1), ultrasonic vibration platform system (3), sensor measurement system (4), internal cooling and lubrication grinding wheel system (5), and cooling and lubrication fluid supply system (6). The cooling and lubricating fluid supply system (6) pumps cooling and lubricating fluid to the internal cooling and lubricating grinding wheel system (5). The internal cooling and lubricating grinding wheel system (6) and the ultrasonic vibration platform system (3) start processing simultaneously. The sensor measurement system (4) and the internal cooling and lubricating grinding wheel system (5) measure the grinding force, workpiece surface roughness, and grinding temperature in real time, respectively. The measured signals are transmitted to the built-in control unit of the power distribution acquisition and control system (2), and then transmitted to the built-in DRL decision-maker of the PC-side deep reinforcement learning system (1). The DRL decision-maker performs self-deep reinforcement learning through the set reward function and target weight to find the optimal input strategy, and then makes a decision. The decision information is transmitted to the built-in control unit of the power distribution acquisition and control system (2). The control unit then adjusts the input process parameters such as spindle speed, coolant flow rate, lubricant flow rate, and ultrasonic vibration frequency to make the processing target (grinding force, workpiece surface roughness, grinding temperature) result close to the optimal solution.

[0032] Although embodiments of the invention have been shown and described, it will be apparent to those skilled in the art that the invention is not limited to the details of the specific embodiments described above, and that the invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the invention. No reference numerals in the drawings should be construed as limiting the scope of the claims.

Claims

1. An ultrasonic vibration internal cooling and lubrication grinding system based on deep reinforcement learning, characterized in that: It includes a PC-based deep reinforcement learning system (1), a power distribution acquisition and control system (2), an ultrasonic vibration platform system (3), a sensor measurement system (4), an internal cooling and lubrication grinding wheel system (5), and a cooling and lubrication fluid supply system (6). The PC-based deep reinforcement learning system (1) is powered by the power distribution acquisition and control system (2) and transmits signals to each other. The power distribution acquisition and control system (2) simultaneously supplies power and transmits signals to the ultrasonic vibration platform system (3), the sensor measurement system (4), the internal cooling and lubrication grinding wheel system (5), and the cooling and lubrication fluid supply system (6). The internal cooling and lubrication grinding wheel system (5) and the cooling and lubrication fluid supply system (6) are connected through a gas-liquid transport pipeline. The PC-based deep reinforcement learning system (1) has a built-in DRL decision-maker; the built-in DRL decision-maker uses a deep reinforcement learning algorithm, and its reward function is: R = α·(1-ΔT / T_max) + β·Ra_min / Ra + γ·F_c / F_c0 (where ΔT is the temperature rise, Ra is the surface roughness, F_c is the cutting force, and α+β+γ=1). The ultrasonic vibration platform system (3) incorporates an embedded micro heat pipe array (315) and a piezoelectric ceramic gradient polarization array (316). The embedded micro heat pipe array (315) is installed parallel to the vibration platform (314) of the ultrasonic vibration platform system (3), and a certain amount of special low-melting-point alloy is loaded inside the micro heat pipe. The piezoelectric ceramic gradient polarization array (316) is installed at the bottom of the vibration platform (314) of the ultrasonic vibration platform system (3) and is distributed in a rectangular array. The polarization intensity of the piezoelectric ceramic gradient polarization array (316) in the edge region of the vibration platform (314) is 20%-35% higher than that in the center region. The internally cooled and lubricated grinding wheel system (5) includes an internally cooled and lubricated grinding wheel (51), which includes a grinding wheel substrate (512) and a MEMS thin film thermocouple (513). The grinding wheel substrate (512) has an embedded asymmetric tree-shaped fractal flow channel (512-1).

2. The ultrasonic vibration internal cooling and lubrication grinding system based on deep reinforcement learning according to claim 1, characterized in that: The asymmetric tree-shaped fractal flow channel (512-1) is a biomimetic tree structure with at least three levels of branches, and the branch angles are designed asymmetrically to utilize centrifugal force to transport fluid. The branch angle range is (θ=15°-75°). The MEMS thin film thermocouples (513) are evenly distributed and installed inside the grinding surface of the grinding wheel substrate (512).

3. The ultrasonic vibration internal cooling and lubrication grinding system based on deep reinforcement learning according to claim 1, characterized in that: The cooling and lubricating fluid supply system (6) has two parts: a cooling fluid supply and a lubricating fluid supply. The two parts can be controlled independently or in tandem, and each is equipped with a separate flow control valve.

Images