Intelligent real-time online monitoring and analyzing system and method for external thread grinding machine

Abstract

The invention discloses an intelligent real-time online monitoring and analyzing system and method for an external thread grinder, and relates to the technical field of intelligent monitoring, the system comprises a force sensor (1), an acoustic emission sensor (2), a vibration sensor (3), a micro-deformation sensor (4), a temperature sensor (5), a pitch diameter measuring instrument (6), a cable (7), a data acquisition device (8) and a computer (9), the data acquisition device (8) is respectively connected with the force sensor (1), the acoustic emission sensor (2), the vibration sensor (3), the micro-deformation sensor (4), the temperature sensor (5), the pitch diameter measuring instrument (6) and the computer (9) through cables (7). The problems that an existing grinding machine is low in machining precision, and fault information cannot be accurately obtained when faults occur are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent monitoring, and particularly relates to an intelligent real-time online monitoring and analysis system and method for an external thread grinding machine. Background Art

[0002] During the grinding process of existing external thread grinding machines, it is difficult to understand the machining accuracy of workpieces and the operating status of the grinding machines. During long-term continuous service, it is also difficult to know the reduction in workpiece machining accuracy caused by wear of the bed, guide rails, grinding wheels, etc. This will result in the inability to obtain fault information when the grinding machine fails. Aiming at improving the reliability and stability of external thread grinding machines, an intelligent real-time online monitoring and analysis system for external threads is proposed. It can monitor the operating status of the grinding machine in real time, analyze and predict the machining status in real time, avoid manual inspection one by one, and save the production stoppage time caused by faults. Summary of the Invention

[0003] Aiming at the above deficiencies in the prior art, an intelligent real-time online monitoring and analysis system and method for an external thread grinding machine provided by the present invention solve the problems of low machining accuracy of existing grinding machines and the inability to accurately obtain fault information when a fault occurs.

[0004] To achieve the above invention purpose, the technical solution adopted by the present invention is: an intelligent real-time online monitoring and analysis system for an external thread grinding machine, including a force sensor, an acoustic emission sensor, a vibration sensor, a micro deformation sensor, a temperature sensor, a pitch diameter measuring instrument, a cable, a data acquisition device and a computer. The data acquisition device is connected to the force sensor, the acoustic emission sensor, the vibration sensor, the micro deformation sensor, the temperature sensor, the pitch diameter measuring instrument and the computer through the cable respectively.

[0005] Further, the force sensor is a strain type force sensor, which is arranged at the tailstock of the external thread grinding machine and is used for measuring the three-direction grinding force received by the workpiece;

[0006] The acoustic emission sensor is a piezoelectric sensor, which is arranged at the grinding wheel spindle of the external thread grinding machine and is used for measuring the acoustic wave signal generated by the high-speed operation of the grinding wheel spindle;

[0007] The vibration sensor is a piezoelectric sensor, which is arranged at multiple steady rests of the external thread grinding machine and is used for measuring the three-direction vibration generated when grinding the workpiece;

[0008] The micro deformation sensor is a contact type displacement sensor, which is arranged at the tailstock of the external thread grinding machine and is used for measuring the micro deformation of the workpiece caused by temperature change;

[0009] The temperature sensor is a K-type thermocouple, which is arranged at multiple places on the bed of the external thread grinding machine and is used for measuring the temperature change caused by the grinding process;

[0010] The mid-diameter measuring instrument is a contact displacement sensor, which is arranged on the outer sheet metal of the grinding wheel spindle or the grinding wheel frame of the external thread grinding machine and is used to measure the mid-diameter of the workpiece after grinding processing;

[0011] The data acquisition device is a device composed of various data corresponding acquisition cards.

[0012] The technical solution adopted by the present invention is also: an intelligent real-time online monitoring and analysis method for an external thread grinding machine, including the following steps:

[0013] S1: Install each sensor at the corresponding position of the external thread grinding machine to obtain the corresponding information of the sensor, cable and data acquisition device;

[0014] S2: Input the processing parameters of the external thread grinding machine, transmit the information collected by each sensor to the data acquisition device, and use the data acquisition device to transmit the received electrical signal to the computer;

[0015] S3: Use the computer to assemble the sensor numerical matrix, and input the sensor numerical matrix into the trained multi-fidelity neural network algorithm to obtain the theoretical operating state of the external thread grinding machine, the theoretical processing accuracy of the lead screw and the theoretical operating state of the grinding wheel;

[0016] S4: Judge whether it is necessary to adjust the processing parameters according to the theoretical processing accuracy of the lead screw. If so, return to step S2. Otherwise, the workpiece processing is completed, the operation of the external thread grinding machine is stopped, and the intelligent real-time online monitoring and analysis of the external thread grinding machine is completed.

[0017] Further, the training of the multi-fidelity neural network algorithm in S3 includes the following sub-steps:

[0018] S31: Input the data of each sensor as low-fidelity model data into the low-fidelity neural network model to obtain new data containing low-fidelity information;

[0019] S32: Input the new data containing low-fidelity information and high-fidelity model data into the high-fidelity neural network model to obtain the output value of the high-fidelity neural network model;

[0020] S33: Input the output value of the high-fidelity neural network model into the physics-informed neural network to obtain the output value of the physics-informed neural network;

[0021] S34: Construct a loss function based on the output value of the physics-informed neural network, and iteratively optimize the weights and biases of each neural network until the overall model loss value converges, and complete the training of the multi-fidelity neural network algorithm.

[0022] Further, the high-fidelity neural network model in S32 includes a neural network with an activation function and a neural network without an activation function.

[0023] Further, the physical information neural network in S33 is a neural network for solving the partial differential equation of the vibration equation of external thread grinding machining, and the vibration equation of external thread grinding machining is as follows:

[0024]

[0025] where, T e is the system kinetic energy, U e is the system potential energy, L is the workpiece length, ρ is the workpiece material density, A is the workpiece cross-sectional area, U, V, and W are the translational displacements of the three directions x, y, and z of the deflection of any cross-section of the workpiece, B and Γ are the rotations of the deflection of any cross-section of the workpiece around the x and z directions, I is the workpiece area inertia matrix, Ω is the workpiece rotational speed, i is the number of clamping devices, N is the number of vibration mode functions, m i is the mass of the i-th clamping device, E is the elastic modulus, κ is the shear shape factor, G is the shear modulus, P0 is the axial compressive force generated by the tailstock, k i is the stiffness of the i-th clamping device, and are the derivatives with respect to time t, U', V', and W' are the first-order derivatives of U, V, and W with respect to the position x respectively, and V'' and W'' are the second-order derivatives of V and W with respect to the position x respectively.

[0026] Further, the loss function L1 in S34 is as follows:

[0027]

[0028] where, MSE yL is the loss value of the low-fidelity neural network model, MSE yH is the loss value of the high-fidelity neural network model, MSE fe is the loss value of the physical information neural network model, λ is the regularization term weight, and β i is the regularization term.

[0029] Further, the weights and biases in S34 are as follows:

[0030]

[0031] where, ω is the neural network weight, α is the learning rate, L1 is the loss function, and b is the bias.

[0032] The beneficial effects of the present invention are as follows:

[0033] (1) The present invention can monitor the operating state of the external thread grinding machine in real time online to ensure the machining stability and safety;

[0034] (2) The present invention can intelligently analyze the machining accuracy of workpieces and the running stability of grinding wheels. With high prediction accuracy by combining sensor values, it reduces machining risks, is applicable to various machining environments, and meets the actual engineering requirements.

[0035] (3) The present invention provides an intelligent real-time online monitoring and analysis system for external thread grinding machines, which not only has the function of online real-time judging the theoretical running state of the grinding machine, the theoretical machining accuracy of the lead screw, and the theoretical running state of the grinding wheel, but also has the function of real-time training according to the collected data and updating variables. Description of the Drawings

[0036] Figure 1 It is a schematic structural diagram of the intelligent real-time online monitoring and analysis system for the external thread grinding machine of the present invention.

[0037] Figure 2 It is a connection diagram of the sensor, cable, and computer in the present invention.

[0038] Figure 3 It is a flowchart of the intelligent real-time online monitoring and analysis method for the external thread grinding machine of the present invention.

[0039] 1. Force sensor; 2. Acoustic emission sensor; 3. Vibration sensor; 4. Micro-deformation sensor; 5. Temperature sensor; 6. Middle diameter measuring instrument; 7. Cable; 8. Data acquisition device; 9. Computer; 10. Grinding wheel carriage; 11. Grinding wheel spindle; 12. Center rest; 13. Bed; 14. Tailstock; 15. Headstock. Detailed Embodiments

[0040] The following further describes the present invention with reference to the drawings and specific embodiments.

[0041] Embodiment 1, as shown in Figure 1 and Figure 2 An intelligent real-time online monitoring and analysis system for an external thread grinding machine, characterized in that it includes a force sensor 1, an acoustic emission sensor 2, a vibration sensor 3, a micro-deformation sensor 4, a temperature sensor 5, a middle diameter measuring instrument 6, a cable 7, a data acquisition device 8, and a computer 9. The data acquisition device 8 is connected to the force sensor 1, the acoustic emission sensor 2, the vibration sensor 3, the micro-deformation sensor 4, the temperature sensor 5, the middle diameter measuring instrument 6, and the computer 9 through the cable 7 respectively.

[0042] The force sensor 1 is a strain type force sensor, which is arranged at the tailstock 14 of the external thread grinding machine and is used to measure the three-direction grinding force applied to the workpiece;

[0043] The acoustic emission sensor 2 is a piezoelectric sensor, which is arranged at the grinding wheel spindle 11 of the external thread grinding machine and is used to measure the acoustic wave signal generated by the high-speed operation of the grinding wheel spindle 11;

[0044] The vibration sensor 3 is a piezoelectric sensor, which is arranged at multiple steady rests 12 of the external thread grinding machine and is used to measure the three-direction vibration generated during grinding of the workpiece;

[0045] The micro-deformation sensor 4 is a contact displacement sensor, which is arranged at the tailstock 14 of the external thread grinding machine and is used to measure the micro-deformation of the workpiece caused by temperature change;

[0046] The temperature sensor 5 is a K-type thermocouple, which is arranged at multiple positions on the bed 13 of the external thread grinding machine and is used to measure the temperature change caused by grinding;

[0047] The pitch diameter measuring instrument 6 is a contact displacement sensor, which is arranged on the outer sheet metal of the grinding wheel spindle 11 or the wheelhead 10 of the external thread grinding machine and is used to measure the pitch diameter of the workpiece after grinding;

[0048] The data acquisition device 8 is a device composed of various data corresponding acquisition cards.

[0049] Embodiment 2, as Figure 3 shown, an intelligent real-time online monitoring and analysis method for an external thread grinding machine includes the following steps:

[0050] S1: Install each sensor at the corresponding position of the external thread grinding machine to obtain the corresponding information of the sensor, cable and data acquisition device;

[0051] S2: Input the processing parameters of the external thread grinding machine, transmit the information collected by each sensor to the data acquisition device, and use the data acquisition device to transmit the received electrical signal to the computer;

[0052] S3: Use the computer to assemble the sensor numerical matrix wherein, is the force sensor data, is the acoustic emission sensor data, is the vibration sensor data, is the micro-deformation sensor data, is the temperature sensor data, is the pitch diameter measuring instrument data; and input the sensor numerical matrix into the trained multi-fidelity neural network algorithm to obtain the theoretical operating state of the external thread grinding machine, the theoretical machining accuracy of the lead screw and the theoretical operating state of the grinding wheel;

[0053] S4: Judge whether it is necessary to adjust the processing parameters according to the theoretical machining accuracy of the lead screw. If so, return to step S2. Otherwise, the workpiece machining is completed, the operation of the external thread grinding machine is stopped, and the intelligent real-time online monitoring and analysis of the external thread grinding machine is completed.

[0054] The training of the multi-fidelity neural network algorithm in S3 includes the following sub-steps:

[0055] S31: Input the sensor data as low-fidelity model data into the low-fidelity neural network model to obtain new data containing low-fidelity information;

[0056] S32: Input the new data containing low-fidelity information and the high-fidelity model data into the high-fidelity neural network model to obtain the output value of the high-fidelity neural network model;

[0057] S33: Input the output value of the high-fidelity neural network model into the physics-informed neural network to obtain the output value of the physics-informed neural network;

[0058] S34: Construct a loss function based on the output value of the physics-informed neural network, and iteratively optimize the weights and biases of each neural network until the overall model loss value converges, completing the training of the multi-fidelity neural network algorithm.

[0059] The low-fidelity model data is sensor data with large environmental noise and temperature variations; the high-fidelity model data is sensor data with small environmental noise and temperature variations.

[0060] The high-fidelity neural network model in S32 includes a neural network with an activation function and a neural network without an activation function, used to identify the linear and non-linear correlations between the low-fidelity data and the high-fidelity data.

[0061] The physics-informed neural network in S33 is a neural network for solving the partial differential equation of the vibration equation of external thread grinding, and the vibration equation of external thread grinding is:

[0062]

[0063] where, T e is the system kinetic energy, U e is the system potential energy, L is the workpiece length, ρ is the workpiece material density, A is the workpiece cross-sectional area, U, V, and W are the translational displacements in the three directions x, y, and z of the deflection of any cross-section of the workpiece, B and Γ are the rotations about the x and z directions of the deflection of any cross-section of the workpiece, I is the workpiece area inertia matrix, Ω is the workpiece rotation speed, i is the number of clamping devices, N is the number of mode shape functions, m i is the mass of the i-th clamping device, E is the elastic modulus, κ is the shear shape factor, G is the shear modulus, P0 is the axial compression force generated by the tailstock, k i is the stiffness of the i-th clamping device, and are the derivatives with respect to time t, U', V', and W' are the first-order derivatives of U, V, and W with respect to position x, and V” and W” are the second-order derivatives of V and W with respect to position x.

[0064] The loss function L1 in S34 is:

[0065]

[0066] Among them, MSE yL is the loss value of the low-fidelity neural network model, MSE yH is the loss value of the high-fidelity neural network model, MSE fe is the loss value of the physics-informed neural network model, λ is the regularization term weight, and β i is the regularization term.

[0067] The weights and biases in S34 are as follows:

[0068]

[0069] Among them, ω is the neural network weight, α is the learning rate, L1 is the loss function, and b is the bias.

[0070] In S3, the specific implementation of outputting the theoretical operating state of the grinding machine, the theoretical machining accuracy of the lead screw, and the theoretical operating state of the grinding wheel is as follows: The trained multi-fidelity neural network algorithm includes a neural network structure and a weight matrix. The weight matrix is a set of parameters trained through existing sensor data that can well fit the theoretical operating state of the grinding machine, the theoretical machining accuracy of the lead screw, and the theoretical operating state of the grinding wheel. When similar sensor data is input into the trained multi-fidelity neural network algorithm again, the algorithm fits the data to output the corresponding operating state of the grinding machine, the machining accuracy of the lead screw, and the operating state of the grinding wheel. Among them, the operating state of the grinding machine is the stable operating stage and the unstable operating stage. In the stable operating stage, the vibration value of the acoustic emission sensor fluctuates within a specific range. In the unstable operating stage, the vibration value of the acoustic emission sensor fluctuates outside the specific range. The operating state of the grinding wheel is the initial wear stage, the stable wear stage, and the severe wear stage. Each stage has a corresponding vibration range. In the initial wear stage, the vibration value fluctuates greatly and the vibration amplitude is small. In the stable wear stage, the vibration value fluctuates slightly and the vibration amplitude is moderate. In the severe wear stage, the vibration value fluctuates greatly and the vibration amplitude is large. The machining accuracy of the lead screw is to theoretically output the corresponding machining accuracy of the lead screw according to a large amount of previously trained sensor data and the accuracy of the corresponding lead screw after machining, combined with the existing sensor data.

[0071] In S4, it is determined whether to adjust the machining parameters according to the theoretical machining accuracy of the lead screw, specifically: According to the output theoretical machining accuracy of the lead screw, it is judged whether the accuracy index meets the machining requirements.

[0072] Those of ordinary skill in the art will realize that the embodiments described herein are provided to assist the reader in understanding the principles of the present invention, and it should be understood that the scope of protection of the present invention is not limited to such specific statements and embodiments. Those of ordinary skill in the art can make various other specific deformations and combinations that do not depart from the essence of the present invention based on these technical revelations disclosed in the present invention, and these deformations and combinations are still within the scope of protection of the invention.

Claims

1. An intelligent real-time online monitoring and analysis system for external thread grinder, characterized in that: The invention comprises a force sensor (1), an acoustic emission sensor (2), a vibration sensor (3), a micro-deformation sensor (4), a temperature sensor (5), a mid-diameter measuring instrument (6), a cable (7), a data acquisition device (8) and a computer (9); the data acquisition device (8) is respectively connected to the force sensor (1), the acoustic emission sensor (2), the vibration sensor (3), the micro-deformation sensor (4), the temperature sensor (5), the mid-diameter measuring instrument (6) and the computer (9) via the cable (7).

2. The intelligent real-time online monitoring and analysis system for external thread grinder according to claim 1 is characterized in that: The force sensor (1) is a strain type force sensor, which is arranged at the tailstock (14) of the external thread grinder and is used to measure the three-dimensional grinding force exerted on the workpiece; The acoustic emission sensor (2) is a piezoelectric sensor, which is arranged at the grinding wheel spindle (11) of the external thread grinder and is used to measure the acoustic wave signal generated by the high-speed operation of the grinding wheel spindle (11); The vibration sensor (3) is a piezoelectric sensor, which is arranged at a plurality of center frames (12) of the external thread grinding machine and is used to measure the three-dimensional vibration generated when grinding a workpiece; The micro-deformation sensor (4) is a contact displacement sensor, which is arranged at the tailstock (14) of the external thread grinder and is used to measure the micro-deformation of the workpiece caused by temperature change; The temperature sensor (5) is a K-type thermocouple, which is arranged at multiple locations on the bed (13) of the external thread grinder and is used to measure the temperature change caused by the grinding process; The center diameter measuring instrument (6) is a contact displacement sensor, which is arranged on the outer sheet metal of the grinding wheel spindle (11) or the grinding wheel frame (10) of the external thread grinder and is used to measure the center diameter of the workpiece after grinding; The data acquisition device (8) is a device composed of a plurality of data corresponding acquisition cards.

3. The analysis method of an intelligent real-time online monitoring and analysis system for an external thread grinder according to any one of claims 1 to 2, characterized in that: The following steps are involved: S1: Install each sensor at the corresponding position of the external thread grinder to obtain corresponding information of the sensor, cable and data acquisition device; S2: Inputting the processing parameters of the external thread grinder, transmitting the information collected by each sensor to the data acquisition device, and using the data acquisition device to transmit the received electrical signals to the computer; S3: Assemble the sensor numerical matrix by computer, and input the sensor numerical matrix into the trained multi-fidelity neural network algorithm to obtain the theoretical operation state of the external thread grinder, the theoretical processing accuracy of the screw and the theoretical operation state of the grinding wheel; S4: Determine whether the processing parameters need to be adjusted based on the theoretical processing accuracy of the screw. If yes, return to step S2. Otherwise, the workpiece processing is completed, the external thread grinder is stopped, and the intelligent real-time online monitoring and analysis of the external thread grinder is completed.

4. The analysis method of the intelligent real-time online monitoring and analysis system for external thread grinder according to claim 3 is characterized in that: The training of the multi-fidelity neural network algorithm in S3 includes the following steps: S31: inputting each sensor data as low-fidelity model data into the low-fidelity neural network model to obtain new data containing low-fidelity information; S32: inputting new data containing low-fidelity information and high-fidelity model data into the high-fidelity neural network model to obtain an output value of the high-fidelity neural network model; S33: inputting the output value of the high-fidelity neural network model into the physical information neural network to obtain the output value of the physical information neural network; S34: Construct a loss function based on the output value of the physical information neural network, and iteratively optimize the weights and biases of each neural network until the overall model loss value converges, completing the training of the multi-fidelity neural network algorithm.

5. The analysis method of the intelligent real-time online monitoring and analysis system for external thread grinder according to claim 4 is characterized in that: The high-fidelity neural network model in S32 includes a neural network with an activation function and a neural network without an activation function.

6. The analysis method of the intelligent real-time online monitoring and analysis system for external thread grinder according to claim 4 is characterized in that: The physical information neural network in S33 is a neural network for solving the partial differential equation of the external thread grinding vibration equation, and the external thread grinding vibration equation is: Among them, T e is the system kinetic energy, U e is the system potential energy, L is the workpiece length, ρ is the workpiece material density, A is the cross-sectional area of ​​the workpiece, U, V and W are the translation displacements of the workpiece in the three directions of x, y and z, respectively, B and Γ are the rotations of the workpiece in the three directions of x and z, I is the workpiece area inertia matrix, Ω is the workpiece rotation speed, i is the number of clamping devices, N is the number of vibration mode functions, m i is the mass of the i-th clamping device, E is the elastic modulus, κ is the shear shape factor, G is the shear modulus, P0 is the axial compression force generated by the tailstock, k i is the stiffness of the i-th clamping device, and is the derivative with respect to time t, U', V' and W' are the first-order derivatives of U, V and W with respect to position x respectively, V'' and W'' are the second-order derivatives of V and W with respect to position x respectively.

7. The analysis method of the intelligent real-time online monitoring and analysis system for external thread grinder according to claim 4 is characterized in that: The loss function L1 in S34 is: Among them, MSE yL is the loss value of the low-fidelity neural network model, MSE yH is the loss value of the high-fidelity neural network model, is the loss value of the physical information neural network model, λ is the regularization term weight, β i is the regularization term.

8. The analysis method of the intelligent real-time online monitoring and analysis system for external thread grinder according to claim 4 is characterized in that: The weights and biases in S34 are: Among them, ω is the neural network weight, α is the learning rate, L1 is the loss function, and b is the bias.

Images