A robot constant force grinding optimization method, system, device and medium

Abstract

The present invention relates to a robot constant force grinding optimization method, system, equipment and medium, which comprises the following steps: establishing a robot constant force grinding optimization model and determining its constraint conditions; based on the determined robot constant force grinding optimization model, establishing a BP neural network based on variable history influencing factors; inputting the grinding parameters under the working conditions to be tested into the trained BP neural network to obtain the surface processing quality prediction results of the composite workpiece corresponding to the grinding parameters. The present invention proposes a method for predicting the roughness and uniformity of the polished surface, which can accurately predict the polishing processing target under specific working conditions. The present invention has important significance for robot-controlled constant force polishing in predicting the processing target accurately and quickly, reducing the processing cost and improving the production efficiency.

Description

Technical Field

[0001] The present invention relates to a robot constant force grinding optimization method, system, equipment and medium, and in particular to a robot constant force grinding optimization method, system, equipment and medium based on deep learning, belonging to the field of composite material processing. Background Art

[0002] For small batches and complex surface composite material processing conditions, grinding, as the last process, has a great impact on the surface properties of the formed body. If errors occur during grinding, the entire workpiece will often be wasted. In order to determine the grinding parameters, experienced engineers are required to conduct a large number of experiments and data fitting to determine them. In order to improve the quality of the polished surface and reduce the cost of the experiment, an efficient method is needed to find the optimal grinding parameters and determine the process optimization plan.

[0003] Different from ordinary grinding, when a robot-controlled end is used to grind composite materials, the robot controls constant force grinding, and the grinding head is a constant force end adsorbed by sandpaper. The workpiece to be polished is a composite material with a complex surface. The grinding parameters that affect the surface roughness and uniformity of the polished surface mainly depend on the grinding speed, feed speed, constant force applied during grinding, and grinding track spacing. At the same time, because the robot-controlled end uses a constant force pneumatic grinding head during grinding, the sandpaper is adsorbed on the rotating part at the end, and sandpaper is used instead of a grinding wheel for grinding. It is difficult to give an exact target function for different materials and surface shapes of the polished body. Many grinding parameters require a lot of experiments and mathematical formula fitting to determine, which undoubtedly wastes time and increases production costs when faced with small-batch production. Summary of the invention

[0004] In view of the above problems, the purpose of the present invention is to provide a robot constant force grinding optimization method, system, equipment and medium. For the optimization model of robot constant force grinding of composite materials, the optimization objective function and constraints are determined, and the deep learning neural network method is adopted to avoid the problem of data fitting or genetic algorithm requiring the calculation of exact parameters, and directly construct a relationship network between grinding parameters and surface processing quality, which provides a scientific and effective method for the grinding of small batches and difficult-to-process workpieces, and can achieve efficient and low-cost robot constant force grinding.

[0005] To achieve the above object, the present invention adopts the following technical solutions:

[0006] In a first aspect, the present invention provides a robot constant force grinding optimization method, which comprises the following steps:

[0007] Establish a robot constant force grinding optimization model and determine its constraints;

[0008] Based on the determined robot constant force grinding optimization model, a BP neural network based on variable history influencing factors is established;

[0009] The grinding parameters under the working conditions to be tested are input into the trained BP neural network to obtain the surface processing quality prediction results of the composite workpiece corresponding to the grinding parameters.

[0010] Furthermore, the method for establishing a robot constant force grinding optimization model and determining its constraint conditions includes:

[0011] The characteristics of robot constant force grinding of composite workpieces are analyzed to determine the factors affecting the surface roughness and grinding uniformity of composite workpieces after forming, including feed speed vf, grinding speed vs, constant pressure F, and grinding track spacing D;

[0012] According to the determined influencing factors, an optimization function with surface roughness and grinding uniformity as the target is established;

[0013] The constraints of the objective optimization function are determined, which include working condition variable constraints and sandpaper durability constraints.

[0014] Furthermore, the surface roughness calculation formula is:

[0015]

[0016] Among them, Ra is the surface roughness; K is the proportionality coefficient; b, c, u are coefficients; v f is the feed speed; v s is the grinding speed; K p is the Preston constant; L′ is the intersection line between the grinding head and the previous grinding track; F is the constant pressure; s′ is the overlapping part of the two tracks;

[0017] The grinding uniformity calculation formula is:

[0018]

[0019] Where R is the radius of the grinding head; S = 2·π·R is the bottom area of ​​the grinding head; l min It is the shortest intersection line between the grinding head and the object being ground during its forward movement; s max It is the maximum contact area between the grinding head and the object being ground during the forward movement when there is an overlapping area; H max and H min It is the highest and lowest point in the grinding track.

[0020] Furthermore, the method of establishing a BP neural network based on a variable history influencing factor based on the determined robot constant force grinding optimization model includes:

[0021] Determine the BP neural network structure;

[0022] Determine the activation function, optimizer, and loss function;

[0023] The experimental data between grinding parameters and machining targets are obtained by orthogonal experimental method, a preset proportion of data is used as a training set, the remaining data is used as a test set, and all data are normalized;

[0024] Based on the obtained normalized training set and test set, the established BP neural network is trained with the determined activation function, optimizer and loss function until the training meets the error requirement, thereby obtaining a trained BP neural network.

[0025] Furthermore, the BP neural network includes an input layer, a hidden layer and an output layer, wherein the input layer is based on a feed rate v f , grinding speed v s , constant pressure F and grinding track spacing D are input vectors; the output layer takes surface roughness Ra and grinding uniformity RE as output vectors; the hidden layer is a single hidden layer, and the number of nodes is set to a preset value.

[0026] Furthermore, the activation function of each layer of the BP neural network adopts the RELU activation function; the loss function adopts the mean square error function, the optimizer adopts the Nadam optimizer, and the calculation formula is:

[0027]

[0028] V (t) =β1V (t-1) +(1-β1)ΔW (t)i

[0029] S (t) =β2S (t-1) +(1-β2)ΔW (t)i ΔW (t)i

[0030] Where W (t)i is the weight value at time t; W (t-1)i is the weight value at time t-1; ΔW (t)i is the change in weight at each iteration; V (t) is ΔW at time t (t) The sum of V (t-1) is ΔW at time t-1 (t) The sum of; β1 and β2 are the historical impact factor coefficients; η is the learning rate; S (t) is the cumulative square gradient; S (t-1) is the cumulative square gradient of the previous moment; ε is a constant; γ is a coefficient; N (t+1)is the weight value at the next moment; t represents the moment.

[0031] Furthermore, the historical impact factor coefficient β1 is determined by the following method:

[0032] First, the gradient descent rate of change G is detected during the gradient descent process, and its calculation formula is:

[0033]

[0034] Secondly, set the gradient to be in a gentle state when the gradient descent rate meets the preset conditions, and record the number of steps G of the continuous gentle gradient state t ;

[0035] Finally, according to the recorded continuous gradient smoothing steps G t Set the β1 parameter, where G t The relationship between and β1 is determined by the critical value 1 / e of the exponentially weighted moving average method.

[0036] In a second aspect, the present invention provides a robot constant force grinding optimization system, comprising:

[0037] Optimization model building module, used to build the robot constant force grinding optimization model and determine its constraint conditions;

[0038] A neural network module is used to establish a BP neural network based on variable history influence factors based on a determined robot constant force grinding optimization model;

[0039] The prediction module is used to input the grinding parameters under the working conditions to be tested into the trained BP neural network to obtain the surface processing quality prediction results of the composite material workpiece corresponding to the grinding parameters.

[0040] In a third aspect, the present invention provides a processing device, which includes at least a processor and a memory, wherein the memory stores a computer program, and when the processor runs the computer program, the computer program is executed to implement the steps of the robot constant force grinding optimization method.

[0041] In a fourth aspect, the present invention provides a computer storage medium having computer-readable instructions stored thereon, wherein the computer-readable instructions can be executed by a processor to implement the steps of the robot constant force grinding optimization method.

[0042] The present invention adopts the above technical solution, which has the following advantages:

[0043] 1. An optimization model is derived for robot constant force grinding, and the grinding workpiece is a composite material with a complex surface. It is analyzed that the grinding parameters that affect the grinding surface roughness and grinding uniformity mainly depend on the grinding speed, feed speed, constant force applied during grinding, and grinding track spacing;

[0044] 2. An optimized gradient descent method with variable historical influence factors based on the exponentially weighted moving average method is proposed, which makes the descent speed faster and effectively reduces the gradient oscillation problem.

[0045] Therefore, the present invention can be widely applied to the field of composite material processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Throughout the accompanying drawings, the same reference numerals are used to represent the same components. In the accompanying drawings:

[0047] Figure 1 is a flow chart of a robot constant force grinding optimization method provided by an embodiment of the present invention;

[0048] Figure 2 is a schematic diagram of calculating the tangent area between two grinding tracks provided by an embodiment of the present invention;

[0049] Figure 3 is a schematic diagram of a robot grinding trajectory provided by an embodiment of the present invention;

[0050] Figure 4 It is a schematic diagram of polishing uniformity provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0051] In order to make the purpose, technical solution and advantages of the embodiment of the present invention clearer, the technical solution of the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings of the embodiment of the present invention. Obviously, the described embodiment is a part of the embodiment of the present invention, not all of the embodiments. Based on the described embodiment of the present invention, all other embodiments obtained by ordinary technicians in this field belong to the scope of protection of the present invention.

[0052] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0053] After analyzing robot constant force grinding and ordinary manual grinding, the present invention finds that the main differences between the two are:

[0054] 1. Use the robot to control the tool path for grinding, and the grinding path is pre-set;

[0055] 2. Grinding is performed by a grinding head clamped at the end of the robot. The grinding head sets a constant pressure instead of a fixed grinding depth. The grinding depth is determined by the constant pressure and the dwell time per unit area.

[0056] 3. The grinding head grinds with sandpaper instead of grinding wheel.

[0057] That is, compared with ordinary manual grinding, robot constant force grinding can accurately control the grinding trajectory and grinding efficiency. However, it is difficult to make dynamic adjustments during robot grinding, and the grinding parameters need to be set in advance, otherwise it will cause uneven grinding, substandard grinding depth, and unqualified surface roughness. Manual adjustments are required, which greatly increases costs and time.

[0058] Based on the above analysis, some embodiments of the present invention disclose a robot constant force grinding optimization method. First, based on the previous research foundation on manual grinding and combined with the characteristics of robot constant force grinding of composite materials, an optimization model of surface roughness and grinding uniformity is established. Surface roughness and grinding uniformity are used as objective functions, and working condition variable constraints and sandpaper durability constraints are determined at the same time; then, the multi-objective problem is optimized based on the BP neural network to obtain the optimal robot constant force grinding process parameters and grinding processing quality.

[0059] Correspondingly, some other embodiments of the present invention provide a robot constant force grinding optimization system, equipment and medium.

[0060] Example 1

[0061] like Figure 1 As shown, this embodiment provides a robot constant force grinding optimization method, comprising the following steps:

[0062] 1) Model establishment: Establish a robot constant force grinding optimization model and determine its constraints. The optimization model is based on the feed speed v f , grinding speed v s , constant pressure F, and grinding track spacing D are grinding parameter variables, surface roughness and grinding uniformity are taken as objective functions, and working condition variable constraints and sandpaper durability constraints are taken as constraints.

[0063] Specifically, the steps include:

[0064] 1.1) Analyze the characteristics of robot constant force grinding of composite workpieces and determine the factors affecting the surface roughness and grinding uniformity of composite workpieces after forming.

[0065] Among them, the image factors of the surface roughness and polishing uniformity of the image composite material workpiece after forming determined in the present invention include the feed speed vf , grinding speed v s , constant pressure F, grinding track spacing D.

[0066] 1.2) Based on the determined influencing factors, an optimization function with surface roughness and grinding uniformity as the target is established.

[0067] Among them, the ideal surface roughness formula for robot constant force grinding is established as follows:

[0068]

[0069] Among them, Ra is the surface roughness; K is the proportionality coefficient; b, c, u are coefficients; v f is the feed speed; v s is the grinding speed; K p is the Preston constant; L′ is the intersection line between the grinding head and the previous grinding track; F is the constant pressure; s′ is the overlapping part of the two tracks, that is, the sectional area between the grinding head and the previous track.

[0070] like Figure 2 As shown, the calculation formula of s′ is:

[0071]

[0072] Wherein, D is the lateral length of the non-overlapping portion between the grinding head and the contact surface; R is the radius of the grinding head; and x is the integral variable.

[0073] The established robot constant force grinding uniformity model, that is, the formula of the grinding error RE between the lowest point and the highest point in the grinding trajectory is:

[0074]

[0075]

[0076]

[0077] From formula (3), (4), (5), we can get:

[0078]

[0079] Where R is the radius of the grinding head; S = 2·π·R is the bottom area of ​​the grinding head; l min It is the shortest intersection line between the grinding head and the object being ground during its forward movement; s max It is the maximum contact area between the grinding head and the object being ground during the forward movement when there is an overlapping area; H max and H min are the highest and lowest points in the grinding track. min and maxThe constant pressure F of the grinding head and the grinding speed v s , feed speed v f Influence.

[0080] 1.3) Determine the constraints of the objective optimization function, which include working condition variable constraints and sandpaper durability constraints.

[0081] Specifically, the working condition variable constraint means that during the grinding process, the grinding speed, feed speed and grinding pressure cannot exceed the limit values ​​of the robot and the end effector, that is:

[0082] Grinding speed v s :The speed is 4500r / min-10000r / min, and 2·π·R·n should be guaranteed min ≤v s ≤2·π·R·n max ,Right now

[0083] 282.6m / s≤v s ≤628m / s (7)

[0084] Among them, n min is the minimum speed of the grinding head, n max The maximum speed of the grinding head.

[0085] Feed speed v f Defined by the robot, i.e.

[0086] 0mm / s<v f ≤5mm / s (8)

[0087] The grinding pressure F is defined by the end effector, i.e.

[0088] 0N<F≤45N (9)

[0089] Usually, the durability of sandpaper is measured by experiment, that is, when the grinding length is L s , sandpaper reaches maximum durability E s , then the grinding removal amount E≤E s Among them, E s The calculation formula is:

[0090]

[0091] Where p is the pressure per unit area.

[0092] 2) Network training: Based on the determined robot constant force grinding optimization model, a BP neural network based on variable historical influencing factors is established. The BP neural network is based on the feed speed v f , grinding speed v s, constant pressure F and grinding track spacing D are taken as input vectors, and surface roughness and grinding uniformity are taken as output vectors.

[0093] According to the relationship between the grinding parameters and the machining target determined by the above objective function and constraint conditions, the grinding parameters that affect the machining target are determined to be the feed speed v f , grinding speed v s , constant pressure F and grinding track spacing D. In view of the constant force grinding working condition of the robot targeted by the present invention, the working condition variable during the grinding process is the feed speed v f , grinding speed v s , constant pressure F and grinding track spacing D. In the optimization model, due to many parameters such as: min and max , K, K p It is difficult to calculate, so this embodiment chooses to adopt the BP neural network method, skips the influence of intermediate parameters, directly builds the relationship model between input and output, and eliminates the predicted values ​​that do not meet the requirements through constraints.

[0094] Specifically, the steps include:

[0095] 2.1) Determine the BP neural network structure.

[0096] The BP neural network constructed in this embodiment includes an input layer, a hidden layer and an output layer, wherein the input layer is based on the feed speed v f , grinding speed v s , constant pressure F and grinding track spacing D are input vectors, which includes 4 neural nodes (X1, X2, X3, X4) T , where X1 is the feed speed v f , X2 is the grinding speed v s , X3 is the constant pressure F, X4 is the grinding track spacing D; the output layer uses the surface roughness Ra and the grinding uniformity RE as the output vector, that is, it includes two neural nodes (O1, O2) T , where O1 is the surface roughness Ra, O2 is the polishing uniformity RE; the hidden layer is a single hidden layer, and the number of nodes is set to 8.

[0097] 2.2) Determine the activation function, optimizer and loss function

[0098] Among them, the activation function is used to calculate the data input into the BP neural network to obtain the corresponding output; the loss function is the objective function of the neural network optimization, and the process of neural network training or optimization is the process of minimizing the loss function, that is, the smaller the loss function value, the closer the corresponding predicted result is to the actual result value; the optimizer is used to update and calculate the network parameters that affect model training and model output.

[0099] In this embodiment, the activation function of each layer adopts the RELU activation function:

[0100] f(x)=max(0,x) (11)

[0101] The optimizer selects the Nadam optimizer:

[0102]

[0103] V (t) =β1V (t-1) +(1-β1)ΔW (t)i (13)

[0104] S (t) =β2S (t-1) +(1-β2)ΔW (t)i ΔW (t)i (14)

[0105] and:

[0106]

[0107] Among them, W (t)i is the weight value at time t; W (t-1)i is the weight value at time t-1; ΔW (t)i is the change in weight at each iteration, and the biased optimization method during training is the same as the weight; V (t) is ΔW at time t (t) The sum of V (t-1) is ΔW at time t-1 (t) The sum of β1 and β2 are historical impact factor coefficients, which are generally set manually; η is the learning rate; S (t) is the cumulative square gradient; S (t-1) is the cumulative square gradient of the previous moment; ε is an arbitrary infinitesimal value, in order to prevent the denominator from being 0; γ is a coefficient, usually set manually; N (t+1) is the weight value at the next moment; t represents the moment.

[0108] The β1 and β2 parameters are generally set by experience, usually β1 = 0.9, β2 = 0.99, but this will lead to problems such as terminal oscillation and convergence oscillation during the gradient descent process. The present invention proposes a new parameter setting method, which can reduce the gradient oscillation problem and speed up the convergence speed by dynamically adjusting β1. The specific method is:

[0109] First, the gradient descent rate of change G is detected during the gradient descent process, and its calculation formula is:

[0110]

[0111] Secondly, set the gradient descent rate to be a gentle gradient state when -1<G<1, and record the number of steps G in the continuous gentle gradient state t , that is, the number of consecutive times -1<G<1;

[0112] Then, according to the continuous gradient smoothing step number G t Set the β1 parameter. The specific setting values ​​are shown in Table 1 below, where G t The relationship between and β1 is determined by the critical value 1 / e (e is a mathematical constant) of the exponentially weighted moving average method, that is, when When , it is determined that the data within the flat step number has a positive impact on the subsequent data.

[0113] Table 1 Continuous gradient smoothing steps G t and β1 parameter relationship

[0114]

[0115] The loss function of the BP neural network adopts the mean square error function, and the formula of the loss value loss is:

[0116]

[0117] Among them, loss and Y K is the predicted value, is the true value, and N is the number of data.

[0118] 2.3) The experimental data between grinding parameters and machining targets were obtained through orthogonal experimental method, 80% of the data were used as training set, 20% of the data were used as test set, and all the data were normalized.

[0119] The purpose of data normalization is to reduce the variance of data before neural network training, prevent gradient oscillation during the convergence process, and improve training accuracy. The column vector normalization process is the same as the row vector normalization process, both using the peak method:

[0120]

[0121] in, is the normalized value, X i is the value before normalization, X min is the minimum value of this type of data, X max This is the maximum value of this type of data.

[0122] 2.4) Based on the obtained normalized training set and test set, the established BP neural network is trained with the determined activation function, optimizer and loss function until the training meets the error requirement, thereby obtaining a trained BP neural network.

[0123] Among them, the method of using activation function, optimizer and loss function to train the established BP neural network is a well-known technology for those skilled in the art, and the present invention will not be repeated here.

[0124] 3) Data prediction: The grinding parameters under the working conditions to be tested are input into the trained BP neural network to obtain the surface processing quality prediction results of the composite workpiece corresponding to the grinding parameters.

[0125] like Figure 3 and Figure 4 As shown, Figure 3 This is an illustration of the trajectory of the grinding head when grinding a flat surface. The grinding path is indicated by the arrow in the figure. Figure 4 This is a simulation diagram of the surface shape after constant force grinding, where the Y-axis direction is the grinding track direction, and the deeper along the Z-axis, the deeper the grinding depth.

[0126] Example 2

[0127] The above-mentioned embodiment 1 provides a robot constant force grinding optimization method, and correspondingly, this embodiment provides a robot constant force grinding optimization system. The system provided in this embodiment can implement a robot constant force grinding optimization method of embodiment 1, and the system can be implemented by software, hardware, or a combination of software and hardware. For example, the system may include integrated or separate functional modules or functional units to perform the corresponding steps in each method of embodiment 1. Since the system of this embodiment is basically similar to the method embodiment, the process described in this embodiment is relatively simple, and the relevant parts can refer to the partial description of embodiment 1. The embodiment of the system provided in this embodiment is only schematic.

[0128] This embodiment provides a robot constant force grinding optimization system, including:

[0129] The optimization model building module is used to establish the robot constant force grinding optimization model and determine its constraint conditions, wherein the optimization model is based on the feed speed v f , grinding speed v s , constant pressure F, grinding track spacing D are grinding parameter variables, surface roughness and grinding uniformity are objective functions, working condition variable parameters and sandpaper durability are constraints

[0130] A neural network module is used to establish a BP neural network based on a variable history influencing factor based on a determined robot constant force grinding optimization model. The BP neural network uses feed speed, grinding speed, constant pressure, and grinding trajectory as input vectors, and uses surface roughness and grinding uniformity as output vectors;

[0131] The prediction module is used to input the grinding parameters under the working conditions to be tested into the trained BP neural network to obtain the surface processing quality prediction results of the composite material workpiece corresponding to the grinding parameters.

[0132] Example 3

[0133] This embodiment provides a processing device corresponding to a robot constant force grinding optimization method provided in this embodiment 1. The processing device can be a processing device for a client, such as a mobile phone, a laptop computer, a tablet computer, a desktop computer, etc., to execute the method of embodiment 1.

[0134] The processing device includes a processor, a memory, a communication interface and a bus, and the processor, the memory and the communication interface are connected through the bus to complete mutual communication. The memory stores a computer program that can be run on the processor, and the processor executes a robot constant force grinding optimization method provided in this embodiment 1 when running the computer program.

[0135] In some embodiments, the memory may be a high-speed random access memory (RAM), and may also include a non-volatile memory, such as at least one disk memory.

[0136] In other embodiments, the processor may be a central processing unit (CPU), a digital signal processor (DSP), or other general-purpose processors of various types, which are not limited herein.

[0137] Example 4

[0138] A robot constant force grinding optimization method of this embodiment 1 can be specifically implemented as a computer program product. The computer program product may include a computer-readable storage medium on which computer-readable program instructions for executing the robot constant force grinding optimization method described in this embodiment 1 are loaded.

[0139] Computer readable storage media can be tangible devices that hold and store instructions used by instruction execution devices. Computer readable storage media can be, for example, but not limited to, electronic storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any combination thereof.

[0140] It should be noted that the flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the systems, methods and computer program products according to multiple embodiments of the present application. Each box in the flowchart or block diagram may represent a module, a program segment or a part of the code, and the module, program segment or a part of the code contains one or more executable instructions for implementing the specified logical functions.

[0141] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

[0142] The above embodiments are only used to illustrate the present invention, wherein the structure, connection mode and manufacturing process of each component may be changed. Any equivalent transformations and improvements based on the technical solution of the present invention should not be excluded from the protection scope of the present invention.

Claims

1. A robot constant force grinding optimization method, characterized in that: The following steps are involved: Establish a robot constant force grinding optimization model and determine its constraints; Based on the determined robot constant force grinding optimization model, a BP neural network based on variable history influencing factors is established; The grinding parameters under the working condition to be tested are input into the trained BP neural network to obtain the surface processing quality prediction result of the composite material workpiece corresponding to the grinding parameters; The method for establishing a robot constant force grinding optimization model and determining its constraint conditions comprises: The characteristics of robot constant force grinding of composite workpieces are analyzed to determine the factors affecting the surface roughness and grinding uniformity of composite workpieces after forming, including feed speed , Grinding speed , constant pressure F, grinding track spacing D; According to the determined influencing factors, an optimization function with surface roughness and grinding uniformity as the target is established; Determine the constraint conditions of the objective optimization function, which include working condition variable constraint and sandpaper durability constraint; The surface roughness calculation formula is: in, is the surface roughness; is the proportionality coefficient; b, c, is the coefficient; is the feed speed; is the grinding speed; is the Preston constant; It is the intersection line between the grinding head and the previous grinding track; is a constant pressure; It is the overlapping part of two trajectories; The grinding uniformity is characterized by the grinding error RE between the lowest point and the highest point in the grinding track, and its calculation formula is: Where R is the radius of the grinding head; To polish the bottom area of ​​the head; It is the shortest intersection line between the grinding head and the object being ground during its forward movement; It is the maximum contact area between the grinding head and the object being ground during the forward movement when there is an overlapping area; The method of establishing a BP neural network based on a variable history influencing factor based on a determined robot constant force grinding optimization model includes: Determine the BP neural network structure; Determine the activation function, optimizer, and loss function; The experimental data between grinding parameters and machining targets are obtained by orthogonal experimental method, a preset proportion of data is used as a training set, the remaining data is used as a test set, and all data are normalized; Based on the obtained normalized training set and test set, the established BP neural network is trained with the determined activation function, optimizer and loss function until the training reaches the error requirement, thereby obtaining a trained BP neural network; The activation function of each layer of the BP neural network adopts the RELU activation function; the loss function adopts the mean square error function, the optimizer adopts the Nadam optimizer, and the calculation formula is: In the formula, is the weight value at time t; is the weight value at time t-1; is the amount of change in weight at each iteration; is the time t sum; is the time t-1 sum; and is the historical impact factor coefficient; is the learning rate; is the cumulative square gradient; is the cumulative square gradient of the previous moment; is a constant; is the coefficient; is the weight value at the next moment; t represents the moment; The historical impact factor coefficient Determined using the following method: First, detect the gradient descent rate of change during the gradient descent process , and its calculation formula is: Secondly, set the gradient to be in a gentle state when the gradient descent rate meets the preset conditions, and record the number of steps in the continuous gentle gradient state. ; Finally, according to the number of recorded continuous gradient smoothing steps set up Parameters, among which, and The relationship between the critical value of the exponentially weighted moving average method Sure.

2. A robot constant force grinding optimization method as claimed in claim 1, characterized in that: The BP neural network includes an input layer, a hidden layer and an output layer. The input layer is based on the feed rate. , Grinding speed , constant pressure F and grinding track spacing D are input vectors; the output layer takes surface roughness Ra and grinding uniformity RE as output vectors; the hidden layer is a single hidden layer, and the number of nodes is set to a preset value.

3. A robot constant force grinding optimization system, characterized in that: include: Optimization model building module, used to build the robot constant force grinding optimization model and determine its constraint conditions; A neural network module is used to establish a BP neural network based on variable history influence factors based on a determined robot constant force grinding optimization model; A prediction module is used to input the grinding parameters under the working condition to be tested into the trained BP neural network to obtain the surface processing quality prediction results of the composite material workpiece corresponding to the grinding parameters; The method of establishing a robot constant force grinding optimization model and determining its constraint conditions includes: The characteristics of robot constant force grinding of composite workpieces are analyzed to determine the factors affecting the surface roughness and grinding uniformity of composite workpieces after forming, including feed speed , Grinding speed , constant pressure F, grinding track spacing D; According to the determined influencing factors, an optimization function with surface roughness and grinding uniformity as the target is established; Determine the constraint conditions of the objective optimization function, which include working condition variable constraint and sandpaper durability constraint; The surface roughness calculation formula is: in, is the surface roughness; is the proportionality coefficient; b, c, is the coefficient; is the feed speed; is the grinding speed; is the Preston constant; It is the intersection line between the grinding head and the previous grinding track; is a constant pressure; It is the overlapping part of two trajectories; The grinding uniformity is characterized by the grinding error RE between the lowest point and the highest point in the grinding track, and its calculation formula is: Where R is the radius of the grinding head; To polish the bottom area of ​​the head; It is the shortest intersection line between the grinding head and the object being ground during its forward movement; It is the maximum contact area between the grinding head and the object being ground during the forward movement when there is an overlapping area; The robot constant force grinding optimization model determined based on the above method establishes a BP neural network based on variable historical influencing factors, including: Determine the BP neural network structure; Determine the activation function, optimizer, and loss function; The experimental data between grinding parameters and machining targets are obtained by orthogonal experimental method, a preset proportion of data is used as a training set, the remaining data is used as a test set, and all data are normalized; Based on the obtained normalized training set and test set, the established BP neural network is trained with the determined activation function, optimizer and loss function until the training reaches the error requirement, thereby obtaining a trained BP neural network; The activation function of each layer of the BP neural network adopts the RELU activation function; the loss function adopts the mean square error function, the optimizer adopts the Nadam optimizer, and the calculation formula is: In the formula, is the weight value at time t; is the weight value at time t-1; is the amount of change in weight at each iteration; is the time t sum; is the time t-1 sum; and is the historical impact factor coefficient; is the learning rate; is the cumulative square gradient; is the cumulative square gradient of the previous moment; is a constant; is the coefficient; is the weight value at the next moment; t represents the moment; The historical impact factor coefficient Determined using the following method: First, detect the gradient descent rate of change during the gradient descent process , and its calculation formula is: Secondly, set the gradient to be in a gentle state when the gradient descent rate meets the preset conditions, and record the number of steps in the continuous gentle gradient state. ; Finally, according to the number of recorded continuous gradient smoothing steps set up Parameters, among which, and The relationship between the critical value of the exponentially weighted moving average method Sure.

4. A processing device, the processing device comprising at least a processor and a memory, the memory storing a computer program, characterized in that: When the processor runs the computer program, the computer program is executed to implement the steps of the robot constant force grinding optimization method according to any one of claims 1 to 2.

5. A computer storage medium, characterized in that: Computer-readable instructions are stored thereon, and the computer-readable instructions can be executed by a processor to implement the steps of the robot constant force grinding optimization method according to any one of claims 1 to 2.

Images