CN108563831B - An optimization method for transmission precision of RV reducer - Google Patents

Abstract

An optimization method for transmission accuracy of an RV reducer comprises the following steps: determining influencing factors of transmission accuracy of the RV reducer, including component error, component matching clearance and working load; constructing an orthogonal test scheme in combination with the influencing factors; using three-dimensional modeling software CREO and multi-body dynamics simulation software ADAMS to construct a virtual prototype experimental group according to the orthogonal test scheme; measuring the transmission error of each virtual prototype experimental group in the simulation software Adams; analyzing the orthogonal test results; and using the orthogonal test result data as input end data of a BP neural network to predict the optimal combination.

Description

An optimization method for transmission precision of RV reducer

technical field

The invention relates to a method for optimizing the transmission accuracy of a high-precision robot RV reducer.

Background technique

Due to the manufacturing errors and assembly errors of the RV reducer parts, as well as the existence of temperature deformation and elastic deformation in the transmission process, the input and output rotation angle errors are unavoidable. Rotation angle error refers to the deviation between the actual rotation angle of the output shaft and the theoretical rotation angle, and is an important indicator for evaluating the transmission accuracy of the RV reducer. The application fields of RV reducer are mostly precision transmission devices that require high transmission accuracy, such as robots, radars, precision machine tools, etc. In order to ensure the accuracy of the position of the transmission device when it completes the same cycle of motion multiple times, RV reducer Must have high transmission accuracy.

With the widespread application of robots, more and more attention has been paid to the research on RV reducers, especially the optimization of transmission accuracy. There are many parts of RV reducer and the requirements for processing accuracy are high, which is limited to the requirements of processing and manufacturing costs. When optimizing the transmission accuracy, it is difficult to optimize the manufacturing error and matching error of each part one by one. The sensitivity to the influence of precision is the premise to reasonably improve the transmission precision of RV reducer. According to the sensitivity of the factors affecting the transmission accuracy of the RV reducer, combined with the production and processing costs, the transmission accuracy of the RV reducer can be improved leanly.

When using manual measurement and experiments to study the transmission accuracy of RV reducers, measurement errors will inevitably occur. With the accumulation of errors, the accuracy of the research results is difficult to guarantee, and the experiment cost is high and the cycle is long. The application of virtual prototyping technology just solved this problem. The three-dimensional modeling software was used to construct the three-dimensional model according to the part error, and the model was imported into the simulation software for motion simulation, which eliminated the influence of other noise factors and improved the accuracy of the experiment. The experiment cost is saved and the experiment cycle is shortened.

Contents of the invention

The present invention aims to solve the problem that it is difficult to distinguish the degree of influence of factors such as part error, part fit clearance, and working load on the transmission accuracy of the RV reducer in the process of optimizing the transmission accuracy of the RV reducer. It is difficult to find out the shortcomings of the optimal combination of transmission accuracy, and an optimization method for RV reducer transmission accuracy based on virtual prototype technology and BP neural network is provided.

In order to solve the above-mentioned technical problems, the present invention provides a method for optimizing the transmission accuracy of the RV speed reducer based on virtual prototype technology and BP neural network, comprising the following steps:

S1. Factors affecting the transmission accuracy of the RV reducer mainly include three aspects: part error, part fit clearance, and working load;

S11. Select the RV40-E type reducer as the research object, and select some part errors as the test factors according to the transmission relationship of the parts;

S12. Analyze the influence relationship between the clearance of parts in the RV reducer and the transmission accuracy, and select the matching clearance of several major parts as the test factor;

S2. Combining the influencing factors to construct an orthogonal test plan;

S3. Using the 3D modeling software CREO and the multi-body dynamics simulation software ADAMS to construct a virtual prototype experiment group according to the orthogonal test plan;

S31. Use the 3D modeling software CREO to establish the virtual prototype model of the RV reducer according to the part error in the orthogonal test plan. Since it is difficult to build a complex 3D model in ADAMS, use CREO to establish the RV reducer virtual prototype experimental group according to the orthogonal test plan , some parts are simplified during the modeling process, and the cycloidal equation is established in CREO to obtain the cycloidal gear profile curve. The cycloidal equation is as follows:

Among them, rz is the radius of the center circle of the pin tooth, zb is the number of pin teeth, rzz is the radius of the pin tooth, za is the number of cycloidal gear teeth, e is the eccentricity, drz is the distance training amount, and drzz is the equidistant practice amount;

The half-tooth profile of the complete cycloidal gear is drawn through the equation, and then the three-dimensional model of the cycloidal gear can be obtained through the mirror image, array, and stretch commands. The establishment method of other solid models is similar to this. Assemble all the solid models of the parts, and then conduct a non-interference inspection on the assembly, and save the assembly as the CREO and ADAMS intermediate file format parasolid (*.x_t) after confirming that the assembly is correct and there is no part interference.

S32. Import the virtual prototype model into the multi-body dynamics simulation software to define the material properties; import the model file in the intermediate format into the ADAMS software through the File/Import command, and define the material of the parts. Based on the ADAMS contact collision theory, the parts The elastic deformation generated during contact and collision will also affect the rotation error of the RV reducer, so the elastic modulus, density, Poisson's ratio and other material properties of each part of the model are added.

S33. Add constraints to the virtual prototype in ADAMS; in order to ensure the correctness of the relative motion of each component, the construction of the virtual prototype also needs to provide constraints or contact relationships according to the trajectory of the components, through the motion and contact analysis of the RV reducer components , to determine the constraints and contact relationships of each component.

S34. Simulate the virtual prototype to verify whether the constructed model is correct; measure the ratio between the input shaft and the planet carrier speed and compare it with the theoretical transmission ratio of the RV reducer to verify whether the model is accurate.

S4, measure the transmission error of each virtual prototype experimental group in the simulation software Adams;

The rotation error is selected as the evaluation index of the transmission accuracy, the transmission ratio of the selected RV reducer test model is 121, according to the calculation formula of the rotation error

In the formula, is the corner error; Enter the rotation angle for the input shaft (i.e. the sun gear); is the actual rotation angle of the output shaft (that is, the planet carrier); i is the transmission ratio of the reducer.

In ADAMS, the rotation angles of the input shaft and the output rack are measured in real time, and the output curves of the rotation angles are shown in Figure 5(a) and (b). Create a measurement function

FUNCTION＝.JOINT_1_MEA_1/121—.JOINT_16_MEA_1 In the formula, FUNCTION is the difference between the actual output rotation angle and the theoretical output rotation angle, that is, the rotation error; JOINT_1_MEA_1 is the rotation angle of the input shaft; .JOINT_16_MEA_1 is the rotation angle of the planet carrier.

According to the orthogonal test table, the virtual prototype model is established and imported into ADAMS for simulation, the rotation error of each group is measured, and the orthogonal test result table is perfected.

S5, analyzing the results of the orthogonal test;

Carry out a range analysis on the results of the orthogonal test. The range analysis uses the data range to analyze the problem. By comparing the average range of the experimental results, the main factors affecting the test indicators are found. The principle is that when considering the influence of a single factor A on the result, the influence of other factors on the result is considered to be balanced, and the differences in the levels of the A factor are caused by the A factor itself. The larger the range, the greater the influence of the factor on the experimental index. The range R of each factor is calculated by the following formula.

R＝max{k _i }-min{k _i }

T=∑k _i

In the formula, R is the range; i is the level number of the factor; k _i is the mean value of the sum of the rotation errors corresponding to the i level; K _i is the sum of the rotation errors corresponding to the i level; The number of occurrences of each level; T is the sum of rotation errors;

According to the extreme difference of each influencing factor, the optimal combination of the sensitivity degree of each factor affecting the dynamic rotation error of RV reducer from large to small sequence and factor level is obtained.

S6. Using the orthogonal test result data as the input end data of the BP neural network to predict the optimal combination.

S61, establishment and training of BP artificial neural network

The main factors affecting the rotation error of the RV reducer in the orthogonal test design are used as the input layer of the built BP neural network, and the output layer of the network includes an output node corresponding to the evaluation index rotation error. Using MATLAB software for programming, select hidden layer transfer function, output layer transfer function, training function; set hidden layer, training accuracy, learning rate and other parameters, and compare the training of networks with different numbers of neurons.

S62. Using BP artificial neural network to predict the optimal combination;

In the previous research, the application of MATLAB software programming, through the training of experimental data samples and the optimization of BP neural network parameters, successfully established a BP neural network model that can accurately describe the functional relationship between the rotation error of the RV reducer and its evaluation index , use the established network model to carry out simulation, take the values of several influencing factors of the orthogonal test as independent variables, and then assign values to them respectively, set an appropriate step size, and use the correlation function in MATLAB to define the domain of each factor The value is programmed in order to find the minimum combined value of its output value slew error.

S63. Optimization result inspection

According to the optimization results in the previous steps, a virtual prototype is established, and imported into ADAMS for multi-body dynamics simulation, the rotation error of the virtual prototype is measured, and compared with the results of the orthogonal test to obtain the optimization effect of transmission accuracy.

A method for optimizing transmission accuracy of an RV reducer based on virtual prototype technology and BP neural network of the present application has the following beneficial effects:

1. Using the method of the present invention solves the problem that in the process of optimizing the transmission accuracy of RV reducers, it is impossible to optimize the design of each factor due to the limitation of processing costs, and it is impossible to distinguish the importance of each influencing factor on the transmission accuracy. Further leanly improve the transmission accuracy of RV reducers , To provide a basis for the development of tolerances for parts.

2. Through modeling and simulation, it avoids the influence of measurement errors and other noise factors in the process of artificial measurement experiments, making the experimental results more accurate.

3. In the method of the present invention, the transmission accuracy of the RV reducer is studied by combining the virtual simulation technology and the BP neural network algorithm, and the accurate prediction of the combination of factors for the optimal transmission accuracy is realized, and no physical prototype test is required, which saves research cost and time , greatly improving the research efficiency.

Description of drawings

Fig. 1 is a flow chart of the method of the present invention.

Figure 2 is a model diagram of the RV reducer assembly

Figure 3 is a virtual prototype (with hidden planet carrier)

Figure 4a is the speed curve of the input shaft, and Figure 4b is the speed curve of the planetary carrier of the output mechanism

Figure 5a is the input rotation angle curve, Figure 5b is the output rotation angle curve, and Figure 5c is the virtual prototype 1 rotation angle error curve

Figure 6 is a comparison chart of the original data and the predicted value of the neural network

Detailed ways

The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

The present invention provides a method for optimizing transmission accuracy of an RV reducer based on virtual prototype technology and BP neural network. The process is shown in Figure 1, including the following steps:

S1. Factors affecting the transmission accuracy of the RV reducer mainly include three aspects: part error, part fit clearance, and working load;

S11. Select the RV40-E type reducer as the research object. According to the transmission relationship of the components, the two-stage cycloidal transmission is the main research part. Combined with the part error part of the "Practical Gear Design Calculation Manual", the crankshaft eccentricity and cycloidal wheel are selected. There are 5 influencing factors: distance training amount, cycloid wheel equidistant training amount, pin tooth radius error, and pin tooth center circle radius error;

S12. Analyze the influence relationship of the parts clearance on the transmission accuracy in the RV reducer. Among them, the parts clearance mainly includes the meshing clearance of the sun gear and the planetary gear, the bearing clearance between the crank shaft hole of the cycloid wheel and the crank shaft, and the crank shaft of the planet carrier. The bearing clearance between the hole and the crankshaft, the bearing clearance between the planet carrier and the frame, the meshing clearance of the cycloid pin teeth, and the matching clearance between the pin teeth and the pin gear disc. In the virtual prototype model, the bearing is replaced by a shaft sleeve, and the pin tooth and the pin tooth disc are consolidated, so the clearance of the parts is selected from the clearance between the inner hole of the cycloid wheel and the sleeve of the rotating arm, the clearance between the crank shaft and the sleeve of the rotating arm, and the clearance of the crank shaft The clearance between the supporting bushing, the clearance between the supporting bushing and the planetary carrier, and the clearance between the supporting bushing and the flange are the test factors, and a total of 11 influencing factors are selected together with the working load;

S2. Combining the influencing factors to construct an orthogonal test plan;

S21. Formulate a factor level table. In order to analyze the influence of factors such as part processing error, part fit clearance, and load on the dynamic rotation error of the RV reducer, a total of 11 factors were selected as test factors, and the orthogonal table L (3 ¹³ ) was selected. The factor level table is shown in Table 1. Show;

Table 1

S22. Combining the factor level table to formulate an orthogonal test plan as shown in Table 2; Table 2

S3. Combining the 3D modeling software CREO and the multi-body dynamics simulation software ADAMS to build a virtual prototype according to the orthogonal test plan;

S31. Use the 3D modeling software CREO to establish 27 sets of RV reducer virtual prototype models according to the part error in the orthogonal test plan; because it is difficult to build a complex 3D model in ADAMS, use CREO to establish the RV reducer virtual prototype according to the orthogonal test plan In the experimental group, some components were simplified during the modeling process, including: bearings were replaced by bushings, fine structures such as chamfered bolts were ignored, and pins, keys, and bolts were replaced by consolidation; sealing rings, gaskets, etc. were removed. Research unaffected parts. Establish the cycloid equation in CREO to obtain the cycloid gear profile curve, the cycloid equation is as follows:

x=(rz+drz)*(sin(360*t)-(k1/zb)*sin(zb*360*t))+(rzz

+drzz)*(-sin(360*t)+k1*sin(zb*360*t))/sqrt(1

+k1*k1-2*k1*cos(za*360*t))

y=(rz+drz)*(cos(360*t)-(k1/zb)*cos(zb*360*t))-(rzz

+drzz)*(cos(360*t)-k1*cos(zb*360*t))/sqrt(1

+k1*k1-2*k1*cos(za*360*t))

k1=e*zb/(rz+drz)

Among them, rz is the radius of the center circle of the pin tooth, zb is the number of pin teeth, rzz is the radius of the pin tooth, za is the number of cycloidal gear teeth, e is the eccentricity, drz is the distance training amount, and drzz is the equidistant practice amount;

The half tooth profile of the complete cycloidal gear is drawn through the cycloidal equation, and then the three-dimensional model of the cycloidal gear can be obtained through mirror image, array, and stretching commands. The establishment method of other solid models is similar to this. Assemble all the solid models of the parts to obtain the assembly model of the RV reducer as shown in Figure 2. Then carry out the non-interference inspection on the assembly, and save the assembly as parasolid (*.

S32. Import the virtual prototype model into the multi-body dynamics simulation software to define the material properties; import the model file in the intermediate format into the ADAMS software through the File/Import command, and define the material of the parts. Based on the ADAMS contact collision theory, the parts The elastic deformation generated during contact and collision will also affect the rotation error of the RV reducer, so the material properties such as elastic modulus, density, and Poisson's ratio of each part of the added model are shown in Table 3. table 3

S33. Add constraints to the virtual prototype in ADAMS; in order to ensure the correctness of the relative motion of each component, the construction of the virtual prototype also needs to provide constraints or contact relationships according to the trajectory of the components, through the motion and contact analysis of the RV reducer components , determine the constraints and contact relationships of each component, as shown in Table 4.

Table 4

S34. Simulate the virtual prototype to verify whether the constructed model is correct; import the intermediate format model file of the assembly established by CREO into the ADAMS system, and add the material properties and constraint relationships of the parts to obtain the virtual as shown in Figure 3 (hidden planet carrier) prototype. The virtual prototype contains 58 parts, 18 constraints and 94 contacts. Set the rotary drive input speed function F(time)=7000d*time*step(time,0,1,0,1), set the load torque according to the test plan, and define the load torque function F(time)=step(time,1, 1.5,0,X), X depends on the data of the experimental group. Self-defined 4S simulation time, 100 steps of simulation steps. The measured speed curves of the input shaft and the planetary carrier are shown in Figure 4a and Figure 4b. After 1.5s, the model motion is stable, the input shaft speed is 7000°/s, the average speed of the planetary carrier is 57.7445°/s, and the ratio between the two is 121.2236. It is consistent with the theoretical transmission ratio of 121, which proves that the model is accurate and reliable.

S4, measure the transmission error of each virtual prototype experimental group in the simulation software Adams;

The rotation error is selected as the evaluation index of the transmission accuracy, the transmission ratio of the selected RV reducer test model is 121, according to the calculation formula of the rotation error

In the formula, is the corner error; Enter the rotation angle for the input shaft (i.e. the sun gear); is the actual rotation angle of the output shaft (that is, the planet carrier); i is the transmission ratio of the reducer.

In ADAMS, the rotation angles of the input shaft and the output frame are measured in real time, and the output curves of the rotation angles are shown in Figure 5a and Figure 5b. Create a measurement function

FUNCTION＝.JOINT_1_MEA_1/121—.JOINT_16_MEA_1 In the formula, FUNCTION is the difference between the actual output rotation angle and the theoretical output rotation angle, that is, the rotation error; JOINT_1_MEA_1 is the rotation angle of the input shaft; .JOINT_16_MEA_1 is the rotation angle of the planet carrier.

According to the orthogonal experiment table, the virtual prototype model was established and imported into ADAMS for simulation, and the rotation errors of each group were measured. Figure 5c shows the rotation error curve of virtual prototype 1. Since there are many experimental groups, the experimental curves of other groups will not be listed one by one.

Since the load and drive are loaded at a uniform speed within 0 to 1.5s, the motion state of the model is unstable, so the mean value of the error curve in the time period of 1.5 to 4s is selected as the evaluation index, and the results of the perfect orthogonal test are shown in Table 2.

S5, analyzing the results of the orthogonal test;

Carry out a range analysis on the results of the orthogonal test. The range analysis uses the data range to analyze the problem. By comparing the average range of the experimental results, the main factors affecting the test indicators are found. The principle is that when considering the influence of a single factor A on the result, the influence of other factors on the result is considered to be balanced, and the differences in the levels of the A factor are caused by the A factor itself. The larger the range, the greater the influence of the factor on the experimental index. The range R of each factor is calculated by the following formula.

R＝max{k _i }-min{k _i }

T=∑k _i

In the formula, R is the range; i is the level number of the factor; k _i is the mean value of the sum of the rotation errors corresponding to the i level; K _i is the sum of the rotation errors corresponding to the i level; The number of occurrences of each level; T is the sum of rotation errors;

The results of the range analysis of the test results are shown in Table 2. According to the extreme difference of each influencing factor, it can be concluded that the sensitivity of each factor affecting the dynamic rotation error of the RV reducer from large to small is the pin tooth center circle radius error, pin tooth radius error, cycloid wheel travel distance practice amount, load size, and crank Clearance between shaft and arm bushing, equidistant distance of cycloidal wheel, clearance between upper support bushing and planetary carrier, eccentric error of crankshaft, clearance between crankshaft and load-bearing bushing, clearance between inner hole of cycloidal wheel and pivoting arm bushing, The gap between the lower support bushing and the flange; the optimal combination of factor levels is A ₁ B ₁ C ₃ D ₃ E ₁ F ₃ G ₂ H ₂ I ₃ J ₁ K ₁ .

S6. Using the orthogonal test result data as the input end data of the BP neural network to predict the optimal combination.

S61, establishment and training of BP artificial neural network

Taking the main factors affecting the rotation error of the RV reducer in the orthogonal test design as the input layer of the built BP neural network, it contains 11 input nodes corresponding to the pin tooth radius error, the pin tooth center circle radius error, the crankshaft eccentricity error, the pendulum Modification amount of wire wheel distance, equidistant modification amount of cycloidal wheel, clearance between inner hole of cycloidal wheel and pivot arm bushing, clearance between crankshaft and pivoting arm bushing, clearance between crankshaft and support bushing, upper support bushing The gap between the planet carrier, the gap between the lower support bushing and the flange, and the working load. The output layer of the network contains an output node corresponding to the evaluation index rotation error. MATLAB (version R2016a, USA) software was used for programming, the hyperbolic tangent transfer function (tansig) was selected as the hidden layer transfer function, the linear transfer function (purelin) was used as the output layer transfer function, and one hidden layer was set. Networks with different numbers of neurons were trained and compared, and 23 neurons were determined to be included. The newly-built BP network was trained using the traindm function, and the network training parameter values were set. The maximum number of training times was 100, the training accuracy was 0.0001, and the learning rate was 0.1, other parameters are default values. Figure 6 is a comparison chart between the original data and the predicted value by the neural network. It can be seen from the figure that the two have a high degree of coincidence, which proves that the BP neural network can be used to predict the transmission accuracy of the RV reducer, and the result is relatively accurate.

S62, BP artificial neural network combined with orthogonal test to optimize factor parameters

In the previous research, the application of MATLAB software programming, through the training of experimental data samples and the optimization of BP neural network parameters, successfully established a BP neural network model that can accurately describe the functional relationship between the rotation error of the RV reducer and its evaluation index , using the established network model to carry out the simulation, taking the values of 11 influencing factors A, B, C, D, E, F, G, H, I, J, K of the orthogonal test as independent variables, and then the 11 Assign a value to each factor, set an appropriate step size, and use the relevant function in MATLAB to program the domain value of each factor to find the minimum combined value of the output value rotation error. After the BP artificial neural network model simulation optimization, the combination of factors for the minimum rotation error of the RV reducer is: pin tooth radius error 0.01mm, pin tooth center circle radius error 0.01mm, crankshaft eccentricity error 0.1mm, cycloid wheel distance modification 0.03mm, equidistant modification of cycloid wheel 0.01mm, gap between cycloid wheel inner hole and arm bushing 0.01mm, crankshaft and pivot arm bushing gap 0.0042mm, crankshaft and support bushing gap 0.0058, upper The gap between the supporting bushing and the planet carrier is 0.01, the gap between the lower supporting bushing and the flange is 0.0012, and the working load is 300N·mm.

S63. Optimization result inspection

Establish a virtual prototype based on the optimization results in the previous steps, and import it into ADAMS for multi-body dynamics simulation. It is obtained that the rotation error of the virtual prototype is 0.0512°, which is 16.2% higher than the optimal 0.0611° in the orthogonal test results. , which proves that the inventive method has a significant effect on optimizing the transmission accuracy of the RV reducer.

A method for optimizing transmission accuracy of an RV reducer based on virtual prototype technology and BP neural network of the present application has the following beneficial effects:

1. The method of the present invention solves the difficult problem that in the process of optimizing the transmission accuracy of the RV reducer, the processing cost cannot optimize the design of each factor, and it is difficult to distinguish the importance of each influencing factor on the transmission accuracy. In order to further improve the RV reducer The transmission accuracy provides a new method.

2. Through modeling and simulation, it avoids the influence of measurement errors and other noise factors in the process of artificial measurement experiments, making the experimental results more accurate.

3. In the method of the present invention, the transmission accuracy of the RV reducer is studied by combining the virtual simulation technology and the BP neural network algorithm, and the accurate prediction of the optimal transmission accuracy factor combination is realized, and no physical prototype test is required, which saves research cost and time , greatly improving the research efficiency.

The content described in the embodiments of this specification is only an enumeration of the implementation forms of the inventive concept. The protection scope of the present invention should not be regarded as limited to the specific forms stated in the embodiments. Equivalent technical means that a person can think of based on the concept of the present invention.

Claims (1)

1. A method for optimizing the transmission precision of an RV reducer comprises the following steps:

s1, determining influence factors of the transmission precision of the RV reducer, wherein the influence factors comprise part errors, part fit clearances and working loads;

s11, selecting an RV40-E type speed reducer as a research object, and selecting a plurality of part errors as influence factors according to the transmission relation of parts;

s12, analyzing the influence relation of the part clearance in the RV reducer on the transmission precision, and selecting a plurality of main part fit clearances as influence factors;

s2, constructing an orthogonal test scheme by combining the influence factors;

s3, constructing a virtual prototype experimental group by adopting three-dimensional modeling software CREO and multi-body dynamics simulation software ADAMS according to an orthogonal experimental scheme;

s31, establishing a virtual prototype model of the RV reducer according to part errors in an orthogonal test scheme by adopting three-dimensional modeling software CREO, and establishing a virtual prototype experimental group of the RV reducer according to the orthogonal test scheme by adopting the CREO because a complex three-dimensional model is difficult to construct in ADAMS, simplifying partial parts in the modeling process, and establishing a cycloid equation in the CREO to obtain a cycloid wheel tooth profile curve, wherein the cycloid equation is as follows:

wherein rz is the radius of the central circle of the pin gear, zb is the number of the pin gear teeth, rzz is the radius of the pin gear teeth, za is the number of the cycloid gear teeth, e is the eccentric distance, drz is the displacement repair quantity, and drzz is the equidistant repair quantity;

drawing a half tooth profile of the complete cycloid wheel through an equation, then obtaining a three-dimensional model of the cycloid wheel through mirror image, array and stretching commands, and establishing methods of other solid models are similar to the three-dimensional model; assembling all part solid models, then carrying out interference-free inspection on the assembly body, and storing the assembly body as a CREO and ADAMS intermediate file format parasolid (. x _ t) after determining that the assembly is correct and no part interference exists;

s32, importing the virtual prototype model into multi-body dynamic simulation software to define material characteristics; importing the model File in the intermediate format into ADAMS software through a File/Import command, defining the material of the part, and based on the ADAMS contact collision theory, the elastic deformation generated during the contact collision of the part also influences the rotation error of the RV reducer, so that the elastic modulus, the density and the Poisson ratio material characteristics of each part of the model are added;

s33, adding a constraint relation to the virtual prototype in the ADAMS; in order to ensure the correctness of the relative motion of each part, a virtual prototype is constructed by providing a constraint or contact relation according to the motion track of the part, and the constraint and contact relation of each part is determined through the motion and contact analysis of the parts of the RV reducer;

s34, simulating the virtual prototype, and verifying whether the constructed model is correct; measuring the ratio of the rotating speed of the input shaft to the rotating speed of the planet carrier, comparing the ratio with the theoretical transmission ratio of the RV reducer, and verifying whether the model is accurate or not;

s4, measuring the transmission error of each virtual prototype experimental group in simulation software Adams;

selecting the rotation error as an evaluation index of the transmission precision, selecting the RV reducer test model with the transmission ratio of 121, and calculating a formula according to the rotation error

In the formula (I), the compound is shown in the specification,is a corner error;inputting a rotation angle for the input shaft;is the actual rotation angle of the output shaft; i is the transmission ratio of the speed reducer;

in ADAMS, the rotation angles of an input shaft and an output frame are measured in real time, the rotation angle curves of the input shaft and the output frame are output, and a measurement function is established

FUNCTION＝.JOINT_1_MEA_1/121—.JOINT_16_MEA_1

In the formula, FUNCTION is a difference value between an actual output rotation angle and a theoretical output rotation angle, namely a rotation error; JOINT _1_ MEA _1 is the rotation angle of the input shaft revolute pair; JOINT _16_ MEA _1 is the rotating auxiliary angle of the planet carrier;

respectively establishing virtual prototype models according to the orthogonal test table, introducing ADAMS simulation, measuring each group of rotation errors, and perfecting the orthogonal test result table;

s5, analyzing the results of the orthogonal test;

performing linear range analysis on orthogonal test results, wherein the linear range analysis is to analyze problems by using data range, and finding out main factors influencing test indexes by comparing average range of each test result; the principle is that when the influence of a single factor A on the result is considered, the influence of other factors on the result is considered to be balanced, and the difference of the levels of the factor A is caused by the factor A; the larger the range is, the larger the influence of the factor on the experimental index is, and the range R of each factor is calculated by the following formula;

R＝max{k_i}-min{k_i}

T＝∑k_i

wherein R is extremely poor; i is the number of levels of the factor; k is a radical of_iIs the average value of the sum of the corresponding rotation errors at the i level; k_iIs the sum of the corresponding rotation errors at the i level; n is the number of occurrences of each level on any column; t is the sum of the rotation errors;

obtaining the optimal combination of the influence sensitivity degree of each factor of the RV reducer dynamic rotation error from a large sequence to a small sequence and the factor level according to the extreme difference of each influence factor;

s6, taking the orthogonal test result data as the input end data of the BP neural network, and predicting the optimal combination;

s61, establishing and training a BP artificial neural network;

the method comprises the following steps of taking main factors influencing the rotation error of the RV reducer in orthogonal test design as an input layer of the established BP neural network, wherein the output layer of the network comprises an output node corresponding to the evaluation index rotation error; programming by adopting MATLAB software, and selecting a hidden layer transfer function, an output layer transfer function and a training function; setting hidden layers, training precision and learning rate parameters, and training and comparing networks containing different neuron numbers;

s62, predicting the optimal combination by adopting a BP artificial neural network;

applying MATLAB software programming, successfully establishing a BP neural network model capable of accurately describing the functional relation between the RV reducer revolution error and the evaluation index thereof through training of an experimental data sample and optimization of BP neural network parameters, applying the established network model to carry out simulation, taking a plurality of influence factor values of an orthogonal test as independent variables, respectively assigning the independent variables, setting a proper step length, and programming a definition domain value of each factor by using a related function in the MATLAB to obtain a minimum combination value of the revolution error of the output value;

s63, checking an optimization result;

and establishing a virtual prototype according to the optimization result in the step, introducing ADAMS for multi-body dynamics simulation, measuring the rotation error of the virtual prototype, and comparing the rotation error with the orthogonal test result to obtain the transmission precision optimization effect.