TWI767368B - Intelligent ultrasonic grinding and polishing aided system and method thereof - Google Patents

Abstract

An intelligent ultrasonic grinding and polishing aided system is proposed. The intelligent ultrasonic grinding and polishing aided system includes a polishing apparatus, a data base and a processor. The data base records a plurality of processing data of the polishing apparatus. The processor includes a processing precision predicting module and a processing parameter optimizing module. The processing precision predicting module receives the processing data from the data base, the processing data are used to establish a precision predicting model according to an interval type-II fuzzy neural network, and produce a surface polishing precision. The processing parameter optimizing module calculates a best processing parameter corresponds to the surface polishing precision. The polishing apparatus polishes the workpiece according to the best processing parameter. Thus, the cost and the material cost by the trial and error method can be reduced, and the beginner can also polish the workpiece with an ideal polishing precision.

Description

Intelligent ultrasonic-assisted grinding system and method

The present invention relates to a grinding processing system and a method thereof, in particular to an intelligent ultrasonic-assisted grinding processing system and a method thereof.

In the existing grinding and polishing processing technology, the selection of processing materials, processing grinding rods, polishing tools and polishing materials will result in different processing accuracy, surface roughness and grinding and polishing time. Grinding and polishing mainly relies on personal experience to set the processing parameters. If you want to grind and polish untouched materials, you cannot set the processing parameters accurately.

It can be seen that there is currently a lack of a grinding processing system on the market that can set the expected surface roughness value according to the user's needs and calculate the optimal processing parameters, so the relevant manufacturers are all looking for the solution.

Therefore, the purpose of the present invention is to provide an intelligent ultrasonic-assisted grinding processing system and method thereof, which establishes a grinding prediction model according to previously recorded processing data to generate surface grinding precision, and calculates the optimal processing parameters corresponding to the surface grinding precision.

According to one embodiment of the structural aspect of the present invention, an intelligent ultrasonic-assisted grinding processing system is provided for predicting the surface grinding accuracy of the workpiece after grinding. The intelligent ultrasonic-assisted grinding processing system includes a grinding machine, a database and a processor. The database signal is connected to the grinding machine, and records the multiple processing data of the grinding machine. The processor signal is connected to the grinding machine and the database, and includes a processing accuracy prediction module and a processing parameter optimization module. The machining accuracy prediction module receives the machining data in the database, and the machining accuracy prediction module trains the machining data according to the interval type 2 fuzzy neural network and establishes a grinding prediction model. The dynamic grouping difference calculation unit is adjusted to produce surface grinding accuracy. The signal of the processing parameter optimization module is connected to the processing precision prediction module and receives the surface grinding precision. The processing parameter optimization module calculates a plurality of optimal processing parameters corresponding to the surface grinding precision according to the optimization of the processing data by the optimization unit. The grinding machine grinds the workpiece according to these optimal machining parameters.

In this way, even an unexperienced operator can set the optimum processing parameters corresponding to the surface grinding precision according to the surface grinding precision to be achieved, and grind or polish the workpiece.

According to an embodiment of the method aspect of the present invention, an intelligent ultrasonic-assisted grinding processing method is provided for predicting the surface grinding precision of the workpiece after grinding. The intelligent ultrasonic-assisted grinding method includes a processing data recording step, a processing accuracy prediction step, a processing parameter optimization step, and a grinding step. The processing data recording step is to record the multiple processing data of the grinding machine in the database. The processing accuracy prediction step is to train these processing data according to the interval type 2 fuzzy neural network model, and establish a grinding prediction model, wherein the interval type 2 fuzzy neural network model is adjusted by the dynamic grouping difference algorithm to generate surface grinding. precision. The machining parameter optimization step is to optimize the machining data according to the optimization rule to calculate a plurality of optimal machining parameters corresponding to the surface grinding precision. The grinding step drives the grinding machine to grind the workpiece according to these optimal processing parameters.

In this way, even an unexperienced operator can set the optimum processing parameters corresponding to the surface grinding precision according to the surface grinding precision to be achieved, and grind or polish the workpiece.

Please refer to Figures 1 to 3 together. FIG. 1 is a schematic block diagram of the intelligent ultrasonic-assisted grinding system 100 according to the first embodiment of the present invention; FIG. 2 is a block diagram of the second type of blurring of the intelligent ultrasonic-assisted grinding system 100 of FIG. 1 A schematic diagram of the neural-like network 320; and FIG. 3 is a block diagram illustrating the dynamic grouping and differential computing unit 330 of the intelligent ultrasonic-assisted polishing system 100 in FIG. 1. FIG. The intelligent ultrasonic-assisted grinding processing system 100 is used to predict the surface grinding accuracy Y after the workpiece is ground, and includes a grinding machine 10 , a database 20 and a processor 30 . The database 20 is connected to the grinding machine 10 with a signal, and records multiple processing data of the grinding machine 10 . The processor 30 is connected to the grinding machine 10 and the database 20 by signals. Specifically, the grinding machine 10 may be a three-axis CNC milling machine, the database 20 may be a memory, and the processor 30 may be a central processing unit (CPU), but the invention is not limited thereto.

The processor 30 includes a machining accuracy prediction module 310 and a machining parameter optimization module 340 . The machining accuracy prediction module 310 receives the machining data from the database 20, trains the machining data according to the interval type 2 fuzzy neural network 320, and establishes a grinding prediction model. Unit 330 is adjusted to produce surface finish Y . The signal of the processing parameter optimization module 340 is connected to the processing precision prediction module 310 and receives the surface grinding precision Y . Specifically, the machining data includes complex machining parameters and corresponding precision values, and the machining parameters include machining material, diamond number, line speed, feed speed, depth of cut, width of cut and ultrasonic power, but the present invention is not limited to this .

The interval type II fuzzy neural network 320 includes a first layer Layer1 , a second layer Layer2 , a third layer Layer3 , a fourth layer Layer4 and a fifth layer Layer5 . The first layer Layer1 stores processing data, and transmits the processing data as input values X ₁ to X _n of the interval type II fuzzy neural network 320 to the second layer Layer 2; Calculate interval type 2 fuzzy sets; the third layer Layer 3 performs cumulative multiplication operation on interval type 2 fuzzy sets; the fourth layer Layer 4 performs order reduction operations on interval type 2 fuzzy sets and linear function weights, and calculates type 1 fuzzy sets; The fifth layer, Layer 5, performs de-blurring operations on the first-type fuzzy set, and calculates the surface grinding accuracy Y , where the interval-type two-type fuzzy set includes Gaussian mean, Gauss standard deviation and Gaussian displacement.

The machining accuracy prediction module 310 further adjusts the parameters in the interval type 2 fuzzy neural network 320 through the dynamic group difference calculation unit 330 . The parameters include Gaussian mean, Gaussian standard deviation, Gaussian displacement and linear function weight. The dynamic grouping and difference computing unit 330 includes an initialization submodule 331 , an adaptive threshold calculation submodule 332 , a grouping submodule 333 , a mutation submodule 334 , an exchange submodule 335 and a selection submodule 336 .

The initialization sub-module 331 encodes the Gaussian mean, the Gaussian standard deviation, the Gaussian displacement and the linear function weight into one body, wherein the interval type II fuzzy neural network 320 further includes a plurality of individuals.

，然後計算距離閥值(

)及適應閥值(

)，其計算規則如式(1)到式(4)所示：

尚未被分群              (1)；

(2)；

尚未被分群              (3)；

(4)； 其中D代表維度，NP為個體總數，

為組別，

則代表第

群的領袖，

代表當前的個體，NI代表當前個體中群組編號為0的個體總數。 The adaptive threshold calculation sub-module 332 is connected to the initialization sub-module 331 with a signal, and the adaptive threshold calculation sub-module 332 calculates the distance threshold and the adaptive threshold of each individual, and defines the plural leaders of the plural groups. Specifically, the adaptation threshold calculation sub-module 332 sets the initial value of the group numbers of these individuals to 0, and sorts all individuals according to their adaptation thresholds. After sorting, the individual with the highest fitness threshold is set as the leader from the individuals whose group number is 0, and the group number is updated to

, and then calculate the distance threshold (

) and the adaptation threshold (

), and its calculation rules are shown in equations (1) to (4):

has not been grouped (1);

(2);

has not been grouped (3);

(4); where D represents the dimension, NP is the total number of individuals,

for the group,

represents the first

group leader,

represents the current individual, and NI represents the total number of individuals whose group number is 0 in the current individual.

)及適應值差(

)，並判斷未分群的個體之群體編號，距離差(

)及適應值差(

)的計算規則如式(5)、式(6)所示：

(5)；

(6)； 若

且

，則表示個體

與第

群的領袖

是相似的，將

的群組編號更新為

，所有個體皆有組別後，分群子模組333終止執行。 The grouping sub-module 333 is connected to the adaptive threshold calculating sub-module 332 with a signal, and divides the individuals into these groups according to the distance threshold and the adaptive threshold of each individual. Specifically, the grouping sub-module 333 calculates the distance difference (

) and the fitness difference (

), and determine the group number of the ungrouped individuals, the distance difference (

) and the fitness difference (

) calculation rules are shown in formulas (5) and (6):

(5);

(6); if

and

, which means that the individual

with the first

group leader

is similar to the

The group ID of is updated to

, after all individuals have groups, the grouping sub-module 333 terminates the execution.

(7)；

(8)；

(9)；

(10)； 其中

為突變向量，

為最佳領袖，

為所有群組中隨機選取的其中一領袖，F為突變權重因子，

為萊維飛行策略，

為萊維飛行策略之飛行指數，

和

為具有均值

與

的常態隨機分布。突變子模組334藉由結合萊維飛行策略，將最佳領袖設為基準向量，並加入兩個隨機個體的差異向量，使得突變後的突變向量圍繞著最好的個體。 The mutant sub-module 334 signal connects to the clustering sub-module 333 and generates a mutation vector from one of the leaders and the best of these leaders according to the Levi flight strategy. In detail, the mutation rules of the mutation sub-module 334 combined with the Levi flight strategy are shown in formula (7), formula (8), formula (9), and formula (10):

(7);

(8);

(9);

(10); wherein

is the mutation vector,

for the best leader,

is one of the leaders randomly selected from all groups, F is the mutation weight factor,

For Levi flight strategy,

is the flight index of Levi's flight strategy,

and

to have the mean

and

a normal random distribution. The mutation sub-module 334 sets the best leader as the reference vector by combining the Levi flight strategy, and adds the difference vector of two random individuals, so that the mutated mutation vector surrounds the best individual.

(11)； 其中

為試驗向量；

為突變向量；

為目標向量；

為每個維度對應的隨機值，其值為0到1之間的亂數；CR為交換率。若交換率CR愈大，則試驗向量與突變向量的相似度愈高；若交換率CR愈小，則試驗向量與突變向量的相似度愈低。 The signal of the exchange sub-module 335 is connected to the mutation sub-module 334, and the mutation vector is exchanged to generate a test vector. The exchange sub-module 335 performs an exchange operation between each mutation vector and the corresponding target vector according to an exchange rule, and generates a new test vector. The exchange operation rule is shown in formula (11):

(11); wherein

is the test vector;

is the mutation vector;

is the target vector;

is a random value corresponding to each dimension, and its value is a random number between 0 and 1; CR is the exchange rate. If the exchange rate CR is larger, the similarity between the test vector and the mutation vector is higher; if the exchange rate CR is smaller, the similarity between the test vector and the mutation vector is lower.

(12)； 其中

為選擇後的目標向量；

為試驗向量；

為當前的目標向量；

為試驗向量的適應值差；

為當前的目標向量的適應值差。 The signal of the selection sub-module 336 is connected to the exchange sub-module 335, and the test vector and the best leader are compared, and the one with the largest adaptive threshold value is selected to be the surface grinding accuracy Y. Specifically, the selection sub-module 336 uses the adaptive threshold to evaluate whether the test vector can become the surface grinding accuracy Y , and the selection formula is shown in formula (12):

(12); wherein

is the selected target vector;

is the test vector;

is the current target vector;

is the fitness value difference of the test vector;

is the fitness value difference of the current target vector.

(13)；

(14)； 其中

是ith粒子的速度；

是當前粒子的位置；

是粒子本身的最佳解；

是整個族群的最佳解；

為慣性權重；

為認知係數；

為社會係數；

及

為介於0到1之間的隨機亂數。 The machining parameter optimization module 340 calculates the optimal machining parameters corresponding to the surface grinding precision Y according to the optimization of machining data by the optimization unit 350 . The optimization unit 350 calculates the optimal machining parameters corresponding to the surface polishing accuracy Y by calculating the complex machining parameters and the surface polishing accuracy Y of each machining data. The calculation rules of the optimization unit 350 are shown in equations (13) and (14):

(13);

(14); wherein

is the velocity of the ith particle;

is the position of the current particle;

is the optimal solution of the particle itself;

is the best solution for the entire population;

is the inertia weight;

is the cognitive coefficient;

is the social coefficient;

and

is a random number between 0 and 1.

The grinding machine 10 grinds the workpiece according to the optimum processing parameters calculated by the processing parameter optimization module 340 .

The processor 30 further includes an update data collection module 360 , and the signal of the update data collection module 360 is connected to the machining accuracy prediction module 310 and the machining parameter optimization module 340 . The update data collection module 360 collects the surface grinding longitude and optimal processing parameters to form an update data set. The machining accuracy prediction module 310 updates the grinding prediction model according to the update data set formed by the update data collection module 360 to generate an update prediction model. In detail, with the long-term operation of the grinding machine 10, the wear of the internal parts makes the surface grinding precision Y predicted by the grinding prediction model and the surface grinding precision Y that can be obtained by the actual grinding of the grinding machine 10 have errors. By updating the prediction The model can make the surface grinding precision Y predicted by the intelligent ultrasonic-assisted grinding processing system 100 closer to the actual grinding precision.

Thereby, the intelligent ultrasonic-assisted polishing processing system 100 of the present invention predicts the surface polishing accuracy Y generated by the combination of different machining parameters through the polishing prediction model, and calculates the optimal machining parameters to obtain the surface polishing accuracy Y. It can effectively reduce the huge material and cost of trial and error method, so that the unexperienced operator can set the corresponding optimal processing parameters according to the surface grinding accuracy Y to be achieved, and grind or polish the workpiece.

Please refer to FIG. 1 to FIG. 5, FIG. 4 is a schematic flow chart of the intelligent ultrasonic-assisted grinding method 200 according to the second embodiment of the present invention; FIG. 5 is the intelligent ultrasonic wave of FIG. 4 A schematic flowchart of the dynamic grouping difference calculation rule S22 of the auxiliary grinding method 200 . As shown in the figure, the intelligent ultrasonic-assisted grinding method 200 is used to predict the surface grinding accuracy Y after the workpiece is ground. The intelligent ultrasonic-assisted grinding method 200 includes a processing data recording step S10, a processing accuracy prediction step S20, a processing parameter optimization step S30, an update data collection step S40, and a grinding step S50.

The processing data recording step S10 records the plurality of processing data of the grinding machine table 10 in the database 20 .

The processing accuracy prediction step S20 is to train the processing data according to the interval type 2 fuzzy neural network model S21, and establish a grinding prediction model, wherein the interval type 2 fuzzy neural network model S21 is adjusted by the dynamic grouping difference calculation rule S22 And produce surface grinding precision Y. The machining accuracy prediction step S20 is performed by the machining accuracy prediction module 310 . The interval type 2 fuzzy neural network model S21 is executed by the interval type 2 fuzzy neural network 320 , and the dynamic group difference calculation rule S22 is executed by the dynamic group difference calculation unit 330 .

The dynamic grouping difference calculation rule S22 includes an initialization sub-step S221, an adaptive threshold calculation sub-step S222, a grouping sub-step S223, a mutation sub-step S224, an exchange sub-step S225, and a selection sub-step S226.

The initialization sub-step S221 encodes the Gaussian mean value, the Gaussian standard deviation, the Gaussian displacement amount and the linear function weight into an individual, wherein the interval type II fuzzy neural network model S21 further includes a plurality of individuals. The initialization sub-step S221 is performed by the initialization sub-module 331 .

The adaptive threshold calculation sub-step S222 calculates the distance threshold and the adaptive threshold of each individual, and defines the plural leaders of the plural groups. The adaptive threshold calculation sub-step S222 is performed by the adaptive threshold calculation sub-module 332 .

The sub-step S223 of grouping is to divide the individuals into these groups according to the distance threshold and adaptation threshold of each individual. The grouping sub-step S223 is performed by the grouping sub-module 333 .

The mutation sub-step S224 generates mutation vectors from one of the leaders and the best leader of these leaders according to the Levi flight strategy. The mutation sub-step S224 is performed by the mutation sub-module 334 .

The swap sub-step S225 performs a swap operation on the mutation vector and generates a test vector. The exchange sub-step S225 is performed by the exchange sub-module 335 .

The selection sub-step S226 compares the test vector and the best leader, and selects the one with the largest remaining adaptation threshold value to be the surface grinding accuracy Y . The selection sub-step S226 is performed by the selection sub-module 336 .

The machining parameter optimization step S30 is to optimize the machining data according to the optimization rule to calculate a plurality of optimal machining parameters corresponding to the surface grinding precision Y. The process parameter optimization step S30 is performed by the process parameter optimization module 340 . The optimization rule calculates the optimal processing parameters corresponding to the surface grinding precision Y by calculating the complex processing parameters of each processing data and the surface grinding precision Y.

The update data collection step S40 is to collect the surface polishing accuracy Y and these optimal processing parameters to form an update data set, and update the grinding prediction model according to the update data set to generate an update prediction model. The update data collection step S40 is performed by the update data collection module 360 .

The grinding step S50 is to drive the grinding machine 10 to grind the workpiece according to the optimal processing parameters.

Thereby, the intelligent ultrasonic-assisted grinding method 200 of the present invention predicts the surface grinding precision Y generated by the combination of different processing parameters through the grinding prediction model, and calculates the best processing parameters to obtain the surface grinding precision Y. It can effectively reduce the huge material and cost of trial and error method, so that the unexperienced operator can set the corresponding optimal processing parameters according to the surface grinding accuracy Y to be achieved, and grind or polish the workpiece.

Although the present invention has been disclosed in the above embodiments, it is not intended to limit the present invention. Anyone skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the protection of the present invention The scope shall be determined by the scope of the appended patent application.

10: Grinding machine 20: Database 30: Processor 100: Intelligent ultrasonic-assisted grinding and processing system 200: Intelligent ultrasonic-assisted grinding and processing method 310: Machining accuracy prediction module 320: Interval type II fuzzy neural network 330: Dynamic grouping and difference calculation unit 331: Initialization sub-module 332: Adaptation threshold calculation sub-module 333: Grouping sub-module 334: Mutation sub-module 335: Swap sub-module 336: Selection sub-module 340: Processing parameters Optimization module 350: Optimization unit 360: Update data collection module S10: Processing data recording Step S20: Machining accuracy prediction Step S21: Interval type II fuzzy neural network model S22: Dynamic grouping difference calculation rule S221: Initialization sub-step S222 : adaptive threshold calculation sub-step S223: grouping sub-step S224: mutation sub-step S225: exchange sub-step S226: selection sub-step S30: processing parameter optimization step S40: update data collection step S50: grinding step Layer1: first layer Layer2: The second layer Layer3: the third layer Layer4: the fourth layer Layer5: the fifth layer X ₁ , X _n : input value Y: surface grinding accuracy

FIG. 1 is a schematic block diagram of an intelligent ultrasonic-assisted grinding system according to a first embodiment of the present invention; Fig. 2 is a schematic diagram of the interval type 2 fuzzy neural network of the intelligent ultrasonic-assisted polishing system of Fig. 1; Fig. 3 is a block diagram of the dynamic grouping and differential computing unit of the intelligent ultrasonic-assisted grinding system of Fig. 1; FIG. 4 is a schematic flow chart showing the intelligent ultrasonic-assisted grinding method according to the second embodiment of the present invention; and FIG. 5 is a schematic flowchart of the dynamic grouping and difference calculation rule of the intelligent ultrasonic-assisted grinding method of FIG. 4 .

100: Intelligent ultrasonic-assisted grinding and processing system 10: Grinding machine table 20:Database 30: Processor 310: Machining accuracy prediction module 320: Interval Type II Fuzzy Neural Networks 330: Dynamic grouping difference calculation unit 340: Processing parameter optimization module 350: Optimization Unit 360: Update data collection module

Claims (8)

An intelligent ultrasonic-assisted grinding processing system is used to predict the surface grinding accuracy of a workpiece after grinding. The intelligent ultrasonic-assisted grinding processing system comprises: a grinding machine; a database, the signal is connected to the grinding machine , and record the plurality of processing data of the grinding machine; and a processor, the signal is connected to the grinding machine and the database, the processor includes: a processing accuracy prediction module, receiving the processing data of the database, The machining accuracy prediction module trains the machining data according to an interval type 2 fuzzy neural network and establishes a grinding prediction model, and the interval type 2 fuzzy neural network is adjusted by a dynamic group difference arithmetic unit to generate the Surface grinding precision; a processing parameter optimization module, the signal is connected to the processing precision prediction module and receives the surface grinding precision, and the processing parameter optimization module calculates the corresponding surface grinding precision according to the optimization of the processing data by an optimization unit a plurality of optimal processing parameters; and an update data collection module, the signal is connected to the processing accuracy prediction module and the processing parameter optimization module, the updated data collection module collects the surface grinding accuracy and the best processing parameters to obtain An update data set is formed; wherein, the grinding machine grinds the workpiece according to the optimal processing parameters, and the machining accuracy prediction module is generated by updating the grinding prediction model according to the update data set formed by the update data collection module An update forecast model type. The intelligent ultrasonic-assisted grinding and processing system according to claim 1, wherein the interval type II fuzzy neural network comprises: a first layer, storing the processing data; a second layer, storing the processing data Perform a fuzzification operation, and calculate an interval type 2 fuzzy set; a third layer, perform a cumulative multiplication operation on the interval type 2 fuzzy set; a fourth layer, the interval type 2 fuzzy set and a linear function The weights are subjected to a first-order reduction operation, and a type-1 fuzzy set is calculated; and a fifth layer, a de-fuzzification operation is performed on the type-1 fuzzy set, and the surface grinding accuracy is calculated; wherein, the interval type-2 fuzzy set Contains a Gaussian mean, a Gaussian standard deviation and a Gaussian shift. The intelligent ultrasonic-assisted polishing system as claimed in claim 2, wherein the dynamic grouping and difference calculation unit comprises: an initialization sub-module, the Gaussian mean value, the Gaussian standard deviation, the Gaussian displacement and the linear function The weight is encoded as an individual, wherein the interval type II fuzzy neural network further includes a plurality of the individual; an adaptive threshold calculation sub-module, the signal is connected to the initialization sub-module, and calculates a distance threshold and An adaptation threshold, and defines the plural leaders of the plural groups; a sub-group sub-module, the signal is connected to the adaptation threshold calculation sub-module, and according to The individuals are divided into the groups according to the distance threshold and the adaptation threshold of each individual; a mutant sub-module, the signal is connected to the grouping sub-module, and according to a Levi flight strategy from one of the leaders and one of the leaders, the best leader, generates a mutation vector; a swap sub-module, the signal is connected to the mutation sub-module to perform a swap operation on the mutation vector and generate a test vector; and a selection sub-module, The signal is connected to the exchange sub-module, the test vector is compared with the best leader, and the one with the largest adaptation threshold is selected to be the surface grinding accuracy. The intelligent ultrasonic-assisted grinding processing system according to claim 3, wherein the optimization unit calculates the optimal processing corresponding to the surface grinding precision by calculating the complex processing parameters of the processing data and the surface grinding precision parameter. An intelligent ultrasonic-assisted grinding processing method is used to predict the surface grinding accuracy of a workpiece after grinding. The intelligent ultrasonic-assisted grinding processing method comprises: a processing data recording step of processing a plurality of grinding machines The data is recorded in a database; a processing accuracy prediction step is to train the processing data according to an interval type II fuzzy neural network model, and establish a grinding prediction model, wherein the interval type II fuzzy neural network model via a dynamic group difference The calculation rule is adjusted to generate the surface grinding precision; a processing parameter optimization step is to optimize the processing data according to an optimization rule to calculate a plurality of optimal processing parameters corresponding to the surface grinding precision; a grinding step is to drive the grinding The machine grinds the workpiece according to the optimal processing parameters; and an update data collection step collects the surface grinding accuracy and the optimal processing parameters to form an updated data set, and updates the grinding prediction according to the updated data set model to generate an updated prediction model. The intelligent ultrasonic-assisted grinding method according to claim 5, wherein the interval type II fuzzy neural network model comprises: a first layer, storing the processing data; a second layer, processing the processing data The data is subjected to a fuzzification operation, and an interval type 2 fuzzy set is calculated; a third layer, a cumulative multiplication operation is performed on the interval type 2 fuzzy set; a fourth layer, the interval type 2 fuzzy set and a linear The function weight performs a first-order reduction operation, and calculates a type-1 fuzzy set; and a fifth layer, performs a de-fuzzification operation on the type-1 fuzzy set, and calculates the surface grinding accuracy; wherein the interval type-2 fuzzy set Contains a Gaussian mean, a Gaussian standard deviation and a Gaussian shift. The intelligent ultrasonic-assisted grinding machine as described in claim 6 The engineering method, wherein the dynamic grouping difference algorithm includes: an initialization sub-step, encoding the Gaussian mean value, the Gaussian standard deviation, the Gaussian displacement amount and the linear function weight into a body, wherein the interval type II fuzzy neural network The network model further includes a plurality of the individuals; an adaptation threshold calculation sub-step, calculates a distance threshold and an adaptation threshold for each of the individuals, and defines the plurality of leaders of the plurality of groups; a group sub-step is based on each of the individuals the distance threshold and the adaptation threshold divide the individuals into the groups; a mutation substep generating a mutation vector from one of the leaders and the best one of the leaders according to a Levi flight strategy; A swap sub-step is to perform a swap operation on the mutation vector to generate a test vector; and a selection sub-step is to compare the test vector with the best leader and select the one that retains the largest adaptation threshold to become the test vector Surface grinding accuracy. The intelligent ultrasonic-assisted polishing method as claimed in claim 7, wherein the optimization rule calculates the optimal machining corresponding to the surface polishing accuracy by calculating the complex machining parameters of the machining data and the surface polishing accuracy. parameter.

Images