A method and device for testing dynamic balance of cutting spindle of dicing machine

Abstract

The present invention discloses a method and device for testing the dynamic balance of a cutting spindle of a dicing machine, and relates to the field of semiconductor processing technology. The method comprises: calling a dicing machine work log to determine a plurality of cutting conditions, building an imbalance prediction branch in combination with imbalance influencing factors, and generating an imbalance predictor. Real-time working conditions and influencing data sets are monitored and acquired, and imbalance position distribution is input into an imbalance predictor for imbalance prediction, and then the dynamic balance test and counterweight adjustment of the spindle are performed. The technical problem that the existing dynamic balance test method has insufficient accuracy and efficiency in dynamic balance test and correction due to the large blindness of position test is solved. By accurately predicting the imbalance position under a variety of working conditions, and performing the dynamic balance test and counterweight adjustment of the spindle based on the predicted imbalance position, the time and resources consumed by blind testing can be reduced, thereby achieving the technical effect of improving test accuracy and efficiency and reducing resource waste.

Description

Technical Field

[0001] The present application relates to the field of semiconductor processing technology, and in particular to a method and device for testing the dynamic balance of a cutting spindle of a dicing machine. Background Art

[0002] In the application scenario of the dicing machine, the dynamic balance problem of the cutting spindle is particularly prominent, and the contradiction between the demand for efficient and accurate dynamic balance test methods is also more prominent. Finding a reliable dynamic balance test method has become a crucial link to ensure the normal operation of the dicing machine. The traditional dynamic balance test method of the cutting spindle of the dicing machine is often blind and inefficient. It only relies on multiple weight adjustments and repeated tests. It lacks in-depth analysis and utilization of the dicing machine operation data, and it is difficult to accurately predict the imbalance state of the spindle. There are unreasonable situations in the allocation of test resources, resulting in a lot of time and resources. The formulation of the dynamic balance adjustment plan is relatively simple and fixed, and it cannot cope well with complex equipment operation conditions.

[0003] In the current related technologies, there is a technical problem that the dynamic balancing test method has insufficient accuracy and efficiency in dynamic balancing test and correction due to the large blindness of position test. Summary of the invention

[0004] The present application provides a method and device for testing the dynamic balance of a cutting spindle of a dicing machine, which uses the method of calling the dicing machine work log to determine multiple cutting conditions, and constructs an imbalance prediction branch in combination with imbalance influencing factors to generate an imbalance predictor. The real-time working conditions and influencing data sets are monitored and acquired, and the imbalance predictor is input to perform imbalance prediction, and the imbalance position distribution is generated, and then the spindle dynamic balance test and counterweight adjustment are performed. The imbalance position under various working conditions is accurately predicted, and the spindle dynamic balance test and counterweight adjustment are performed based on the predicted imbalance position, which can reduce the time and resources consumed by blind testing, thereby achieving the technical effect of improving test accuracy and efficiency and reducing resource waste.

[0005] The present application provides a method for testing the dynamic balance of a cutting spindle of a dicing machine, comprising:

[0006] Call the work log of the dicing machine to perform cutting condition analysis and determine multiple cutting conditions; obtain the imbalance influencing factors of the cutting spindle of the dicing machine, and construct multiple imbalance prediction branches according to the imbalance influencing factors and multiple cutting conditions; establish a mapping association between the cutting condition and the imbalance prediction branch, and based on the mapping association, generate an imbalance predictor according to the multiple cutting conditions and multiple imbalance prediction branches; monitor and obtain real-time conditions and influencing data sets, input the real-time conditions and influencing data sets into the imbalance predictor, perform imbalance prediction for a predetermined period, and generate an imbalance position distribution; perform spindle dynamic balancing test and counterweight adjustment based on the imbalance position distribution.

[0007] The present application also provides a dicing machine cutting spindle dynamic balance test device, comprising:

[0008] A cutting condition determination module, the cutting condition determination module is used to call the work log of the dicing machine to perform cutting condition analysis and determine multiple cutting conditions; an imbalance prediction branch construction module, the imbalance prediction branch construction module is used to obtain the imbalance influencing factors of the cutting spindle of the dicing machine, and construct multiple imbalance prediction branches according to the imbalance influencing factors and multiple cutting conditions; an imbalance predictor generation module, the imbalance predictor generation module is used to establish a mapping association between the cutting condition and the imbalance prediction branch, and based on the mapping association, generate an imbalance predictor according to the multiple cutting conditions and multiple imbalance prediction branches; an imbalance position distribution generation module, the imbalance position distribution generation module is used to monitor and obtain real-time conditions and influencing data sets, input the real-time conditions and influencing data sets into the imbalance predictor, perform imbalance prediction for a predetermined period, and generate an imbalance position distribution; a counterweight adjustment module, the counterweight adjustment module is used to perform spindle dynamic balancing test and counterweight adjustment based on the imbalance position distribution.

[0009] The proposed method and device for dynamic balancing test of the cutting spindle of a dicing machine proposed in this application first calls the dicing machine work log to determine multiple cutting conditions, builds an imbalance prediction branch in combination with imbalance influencing factors, and generates an imbalance predictor. Monitor and obtain real-time working conditions and influencing data sets, input the imbalance predictor for imbalance prediction, generate imbalance position distribution, and then perform spindle dynamic balancing test and counterweight adjustment. By accurately predicting the imbalance position under various working conditions, and performing spindle dynamic balancing test and counterweight adjustment based on the predicted imbalance position, the time and resources consumed by blind testing can be reduced, thereby achieving the technical effect of improving test accuracy and efficiency and reducing resource waste. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solution of the embodiment of the present invention, the accompanying drawings of the embodiment of the present invention will be briefly introduced below. A flow chart is used in the present application to illustrate the operations performed by the device according to the embodiment of the present application. It should be understood that the previous or following operations are not necessarily performed precisely in order. On the contrary, various steps can be processed in reverse order or simultaneously as needed. At the same time, other operations can also be added to these processes, or one or more operations can be removed from these processes.

[0011] Figure 1 A schematic diagram of a flow chart of a method for testing the dynamic balance of a cutting spindle of a dicing machine provided in an embodiment of the present application;

[0012] Figure 2 A schematic structural diagram of a dicing machine cutting spindle dynamic balance test device provided in an embodiment of the present application.

[0013] Explanation of reference numerals: cutting condition determination module 10 , imbalance prediction branch construction module 20 , imbalance predictor generation module 30 , imbalance position distribution generation module 40 , counterweight adjustment module 50 . DETAILED DESCRIPTION

[0014] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below.

[0015] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limiting the present application. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of this application.

[0016] In the following description, reference is made to "some embodiments", which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict, and the terms "first\second" involved are merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "including" and "having" and any variations are intended to cover non-exclusive inclusions, for example, a process, method, device, product, or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by technicians in the technical field of this application. The terms used herein are for the purpose of describing the embodiments of the present application only.

[0017] The present application embodiment provides a method for testing the dynamic balance of a cutting spindle of a dicing machine. Figure 1 As shown, the method includes:

[0018] Step S100, call the work log of the dicing machine to perform cutting condition analysis and determine multiple cutting conditions. Specifically, first obtain the work log of the dicing machine, which contains information such as spindle speed, cutting load, material properties and ambient temperature. Extract the spindle speed, cutting load, material properties and ambient temperature from the work log as working condition factors, and then perform feature extraction on the work log based on the working condition factors to determine the spindle speed threshold, cutting load threshold, material property threshold and ambient temperature threshold. Then, segment the threshold according to the predetermined division scalar, and randomly combine the segmentation results to generate multiple cutting conditions. For example, the spindle speed is divided into high, medium and low intervals, the cutting load is divided into heavy, medium and light levels, the material properties are divided into different types, and the ambient temperature is divided into high, medium and low ranges, and then randomly combined to obtain multiple cutting conditions with different characteristics, which provides a basis for subsequent imbalance prediction and dynamic balancing tests.

[0019] In a possible implementation, the work log of the dicing machine is called to perform cutting condition analysis, and multiple cutting conditions are determined. Step S100 further includes step S110, obtaining the condition factor, wherein the condition factor includes at least the spindle speed, cutting load, material properties and ambient temperature. Specifically, it is first clarified that the condition factor includes at least the spindle speed, cutting load, material properties and ambient temperature. During the working process of the dicing machine, the condition factor has an important influence on its performance and state. The spindle speed determines the efficiency and accuracy of cutting. Different speeds will produce different centrifugal forces and vibrations. For example, a higher spindle speed will increase the cutting speed, but it will also cause greater vibration and higher temperature. The cutting load reflects the pressure that the dicing machine is subjected to during the cutting process. The change in load will affect the stability of the equipment and the cutting quality. For example, when cutting harder materials, the cutting load will increase. Material properties include aspects such as hardness, toughness, and conductivity of the material. Different material properties require different cutting parameters and processes. For example, for materials with higher hardness, a lower spindle speed and a larger cutting load are required. The ambient temperature will affect the heat dissipation of the dicing machine, the performance of the material and the accuracy of the equipment. Under different ambient temperatures, the working state of the dicing machine may be different.

[0020] Step S120, based on the spindle speed, cutting load, material properties and ambient temperature, feature extraction is performed on the work log of the dicing machine to determine the spindle speed threshold, cutting load threshold, material property threshold and ambient temperature threshold. Specifically, feature extraction is performed on the work log of the dicing machine based on the spindle speed, cutting load, material property and ambient temperature, and by analyzing a large amount of data in the work log, the operation of the dicing machine under different working conditions is understood, and the spindle speed threshold is determined. For example, by observing the vibration of the dicing machine at different spindle speeds, cutting quality and other indicators, a suitable speed range boundary is found. If the speed exceeds this threshold, it may cause the equipment to malfunction. Assuming that it is found in the work log that when the spindle speed exceeds 12,000 rpm, the vibration of the dicing machine increases significantly and the cutting quality decreases, then 12,000 rpm can be determined as the spindle speed threshold. Determine the cutting load threshold, which is determined based on the performance of the equipment under different loads in the work log. When the load exceeds a certain value, it may cause damage to the equipment or affect the cutting quality. For example, when the cutting load exceeds 80 Newtons, the tool wear of the dicing machine increases and the cutting accuracy decreases. In this case, 80 Newtons can be used as the cutting load threshold. Determine the material property threshold, which is divided according to the different properties of the cut material. For example, for materials with higher hardness, a higher cutting load threshold and a lower spindle speed threshold can be set. By analyzing the cutting conditions of different materials in the work log, the boundaries of different material property intervals can be determined. Determine the ambient temperature threshold, considering the impact of ambient temperature on the dicing machine. When the ambient temperature is too high or too low, it will affect the heat dissipation and performance of the equipment. Assume that it is found in the work log that when the ambient temperature exceeds 30°C, the heat dissipation effect of the dicing machine is significantly reduced, then 30°C can be used as the ambient temperature threshold.

[0021] Step S130, dividing the spindle speed threshold, cutting load threshold, material property threshold and ambient temperature threshold according to a predetermined division scalar, and randomly combining the division results to generate the plurality of cutting working conditions. Specifically, the spindle speed threshold, cutting load threshold, material property threshold and ambient temperature threshold are divided according to a predetermined division scalar. For the spindle speed threshold, it can be divided into three intervals of high, medium and low. For the cutting load threshold, it is also divided into three levels of heavy, medium and light. For the material property threshold, it is classified according to different material properties, such as metal materials, non-metallic materials, composite materials and other intervals. For the ambient temperature threshold, it is divided into three ranges of high, medium and low. Then, the segmentation results are randomly combined to generate multiple cutting conditions. Each cutting condition includes a spindle speed interval, a cutting load interval, a material property interval and an ambient temperature interval. For example, a high speed interval, a heavy load level, a metal material interval, and a high temperature range are combined into one cutting condition; a medium speed interval, a medium load level, a non-metallic material interval, and a medium temperature range are combined into another cutting condition, etc., to obtain multiple cutting conditions with different characteristics, which provides rich condition data for subsequent imbalance prediction and dynamic balancing tests.

[0022] Step S200, obtain the imbalance influencing factors of the cutting spindle of the dicing machine, and construct multiple imbalance prediction branches according to the imbalance influencing factors and multiple cutting conditions. Specifically, first obtain the imbalance influencing factors of the cutting spindle of the dicing machine, including amplitude distribution, spindle speed, cutting load, spindle temperature and coolant temperature, then obtain the equipment state characteristics of the dicing machine, randomly select the first working condition from multiple cutting working conditions, perform spindle dynamic balance association retrieval with the equipment state characteristics and the first working condition as similar constraints, obtain the retrieval data set, filter the retrieval data set according to the imbalance influencing factors, obtain the sample data set, and then use the sample data set to supervise the BP neural network training, obtain the imbalance prediction branch that meets the expected convergence conditions and add it to multiple imbalance prediction branches, which provides a basis for the subsequent generation of imbalance predictors and imbalance prediction.

[0023] In a possible implementation, unbalance influencing factors of the cutting spindle of the dicing machine are obtained, and multiple unbalance prediction branches are constructed according to the unbalance influencing factors and multiple cutting conditions, and step S200 further includes step S210, wherein the unbalance influencing factors at least include amplitude distribution, spindle speed, cutting load, spindle temperature, and coolant temperature. Specifically, the unbalance influencing factors of the cutting spindle of the dicing machine at least include amplitude distribution, spindle speed, cutting load, spindle temperature, and coolant temperature. The amplitude distribution reflects the vibration situation and can intuitively reflect the imbalance state; the increase in spindle speed will increase the centrifugal force, making the imbalance problem more prominent, and different speed ranges may cause different types of imbalance; the cutting load size affects the spindle balance. A larger load will cause the spindle to be subjected to greater pressure and torque, which is easy to cause imbalance, and the load change will affect the vibration characteristics and temperature distribution; the spindle temperature change causes thermal expansion, affecting the equilibrium state. Long-term operation or high-speed cutting will cause the spindle to heat up and deform, destroying the balance. Uneven temperature will also aggravate the imbalance; the coolant temperature affects the cooling effect, and then affects the spindle temperature distribution and thermal expansion. Too high or too low coolant temperature may aggravate the imbalance, and also affect the ambient temperature around the spindle, indirectly affecting the balance.

[0024] In a possible implementation, the unbalance influencing factors of the cutting spindle of the dicing machine are obtained, and multiple unbalance prediction branches are constructed according to the unbalance influencing factors and multiple cutting conditions. Step S200 further includes step S220, obtaining the equipment status characteristics of the dicing machine, wherein the equipment status characteristics at least include equipment type, equipment specifications, tool wear status, and bearing wear status. Specifically, it is first clarified that the equipment status characteristics at least include equipment type, equipment specifications, tool wear status, and bearing wear status. Different equipment types have differences in structural design, performance characteristics, etc., which will affect the unbalance performance of the spindle. For example, some models of dicing machines are more prone to specific types of imbalance problems due to their design. Equipment specifications include parameters such as size and power. Larger or higher-powered equipment may have different dynamic characteristics during operation, which may affect the imbalance. The wear state of the tool is critical. As the tool wears, the cutting force may change, which in turn affects the load and vibration of the spindle. If the tool wears severely, it will lead to uneven cutting load and increase the possibility of spindle imbalance. The wear state of the bearings will also have a direct impact on the spindle balance. Worn bearings will make the spindle run unstable, produce additional vibration and offset, and aggravate the imbalance problem.

[0025] Step S230, randomly select a first working condition from the multiple cutting working conditions, use the equipment state characteristics and the first working condition as similarity constraints, perform spindle dynamic balance association retrieval, and obtain a retrieval data set. Specifically, randomly select one of the multiple cutting working conditions that have been determined as the first working condition, and the first working condition includes a specific spindle speed range, a cutting load range, a material property range, and an ambient temperature range. Then, use the equipment state characteristics (equipment type, equipment specifications, tool wear state, and bearing wear state) and the first working condition as similarity constraints to perform spindle dynamic balance association retrieval, and search for situations similar to the current equipment state and the first working condition in historical data. Through the retrieval, a retrieval data set can be obtained, which contains historical operating data similar to the current situation, such as information such as amplitude distribution, spindle speed, cutting load, spindle temperature, and coolant temperature under similar equipment states and cutting conditions, as well as corresponding unbalanced positions.

[0026] Step S240, the retrieval data set is screened according to the imbalance influencing factors to obtain a sample data set. Specifically, the retrieval data set is screened according to the imbalance influencing factors (amplitude distribution, spindle speed, cutting load, spindle temperature and coolant temperature), and the retrieval data set is preliminarily screened according to the imbalance influencing factors to obtain an initial sample data set, the initial sample data set includes information such as sample amplitude distribution, sample spindle speed, sample cutting load, sample spindle temperature, sample coolant temperature and sample imbalance position, then, a first initial sample data is randomly selected from the initial sample data set, and a similarity comparison is performed between the first sample amplitude distribution and other sample amplitude distributions in the initial sample data set to obtain the number of data with a similarity less than a similarity threshold, which is set as a first difference coefficient. If the difference coefficient is greater than a predetermined difference threshold, it means that the amplitude distribution of the sample is quite different from that of the other samples, and further comparison may be required. The first step is to analyze other factors. Then, if the first difference coefficient is greater than the predetermined difference threshold, the deviation of the first sample spindle speed and other sample spindle speeds in the initial sample data set is calculated, and the number of data with deviation values ​​greater than the speed deviation threshold is counted and set as the second difference coefficient. If the second difference coefficient is greater than the predetermined difference threshold, the first sample cutting load, the first sample spindle temperature and the first sample coolant temperature are analyzed in turn to determine the third difference coefficient, the fourth difference coefficient and the fifth difference coefficient. If the third difference coefficient, the fourth difference coefficient and the fifth difference coefficient are all greater than the predetermined difference threshold, it means that the sample and other samples have large differences in multiple imbalance influencing factors, and they can be added to the sample data set. Through the screening process, a more accurate and representative sample data set can be obtained.

[0027] Step S250, supervise the BP neural network using the sample data set to obtain a first imbalance prediction branch that meets the expected convergence conditions, and add it to the multiple imbalance prediction branches. Specifically, the sample data set is divided into a training set and a validation set according to a preset ratio, usually the training set accounts for a larger proportion, such as 80%, and the validation set accounts for 20%. The sample amplitude distribution, sample spindle speed, sample cutting load, sample spindle temperature and sample coolant temperature are used as input data, and the sample imbalance position is used as supervision. Based on the gradient descent algorithm, the BP neural network is iteratively trained using the training set. During the training process, the weights and biases of the neural network are continuously adjusted so that the output of the network gradually approaches the actual imbalance position. After a certain number of iterative trainings, the validation set is used for validation training to check the generalization ability of the network, and the training and validation are continued until the output results meet the expected convergence conditions, that is, the prediction error of the network is within an acceptable range. The obtained network model is the first imbalance prediction branch. Finally, the first imbalance prediction branch is added to the multiple imbalance prediction branches to provide a basis for the subsequent generation of imbalance predictors and imbalance predictions.

[0028] In a possible implementation, the retrieval data set is screened according to the imbalance influencing factors to obtain a sample data set, and step S240 further includes step S241, screening the retrieval data set according to the imbalance influencing factors to obtain an initial sample data set, wherein the initial sample data set includes a sample amplitude distribution, a sample spindle speed, a sample cutting load, a sample spindle temperature, a sample coolant temperature, and a sample imbalance position. Specifically, first, the retrieval data set is screened according to the imbalance influencing factors (amplitude distribution, spindle speed, cutting load, spindle temperature and coolant temperature), because the factors have an important influence on the imbalance state of the cutting spindle of the dicing machine. Through screening, the initial sample data set is obtained, which includes information such as sample amplitude distribution, sample spindle speed, sample cutting load, sample spindle temperature, sample coolant temperature and sample imbalance position. The information covers all aspects that affect the spindle imbalance. For example, the sample amplitude distribution can reflect the vibration of the spindle, and different amplitude distributions may mean different degrees of imbalance; the sample spindle speed determines the size of the centrifugal force and has a direct impact on the imbalance; the sample cutting load reflects the pressure on the spindle during operation, which will also affect the imbalance state; the sample spindle temperature and coolant temperature may cause thermal expansion or contraction of the spindle, thereby affecting the balance.

[0029] Step S242, randomly select the first initial sample data from the initial sample data set, perform a similarity comparison on the amplitude distribution of the first sample and other sample amplitude distributions in the initial sample data set, obtain the number of data with a similarity less than a similarity threshold, and set it as a first difference coefficient. Specifically, randomly select a data from the initial sample data set as the first initial sample data, then perform a similarity comparison on the amplitude distribution of the first sample and other sample amplitude distributions in the initial sample data set. The similarity comparison may adopt a certain similarity calculation method, such as calculating the correlation coefficient or distance measurement between the two amplitude distributions. If the similarity is less than a preset similarity threshold, it is considered that there is a large difference in the amplitude distribution of the two samples. The number of data with a similarity less than the similarity threshold is counted and set as the first difference coefficient. The coefficient reflects the degree of difference in amplitude distribution between the first initial sample data and other samples.

[0030] Step S243, if the first difference coefficient is greater than the predetermined difference threshold, the deviation calculation is performed on the first sample spindle speed and the other sample spindle speeds in the initial sample data set, and the number of data with a deviation value greater than the speed deviation threshold is counted and set as the second difference coefficient. Specifically, if the first difference coefficient is greater than the predetermined difference threshold, it means that the first initial sample data and the other samples have a large difference in amplitude distribution, and other factors need to be further analyzed. At this time, the deviation calculation is performed on the first sample spindle speed and the other sample spindle speeds in the initial sample data set. The difference between the two spindle speeds can be calculated and compared with the preset speed deviation threshold. If the deviation value is greater than the speed deviation threshold, it is considered that there is also a large difference in the spindle speed between the two samples. The number of data with a deviation value greater than the speed deviation threshold is counted and set as the second difference coefficient. The coefficient reflects the degree of difference in spindle speed between the first initial sample data and the other samples.

[0031] Step S244, if the second difference coefficient is greater than the predetermined difference threshold, the first sample cutting load, the first sample spindle temperature and the first sample coolant temperature are analyzed in turn to determine the third difference coefficient, the fourth difference coefficient and the fifth difference coefficient. Specifically, if the second difference coefficient is greater than the predetermined difference threshold, it means that the difference in spindle speed between the first initial sample data and other samples is also large, and it is necessary to continue to analyze other factors, and analyze the first sample cutting load, the first sample spindle temperature and the first sample coolant temperature in turn. For the cutting load, the difference between the loads can be calculated and compared with the corresponding load deviation threshold to determine the third difference coefficient; for the spindle temperature, the difference between the temperatures can be calculated and compared with the temperature deviation threshold to determine the fourth difference coefficient; for the coolant temperature, a similar analysis can be performed to determine the fifth difference coefficient.

[0032] Step S245, if the third difference coefficient, the fourth difference coefficient and the fifth difference coefficient are all greater than the predetermined difference threshold, then the first initial sample data is added to the sample data set. Specifically, if the third difference coefficient, the fourth difference coefficient and the fifth difference coefficient are all greater than the predetermined difference threshold, it means that there are also large differences between the first initial sample data and other samples in terms of cutting load, spindle temperature and coolant temperature. In this case, the first initial sample data is added to the sample data set, because the sample is greatly different from other samples and has uniqueness, which can provide more information and diversity for subsequent analysis and training. Through the above steps, the initial sample data set can be further screened and optimized to obtain a more accurate and representative sample data set, which provides a better basis for the subsequent use of the sample data set for supervised training of the BP neural network.

[0033] In a possible implementation, supervised training is performed on the BP neural network using the sample data set to obtain a first unbalanced prediction branch that meets the expected convergence conditions, which is added to the multiple unbalanced prediction branches. Step S250 further includes step S251, dividing the sample data set into a training set and a validation set according to a preset ratio, wherein the training set accounts for 80% and the validation set accounts for 20%. Specifically, first, the sample data set is divided into a training set and a validation set according to a preset ratio, and the training set accounts for 80% and the validation set accounts for 20%. The purpose of the division is to have enough data for model learning and parameter adjustment (training set) during the model training process, and to have a part of independent data for evaluating the performance and generalization ability of the model (validation set). For example, assuming there are 1,000 sample data, 800 of them will be divided into the training set for iterative training of the BP neural network; the remaining 200 data will be divided into the validation set for subsequent validation training.

[0034] Step S252, using sample amplitude distribution, sample spindle speed, sample cutting load, sample spindle temperature and sample coolant temperature as input data, using sample unbalanced position as supervision, and using the training set to iteratively train the BP neural network based on the gradient descent algorithm. Specifically, using sample amplitude distribution, sample spindle speed, sample cutting load, sample spindle temperature and sample coolant temperature as input data, the input data covers the main factors that may affect the unbalance of the cutting spindle of the dicing machine, using the sample unbalanced position as supervision, that is, taking the sample unbalanced position as the desired output result, allowing the BP neural network to adjust towards this goal during the learning process, and using the training set to iteratively train the BP neural network based on the gradient descent algorithm. In each iteration, the error between the output of the network and the actual unbalanced position is calculated, and then the weight and bias of the network are adjusted according to the direction of the gradient descent, so that the error is gradually reduced. For example, in the first iteration, a set of input data (such as a specific amplitude distribution, spindle speed, etc.) is input into the network to obtain a predicted unbalanced position. Then the error between this predicted value and the actual unbalanced position is calculated, and the parameters of the network are adjusted according to the error back propagation algorithm. Then proceed to the next iteration, and repeat this process until the performance of the network gradually improves.

[0035] Step S253, after the preset number of times is met, the verification training is performed using the verification set until the output result meets the expected convergence condition, and the first imbalance prediction branch is output. Specifically, after the preset number of iterative training is met, the verification training is performed using the verification set. The verification set is independent during the training process and does not participate in the adjustment of network parameters. The input data in the verification set is input into the network that has undergone a certain number of iterative training to obtain the predicted imbalance position, and then the difference between these predicted values ​​and the actual imbalance position is compared to evaluate the performance of the network. If the output result does not meet the expected convergence condition, continue to train and adjust, and repeat this process until the output result meets the expected convergence condition. The expected convergence condition can be that the error is less than a certain value, the accuracy reaches a certain level, etc. When the output result of the network meets the expected convergence condition, the first imbalance prediction branch is output. The branch is obtained through training and verification, and can be used to predict the imbalance position of the cutting spindle of the dicing machine.

[0036] Step S300, establish a mapping association between the cutting condition and the unbalanced prediction branch, and based on the mapping association, generate an unbalanced predictor according to the multiple cutting conditions and the multiple unbalanced prediction branches. Specifically, first clarify the characteristics of the cutting condition and the unbalanced prediction branch, and establish a mapping association between the cutting condition and the unbalanced prediction branch by determining the combination of unbalanced influencing factors corresponding to each cutting condition, which can be recorded by establishing a mapping table or the like. Then, based on this mapping association, multiple cutting conditions and corresponding unbalanced prediction branches are integrated to generate an unbalanced predictor. The unbalanced predictor can call the corresponding unbalanced prediction branch to predict the unbalanced position according to the input real-time working condition information, determine its input as the working condition information such as the real-time spindle speed, cutting load, material properties and ambient temperature, and output as the predicted unbalanced position distribution. By integrating the prediction results of each unbalanced prediction branch, such as weighted average, the final unbalanced position distribution is obtained, which provides a powerful tool for the dynamic balance test and counterweight adjustment of the cutting spindle of the dicing machine.

[0037] Step S400, monitor and obtain real-time working conditions and impact data sets, input the real-time working conditions and impact data sets into the imbalance predictor, perform imbalance prediction for a predetermined period, and generate imbalance position distribution. Specifically, first, monitor and obtain real-time working conditions (including spindle speed, cutting load, material properties and ambient temperature, etc.) and impact data sets (such as amplitude distribution, spindle temperature and coolant temperature, etc.) by installing sensors at key parts of the dicing machine, and process the collected data to ensure accuracy and reliability. Then, the processed real-time working conditions and impact data sets are input into the imbalance predictor, which consists of multiple imbalance prediction branches, and can find the corresponding branches for processing according to the mapping association. Then determine the predetermined period for imbalance prediction, and the predictor calls the corresponding branch for prediction according to the input data, and calculates the possible imbalance position and the probability of occurrence. Finally, the prediction results of each branch are integrated to obtain the imbalance position distribution, and output in the form of graphics, tables, etc., to provide a basis for subsequent dynamic balancing tests and counterweight adjustments.

[0038] In one possible implementation, real-time working conditions and influencing data sets are monitored and acquired, the real-time working conditions and influencing data sets are input into the imbalance predictor, imbalance prediction is performed for a predetermined period, and imbalance position distribution is generated. Step S400 further includes step S410, configuring a factor monitoring network according to the imbalance influencing factors, wherein the factor monitoring network includes a plurality of laser vibrometers installed on the cutting spindle of the dicing machine, and the laser vibrometers have position identifiers. Specifically, first, it is necessary to clarify the factors affecting imbalance, including amplitude distribution, spindle speed, cutting load, spindle temperature and coolant temperature. In order to accurately monitor the impact of these factors on the cutting spindle of the dicing machine, it is necessary to configure a corresponding factor monitoring network. The factor monitoring network includes multiple laser vibrometers installed on the cutting spindle of the dicing machine. The laser vibrometer is a high-precision measuring device that can accurately detect the vibration of the spindle. By installing multiple laser vibrometers, the spindle can be monitored from different positions to obtain more comprehensive vibration information. Each laser vibrometer has a position identifier, which can be a number or coordinate used to distinguish different laser vibrometers and determine their position on the spindle. In subsequent data processing and analysis, it is possible to accurately know where each vibration data comes from, which helps to more accurately analyze the imbalance of the spindle.

[0039] Step S420, using the factor monitoring network to perform fixed-point monitoring on the dicing machine, and obtaining the impact data set, wherein the impact data set includes real-time amplitude distribution, and the real-time amplitude distribution includes vibration amplitudes at multiple positions. Specifically, the configured factor monitoring network is used to perform fixed-point monitoring on the dicing machine. Fixed-point monitoring means monitoring the dicing machine at a specific position and time to obtain an accurate impact data set, and the impact data set includes real-time amplitude distribution. Real-time amplitude distribution refers to the vibration amplitude at different positions on the spindle at the monitoring moment. Through the measurement of multiple laser vibrometers, the vibration amplitudes at multiple positions can be obtained, thereby constructing a real-time amplitude distribution. For example, assuming that five laser vibrometers are installed on the cutting spindle of the dicing machine, which are located at different positions, each laser vibrometer will measure a vibration amplitude value when monitoring at a certain moment. These values ​​are sorted according to the position identifier to obtain a real-time amplitude distribution, which includes the vibration amplitudes at five positions. The factor monitoring network is used to perform fixed-point monitoring on the dicing machine, and an impact data set including real-time amplitude distribution is obtained, which provides important data support for subsequent imbalance prediction and dynamic balancing tests.

[0040] In a possible implementation, real-time working conditions and impact data sets are monitored and acquired, and the real-time working conditions and impact data sets are input into the imbalance predictor, imbalance prediction of a predetermined period is performed, and imbalance position distribution is generated. Step S400 further includes step S430, in which the real-time working conditions are input into the imbalance predictor, and the associated imbalance prediction branch is called. Specifically, first, the real-time working conditions are input into the imbalance predictor, and the real-time working conditions include information such as the current spindle speed, cutting load, material properties, and ambient temperature of the dicing machine. The information reflects the working state of the dicing machine at the current moment. The imbalance predictor is a tool generated based on multiple cutting working conditions and multiple imbalance prediction branches, and can predict the imbalance position distribution based on the input working condition information. When the real-time working conditions are input, the imbalance predictor will find the imbalance prediction branch related to the working condition according to the mapping association. For example, the real-time working conditions may show that the spindle speed is in the medium speed range, the cutting load is medium, the material is a specific metal, and the ambient temperature is in a moderate range. The imbalance predictor determines the corresponding imbalance prediction branch based on this information, and calls the branch for subsequent imbalance prediction.

[0041] Step S440, using the associated imbalance prediction branch to perform N imbalance predictions on the impact data set for a predetermined period, and determine multiple predicted imbalance positions and multiple position occurrence frequencies, wherein N is greater than or equal to 50. Specifically, using the called associated imbalance prediction branch to perform imbalance prediction on the impact data set, the impact data set usually includes real-time amplitude distribution and other information related to imbalance, and performing N imbalance predictions for a predetermined period, wherein N is greater than or equal to 50, is to improve the accuracy and reliability of the prediction. Through multiple predictions, the influence of accidental factors can be reduced, and a more stable prediction result can be obtained. In each prediction, the associated imbalance prediction branch determines a possible imbalance position according to the input impact data set. After N predictions, multiple predicted imbalance positions and the frequency of occurrence of each position are obtained. For example, after 50 predictions, three different predicted imbalance positions may be obtained, wherein position A appears 20 times, position B appears 15 times, and position C appears 15 times.

[0042] Step S450, set the predicted unbalanced position corresponding to the position occurrence frequency greater than the predetermined frequency threshold as the target unbalanced position, and construct the unbalanced position distribution according to the target unbalanced position. Specifically, set a predetermined frequency threshold to screen out important predicted unbalanced positions. The predetermined frequency threshold can be set according to actual conditions, usually determined based on experience or data analysis. Set the predicted unbalanced position corresponding to the position occurrence frequency greater than the predetermined frequency threshold as the target unbalanced position. For example, if the predetermined frequency threshold is 18 times, then in the above example, position A will be set as the target unbalanced position, and construct the unbalanced position distribution according to the target unbalanced position. The unbalanced position distribution is a description of the position where the cutting spindle of the dicing machine may be unbalanced. By determining the target unbalanced position, a more accurate unbalanced position distribution can be constructed to provide guidance for subsequent spindle dynamic balancing tests and counterweight adjustments. For example, according to the target unbalanced position A, the position can be clearly marked in the unbalanced position distribution, and information such as the probability of its occurrence or the degree of importance can be given.

[0043] Step S500, perform a spindle dynamic balance test and counterweight adjustment based on the unbalanced position distribution. Specifically, before performing a spindle dynamic balance test and counterweight adjustment based on the unbalanced position distribution, the spindle of the dicing machine and related equipment must be inspected and prepared, including ensuring that the spindle is firmly installed and that the dynamic balance test equipment is ready. Then carefully analyze the unbalanced position distribution to determine the areas that need to be focused on, then install a sensor on the spindle, start the dicing machine to let the spindle rotate at a certain speed, and the sensor collects vibration data and transmits it to the analysis equipment to determine the imbalance amount and position. After that, determine the counterweight scheme according to the imbalance situation, add counterweight blocks, and then perform a dynamic balance test. If the imbalance amount is still large, continue to adjust the position and weight of the counterweight blocks until the spindle reaches a balanced state. After completing the counterweight adjustment, verify the balancing effect. If it is not ideal, optimize the adjustment. Finally, record the entire process data and results and summarize and analyze them to improve the working performance and stability of the dicing machine.

[0044] The embodiment of the present application uses the method of calling the work log of the dicing machine to determine multiple cutting conditions, constructing an unbalance prediction branch in combination with unbalance influencing factors, and generating an unbalance predictor. The real-time working conditions and influencing data sets are monitored and acquired, and the unbalance predictor is input for unbalance prediction, and the unbalance position distribution is generated, and then the spindle dynamic balance test and counterweight adjustment are performed. By accurately predicting the unbalance position under various working conditions, the spindle dynamic balance test and counterweight adjustment are performed based on the predicted unbalance position, which can reduce the time and resources consumed by blind testing, thereby achieving the technical effect of improving test accuracy and efficiency and reducing resource waste.

[0045] In the above, refer to Figure 1A method for testing the dynamic balance of a dicing machine cutting spindle according to an embodiment of the present invention is described in detail. Figure 2 A dicing machine cutting spindle dynamic balance testing device according to an embodiment of the present invention is described.

[0046] A dicing machine cutting spindle dynamic balance test device according to an embodiment of the present invention is used to solve the technical problem that the existing dynamic balance test method has insufficient accuracy and efficiency in dynamic balance test and correction due to the large blindness of position test. By accurately predicting the unbalanced position under various working conditions, the spindle dynamic balance test and counterweight adjustment are performed based on the predicted unbalanced position, which can reduce the time and resources consumed by blind testing, thereby achieving the technical effect of improving test accuracy and efficiency and reducing resource waste. A dicing machine cutting spindle dynamic balance test device includes: a cutting condition determination module 10, an unbalance prediction branch construction module 20, an unbalance predictor generation module 30, an unbalance position distribution generation module 40, and a counterweight adjustment module 50.

[0047] The cutting condition determination module 10 is used to call the work log of the dicing machine to perform cutting condition analysis and determine multiple cutting conditions.

[0048] The imbalance prediction branch construction module 20 is used to obtain the imbalance influencing factors of the cutting spindle of the dicing machine, and construct multiple imbalance prediction branches according to the imbalance influencing factors and multiple cutting conditions.

[0049] The unbalance predictor generation module 30 is used to establish a mapping association between cutting conditions and unbalance prediction branches, and based on the mapping association, generate an unbalance predictor according to the multiple cutting conditions and the multiple unbalance prediction branches.

[0050] The imbalance position distribution generating module 40 is used to monitor and obtain real-time working conditions and impact data sets, input the real-time working conditions and impact data sets into the imbalance predictor, perform imbalance prediction for a predetermined period, and generate imbalance position distribution.

[0051] The counterweight adjustment module 50 is used to perform a spindle dynamic balance test and counterweight adjustment based on the imbalance position distribution.

[0052] Below, the specific configuration of the cutting condition determination module 10 will be described in detail. As described above, the work log of the dicing machine is called to perform cutting condition analysis to determine multiple cutting conditions. The cutting condition determination module 10 further includes: a condition factor acquisition unit, the condition factor acquisition unit is used to acquire condition factors, wherein the condition factors at least include spindle speed, cutting load, material properties and ambient temperature; a feature extraction unit, the feature extraction unit is used to extract features from the work log of the dicing machine based on the spindle speed, cutting load, material properties and ambient temperature, and determine the spindle speed threshold, cutting load threshold, material property threshold and ambient temperature threshold; a cutting condition generation unit, the cutting condition generation unit is used to segment the spindle speed threshold, cutting load threshold, material property threshold and ambient temperature threshold according to a predetermined segmentation scalar, and randomly combine the segmentation results to generate the multiple cutting conditions.

[0053] The specific configuration of the unbalanced prediction branch construction module 20 will be described in detail below. As described above, the unbalanced influencing factors of the cutting spindle of the dicing machine are obtained, and multiple unbalanced prediction branches are constructed according to the unbalanced influencing factors and multiple cutting conditions. The unbalanced prediction branch construction module 20 further includes: an unbalanced influencing factor composition unit, and the unbalanced influencing factor composition unit is used for the unbalanced influencing factors to at least include amplitude distribution, spindle speed, cutting load, spindle temperature and coolant temperature.

[0054] Among them, the imbalance prediction branch construction module 20 further includes: a state feature acquisition unit, the state feature acquisition unit is used to acquire the equipment state features of the dicing machine, wherein the equipment state features at least include equipment type, equipment specifications, tool wear status and bearing wear status; a dynamic balance association retrieval unit, the dynamic balance association retrieval unit is used to randomly select a first working condition from the multiple cutting working conditions, and use the equipment state features and the first working condition as similarity constraints to perform spindle dynamic balance association retrieval to obtain a retrieval data set; a sample data set acquisition unit, the sample data set acquisition unit is used to screen the retrieval data set according to the imbalance influencing factors to obtain a sample data set; a supervised training unit, the supervised training unit is used to use the sample data set to perform supervised training on the BP neural network to obtain a first imbalance prediction branch that meets the expected convergence conditions, and add it to the multiple imbalance prediction branches.

[0055] The retrieval data set is screened according to the imbalance influencing factors to obtain a sample data set, and the sample data set acquisition unit further includes: an initial sample data set acquisition subunit, the initial sample data set acquisition subunit is used to screen the retrieval data set according to the imbalance influencing factors to obtain an initial sample data set, wherein the initial sample data set includes a sample amplitude distribution, a sample spindle speed, a sample cutting load, a sample spindle temperature, a sample coolant temperature and a sample imbalance position; a similarity comparison subunit, the similarity comparison subunit is used to randomly select a first initial sample data from the initial sample data set, perform a similarity comparison on the first sample amplitude distribution and other sample amplitude distributions in the initial sample data set, obtain the number of data with a similarity less than a similarity threshold, and set it as a first difference coefficient; a deviation calculator An operator unit, wherein the deviation calculation subunit is used to calculate the deviation of the first sample spindle speed and other sample spindle speeds in the initial sample data set if the first difference coefficient is greater than a predetermined difference threshold, and to count the number of data with deviation values ​​greater than the speed deviation threshold, and set it as the second difference coefficient; a difference coefficient determination subunit, wherein the difference coefficient determination subunit is used to analyze the first sample cutting load, the first sample spindle temperature and the first sample coolant temperature in turn if the second difference coefficient is greater than the predetermined difference threshold, and determine the third difference coefficient, the fourth difference coefficient and the fifth difference coefficient; a sample data adding subunit, wherein the sample data adding subunit is used to add the first initial sample data to the sample data set if the third difference coefficient, the fourth difference coefficient and the fifth difference coefficient are all greater than the predetermined difference threshold.

[0056] Among them, the BP neural network is supervisedly trained using the sample data set to obtain a first imbalance prediction branch that meets the expected convergence conditions and is added to the multiple imbalance prediction branches. The supervised training unit further includes: a sample data set division subunit, the sample data set division subunit is used to divide the sample data set into a training set and a verification set according to a preset ratio, wherein the training set accounts for 80% and the verification set accounts for 20%; an iterative training subunit, the iterative training subunit is used to use the sample amplitude distribution, sample spindle speed, sample cutting load, sample spindle temperature and sample coolant temperature as input data, and the sample imbalance position as supervision, based on the gradient descent algorithm, and use the training set to iteratively train the BP neural network; a first imbalance prediction branch output subunit, the first imbalance prediction branch output subunit is used to use the verification set for verification training after meeting the preset number of times, until the output result meets the expected convergence conditions, and output the first imbalance prediction branch.

[0057] The specific configuration of the imbalance position distribution generation module 40 will be described in detail below. As described above, the real-time working condition and the impact data set are monitored and acquired, and the real-time working condition and the impact data set are input into the imbalance predictor to perform imbalance prediction for a predetermined period and generate the imbalance position distribution. The imbalance position distribution generation module 40 further includes: a factor monitoring network configuration unit, the factor monitoring network configuration unit is used to configure a factor monitoring network according to the imbalance impact factor, wherein the factor monitoring network includes a plurality of laser vibrometers installed on the cutting spindle of the dicing machine, and the laser vibrometer has a position mark; an impact data set acquisition unit, the impact data set acquisition unit is used to perform fixed-point monitoring of the dicing machine using the factor monitoring network, and acquire the impact data set, wherein the impact data set includes a real-time amplitude distribution, and the real-time amplitude distribution includes the vibration amplitudes of multiple positions.

[0058] Among them, the unbalanced position distribution generating module 40 further includes: a prediction branch calling unit, the prediction branch calling unit is used to input the real-time working condition into the unbalanced predictor, and call the associated unbalanced prediction branch; a predetermined period prediction unit, the predetermined period prediction unit is used to use the associated unbalanced prediction branch to perform N imbalance predictions on the influencing data set for a predetermined period, and determine multiple predicted unbalanced positions and multiple position occurrence frequencies, wherein N is greater than or equal to 50; an unbalanced position distribution constructing unit, the unbalanced position distribution constructing unit is used to set the predicted unbalanced position corresponding to the position occurrence frequency greater than the predetermined frequency threshold as the target unbalanced position, and construct the unbalanced position distribution according to the target unbalanced position.

[0059] A dicing machine cutting spindle dynamic balance test device provided in an embodiment of the present invention can execute a dicing machine cutting spindle dynamic balance test method provided in any embodiment of the present invention, and has corresponding functional modules and beneficial effects of the execution method.

[0060] Although the present application makes various references to certain modules in the device according to the embodiments of the present application, any number of different modules may be used and run on the user terminal and / or server, and the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention.

[0061] The above specific implementation manner does not constitute a limitation to the protection scope of the present application. It should be understood by those skilled in the art that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application should be included in the protection scope of the present application. In some cases, the actions or steps recorded in the present application can be performed in an order different from that in the embodiment and can still achieve the desired results. In addition, the process depicted in the accompanying drawings does not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A method for testing the dynamic balance of a cutting spindle of a dicing machine, characterized in that: include: Call the work log of the dicing machine to analyze the cutting conditions and determine multiple cutting conditions; Obtaining unbalance influencing factors of a cutting spindle of a dicing machine, and constructing multiple unbalance prediction branches according to the unbalance influencing factors and multiple cutting conditions; Establishing a mapping association between cutting conditions and imbalance prediction branches, and generating an imbalance predictor according to the plurality of cutting conditions and the plurality of imbalance prediction branches based on the mapping association; Monitor and obtain real-time working conditions and impact data sets, input the real-time working conditions and impact data sets into the imbalance predictor, perform imbalance prediction for a predetermined period, and generate imbalance position distribution; Performing a spindle dynamic balance test and counterweight adjustment based on the imbalance position distribution; The imbalance influencing factors at least include amplitude distribution, spindle speed, cutting load, spindle temperature and coolant temperature; According to the imbalance influencing factors and multiple cutting conditions, multiple imbalance prediction branches are constructed, including: Acquire equipment status characteristics of the dicing machine, wherein the equipment status characteristics at least include equipment type, equipment specification, tool wear status, and bearing wear status; Randomly selecting a first working condition from the multiple cutting working conditions, taking the equipment state characteristics and the first working condition as similarity constraints, performing spindle dynamic balance association retrieval to obtain a retrieval data set; Screening the search data set according to the imbalance influencing factors to obtain a sample data set; The sample data set is used to perform supervised training on the BP neural network to obtain a first imbalance prediction branch that meets the expected convergence condition, and the first imbalance prediction branch is added to the multiple imbalance prediction branches.

2. A dicing machine cutting spindle dynamic balance test method according to claim 1, characterized in that: Call the work log of the dicing machine to analyze the cutting conditions and determine multiple cutting conditions, including: Acquiring operating condition factors, wherein the operating condition factors at least include spindle speed, cutting load, material properties and ambient temperature; Perform feature extraction on the work log of the dicing machine based on the spindle speed, cutting load, material properties and ambient temperature to determine a spindle speed threshold, a cutting load threshold, a material property threshold and an ambient temperature threshold; The spindle speed threshold, cutting load threshold, material property threshold and ambient temperature threshold are segmented according to a predetermined segmentation scalar, and the segmentation results are randomly combined to generate the multiple cutting working conditions.

3. A dicing machine cutting spindle dynamic balance test method according to claim 1, characterized in that: Get a sample dataset, including: Screening the search data set according to the imbalance influencing factors to obtain an initial sample data set, wherein the initial sample data set includes a sample amplitude distribution, a sample spindle speed, a sample cutting load, a sample spindle temperature, a sample coolant temperature, and a sample imbalance position; Randomly selecting first initial sample data from the initial sample data set, performing a similarity comparison between the first sample amplitude distribution and other sample amplitude distributions in the initial sample data set, and obtaining the number of data with a similarity less than a similarity threshold, which is set as a first difference coefficient; If the first difference coefficient is greater than a predetermined difference threshold, performing deviation calculation on the first sample spindle speed and other sample spindle speeds in the initial sample data set, and counting the number of data with deviation values ​​greater than the speed deviation threshold, which is set as a second difference coefficient; If the second difference coefficient is greater than a predetermined difference threshold, analyzing the first sample cutting load, the first sample spindle temperature, and the first sample coolant temperature in sequence to determine a third difference coefficient, a fourth difference coefficient, and a fifth difference coefficient; If the third difference coefficient, the fourth difference coefficient and the fifth difference coefficient are all greater than a predetermined difference threshold, the first initial sample data is added to the sample data set.

4. A dicing machine cutting spindle dynamic balance test method according to claim 3, characterized in that: Construct the first imbalance prediction branch, including: Divide the sample data set into a training set and a validation set according to a preset ratio, wherein the training set accounts for 80% and the validation set accounts for 20%; Using sample amplitude distribution, sample spindle speed, sample cutting load, sample spindle temperature and sample coolant temperature as input data, using sample unbalanced position as supervision, and using the training set to iteratively train the BP neural network based on a gradient descent algorithm; After the preset number of times is met, verification training is performed using the verification set until the output result meets the expected convergence condition, and the first imbalance prediction branch is output.

5. A dicing machine cutting spindle dynamic balance test method according to claim 1, characterized in that: Monitoring captures impact data sets, including: configuring a factor monitoring network according to the imbalance influencing factors, wherein the factor monitoring network includes a plurality of laser vibrometers installed on a cutting spindle of a dicing machine, and the laser vibrometers have position marks; The factor monitoring network is used to perform fixed-point monitoring on the dicing machine to obtain the impact data set, wherein the impact data set includes a real-time amplitude distribution, and the real-time amplitude distribution includes vibration amplitudes at multiple positions.

6. A dicing machine cutting spindle dynamic balance test method according to claim 1, characterized in that: Generate an unbalanced position distribution, including: Inputting the real-time operating condition into the imbalance predictor and calling the associated imbalance prediction branch; Using the associated imbalance prediction branch to perform N imbalance predictions on the impact data set for a predetermined period, and determining a plurality of predicted imbalance positions and a plurality of frequencies of occurrence of the positions, wherein N is greater than or equal to 50; The predicted unbalanced position corresponding to the position occurrence frequency greater than a predetermined frequency threshold is set as a target unbalanced position, and the unbalanced position distribution is constructed according to the target unbalanced position.

7. A dicing machine cutting spindle dynamic balance test device, characterized in that: The device is used to implement a dicing machine cutting spindle dynamic balance test method according to any one of claims 1 to 6, and the device comprises: A cutting condition determination module, the cutting condition determination module is used to call the work log of the dicing machine to perform cutting condition analysis and determine multiple cutting conditions; An unbalanced prediction branch building module, the unbalanced prediction branch building module is used to obtain unbalanced influencing factors of the cutting spindle of the dicing machine, and build multiple unbalanced prediction branches according to the unbalanced influencing factors and multiple cutting conditions; An unbalance predictor generation module, the unbalance predictor generation module is used to establish a mapping association between a cutting condition and an unbalance prediction branch, and based on the mapping association, generate an unbalance predictor according to the plurality of cutting conditions and the plurality of unbalance prediction branches; an unbalanced position distribution generating module, the unbalanced position distribution generating module is used to monitor and obtain real-time working conditions and impact data sets, input the real-time working conditions and impact data sets into the unbalance predictor, perform unbalance prediction for a predetermined period, and generate unbalanced position distribution; A counterweight adjustment module is used to perform a spindle dynamic balance test and counterweight adjustment based on the imbalance position distribution.

Images