Application of Intelligent Manufacturing Technology in Machining
The modern world is in the boom period of the new industrial revolution, and intelligent manufacturing technology is gradually becoming the main driving force.
The rapid development of information technology, the development of digital network and the integration of intelligent innovation technology are the three core power sources for the development of Industry 4.0.
The article takes intelligent CNC machine tools as a case study, and analyzes its intelligent application in the field of machining in detail.
By improving the stability, processing accuracy and product quality of machining equipment, it promotes the manufacturing industry to advance in the direction of digitization and intelligence, enhances the competitiveness of enterprises, and realizes the high-quality development of the machining field.
Based on industry 4.0 under the use of intelligent manufacturing technology in-depth study of CNC machine tools in the field of machining applications, for the realization of intelligent manufacturing, improve industrial production level is of great significance.
Theory Overview
Intelligent manufacturing technology
Industry 4.0 is a high-tech strategic plan put forward by the German government in 2011, aimed at promoting the digital transformation of the manufacturing industry through technological innovation, whose main features include the interconnectivity of the equipment, the integration of production, the process of real-time and production flexibility.
Intelligent manufacturing technology is a kind of advanced manufacturing technology based on computer simulation and data analysis, aiming to realize the intelligence and high integration of manufacturing process through the collection, storage, improvement, sharing, inheritance and development of intelligent information.
It uses computer simulation expert analysis, judgment, reasoning, conception and decision-making and other intelligent activities, so as to realize the highly flexible and highly integrated operation of the whole manufacturing enterprise.
Globally, many enterprises have begun to practice the concept of intelligent manufacturing and Industry 4.0.
For example, some enterprises through the introduction of automation equipment and robotics, reduce manual operation, improve production efficiency; through the Internet of Things technology to realize the equipment between the
Through the Internet of Things (IoT) technology to realize the interconnection between equipment, remote monitoring and control.
The use of artificial intelligence technology for predictive maintenance and fault diagnosis, the article is based on the development of intelligent manufacturing technology under Industry 4.0, combined with the application of CNC machine tools, to build a machining state diagnosis model based on the Adaboost algorithm, through the study of algorithms in the model training optimization strategy to build an effective state features, so as to improve the accuracy of the model in the state of diagnosis of CNC machine tools.
Machining
Machining refers to the process of changing the external dimensions or performance of the workpiece through mechanical equipment.
According to the different processing methods, machining can be divided into cutting and pressure processing.
Among them, cutting processing includes turning, milling, etc.; pressure processing includes forging, stamping, extrusion, etc..
Application of intelligent manufacturing technology in CNC machine tool machining
Machining state diagnosis
Adaboost algorithm
Adaboost algorithm of intelligent manufacturing technology is a typical enhancement of integrated learning method, which realizes the training of multiple weak classifiers with the help of repeated iterations and adjusts the sampling weights according to the performance of each classifier, and finally combines these weak classifiers into a strong classifier.
The Adaboost algorithm repeats the sample pattern prediction process, increasing the weight of the incorrectly predicted samples and decreasing the weight of the correctly predicted samples.
Throughout the training process in the field of machining, the set of data resources is fixed, and only the sample weights are adjusted singularly, and finally all the weak classifiers are integrated into one strong classifier.
The Adaboost integrated classifier algorithm works as follows.
(1) Given a training dataset π={(π₯1,π¦1),(π₯2,π¦2),β¦ ,(π₯π,π¦π)}, the weight distribution of the training set at the πth base classifier, is given by the following equation.
(1)
where: ππ is the weight distribution of the samples; π€ππ is the weight distribution of the training set at the πth iteration of the base classifiers; Eq.
π€ππ is the weight value corresponding to the π-th training sample at k iterations.
(2) At the πth iteration, the kth base classifier πΊπ(xπ) is trained on the training dataset according to the weight distribution of the sample π·π.
At this time, the weighted error rate on the training data set is calculated as follows.
(2)
Where: eπ is the πth base classifier; π is the weighting; π¦π is the measure of base classifier πΊπ(xi);
πΌ is the indicator function.
(3) Calculate the extreme weight coefficient of πΊπ(x), the specific formula is as follows.
(3)
where: ππ is the coefficient of the kth base classifier.
(4) Update the sample weight, the sample weight coefficient of the πth base classifier is π·π=(π€π1,π€π2..,π€πm), then the sample weight coefficients of the corresponding π+1th base classifiers are given by the following formula.
(4)
Where: ππ is the normalization factor; π€π+1, π is the corresponding weight of the πth training sample at π+1 iterations.
If the πth sample is misclassified, the weight of the sample in the π+1th base classifier will be increased; if it is correctly classified, the weight of the sample in the π+1th base classifier will be decreased.
(5) The ensemble strategy used for Adaboost classification is the weighted voting method, which constructs a linear combination of classifiers with the following formula.
(5)
The final strong classifier is formulated as follows.
(6)
Particle swarm optimization algorithm
In intelligent manufacturing technology, particle swarm optimization (PSO) is a kind of heuristic optimization algorithm, which repeatedly searches for the optimal solution in the search space, and simulates the cooperation and competition between individuals in the group to achieve the best value.
In PSO, the solution of the problem is regarded as a point in the search space and is called a particle. Each particle has its own position (solution vector) and velocity (size and direction of the update step of the solution vector).
The position of the particle represents the possible solutions of the problem, while the velocity represents the direction and speed of the solution towards the optimal solution.The operation of the PSO algorithm can be summarized in the following steps.
(1) Initialize the particle swarm, generate a set of random particles, and randomly initialize the positions and velocities of the particles.
(2) Evaluate each particle position and calculate its fitness value, i.e., the value of the objective function, the fitness value reflects the superiority of the particle position.
(3)Update the velocity and position of the particle using the current position and velocity of the particle, the historical optimal position of the population, and the global optimal position to update the velocity and position of each particle so as to avoid the dilemma of local optimal solution.
(4) The iterative process (the second and third steps) is repeated until the specified number of iterations is reached or the stop condition is satisfied, and the global optimal position found at last is the optimal solution of the problem.
Bearing vibration signal feature extraction
The development of Industry 4.0 promotes the continuous upgrading of intelligent manufacturing technology, so that it can be applied to the field of machining in various forms. The feature extraction of the bearing vibration signal in the machining process of intelligent CNC machine tools is mainly to analyze the statistical characteristics of the signal in the time domain and extract the key parameters reflecting the characteristics of the signal.
The article proposes a time-domain feature extraction method, using these parameters to judge the operating state of CNC machine tools in the machining process, control the quality of machining, and predict possible failures.
Diagnostic model construction of CNC machine tool health state
In the construction of CNC machine tool health diagnostic model, support vector machine (support vector machine, SVM) performance is affected by the penalty coefficient and Gaussian kernel function parameters.
In this paper, PSO optimization is used to determine the penalty coefficients and Gaussian kernel function parameters, which eliminates the traditional method of evaluating the classification results of the SVM model based on experience and classification results in the machining process.
In order to overcome the instability of the traditional SVM in diagnosis, the article uses the Adaboost integration algorithm to improve the SVM and enhance its performance significantly.
The Adaboost algorithm improves the accuracy of the state diagnosis model of CNC machine tools in the machining process to a certain extent by considering the weights of the sub-models and dynamically adjusting these weights.
The precision, recall and F1-score of the traditional binary classification model cannot be used as the evaluation index of the model in multiple classification tasks.
For this reason, we can introduce macro-average to predict the number of samples with large variations, and measure the accuracy, recall and F1-score of each category model by averaging, in order to obtain the macro-precision, macro-recall and macro-F1-score of the whole sample, where the macro-F1-score is the same as the macro-recall, and the macro-F1-score. Macro-F1 is a metric for evaluating the performance of a binary classification model by calculating the harmonic mean of precision and recall for each category and then taking the average of these values.
In addition, the article will choose Accuracy as the strategy index of the CNC machine tool condition prediction model with the following formula.
(7)
(8)
In the formula.
TP is that the model correctly classifies the positive category samples into positive category; and
FP is that the model incorrectly categorizes negative samples as positive samples; FN is that the model incorrectly categorizes positive samples as positive samples.
FN is the model incorrectly categorizing positive samples into negative categories; and
TN is the model correctly categorizing negative samples into negative categories; n is the number of categories; and
n is the number of categories.
Pi is the iterative coefficient of macro-accuracy; Ri is the iterative coefficient of macro-recall.
Ri is the iterative coefficient of macro recall rate.
Key points of CNC machine tool health state diagnosis
In the prediction model of CNC machine tool health state diagnosis in machining, the number of weak classifiers has a significant impact on the accuracy of CNC machine tool state diagnosis based on intelligent manufacturing technology in machining.
Adaboost algorithm is an integrated patterned training strategy, which improves the performance of weak classifiers by sequentially training multiple weak classifiers and effectively adjusting the sampling weights of CNC machine tools according to the performance of each classifier, and finally combining these weak classifiers into a strong classifier.
By increasing the number of weak classifiers, the interpretation ability of the CNC machine tool health diagnosis and prediction model is greatly enhanced, and it becomes more flexible and better adapts to the complex data generated during the operation of CNC machine tools, thus improving the ability to capture complex patterns and subtle features.
Since changes in machine state can be influenced by a variety of factors, a model with high accuracy is needed to capture these changes.
In machining, the CNC machine tool health diagnostic model can better capture the characteristics of the data and improve the overall performance of machining operations. The model has significant advantages in the machine tool state classification task, especially for complex data, and improves the model’s generalization ability and prediction accuracy.
Conclusion
With the continuous promotion of Industry 4.0, intelligent manufacturing technology is widely used in the field of machining, especially in the diagnosis and inspection of CNC machine tools.
Intelligent CNC machine tool monitoring technology has important research and application value in the field of machining.
By improving the stability of equipment, machining accuracy and product quality, the diagnostic model of CNC machine tool health status in machining helps to promote the manufacturing industry to the direction of digitalization and intelligence, enhance the competitiveness of enterprises, and promote the high-quality development of the manufacturing industry.