Turning center chip winding visual identification system and method

Abstract

The invention relates to the technical field of turning machining, in particular to a turning center chip winding visual recognition system and method.The recognition method comprises the following steps that S10, on-site machining pictures are collected, videos are recorded, frames are extracted into pictures, and a data set is established in combination with a web crawler; s20, designing a chip winding detection network model according to the characteristics of the data set; step S30, training a chip winding detection network model by using the data set; step S40, inputting the obtained video stream into the trained chip winding detection network model, and outputting whether chip winding is detected; and S50, judging whether an alarm is needed or not according to the output signal, the problems that in the prior art, the prevention effect is limited, installation is inconvenient and the like are solved, and whether the turning center has the cutting chip winding fault or not can be judged only by obtaining real-time machining information in the turning center.

Description

Technical Field

[0001] The invention relates to the technical field of turning processing, and in particular to a chip winding visual recognition system and method for a turning center. Background Art

[0002] In the modern manufacturing process, the realization of intelligence is particularly important. In CNC machine tool processing, iron chips are difficult to break and have become a common problem in mechanical processing. Especially when processing plastic metal materials, the strip chips formed are sometimes entangled on parts or tools. Severe chip entanglement will also scratch the processed surface of the parts, deteriorate the quality of the processed surface, and affect the service life of the parts.

[0003] Existing solutions include: manual inspection method, process improvement, chip breaker design, high-pressure cooling method, and external chip breaking device.

[0004] However, the manual inspection method is time-consuming and labor-intensive, and requires stopping during the processing of each workpiece, which seriously affects the continuity of processing.

[0005] Improved process: low applicability, only individual processes can be improved.

[0006] Design chipbreakers: Chipbreakers are effective in roughing or semi-finishing conditions with relatively large back-cutting and feed rates, but they completely lose their effectiveness when machining highly plastic materials under precision turning conditions.

[0007] High-pressure cooling method: In the actual production process, the cutting fluid system of the automated equipment often does not have a pressurization system, so the cutting fluid can only be pressurized by an external pressure pump, and the continuity of the high-pressure cooling system cannot be guaranteed. In addition, with the continuous development of machine tool technology, the internal structure of new automated CNC lathes is becoming more and more complex. The processing process will be accompanied by the movement of the tool holder and the mechanical arm, and the strong impact of the cutting fluid will cause the broken chips to accumulate at a certain position of the equipment, affecting the continuous processing process.

[0008] Additional chip breaking device: The chip breaking structure usually needs to be driven by transmission equipment such as gears, which requires a large space. However, the internal structure of new automation equipment is complex and the space is small, which usually makes it difficult to support the installation of such tools. In addition, there are many types of external turning tools with different tool head structures, and it is difficult for a certain tool head structure design to have universal applicability. Summary of the invention

[0009] The object of the present invention is to provide a chip winding visual recognition system and method for a turning center to solve the problems raised in the above-mentioned background technology.

[0010] To achieve the above object, the present invention provides the following technical solution: a visual identification method for chip winding in a turning center, the identification method comprising the following steps:

[0011] Step S10, collect on-site processing pictures, record videos, extract frames into pictures, and establish a data set in combination with a web crawler;

[0012] Step S20, designing a chip winding detection network model according to the characteristics of the data set;

[0013] Step S30, using the data set to train a chip winding detection network model;

[0014] Step S40, inputting the acquired video stream into the trained chip winding detection network model, and outputting whether chip winding is detected;

[0015] Step S50: Determine whether an alarm is required based on the output signal.

[0016] Preferably, in step S10, a certain number of internal processing images of the turning center are obtained by taking pictures with a camera, and more data pictures are crawled by a web crawler. The objects on the pictures are annotated with a picture annotation tool, and the data set is divided into a training set, a validation set and a test set according to a certain ratio. Data collection in various environments is simulated by rotation, random cropping, scale scaling, Gaussian noise, brightness adjustment and flip adjustment, so as to enrich data diversity and improve model robustness.

[0017] Preferably, in step S20, the chip winding detection network model includes detection confidence, and the Mosaic algorithm and the EasyEnsemble algorithm are added to the pre-trained model of YOLOv8n.

[0018] Preferably, in step S30, the training set and the validation set are input into the chip winding detection network model for training. After the training is completed, the test set is input into the chip winding detection network model, the detection accuracy is calculated, and it is determined whether the requirements are met. If so, the rotating target detection network model and training parameters are saved.

[0019] Preferably, in step S50, the obtained confidence is compared with a set confidence threshold, and if the offset angle exceeds the set threshold, it is determined that a chip winding fault occurs.

[0020] Preferably, in the step S50, in a specific implementation of the detection method, the threshold is set according to the best effect achieved at a certain confidence level during the training process.

[0021] A chip winding visual recognition system for a turning center includes a detection device, which includes an image acquisition module, a control module, an alarm module, and a power module. The image acquisition module is connected to the control module, the control module is connected to the alarm module, and the power module is connected to the image acquisition module, the control module, and the alarm module.

[0022] Compared with the prior art, the present invention has the following beneficial effects:

[0023] First, it overcomes the problems of limited prevention effect and inconvenient installation in the existing technology. Specifically, by arranging a camera inside the turning center, the internal processing screen of the turning center is obtained in real time, and the obtained video stream is input into the trained chip winding detection network model. The model automatically detects whether a chip winding fault occurs. If a chip winding fault is detected, an alarm message is issued;

[0024] Second, this method only needs to obtain the real-time processing information inside the turning center to judge whether the turning center has a chip winding fault. The equipment is relatively simple and can be deployed in a relatively safe location, which reduces the impact of harsh environments on the equipment and improves the stability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 A flow chart of the main steps of the chip winding identification system and method provided by the present invention.

[0026] Figure 2 This is a schematic diagram of the Mosaic algorithm provided by the present invention.

[0027] Figure 3 Schematic diagram of the EasyEnsemble algorithm provided by the present invention.

[0028] Figure 4 This is a logic diagram of the working principle of the device provided by the present invention.

[0029] Figure 5 A schematic diagram of a fault detection device provided by the present invention. DETAILED DESCRIPTION

[0030] In order to deepen the understanding and recognition of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described and introduced in combination with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments, and are not intended to limit the embodiments in any form. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0031] See also Figure 1-5The present invention provides a technical solution: a visual identification method for chip winding in a turning center, the identification method comprising the following steps:

[0032] Step S10, collect on-site processing pictures, record videos, extract frames into pictures, and establish a data set in combination with a web crawler;

[0033] Step S20, designing a chip winding detection network model according to the characteristics of the data set;

[0034] Step S30, using the data set to train a chip winding detection network model;

[0035] Step S40, inputting the acquired video stream into the trained chip winding detection network model, and outputting whether chip winding is detected;

[0036] Step S50: Determine whether an alarm is required based on the output signal.

[0037] In step S10, a certain number of internal processing images of the turning center are obtained by using a camera, and more data images are crawled through a web crawler to ensure that the data images contain relatively rich features as much as possible. The target on the image is annotated using an image annotation tool. Each internal data image of the turning center needs to be annotated to obtain the position and size information of the minimum circumscribed rectangle of the chip winding fault. The LabelMe annotation tool is used for annotation, and the data set is divided into a training set, a validation set, and a test set according to a certain ratio. The data collection in various environments is simulated by rotation, random cropping, scale scaling, Gaussian noise, brightness adjustment, and flip adjustment to enrich data diversity and improve the robustness of the model. The ratio of the validation set to the test set is 8:1:1.

[0038] In step S20, the chip winding detection network model includes detection confidence, and the Mosaic algorithm and the EasyEnsemble algorithm are added to the pre-trained model of YOLOv8n. The structure diagram of the Mosaic algorithm is as follows: Figure 2 The Mosaic data enhancement algorithm combines multiple images into one image according to a certain ratio, so that the model can identify targets in a smaller range. The specific implementation method is: first, randomly select the coordinates (xc, yc) of the image splicing reference point, and randomly select four other images; then, according to the reference points, the four images are resized and scaled, and placed in the upper left, upper right, lower left, and lower right positions of the large image of the specified size; then, according to the size transformation method of each image, the mapping relationship is mapped to the image label; finally, according to the specified horizontal and vertical coordinates, the large image is spliced, and the detection box coordinates that exceed the boundary are processed. The advantage of this is that the model robustness can be enhanced, while reducing the batch_size during training and improving the model training efficiency;

[0039] In order to improve the problem of sample imbalance during model training and avoid overfitting, the present invention adds the EasyEnsemble algorithm to the YOLOv8 algorithm. The structure diagram of the EasyEnsemble algorithm is shown in the figure. Figure 3 Show;

[0040] The core idea of ​​the EasyEnsemble algorithm is to construct multiple balanced data sets by repeatedly performing random sampling without replacement from the majority class, and then train the classification models on these data sets respectively. At the same time, since the average random sampling of data from multiple categories in an unbalanced data set will significantly reduce the number of training sets for each classification model, resulting in insufficient learning of sample features for each classification model and weak model expression, after training each classifier using each balanced data set, the algorithm will integrate the prediction results of the classification models to enhance the final prediction performance. The integration method is usually average probability integration or majority voting integration.

[0041] In step S30, the training set and the validation set are input into the chip winding detection network model for training. After the training is completed, the test set is input into the chip winding detection network model to calculate the detection accuracy and determine whether it meets the requirements. If so, the rotating target detection network model and training parameters are saved.

[0042] In step S50, the obtained confidence is compared with the set confidence threshold. If the offset angle exceeds the set threshold, it is determined that a chip winding fault occurs. In step S50, in a specific implementation of the detection method, the threshold is set according to the best effect achieved at a certain confidence during the training process. In step S40, the acquired video stream is input into the trained chip winding detection network model, the chip winding detection result is output, and a signal indicating whether an alarm is needed is output. The real-time video stream inside the turning center collected by the image acquisition device is input into the trained chip winding detection network model, and the detection frame information of the scraper in the video stream is output, wherein the detection frame information includes (x, y, w, h, s), that is, the center point coordinates, width, height and confidence of the detection frame. According to the above information, a basis is provided for whether to issue an alarm message; when the confidence s>0.8, an alarm message is issued. According to the results after model training, the average accuracy of the entire model is the highest when the confidence threshold is 0.8, so the confidence threshold of the entire system is set to 0.8. If the confidence exceeds the set threshold, it is determined that chip winding occurs.

[0043] like Figure 5As shown, the chip winding visual recognition system of the turning center includes a detection device, which includes an image acquisition module, a control module, an alarm module, and a power module. The image acquisition module is composed of an industrial camera, a lens or a camera, and is used to obtain real-time image information inside the turning center, connect to the control module, and transmit the image information to the control module. The control module is composed of an embedded platform or a PC. A chip winding detection network model is deployed in the control module to perform image processing, and then the occurrence of the fault is judged and processed in combination with the obtained confidence and the set confidence threshold. The alarm module is composed of an alarm and is connected to the control module. After the control module detects the occurrence of a chip winding fault, the alarm module sends a signal to the outside to remind the staff to deal with it in time. The power module is connected to the image acquisition module, the control module and the alarm module, and is used to supply power to various components in the device.

[0044] Although the embodiments of the present invention have been shown and described, it should be emphasized that the above description is only an introduction and description of the use of the embodiments of the present invention, and does not limit the present invention in any form. For those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the attached claims and their equivalents.

Claims

1. A visual identification method for chip winding in a turning center, characterized in that: The identification method comprises the following steps: Step S10, collect on-site processing pictures, record videos, extract frames into pictures, and establish a data set in combination with a web crawler; Step S20, designing a chip winding detection network model according to the characteristics of the data set; Step S30, using the data set to train a chip winding detection network model; Step S40, inputting the acquired video stream into the trained chip winding detection network model, and outputting whether chip winding is detected; Step S50: Determine whether an alarm is required based on the output signal.

2. The chip winding visual identification method of a turning center according to claim 1, characterized in that: In step S10, a certain number of internal processing images of the turning center are obtained by using a camera, and more data images are crawled through a web crawler. The objects on the images are annotated using an image annotation tool, and the data set is divided into a training set, a validation set, and a test set according to a certain ratio. Data collection in various environments is simulated by rotation, random cropping, scale scaling, Gaussian noise, brightness adjustment, and flip adjustment, so as to enrich data diversity and improve model robustness.

3. The chip winding visual identification method of a turning center according to claim 1, characterized in that: In step S20, the chip winding detection network model includes detection confidence, and the Mosaic algorithm and the EasyEnsemble algorithm are added to the pre-trained model of YOLOv8n.

4. The chip winding visual identification method of a turning center according to claim 1, characterized in that: In step S30, the training set and the validation set are input into the chip winding detection network model for training. After the training is completed, the test set is input into the chip winding detection network model to calculate the detection accuracy and determine whether the requirements are met. If so, the rotating target detection network model and training parameters are saved.

5. The chip winding visual identification method of a turning center according to claim 1, characterized in that: In the step S50, the obtained confidence is compared with a set confidence threshold, and if the offset angle exceeds the set threshold, it is determined that a chip entanglement fault occurs.

6. The chip winding visual identification method of a turning center according to claim 5, characterized in that: In the step S50, in a specific implementation of the detection method, the threshold is set according to the best effect achieved at a certain confidence level during the training process.

7. A chip winding visual recognition system for a turning center according to any one of claims 1 to 6, comprising a detection device, characterized in that: The detection device includes an image acquisition module, a control module, an alarm module, and a power module. The image acquisition module is connected to the control module, the control module is connected to the alarm module, and the power module is connected to the image acquisition module, the control module, and the alarm module.

Images