The iron and steel industry has always been a typical representative of heavy industry and traditional industries. However, with the advent of Industry 4.0 and the era of intelligence, and with the proposal and gradual implementation of the “dual carbon policy”, the iron and steel industry is also moving towards energy saving, low carbon, high efficiency and intelligence. transformation. Moreover, safety, energy saving, and high efficiency are the eternal focus of large-scale production enterprises and raw material processing industries. In industrial upgrading and production line transformation, it is natural to be inseparable from the multi-faceted assistance of machine vision.
As we know, machine vision is still a relatively new technology and is still in the development stage. Although it has been widely used in 3C electronics, automobiles, photovoltaics and other fields, due to the huge differences in working conditions and technical requirements, the experience of machine vision in laboratories or assembly lines cannot be directly applied to the steel industry. The application is still single and limited. In order to enable machine vision to better assist the transformation and upgrading of the steel industry, it is necessary to give full play to the advantages of machine vision technology and continuously overcome the challenges brought about by unfamiliar working conditions.
1. Large field of view imaging, improve efficiency
In the full-chain production process of iron and steel products, there will be a large number of intermediate products that need to be stored and transported efficiently. For bulky and heavy products such as steel billets, steel ingots, steel coils (plates), and rolls (bars), if manual inventory is used and on-site guidance is used for handling, it requires the cooperation of multiple people, and the efficiency is very low; it is difficult to operate continuously for 24 hours; It also comes with security risks. Usually, it is necessary to customize OCR coding reading, QR code recognition, visual counting and other solutions according to different site working conditions and different types of products, to track the location of large steel products, guide mechanical handling, and manage statistics of inbound and outbound information, etc.
With the continuous improvement of machine vision image resolution, changes in clarity and detection accuracy are one aspect. At the same time, the doubling of the field of view that can be covered under similar accuracy is a very obvious improvement. Intuitively speaking, with the same precision, the field of view of a 20M pixel camera is 4 times that of a 5M camera. If 5M pixel cameras need to be erected at the same time 4 or move 4 times to shoot, only one 20M pixel camera can be replaced. The high-resolution and large-field-of-view imaging solution can better improve the low efficiency of manual positioning, and the design and deployment of the visual system are also simpler; usually, considering the matching lens, data transmission cable, and acquisition card of the camera, a single high-resolution The cost of the high-speed camera solution will also be significantly reduced.
For example, in the warehouse of electro-slag billets, the products are multi-layered and staggered. The size of a single electro-slag billet is between 600~800mm in diameter and 4~6m in length. In addition, the materials are different, and there are many types of products, which cannot be placed regularly. In the past, it was determined by spraying the model, and the manual search was extremely slow. When the stack reached a certain height, it was very inconvenient for personnel to operate.
Nowadays, high-resolution industrial cameras and high-power industrial light sources can be used, and machine vision can be used to read and locate the two-dimensional code pasted on the end surface of the product. Each high-resolution camera can cover a field of view of about 3 to 10 meters, and 10 cameras can cover a storage area of 1,000 square meters. Using the image HDR algorithm and cooperating with light source scheduling, the positioning system can run stably and continuously, greatly saving manpower, and the search and positioning time is doubled.
2. Continuous monitoring to eliminate hidden dangers
The basic raw material of iron and steel products is iron powder processed from iron ore. The iron powder production process can basically be divided into mining, coarse crushing, fine crushing, roughing, grinding, and beneficiation. The process involves jaw crushers, ball mills, spiral classifiers, magnetic separators, etc., and then concentrates the tailings and transports them for refining. In the process of transportation and circulation in different processes, there are monitoring requirements for the size control of mineral materials, whether there are foreign objects and the status of conveyor belts. Abnormal dimensions, foreign objects in tools, torn conveyor belts, etc. can all cause damage to the safety of the production line.
Iron powder raw materials are mostly dark black and irregularly spherical, with low reflectivity and unobvious contours. It is difficult to extract effective features by ordinary algorithms, and it is impossible to distinguish the size of ore balls with similar surfaces. If line laser 3D scanning is used, the cost is high, and the speed is difficult to apply to high-speed transportation. Machine vision technology adopts a high-brightness diffuse surface light source + AVS depth learning algorithm, which can accurately capture and count over-limit mineral balls with only a small number of image samples. Therefore, real-time feedback through equipment communication can improve the ratio of raw materials in the process.
Compared with the manual monitoring method judged based on experience, machine vision monitoring has stronger stability and certainty, and is easy to carry out data summary and statistics; for harsh working conditions and full-time continuous monitoring, machine vision has stronger Adaptability; thereby effectively preventing and reducing the damage to the production line caused by abnormal conditions and the time loss caused by shutdown inspections.