With machine vision as the core, DBM can provide intelligent equipment R&D, manufacturing, and transformation services such as unmanned factories in the metallurgical industry and other industries, machine vision, and robot application design.
Tell us your needs and a technician will contact you immediately.
The core of intelligent manufacturing is the machine vision system. Compared with the human eye, the machine vision system has significant advantages in detection accuracy, detection efficiency, and detection speed. Machine vision can replace the human eye to achieve various functions in various scenarios. Large area is suitable for industrial manufacturing fields including semiconductors, automobiles, new energy, home appliances and other industries.
At present, traditional machine vision still has certain application limitations in the field of industrial manufacturing, and it is impossible to perform more detailed inspections on products. At this time, the introduction of AI vision technology into traditional machine vision can not only greatly improve the production accuracy and quality inspection of industrial manufacturing The accuracy of the results can not only reduce the production cost of the enterprise, but also help the enterprise to achieve efficient production.
AI+ machine vision technology can automatically extract and learn high-level semantic information from original images; extract multiple feature points from acquired images to more accurately describe objects or scenes; it can also identify and analyze complex background, Complex scenes such as illumination changes and attitude changes; fusion of multiple sensor information to obtain more comprehensive and accurate information; efficient processing of image data, so that useful information and patterns can be extracted from the data.
Through machine vision + AI, real-time automatic belt intelligent detection can be realized 7×24 hours, the frequency and intensity of manual inspection can be reduced, and intelligent operation and maintenance can be realized. The abnormal problem will be automatically alerted at the first time, which can effectively reduce the risk of downtime caused by belt failure.
Through machine vision + AI, unified, objective and accurate grading can be achieved, the human dependence on grading can be reduced, and the loss of grading deviation caused by human influence can be reduced. After deploying the scrap steel intelligent judgment application, the remote centralized control management of scrap steel can be realized, and the safety risk of personnel working on-site can be reduced.
Comprehensively consider the production plan of each link of steelmaking, the maintenance of the crane and equipment abnormalities, the real-time location information of the crane/steel ladle, various business rules, etc., make overall arrangements for the crane operation plan, and intelligently generate the traffic scheduling plan. For the dynamic changes in production, it can Complete the decision-making for the next 30 minutes within 1 minute and complete the issuance of instructions, effectively improving the turnover rate of ladles, reducing the waiting time for each furnace, reducing the temperature drop in the process, and reducing the cost per ton of steel in the steelmaking process of steel enterprises.
The rough rolling and steel transfer process of wide and thick plates originally relied on manual operations. The automatic steel transfer solution collects the real-time position and angle of the billet visually, analyzes it in real-time with AI, controls the speed of the roller table in real-time, and realizes accurate and real-time automatic steel plate rotation control. Combined with the real-time control of the electronic fence to ensure that the billet is always in the steel transfer area during the rotation process, ensuring production safety. Through a lot of practice, the average time for each steel transfer is shortened by nearly 3 seconds, a drop of 30%. This solution realizes the complete automation of the steel transfer process and greatly improves production efficiency.
Through big data, cloud, AI and other technologies, based on the principles of physical balance and thermal balance, and combining mechanism + data model, automatic calculation of alloy auxiliary materials addition, automatic planning of LF furnace electric argon blowing, real-time prediction of molten steel composition, avoiding manual operation. randomness and uncertainty. Taking a steelmaking plant with an annual output of 4 million tons of steel as an example, the cost of steelmaking can be saved by 8 million yuan per year.
The traditional steel plate defect detection method needs to wait for the steel cast slab to cool down before the quality inspector observes with the naked eye, and can only watch one surface at a time, which is not only time-consuming, but also difficult to quantify and standardize. Fatigue, negligence, will cause defective products to flow out.
AI continuous casting slab image thermal inspection system can quickly detect transverse cracks, longitudinal cracks, scratches and rejoining on steel cast slabs with a roller table speed of 0~60m/min at a high temperature of 600-1100°C defects, the width of longitudinal cracks and scratches is about 1mm. Moreover, the generated data can be saved and mined, which greatly improves the efficiency of quality inspection, ensures product quality, and improves productivity.
With its advantages of high precision, high efficiency, and high stability, AI technology provides the steel industry with important technical support and competitive advantage for intelligent transformation and upgrading. From the perspective of the industrial chain and value chain of the iron and steel industry, each link can be intelligent and digitalized with the help of various intelligent technologies to improve operation and management efficiency, such as smelting, monitoring, pouring steel and other links at the production end can use super automation Technology (robot process automation-RPA, machine learning-ML and artificial intelligence-AI and other tools work together) and sensor technology to achieve unmanned scheduling, real-time detection and early warning, and build customer self-service platforms and new channels at the downstream end The sales platform empowers steel enterprises with digital value.
Practical case: at the end of 2022, the stainless steel production line and ordinary carbon steel production line of a steel factory in Shandong
Through machine vision, measure the length of the billet and generate the billet number, track the position of the billet on the roller table in real-time, and finally control the production process through linkage with the crane system and process management system.
Since the on-site ambient temperature is kept above 40° all year round, and the ambient temperature of some installation points can reach above 60° when the steel billet passes by, so ordinary security cameras must first be excluded in camera selection. Then, among the few high-temperature-resistant cameras available, factors such as explosion-proof, field of view, pixels, etc. must be considered, and of course, the cost must be controlled in the end.
The process of going through the construction procedures is cumbersome, and then factors such as circuit routing, network routing, and node aggregation have to be considered. In addition, the choice of camera installation points is also very limited. Considering the factors of the crane movement, it is impossible to choose an ideal point to install the camera on both sides of the roller table. It can only be attached to the existing facilities, such as the control room. Walls, crossing columns.
All the information of AI vision comes from the eyes (camera), if the eyes are covered, everything is empty talk. There are many blocking interferences on the roller table on site, the most typical one is the roller table overpass.
In order to solve the problem of field-of-view occlusion, a cross-field-of-view installation method is adopted when installing the camera, that is, two cameras look at each other crosswise, and the side with better imaging effect should be selected as far as possible for the occluded area, which can effectively reduce the effect of field of view occlusion. interference.
The longest roller table on the production site is 130m long. According to the project requirements, it is necessary to track the billets on the entire roller table without dead ends. To meet the demand, it is difficult for a single camera. Even if there is a camera with an ok field of view and pixel accuracy, the site does not have ideal installation conditions. In this way, you can only choose multiple cameras for panoramic stitching.
However, the problem of panoramic stitching also arises. The effect of panoramic stitching is directly related to the accuracy of camera extrinsic calibration, but it is impossible to have an environment for accurate calibration of camera extrinsic parameters on the production site. The error of pairwise stitching of camera views will be transmitted and accumulated as the number of cameras increases. Simply put, the more cameras used for stitching, the worse the effect of panoramic stitching. For the 130m-long roller table to achieve full-view coverage, considering construction factors, at least 7 cameras are needed for splicing, and the final effect can be imagined.
The problem of field of view stitching is the core problem of this solution. Through the continuous simulation of various production scenarios through the data collected on site, the standard panorama stitching solution was finally abandoned, and a series of methods such as calibration of reference objects, screen cropping, and visual transformation were used. The method realizes the “logical” panorama stitching of the screen. In layman’s terms, each lens is only responsible for its designated field of view, and calculates separately on the data (including recognition, tracking, etc.), and finally summarizes all the results.
After more than 4 months, the slab roller track tracking system was finally officially put into trial operation in April 2023, bringing great economic benefits to the steel plant, and only using AI technology to replace front-line workers to perform some high-risk production operations This item is also very meaningful.
