Recently, the “CERI SmartEye” machine vision platform independently developed by CERI Digital Technology (Beijing) Co., Ltd. (hereinafter referred to as “CERI Digital Technology”) was successfully deployed and is now operating stably on a bar and wire production line at a steel mill, marking a significant transition of the rotary shear system from traditional open-loop control to AI-driven closed-loop adaptive control.

The platform establishes a full-chain AI vision loop encompassing "recognition–decision-making–execution–feedback", effectively addressing industry pain points associated with traditional rotary shears, such as poor shearing accuracy and material waste arise from the reliance on fixed parameters or empirical settings for shearing length under complex working conditions, including irregular shapes at the heads of rolled pieces, surface defects (such as cracks and rolling skin), and low-temperature areas (manifested as dark black sections).

Breakthrough in core technology: from "experience-based shearing" to "visual intelligence"

Traditional rotary shear systems are difficult to dynamically adapt to complex and changing field conditions. The “CERI SmartEye” platform uses a high-precision vision system to capture real-time images of heads of rolled pieces before shearing. It accurately identifies key features such as changes in the head geometry of rolled pieces, black defective areas on the surface, and dark, low-temperature sections, dynamically calculates the optimal head shearing length, and transmits the data in real time to the PLC control system, achieving truly precise shearing that is tailored to each steel grade. After shearing, non-contact measurement and real-time feedback are achieved through visual length measurement, forming a complete "set–execute–measure–correct" closed-loop system that drives the continuous self-optimization of shearing commands.

Significant application results: great improvement in accuracy and yield

After the visual length measurement and feedback functions are enabled, the average deviation in head shearing length is decreased from 23 mm under the reference working conditions to 5 mm, with fluctuations reduced by approximately 80%, resulting in significantly improved stability. The system achieves a recognition accuracy of 99.99% for features such as head defects, low-temperature sections, and scale. Combined with pre-recognition and closed-loop feedback, the measured head shearing length closely matches the optimized set value, effectively eliminating quality risks and significantly improving metal recovery rates.