Maintenance & Reliability
Caster work rolls have historically been difficult to change out in a timely manner, presenting equipment setup and safe roll handling challenges. At Steel Dynamics Inc. (SDI) – Flat Roll Group Butler Division, a near-miss incident occurred that involved changing out a bending unit work roll. During the roll change process, roll bearing blocks must be repositioned to align with the guide rails of the track. During repositioning, the roll became derailed and fell to the floor. During the incident review, the problems with roll handling were obvious. However, other problems were also identified. Setting the roll track was done with a forklift without any good way to secure the track to the forks. The removal and installation of the rolls were accomplished with a cable and winch system, which required the operators to subject themselves to pinch points and fall exposures when installing the cable. In the event of replacing rolls in the withdrawal straightener unit, a forklift was used to pull the rolls out and push the rolls in using the forks against the bearing blocks. This created the potential for equipment damage.
Predictive maintenance, and specifically asset condition monitoring and analysis in metal production plants, continues to see significant advances in cloud computing, sensors, and machine learning technologies. Primetals Technologies and ITR have partnered to provide the latest predictive maintenance services and solutions and ongoing research and development in waveform analytics. After decades of using human and statistical techniques, machine learning methods are now being applied to further improve efficiency and accuracy in the diagnosis and prognosis of potential failure modes of rotating machinery. Machine learning methods allow monitoring systems to react even faster and more precisely than traditional tools and methods. This paper discusses how machine learning was successfully applied to route-based data collection and analysis and condition monitoring and analysis systems for machine vibration data to improve process efficiencies and system response times. Additionally, it demonstrates how machine learning further augments even the most experienced professional analysts to ensure no missed problems and false alarms.
The performance of caster bearings highly influences the overall caster performance. These bearings are subjected to high loads and high temperature along with a high amount of water ingresses in the grease and bearings. The compromised grease within the bearing affects the bearing life due to lubricant starvation and corrosion damage. A poor grease selection for caster bearings accelerates bearing damage, lowers bearing life and increases grease consumption. This paper will discuss grease selection for caster rolls. The following grease performance attributes were included in this investigation: load-carrying capacity, wear protection, water resistance, corrosion resistance, thermal stability and bearing performance. These properties are crucial for the service life of the bearings and are very much dependent on the grease formulation.
A system for fully automated, intelligent operation of a spray gunning manipulator is presented. The prototype features motion tracking, autonomous gunning, an in-furnace camera for live monitoring and data analysis, automatic delivery of maintenance reports, and remote control via mobile app. Video and sensor data can be uploaded to a remote storage using an onboard internet connection. The data is used to analyze and optimize gunning. The prototype is retrofitted to an existing manipulator with the vision to get machines of previous generations ready for the challenges of the Industry 4.0 era. Furthermore, integration into new machines is supported.
Repetitive loading of factory crane girders causes deformation and/or failure. In the presented method, light detection and ranging points over one face of a girder are processed to create deformation maps of panels: web area between stiffeners and flanges. Web surface profiles at top, mid and bottom locations are created by matching a standard I-section template to point data. Root mean square error (RMSE) at each girder slice is computed by comparing the nearest girder points and inflection points from the template. The girder point with the lowest RMSE is selected as the correct match and three locations are identified