Root Cause Analysis – Tracing a Problem to Its Origins
2018-12-06
Simulation & Modelling
2018-12-07
Real life experiences have shown that software-based condition monitoring models provide early warning of deteriorating performance
Model Based Condition Monitoring
Utilize IoT and Advanced Analytics to deliver
•Increased asset efficiency and effectiveness
•Reduce operational risk exposure
•Develop people
Time-in-State® Application
- FINANCIAL PERFORMANCE
- PEOPLE AND DECISION-MAKING
- REAL-TIME INTERPRETATION OF DATA
- SHARE HOLDER VALUE
- BIG DATA & IIoT
- ANALYTICS
The argument has been heard may times that control systems are already configured to detect out-of-bound signals – what is the benefit of adding yet another condition monitoring system?
However, real life experiences have shown that software-based condition monitoring models provide early warning of deteriorating performance:
Case 1: Gearbox – Abnormal behaviour was detected on the mill gearbox >30 days prior to failure. The control system registered alarm conditions only four days prior to the failure. The condition monitoring model utilized lubrication oil temperature, bearing temperature, and online vibration measurements as input parameters. A high vibration alarm was generated by the PLC / SCADA four days prior to the failure. Irrespective of the alarm conditions, the client was forced to run-to-failure because of constraints. Having experienced the actual failure provided a firm measure of time between first detection of an anomaly and the actual failure.
Case 2: Electrical motor (mechanical condition)- A condition monitoring model was used to detect abnormal behaviour. Specialists were brought in to execute vibration analysis. Vibration analysis indicated abnormalities on the motor and it was replaced during planned maintenance (subsequent inspection confirmed degradation of the drive-end bearing). In this case, the condition monitoring model provided warning of degradation 20 days prior to the planned motor change-out. An unplanned failure could have resulted in losses exceeding $1million. The condition monitoring model utilized bearing temperatures, winding temperatures, motor current and on-line vibration (accelerometers) measurements as input parameters. At no stage was an alarm condition triggered by the PLC/SCADA i.e. all parameters remained within alarm limits. Time-in-State detected the anomaly based on a deviation within the data pattern representing the motor’s baseline.
Case 3: Electrical Motor (electrical condition) Repeated failures associated with excessive brush wear and slipring flashes initiated the development of a model-based condition monitoring solution. “Major production losses were incurred because of mill motor failures – these failures were associated high brush wear and slip ring damage. The monetary value of this type of loss event can exceed $1.5m. Nkomati Mine utilized Time-in- State to analyse the conditions leading up to failures. Time-in-State is now doing real-time data interpretation / monitoring to provide early warning of high risk conditions. This solution has prevented at least two motor failures in the past four months.” – Braam van der Westhuizen, MMZ Engineering Manager. (The article provides further detail of this solution).
Rotor Phase Current Monitoring
A condition monitoring solution was co-developed between SDG Technologies and IME Solutions. SDG Technologies installed a rotor current transducer panel and measurements are integrated with the existing PLC.
Time-in-State® condition monitoring models were developed to monitor four individual motors. Data from the transducer panel are retrieved from the PLC via OPC and processed by the condition monitoring models.
WEB APPLICATION
Process data is interpreted in real-time and visualization takes place via a web based application. Intuitive and actionable information is presented.
Figure 1 illustrates an anomaly that was detected on one of the motors. The red arrow (top graph) shows the point when the health index value exceeded the upper limit. The green bands on the bottom three graphs represent the parameter index range associated with ‘normal’ conditions. Information contained in these graphs provides insight into the source or ‘root cause’. Stator Phase Current displayed on the right, as the digital read-out, appeared to be normal during the anomaly.
This example clearly proves the added value of software based condition monitoring.
BUSINESS APPLICATION
- FORENSIC ANALYSIS
- CONTINOUS IMPROVEMENT
- PRO-ACTIVE RISK MANAGEMENT
- OPTIMUM PERFORMANCE
- ASSET LIFE CYCLE COST MANAGEMENT
- TRAINING AND DEVELOPMENT
Integration into Asset Management Solutions
In collaboration with Vetasi Africa, Time-in State® condition monitoring models have been further enhanced to feed into the Enterprise Asset Management application. In this configuration Time-in-State® models are associated with specific equipment contained in the asset hierarchy. When an asset’s health index reaches the upper limit, it will trigger a predefined action in the asset management application. These actions can include investigations, inspections, work requests or work orders. Through predefined work flow processes these actions will be escalated, routed and assigned to resources in the organization. Dashboards provide insight on criticality, progress and risks of each of the actions initiated.
Time-in-State® delivers insight into issues and delivers answers to difficult questions through effective use of data, analytics, technology and people.
