Though digitalization has been galvanising business processes to a great extent, new, breakthrough technologies such as Predictive Analytics will further help realize the value of digitalization.
Predictive Analytics is, in fact, making waves in the Energy Industry and is becoming indispensable for the industry to maintain its multi-billion dollar assets by preventing expensive and unexpected failure/downtime in the operating machines.
Adoption of right Predictive Analytic System for Energy Industry
Predictive Analytics helps detect faults and give early warnings in critical rotary equipment such as compressors, pumps, motors, turbines, fans, blowers, and gearboxes.
Merely having the data on the dashboard and analysing it to make decision is not enough. The data should be used to generate alerts and early warnings automatically to operators so that they may take the equipment/components for maintenance and thus avoid unexpected failures of the equipment and the cascading effect on the entire operation of the plant, which would result in huge losses. By helping in avoiding unexpected failure of equipment, unnecessary maintenance can be reduced drastically.
For an organization to choose and implement the right Predictive Analytics solution, it is essential to conduct a study on the Predictive Analytic system and know how to leverage the technological advancements.
Key elements that can determine successful implementation of Predictive Analytics-based maintenance system:
- Compatibility to fetch the live data from historian and import of data from various sources such as TAS/ DCS/SCADA/ PLC/ OPC/ Data base files etc.
- Modern and user-friendly interface with heavy usage of application using desktop and mobile Apps with dashboards.
- Commercially, the off-the-shelf (COTS) system should allow users to configure the system and deploy on the premises with standard libraries and easy to modify models.
- The system should generate notifications and alerts based on data driven techniques like statistical model and advance analytics using Artificial Intelligence (AI) and Machine Learning (ML).
- Compatible to your organizational Data Encryption/Data Protection policy to protect the data to be used for Predictive Analytics. The data is your Digital Asset.
If a system continues to generate numerous or false alerts, then the reliability of Predictive Analytics will be at stake. Here, Machine Learning (ML) and Artificial Intelligence (AI) can play an important role to recognize the pattern from data points and generate reliable and actionable alerts.
So, before you choose a system for predictive analytics, it is essential to evaluate the standard libraries for various types of equipments and the OOTB model available for various use cases. You can simulate with historian data to ascertain that the system functions as per its purpose, and only then deploy the system to work with live data coming from various sources. Developments of open sources are limited and not proven, while there are standard industry software such as AVEVA Predictive Analytics (earlier known as PRiSM), Hexagon EAM, and Baker Hughes Advanced ESP Predictive Failure Analytics, which are some of the globally adopted solutions by the Energy Industry to deploy Predictive Analytics to protect critical assets.
Hardware and Software Selection
The success of the system you choose and deploy for predictive analytics depends on the right combination of hardware and software. Like any analytical software that requires a high processing capacity, Predictive Analytics, when powered by machine learning, the system should be capable of recognizing the patterns generated by data points coming from sensors every day, week, and month. In order to identify the deviation, it really requires a robust hardware and supporting OS and other allied software applications for fetching, integrating, and synthesizing the data to be used for analysing. Do not select the hardware that is nearing its end of life and the same applies to the OS and other applications. Be conscious about compatibility and capability of the predictive analytics system to integrate with your existing systems from where the data is to be fetched. Before you implement it, a thorough investigation on the usability of the existing data available historically should be conducted to ascertain that the data available can be relied upon to build the models.
An illustration where the predictive model developed using historic data for (nearly) normal operation of an asset, is compared with actual data coming from an asset through sensors. Leveraging pattern recognizing algorithms and AI and ML techniques Alarm/Alerts will start generating when there is a deviation from the Predictive Model. Also, the traditional alarm, which is set based on threshold value and which is still far from the actual and predicted model, allows the operator to look at the root cause for this deviation well in advance, before it actually reaches the threshold level and prescribes the action to be performed on the asset being monitored and thereby prevent costly damage that might occur to the asset.
Deployment Methodology and KPI of System
The more you go digital, the more vulnerable you are to the risks of your assets. An idle predictive analytics solution should be deployed on the premises if an organization would like to avert any incident pertaining to data security. The security testing for the system should be conducted using VAPT tool or through CERT. The system performance should be linked to various KPI’s for its functionalities to realize the benefits.
