An IoT powered predictive analysis and maintenance product tailored to automate feedback loops in the manufacturing processes for proactive maintenance and monitoring.
After years of relying on preventive and damage control maintenance, the power generation industry is moving into the new proactive and predictive maintenance territory. Powerful predictive maintenance technologies and services are changing the face of the industry and shaping the future of the maintenance practices. Advancement in data-centric technology is leading the charge from remote monitoring and diagnostics tools to big data analytics and management.
Earlier industries focused on preventive maintenance, relied on past performance data to develop scheduled maintenance and outage service schedules for a specific piece of equipment. However, this approach does not factor in dynamic variations in equipment behavior and its immediate environment that might affect it’s maintenance schedules. Enter predictive monitoring analytics, which can uncover an anomaly before a real problem or unplanned event occurs.
For utilities and equipment service providers, the change in philosophy represents a paradigm shift in how the energy industry manage its assets.The ability to foresee and forestall issues—until an optimal planned maintenance outage date can be selected—is at the very heart of predictive maintenance.
Here are some loopholes predictive analysis plugs in its approach:
Failures between data collection rounds: In the walk around programs, even in well-run ones, technicians only collect data once every 30 or 90 days. The problem is that any defects that occur after the last data collection cycle remain undetected until the next data collection cycle. As a result, machines remain prone to failures which can result in unplanned downtime.
Variable operating conditions: One of the extremely crucial aspects to consider during data collection is that the operating conditions of the equipment need to be constant during every iteration. However, this is hard to achieve. So when the analyst observes an alert such as “HIGH” by sensor they tend to inaccurately conclude it to be machine’s fault. However, it could have just spiked due to increased loads!
Inaccessible Machinery: Often machines are not easily accessible, either due to safety concerns or protective encapsulation. This leads to technicians not collecting complete data rendering their prediction inaccurate.
Manual Analysis: All the collected data needs to be analyzed manually making real-time analysis practically impossible. It’s hard to scale manual analysis when you are dealing with hundreds and thousands of machines concurrently.
Up until recently, most industries relied on the walk around programs because continuous monitoring was extremely expensive. But not anymore!
Solution - Predictive Analysis powered by IoT
Image: High-level description of data flow in predictive analysis systems
How does it work?
The entire power plant data is released and collected into Data Centers through sensors employed at various positions. This data is collected by Predictive Maintenance Optimisation System (PMOS) through OPC protocol. The data is collected over the configurable time period.
Once the data is collected, it is checked and filtered through following logical pipes. The filtered data is then used for subject analysis.
- Plausibility Checks
Plausibility check is conducted to replace values which are crossing a min/max range for a particular sensor.
E.g: If SuperHeater spray should reside between 100-300 and the recorded value in the file is not in the identified range, plausibility check would replace the value as per the current plant load from history. It will also check for NaN values.
- Statistical Algorithm Checks
After collecting and calculating the data form a single cycle, the processed data is stored in a local data machine and pushed to cloud storage system. Cloud storage is a model of data storage in which the digital data is stored across logical pools, the physical storage spans multiple servers (and often locations), and the physical environment is typically privately owned and is responsible for keeping the data available & accessible and, the physical environment protected and running.
Now the real-time monitoring and analysis can be done on the processed data. The system can monitor the situation of the power plant production of the whole group and each branch through the production process, trend curve, the argument list, cards and alarms in real time.
Image: The block diagram of Predictive Maintenance Optimisation System
Image: Predictive Maintenance Optimisation System Interface
Predictive Maintenance Optimisation System is a Product that provides you an interface to facilitate features like Property Manager, Create/Map Event, Configuration, Future Behaviour, Unit Manager, Fault Tree Builder, Condition Builder, Data Browsing, Modeller, Live Monitoring to keep System status updated with Alarms, Notifications, and Cards. These listed features individually refer to system modules which are now explored in detail.
Property Manager: Property Manager as part of the tools forms the underlying base of the system. This module avails creation, mapping, and viewing of properties of the components in the system.
Image: Property Manager Interface
Modeller: System administrator can build a Virtual Powerplant set up by placing virtual components, connecting them and configuring their ports. It is then mapped over properties with DCS tags.
Image: Modeller Interface to build virtual plant simulations
Live Monitoring: System administrator can supervise and overview the entire plant in a single screen with real time data display and monitoring.
Image: Real-time data monitoring in PMOS
Card Manager: System administrator can create cards to present data such as timeline to keep an eye out for condition failures and to observe changes in collected and calculated data.
Image: Card Management Interface
Dashboard Manager: The Dashboard is a collection of cards created by the system administrator. With this interface, the system administrator can add cards according to his need & requirements.
Tools: This block consists of important tools like Regression Plotter, Trend Plotter, Fuel Analysis, TreeMap, Data Browsing Alarm Analysis.
Trend plotter: It’s a tool to plot trend of various properties of the plant by calculating and displaying properties and time in a graph format.
Image: Trend Plotter Tool
Regression Plotter: It’s a tool to predict the future trend and plot them based on multiple regression equations generated by the system automatically by finding nearest reference line from all the points.
Image: Regression Plotter Tool
Alarm Analysis:- It’s a tool to show all faulted conditions with severity on a single screen in reference with time.
Image: Alarm Analysis Tool
Neural Networks:- By leveraging deep machine learning (artificial intelligence) to predict the future behavior of power plants, we can predict up to 50 values of any property based on historical data.
Image: Neural Networks powered by Artificial Intelligence
Predictive Maintenance and Optimisation System as a product configures a plant layout, marks data collection points, collects plant data from DCS at regular configurable intervals, and loads the same into the data mart. Values of all specified measured and calculated points are then checked across various warning and alarming indicators. Collected data is then, run through logic pipes that are built on pre-loaded conditions. Violation of one or more conditions produces notifications and alarms. The root cause for conditional violation is notified to the user.
Hence, using our extensive experience and research we were able to build a predictive analysis solution which now supports and powers TUV SUD having a market capitalization of over 8 billion euros.