Maintaining engineering expertise in hydro plants

Efficient automatic fault monitoring and system diagnostics for hydropower plants can be achieved through a platform-based approach that integrates knowledge management with predictive maintenance, writes Roberto Piacentini

à‚ 

The control room at Itaipu Dam in Brazil

Credit: GE

In recent years, the importance of asset monitoring systems for power utilities has increased significantly due to the pressing challenges of an aging workforce and, with the advent of the Internet of Things megatrend, an exponential growth of devices connected to the grid.

These two factors have driven power utilities to strive for higher levels of power system reliability and availability through enhanced equipment performance and productivity using predictive maintenance methodologies and the effective projection of asset failures.

Typically, a medium-to-large power utility has more than 100 large machines that require monitoring systems in multiple, geographically dispersed sites covering over 15,000 measurement points. These systems continuously stream data using specialized algorithms that convert the data into information that ultimately results in corrective action.

This represents over 2 GB of acquired and analyzed data per second. Acquiring, analyzing, and managing these massive amounts of data efficiently and in a timely manner is a complex task as machines evolve and require additional measurement points and real-time calculations to determine the actual state of the asset being monitored.

So how do executives and operations & maintenance management staff at power utilities handle the transfer of knowledge? Ultimately, how do they not lose the engineering expertise they face with a retiring workforce?

An asset monitoring system can provide the necessary insight into production system health to help increase production, reduce downtime and achieve higher production output. Development of advanced maintenance strategies such as predictive maintenance involves efficient information and data management methodologies.

Predictive maintenance requires the ability to process a lot of information about sensor types, measurement equipment, measurement points, measurement frequency, real-time calculations for health state estimation, and so on. Additionally, implicit operator/technician knowledge is often necessary to diagnose the operational condition of the equipment, conduct fault monitoring, and predict failure criticality.

With deployments throughout Latin America, M&D Monitoring and Diagnostics in Rio de Janeiro, Brazil has successfully addressed these challenges by developing an advanced asset monitoring solution for hydropower generators and substations. The company created it to considerably reduce the need for human intervention and to offer intelligent decision support with fault monitoring, diagnosis and prognosis capabilities.

Open and flexible platforms

To meet stringent power utility requirements for easy integration, interoperability with existing/legacy systems, and high reliability, the M&D approach uses open and flexible commercial off-the-shelf (COTS) hardware platforms and graphical software tools to alleviate the complexity of developing, deploying and managing monitoring systems.

M&D automatic monitoring systems rely on graphical system design with NI LabVIEW software and CompactRIO FPGA-based hardware to automatically acquire and compute math calculations and analyze data in real time using various predictive techniques. These systems work online or offline and take measurements such as vibration (acceleration and velocity); displacement (radial and axial); pressure pulsation and level; flow; air gap and position; temperature (direct measurement and thermographic analysis); pump intermittence times; electrical (voltage, current, power), and the amount of physical and chemical lubricant.

The M&D solution uses existing SCADA/historian systems to extract recent data from multiple sources and operational parameters and then automatically correlates and analyzes the data. The solution transforms this analysis into graphs, reports and other real-time analysis tools to provide accurate information. These tools can be accessed by email or web from tablets and mobile phones to conduct faster and more convenient decision-making in the field.

Knowledge management

Additionally, this intelligent decision support solution works as the organizational memory of the company by formalizing the expert’s knowledge of maintaining equipment over the years through database storage and management.

M&D’s knowledge management methodology digitally captures domain experts’ day-to-day decisions and feeds them into machine-learning algorithms that automatically diagnose and recommend maintenance actions to the operations team. This approach greatly helps mitigate power utilities’ growing concern that, without an intelligent monitoring solution to capture their domain knowledge, retiring, experienced workers would have no time to train new personnel entering the workforce.

As these workers use the M&D asset monitoring solution, they can review, add or update the machine-learning rules (in other words, the digital representation of actual domain experts’ knowledge) for each asset or piece of equipment using a graphical and intuitive interface that they can access with any web browser. Additionally, the M&D solution facilitates knowledge sharing by creating a formal and digital organization/domain memory backed by a database system that senior domain experts can use as a training tool for novice workers and technicians.

Analysis techniques

At the core of the M&D advanced asset monitoring solution, automated diagnostic functions use fuzzy logic algorithms to determine in real time the equipment’s current state. Based on the knowledge previously captured and digitized from domain experts, these algorithms identify faulty equipment. For analysis and reporting, the solution transforms hundreds of gigabytes of data into a single-page report that may be generated automatically or by user request. The solution uses the following traditional and custom signal processing and analysis techniques:

ࢀ¢ Trend analysis (parameters and faults);

ࢀ¢ Orbit analysis with 3D animation;

ࢀ¢ Comparison analysis in the time and spectral domains;

ࢀ¢ Cascade diagrams;

ࢀ¢ 3D graphics;

ࢀ¢ Temperature profiles;

ࢀ¢ Correlation analysis;

ࢀ¢ Intelligent assistant for balancing;

ࢀ¢ Bearing fault analysis (envelope: more than 8000 pre-recorded bearings).

The M&D solution stores online and offline data in a single database for primary (critical) as well as secondary (balance of plant) equipment. This integrates multiple sources of data and increases the system’s automatic diagnosis capabilities. The online DAQ strategy avoids storing irrelevant or repeated information, which eliminates uncontrolled growth of data over time. Furthermore, using spectral signature comparison of online data and decision algorithms based on characteristics of the equipment operating point reduces the risk of false alarms and nondetection events. Recommended for noncritical equipment, offline DAQ uses mobile data collectors (route-based) with simplified operating characteristics and features for the DAQ process, equipment health condition assessment, and their respective warnings and alarms associated with the parameters monitored.

A medium-to-large utility has over 100 large machines that require monitoring systems

Credit: Alstom

Deployment and benefits

M&D advanced asset monitoring systems for the predictive maintenance of plant equipment have been adopted by power utilities due to their modularity, performance and easy integration with monitoring or supervisory software. The M&D solution based on NI hardware and software platforms monitors 38 per cent of Brazil’s installed hydropower: more than 50 plants; more than 230 hydro generators (over 33 GW of power); more than 34,000 measurement points every minute all over the country; and more than 2 TB of data automatically analyzed daily.

The adoption of a graphical programming language and an FPGA-based hardware architecture has resulted in a complete, sensors-to-enterprise asset monitoring solution with greater reliability, flexibility, and performance as well as simpler maintenance. The key benefits of the M&D asset monitoring system include:

ࢀ¢ Reduced operational costs (automatic predictive maintenance analysis);

ࢀ¢ Improved equipment safety (early fault detection);

ࢀ¢ Correlation of multiple data sources (multiple parameter monitoring);

ࢀ¢ Easier maintenance, updates, and expansion (future-ready modular and distributed architecture);

ࢀ¢ More accurate diagnostics (self-adapting monitoring, high-fidelity hardware);

ࢀ¢ Decision-making support (automatic diagnostics based on the previous inputs of specialists/technicians, knowledge management techniques);

ࢀ¢ Easy access (web and mobile phone, reports by email and SMS);

ࢀ¢ Full configuration, from the sensors to the existing knowledge about fault modes, allowing its adaptation to any type of industrial equipment;

ࢀ¢ Support for various languages (English, Spanish and Portuguese).

An advanced asset monitoring system significantly reduces the need for human intervention

Credit: GE

à‚ 

The continual use of advanced technologies to address common asset maintenance challenges allows organizations to collect more accurate and detailed performance data while applying advanced analytics to improve decision-making, generate competitive leverage, and increase productivity growth for their operations.

Roberto Piacentini is Principal Manager at National Instruments. www.ni.com

No posts to display