Wising up to get Smart30 April 2001
The Nuclear Energy Research Initiative’s (NERI’s) Smart programme is generating much interest with component manufacturers, specifically for pumps, and with nuclear industry organisations.
The NERI programme, “Smart equipment and systems to improve reliability and safety in future nuclear power plant operations”, has reached its halfway point. The 36-month collaboration brings together the technical capabilities of Westinghouse, Sandia National Laboratories, Duke Engineering and Services, Massachusetts Institute of Technology and Pennsylvania State University. Its purpose is to design, develop and evaluate an integrated set of tools and methodologies that can improve the reliability and safety of advanced power plants.
Smart equipment embodies elemental components such as sensors, data transmission devices, computer hardware and software and human-machine interface (HMI) devices that continuously monitor and predict the failure modes and remaining useful life of the equipment. The programme will be useful to all reactor technologies, including possible future technologies such as proliferation-resistant or low-output reactors.
Smart predictive maintenance technology will significantly reduce costs associated with design, unit unavailability and maintenance. By combining plant maintenance information with real-time sensor data, the programme has the potential to substantially change the way that nuclear plants are designed and operated. Plant design is often constrained by the need for frequent access to equipment for inspection and repair. Furthermore, redundancy and diversity of equipment are required to ensure safety and reliability under a variety of conditions.
Along with the NERI Risk-Informed programme results that move to a risk-based regulatory approach, the introduction of Smart equipment and systems will allow plant designers to simplify designs without compromising reliability and safety. For example, normal operating systems employing Smart components may supplement, or even replace, traditional safety systems such as emergency core cooling or emergency feedwater. The Smart features of the components may provide the basis for assuring that a non-safety system’s availability is sufficient to meet probabilistic risk assessment (PRA) goals and the regulators’ demands.
Practices that heavily rely on time consuming procedures are a major contributor to high operations and maintenance (O&M) costs. These include periodic overhaul or replacement of parts based primarily on maintenance records, without regard to the actual ‘health’ of those components or systems. The Smart programme results are providing a blueprint for creating the capability to predict their time to failure. This requires understanding how, given specific operating conditions, an entire history or profile of sensor information relates to the component’s or system’s wear and age. Such practices allow overhaul and repair to be performed only when necessary, and provide a capability for assessing the risk of delaying indicated tasks. An additional O&M benefit results from reduced surveillance testing requirements due to the overall reduction of the plant safety equipment.
System evaluation and prioritisation study
The initial task of the Smart programme has been to:
•Develop a methodology for evaluating plant structures, systems and components (SSCs) to identify those that would benefit most from application of Smart equipment concepts.
•Select a demonstration component.
•Determine an optimum health monitoring plan for the selected component, including identification of its failure modes.
The main selection criteria include:
•A high failure rate.
•Well-known failure modes.
•Availability of accessible locations allowing sensor installation and data acquisition.
•Sufficiently long repair time to cause significant lost generation.
A study of failure rates and modes analysed data of SSC contirbutions to forced outages. This study used the NRC MORP 2 database for monthly reports between 1990 and 1999 for 14 PWR and 13 BWR units. SSCs were ranked on their fraction of the total forced outage time (based on occurrence frequency and mean outage duration). Individual failure modes were similarly ranked for the SSCs with the highest forced outage contributions. The most significant result was the identification of rotating machinery, including pumps, as the primary contributors to forced outages in LWRs. Along with their application in both charging and feedwater systems, this result led to the selection of a high energy, horizontal, centrifugal pump as the demonstration component for the project.
The other aspect of this initial task explored the industry’s transition from traditional time-based and corrective maintenance methods to reliability centered maintenance (RCM), including application of condition based maintenance (CBM). Methods for monitoring component health being developed in the Smart programme directly support the transition to CBM.
Current pump diagnostics are often limited to characterising casing vibration with portable sensors. Integration of advanced diagnostic methods, including vibration analysis, rotor dynamics modeling, infrared thermography, motor monitoring, lubrication assessment, acoustic monitoring and performance parameter measurement, will be critical to developing an optimum health monitoring system (HMS) for a pump.
Sensor technology and installation analyses
Criteria for sensor selection have been developed for both sensor elements and sensor system architectures. For sensor elements these include:
•The ability to indicate component state based on either the physics of failure mechanisms or a failure effects and modes analysis.
•The ability to withstand temperature or radiation effects of the local environment.
•Accuracy and reliability.
For a sensor system architecture:
•A web-based design including compatibility with the IEEE 1451 standard.
•A wireless data communications network.
Wireless data communications have the potential to minimise concerns associated with installation feasibility and the cost of wired communication networks. Based on current industry direction, it is recommended that Smart equipment networks be compatible with the Bluetooth wireless protocol, which is emerging as an industrial standard.
To support development of Smart equipment, the identified failure modes were addressed using rotor/bearing dynamics modeling. This is resulting in recommended enhancements in sensor placement and sensor development. Additionally, dynamic modeling is being calibrated with pump operating data to provide an array of “virtual” sensors that can aggressively assess the condition of equipment and supply input data to the HMS Bayesian belief networks (BBNs).
The HMI technology assessment investigated Smart equipment applications in other industries for potential use in nuclear plants. Various techniques for presentation of Smart equipment and predictive maintenance information were identified.
For the second project year the sensor technology and installation task is focusing on evaluating advanced sensor technology for applications supporting Smart equipment use in nuclear plants. A methodology for performing sensor installation feasibility studies will be developed and applied to the horizontal, centrifugal pump. Future HMI work will concentrate on developing a Smart equipment display set and display features. The final result will be a prototype display set for the pump demonstration facility.
Equipment maintenance and reliability simulation
The third task is to develop a virtual machine for the centrifugal pump with the capability to simulate equipment behaviour such as failures, maintenance and user-defined sensor signals. The virtual machine will support design and testing of the HMS, allow evaluation of the benefits of incorporating Smart features and provide a platform for realistic demonstrations.
As shown in the Figure above, the virtual machine will consist of three primary components: a reliability module, a scheduling module and a simulation engine. The reliability model will identify failure modes and their relationships including maintenance impact and effects of ageing, based on historical data supplemented with engineering judgement. The scheduling module will define schedules for equipment use and maintenance. The simulation engine will generate the components behaviour based on inputs from the scheduling module and reliability model and will provide the behaviour as input to the computerised maintenance management system (CMMS) and the HMS software.
Smart equipment health monitoring system
Equipment health can include information about predicted lifetime of the equipment, estimated percentage wear out on various components, recommendations for preventative maintenance activites, predictions of likely failure modes and causes and cost impact of maintenance-related decisions.
Fault trees have been constructed providing a functional decomposition of the centrifugal pump. Starting at the highest level of “pump failure” the fault trees break down pump subsystems until individual cause-consequence branches are identified.
BBNs are selected as the engine required to capture the expertise relating sensor data to system states through the use of conditional probabilities. The BBN approach was selected because:
•It has been shown to work better than rule-based and neural network systems.
•It is flexible and tolerant of complexity.
•It is available on personal computers with a convenient user interface.
The HUGIN BBN shell has been selected for use on the project, and an initial canned demonstration of its application has been completed. Population of the conditional probabilities based on input from pump and maintenance experts is underway.
Sample application of health monitoring system
The Smart programme has selected a high energy, horizontal, centrifugal pump as demonstration component. This pump is used in both charging and feedwater systems for PWRs and was selected based on the criteria established in the first task.
Instrumentation and testing of an actual subsystem typical of the selected centrifugal pump will be carried out on a pump lube oil test system at Pennsylvania State University. The virtual machine will simulate the remainder of the pump to allow testing of a HMS for the entire component. The goal is to make this demonstration system a web interface to allow testing and demonstration of the HMS at a variety of locations.
The HMS demonstration will help develop the methodology for systematically evaluating equipment to determine how best to improve its reliability. It will also provide an opportunity to evaluate and optimise Smart equipment and predictive maintenance strategies and support the HMI validation.
Enterprise level health monitoring
This final task will develop a methodology that combines equipment health information from individual components into overall plant health information. It will expand the health monitoring concept to system and plant levels, allowing communication and integration of data among the Smart equipment, as well as control room systems and plant operators. An advanced information system architecture will be designed to support data transfer and storage at the enterprise scale. The system will be designed to:
•Provide data and configuration information required for interpreting and displaying real-time sensor and health data at the component system and plant levels.
•Provide historical performance and maintenance data required for analysing reliability, spares and maintenance conditions.
•Store component, system and plant configuration models and simulation data.
•Support data requirements of selected reliability and maintenance analysis techniques.
This task is not scheduled until the third project year, but component level work is being carried out.