Motion Control Resources
- This application case study is filed under:
- Automated Inspection
- Electrical Engineering
- Fabricated Metal
- Material Handling
- Medical Devices
- Metal Working
- Plant Engineering
- Solar Power
- Water Purification & Sanitation
- Electric Power Generation
- Electronic Measurement
- Factory Automation
- Off Road/Heavy Equipment
- Oil and Gas
- Paper and Pulp
- Electronic Assembly
Developing an Out-of-the-Box Novelty-Detection Solution Using NI LabVIEW
National Instruments Posted 11/05/2008
Cova González-Quirós - Alstom Aerospace
V.C. Patel - Alstom Aerospace
P.H. Knight - Alstom Aerospace
Developing a package that engineers can easily configure and use in applications where they are monitoring and rapidly processing large amounts of data to highlight new or unusual events that may signify a problem.
Using NI LabVIEW to develop Novus, a data-driven novelty-detection package that combines innovative data processing and statistical techniques in an intuitive engineer-friendly environment.
"Alstom Aerospace now has a powerful, user-friendly condition monitoring package based on LabVIEW software that helps customers identify impending problems sooner. "
At Alstom Aerospace, we developed the new Novus novelty-detection solution using LabVIEW to help engineers conduct condition monitoring and keep processes and systems running while identifying potential problems. Engineers can use the l wherever they need to monitor and process large amounts of data to highlight new or unusual events and schedule repair and preventive maintenance accordingly. Data-driven and easily configurable for different applications, Novus offers powerful, engineer-friendly features that users can apply systematically to stay informed and in charge.
Why Choose LabVIEW?
We developed the first version of Novus, including all the algorithms and the graphical user interface (GUI), using The MathWorks, Inc. MATLAB® software. Although this version did detect novelties, it was not user-friendly and required the engineer setting up the tool to possess extensive process knowledge and an in-depth technical understanding of the algorithms used. Adding new functionality or fixing bugs in this version was very time-consuming and complex because the engineer had to make explicit links between the functions and the inputs and outputs on the screens.
For the new version of the software, we considered several options including a code rewrite using MATLAB software, LabVIEW, C++, and Visual Basic. We chose LabVIEW for the following reasons:
- GUIs with the latest Windows look and feel that is familiar to users
- Automatic linking of objects in the GUIs to code
- Quick coding
- Many complex algorithms created using MATLAB software that can be reused in LabVIEW in the form of .m file scripts, which saves development time because the algorithms have already been validated
- Easier implementation in future releases of new functionality
- LabVIEW support for reading different data file formats
- Reports that can easily be generated with built-in LabVIEW reporting functions
Developing a New Version of Novus Using LabVIEW
When developing the commercial version of Novus, we focused on the ease of use for the engineer preparing the tool. We realized that the engineer must be able to set up the solution without detailed knowledge of the underlying algorithms. Therefore, we developed a workflow that leads the engineer through an intuitive step-by-step procedure for preprocessing and training to establish benchmark behavior for a plant.
The first phase of the workflow is data acquisition. Novus includes a simple configuration wizard, designed with LabVIEW GUIs, that accepts logged data in a variety of formats. We extensively used LabVIEW functionality, such as the VIs within the statistics toolbox, to create the powerful preprocessing tools. These tools allow the data to be interrogated, filtered, and refined to precisely capture the characteristics of normal behavior and optimize the ability of Novus to discriminate abnormalities. Finally, the workflow ends with the training and validation phase. We converted our MATLAB code, which was already optimized and validated, to .m file scripts for the core algorithm and embedded it within the LabVIEW code by using LabVIEW MathScript blocks.
During the training phase, Novus is programmed to recognize the identified normal behavior and embody it in a unique data model. Comparing the model against unseen normal data allows it to be tuned and validated. During validation, engineers can extend the model, if necessary, to encompass a wider description of normality.
Once a data model has been established, the Novus workflow takes the engineer though a logical testing and analysis phase, where any novelty in data is uncovered. Again, we heavily used the LabVIEW functionality to create easily understood screens that can help engineers determine the exact variables responsible for any identified novelties. Engineers use LabVIEW reporting features, as well as the capability to generate Web page-compatible reports, to analyze the data.
Novus for Industrial Gas Turbine Condition Monitoring
Novus has been used on an industrial gas turbine to detect combustion problems early. The exhaust gas temperature (EGT) is used in the engine control loops to limit fuel flow and prevent overheating. The EGT measured at the combustor locations around the engine lends itself to interpretation as a polar plot pattern.
Conventional gas turbine monitoring systems set a hard temperature limit that encompasses a normal working range, and they usually shut the engine down if this range is exceeded. While there is a natural spread of EGT around the engine, this limit must be based on a suitable margin above the highest allowable EGT. This means that subtle changes in EGT, due to events such as a combustor running cooler or hotter than usual, are impossible to detect until a gross error occurs.
Novus analyzes the EGT measurements in a multivariate manner; therefore, it is able to impose tighter threshold limits based on normal behavior. These tighter limits allow Novus to discriminate abnormalities before they become significant. Unlike conventional monitoring systems, Novus is able to identify the onset of faults even when all variables are within the normal working range.
Benefits of Developing a Solution Using LabVIEW
The original Novus tool developed using MATLAB software has been successfully implemented within the LabVIEW platform. We chose LabVIEW over other programming languages because it helped us decrease development time, develop familiar Windows-style GUIs, use LabVIEW MathScript functions to embed our .m file scripts in the solution, and adapt system functionality for future releases.
Alstom Aerospace now has a powerful, user-friendly condition monitoring package based on LabVIEW software that helps customers identify impending problems sooner.
MATLAB® is a registered trademark of The MathWorks, Inc.
For more information on this Case Study, contact:
Whetstone Leicester LE8 6LH
Tel: +44 (0)116 284 5786