Our n-variable Statistical Process Control Tool (nSPCT™) analyzes data sets for anomalies. nSPCT is a multivariate statistical process control software tool used to improve anomaly detection rates, significantly reduce analysis man-hours, reduce system down-time, and increase the cost-effectiveness of complex systems.
nSPCT is an advanced application of multivariate statistical process control theory and incorporates newly developed statistical techniques that increase its effectiveness. nSPCT is based on, but significantly different from, conventional statistical process control techniques due to its high degree of automation.
nSPCT employs advances in multivariate statistical process control theory, including:
- Advanced detection of mean shifts
- Variance shifts
- Distinguishing between mean shifts and variance shifts
- Pattern recognition algorithms
- Rapid decompositions
Prior to the development of nSPCT, applying these techniques required a high-level of statistical knowledge and training. nSPCT removes this requirement by providing comprehensive, automated assumption checks using newly developed, quantifiable metrics. nSPCT provides automated baseline formation and allows for fine tuning the algorithms to meet specific customer needs based on the amount and type of data available. Prompt baseline formation is key to establishing a new online system or rebaselining a system after a major repair or overhaul.
Not only does nSPCT provide better detection rates than competing software, detection rates are provided in an automated fashion. Automating the multivariate statistical analysis process also allows for rapid deployment of new algorithms, which greatly reduces development and implementation costs for the customer.
In addition to providing system-level anomaly detection capabilities, the algorithms are flexible enough to provide alerts for predefined faults that may be common to a system. This greatly increases the diagnostic capability for the end user by automating the early diagnostic steps.
Common Maintenance Challenges
The most common maintenance practices in industry often do not include effective monitoring systems. Even with an active monitoring system installed, frequent and/or catastrophic failures occur and cause significant costs in reactive maintenance and lost production. Typically, monitoring systems are not sensitive enough to detect changes in performance and are frequently complex displays requiring an expert to understand. Some of the more capable systems work only because they are highly tailored to the application, which can be very costly to develop and only work with one system. Many monitoring systems use limits developed for a large population of systems. These population-wide limits mask significant changes in performance for individual systems in the population.
In addition, most monitoring systems are less reliable than the equipment monitored. The result is much time spent diagnosing and repairing monitoring system failures, negatively impacting maintenance costs and revenues.
nSPCT automates the monitoring of complex systems by analyzing each parameter and parameter relationship from individual systems without needing prior knowledge of a system’s expected failure modes. nSPCT detects and isolates individual sensor failures. This eliminates the cost of maintaining redundant sensors and troubleshooting faulty sensors.
nSPCT has been demonstrated on many different systems including jet engines, ground-based gas turbines, diesel engines, natural gas compressors, and wind turbines. It can be applied to any complex data set where the goals are:
- Better measures of system condition
- Improved anomaly detection rates
- Significantly reduced analysis man-hours
- Reduced system down-time
- Reduced operation and maintenance costs
- More effective maintenance practices
- Detection of monitoring system failures
The SwRI-developed nSPCT technology has shown extraordinary performance and has been proven superior to current state-of-the-art technologies in field testing. nSPCT makes excellent use of data currently available without the installation of additional sensors. During field testing, nSPCT detected failures unnoticed by currently in-use detection software and the manual analysis previously required for such complex systems.
nSPCT was designed to allow maintenance personnel, without statistics or engineering backgrounds, to use the tool in an effective manner. The automation of such powerful detection algorithms, combined with the capability to analyze individual system performance, represents a leap in technology surpassing other detection systems.