This dissertation is an account of present and past research and development (R&D) efforts conducted by the author to develop and implement new technology for predictive maintenance and equipment condition monitoring in industrial processes. In particular, this dissertation presents the design of an integrated condition-monitoring system that incorporates the results of three current R&D projects with a combined funding of $2.8 million awarded to the author by the U.S. Department of Energy (DOE). This system will improve the state of the art in equipment condition monitoring and has applications in numerous industries including chemical and petrochemical plants, aviation and aerospace, electric power production and distribution, and a variety of manufacturing processes.
The work that is presented in this dissertation is unique in that it introduces a new class of condition-monitoring methods that depend predominantly on the normal output of existing process sensors. It also describes current R&D efforts to develop data acquisition systems and data analysis algorithms and software packages that use the output of these sensors to determine the condition and health of industrial processes and their equipment. For example, the output of a pressure sensor in an operating plant can be used not only to indicate the pressure, but also to verify the calibration and response time of the sensor itself and identify anomalies in the process such as blockages, voids, and leaks that can interfere with accurate measurement of process parameters or disturb the plant's operation, safety, or reliability.
Today, process data are typically collected at a rate of one sample per second (1 Hz) or slower. If this sampling rate is increased to 100 samples per second or higher, much more information can be extracted from the normal output of a process sensor and then used for condition monitoring, equipment performance measurements, and predictive maintenance. A fast analog-to-digital (A/D) converter can bring this about when it is combined with a large data storage unit to save the massive volume of data that will result from fast data sampling.
Due to recent advances in electronics and computer technologies, fast A/Ds and large data storage units are now readily available and very affordable. Furthermore, advanced signal processing and interpretation techniques are readily available in commercial packages that provide fast Fourier transform (FFT) and wavelet analysis, correlation and cross-correlation results, application-specific neural networks, fuzzy data clustering, and other techniques to help arrive at test results quickly. When existing process sensors are not available to provide the necessary data, wireless sensors can be deployed to fill the gap. It is understood that wireless sensors are still evolving, but an assessment of these sensors performed under one of the R&D projects described in this dissertation shows that they are ready to play a positive role in equipment and process condition monitoring in industrial installations.
Another class of predictive maintenance and condition-monitoring technologies now available is called by a number of names, such as "non-destructive examination," "non-destructive testing," "non-destructive inspection," or "in-service inspection" methods. These methods are not described in this dissertation as they are a separate discipline of their own and require a different set of skills to be implemented in an industrial process. They are nevertheless mentioned here to acknowledge their availability for predictive maintenance and to recognize their prominence as an important class of condition-monitoring techniques. These methods are used for detecting defects such as cracks, corrosion, and wear in metals, plastics, composites, ceramics, and other material except for wood and paper products. Some of these techniques are also used in medical diagnostics.