Manufacturers with equipment in use worldwide commonly grapple with the challenges of remote equipment maintenance. Every piece of expensive capital equipment in the field requires, at the very least, routine servicing. The maintenance—and particularly the repair—of equipment in remote locations presents extraordinary challenges for manufacturing professionals.

OEMs require large staffs of field service engineers (FSEs). Determining how many are needed, training them to use the latest software maintenance tools, and keeping them up to speed on equipment in the field and what can go wrong with it, represents a significant expense. Containing those expenses and managing the field service and maintenance function for optimal productivity requires the effective collection and use of data from and about equipment in the field.

Software tools

Traditionally, this meant gathering and retaining as much fault information as possible. In 1993, software tools that redefined how data are gathered and, more important, how they are used to help FSEs do their jobs, became available.

Expert systems for maintenance support have generally relied on case-based reasoning (CBR). Without rigorous attention to data collection, CBR cannot work. When a machine breaks down, FSEs relying on CBR are obliged to access a vast, encyclopedic database of past problems in order to find a specific symptom. When he finds the symptom, he hopes that the accompanying repair strategy, recorded by other technicians who’ve worked on problems in the past, is correct for the machine he is trying to fix.

There is nothing targeted about the CBR approach. Rather, it is comparable to using an Internet search engine and getting tens of thousands of hits when searching for something very specific. The one answer may be among all those hits, but there is still the unenviable task of isolating it.

While the FSE may find the correct solution right away, he is just as likely to spend a lot of time—possibly hours or days—pouring over records of past problems in an effort to find the solution. At the very least, solving the problem is likely to be time-consuming and labor-intensive.

And it can be much worse. Suppose the problem with the machine has never happened before. Then, it can take a lot of troubleshooting before the technician eventually ferrets out the problem and the solution. In the process, expensive components—many of which have nothing wrong with them—might be replaced in a trial-and-error approach. All the while, the machine is down and costing its owner money.

Model-based reasoning

During the 1990s, a model-based reasoning (MBR) methodology was developed. MBR software relies on built-in knowledge-based models of the machines being serviced and algorithm-driven systems to discern what is wrong and help FSEs quickly fix the problems. Rather than being trial-and-error, the MBR software is analytical, rapidly considering options based on FSE input or, in many instances, information gathered directly from the machine. It also automates the data collection process and streamlines it by gathering only the information that is pertinent to diagnostics and maintenance, thus reducing the effort and storage required to manage such data. The Web-enabled client-broker architecture of the solution automates the collection of health-monitoring and troubleshooting data from equipment in remote locations to a centrally maintained knowledge database which can be used to refine and improve the equipment models used by the reasoning software.

MBR tools facilitate ease of maintenance and diagnostic performance simultaneously, making it possible for system designers to assess equipment performance during the design phase. The key to their utility is a modeling approach that employs multi-signal flow graphs for simple, efficient modeling representations. These models capture the structural and functional relationships of any system, and it is those relationships that are at the heart of the tools’ diagnostic capabilities.

Evolving rapidly over the past decade, MBR tools have been enhanced to include real-time monitoring and diagnosis in the presence of imperfect tests and temporary failures; ranking and selection of available tests enabling quick and accurate fault isolation; and real-time diagnostics that can be embedded in existing computing resources. This makes it possible to monitor equipment in use—including airplane engines—and diagnose problems or potential on-board problems under actual operational conditions.

While software-enabling remote diagnostics were originally developed principally for military and aerospace applications, many commercial companies now take advantage of the myriad benefits these tools offer. With expensive equipment deployed all over the globe, more and more OEMs are discovering that MBR tools can improve efficiency of data collection and management while dramatically reducing the expenses associated with training field service engineers, having equipment off-line for repair and maintaining huge parts inventories in many remote locations.

Maintenance knowledge bases

MBR tools look to evolve and become even more sophisticated in the future. New programs, currently being developed for military clients, will have the capability of selectively expanding their own knowledge. Able to adapt for multiple environments where military aircraft may be deployed and anticipate the variable impacts of those environments, self-evolving maintenance knowledge bases are being developed with the goal of automatically improving the accuracy of the models that underlie troubleshooting, thereby significantly enhancing the quality and speed of troubleshooting.

The models for the same system are no longer one-size-fits-all. They are location-specific, usage-specific and specific to the expertise of the FSE. Self-evolving maintenance knowledge bases in the future will have built-in learning capability to update the knowledge models, using collected data to reduce the uncertainty inherent in empirical knowledge; to adapt to changes in equipment design, behavior and operational environment; and to discover previously unknown facts about the system.

Starting with large initial databases of maintenance equipment, they will then learn and become smarter as they address problems. They are expected to quickly identify problems and solutions for those problems. But as they gather and analyze data from the field, they should also become prognosticators, anticipating problems before they occur, so that maintenance will become, increasingly, just in time. That, in turn, will address inventory issues along with many of the other problems that have always made remote diagnosis and maintenance a significant expense. Q

Sudipto Ghoshal is a principal investigator for Qualtech Systems Inc. (Wethersfield, CT). For more information, call (860) 257-8014 or visit www.teamqsi.com.

Quality Tech tips

• The repair of equipment in remote locations presents extraordinary challenges for manufacturing

professionals.

• Model-based reasoning (MRB) software relies on built-in models of the machines being serviced and algorithm-driven systems to discern what is wrong.

• MBR software is analytical, and it considers options based on technician input or information gathered directly from the machine.

• Self-evolving maintenance knowledge bases are diagnostic software programs that will effectively “teach” themselves, acquiring data and building databases that expand and become increasingly sophisticated over time.