We all know that corrosion is inevitable, and costs industry billions of dollars annually in component replacement, downtime, and sometimes serious consequences if not properly diagnosed and remedied. This includes preventive measures, such as special coatings, anodic protection, etc. Equipment deterioration, like corrosion, also is inevitable, and requires continual maintenance to keep it running as long as possible.

The Heat Treat Doctor in this issue notes that maintenance "...is a fact of life for industrial machinery, and is far more complex for heat treating equipment." Proper maintenance maximizes uptime productivity and preventative maintenance results in better equipment reliability. Detailed checklists that cover all the bases are part of a good maintenance program to ensure the equipment is kept at peak operating efficiency. However, check lists must be followed to the letter and appropriate actions taken, and it requires a conscientious effort to stay ahead of the game.

Scientists at CSIRO's Manufacturing & Infrastructure Technology in Australia are addressing this situation with an interesting development aimed at creating thinking, talking machines to take equipment-maintenance and reliability and safety issues to new heights. The system, developed by Dr. John Mo and his team of scientists, is called RDMS (Remote Diagnostic and Maintenance System).

Dr. Mo notes that engineers over the past century have strived to improve their methods of machine operation diagnosis and maintenance using increasingly sophisticated methods of statistical and frequency analysis, which required a detailed understanding of the characteristics of each machine. By comparison, the RDMS's feature-recognition algorithms greatly reduce the need for detailed understanding of machine dynamics, and do not rely on statistical or frequency analysis. With some human input, the RDMS learns from the machine's history and continues to learn as it monitors machine operation. It produces time signatures, generated through complex relationships by the RDMS from multiple data signal streams, to compare with acceptable signatures, and makes decisions for action.

So what kind of actions might the RDMS take? Decisions could range from emergency shutdown to sending a fine-tuning feedback to the machine to bring it back into the normal operating regime. And between the extremes, it could provide early warning of maintenance needs to managers or operators, or indicate that they should modify some operating parameters.

The core of the new development is an algorithm that enables this learning and diagnosis to be performed in a relatively machine-independent manner. Also, its integration with the Internet or Intranet services enables very easy remote management of machines. The RDMS has a scalable global system architecture that enables its application to one machine or thousands of machines located around the world. A Windows-based integrated user interface and RDMS compatibility with most major database systems ease the integration of the RDMS with clients' existing hardware and software.

While equipment deterioration will continue to be a fact of life due to the inherent physical forces working on it continuously, the capability for a machine to advise when it isn't "feeling well" will be a giant step forward in keeping machines running smoothly. IH