Let's assume you've installed condition monitoring sensors and small data gathering devices on your machines in the field, so that now you have hundreds of machines sending in buckets of data to your central server every day. Each machine phones in every day, and sends you a flat file or a SQL-compatible file containing a few thousand bytes of vibration spectral data, ambient temperatures, motor rpm, on/off cycles, pressures, flow, etc. Now what do you do with it?
Each machine's normal operating conditions can be analyzed by SPC or similar software to look for trends. SPC can detect problems before they become problems. It is applicable to monitoring any parameter, such as measuring shaft load and bearing wear. SPC differentiates signals from noise, is more sensitive to equipment and process deterioration than simple inspection systems, and enables a dynamic, intelligent system response.
Consequently, specific problems can be solved, such as a suspected vibration. The more advanced software at the central server can fully analyze spectral data coming from the remote machine and render a diagnosis. Your service engineers can also access a machine's complete history, and run specific analyses to help troubleshoot a user's machine.
Problems in multiple machines also can be analyzed. For example, let's say that motors are failing on a regular basis, but only in certain machines. An SPC analysis can find the problem. The analysis is trivial, the challenge is gathering the right kind of data from the field. If each failure is recorded, the system, using Pareto charts and simple run charts, is able to establish a personality for each machine. To see if any one machine was acting differently than similar machines, we can add multiple machines to the study, and organize the graphical representation to display machine-to-machine differences. Meaningful correlation studies require tracking specific suspect parameters with specific failure modes.
It is possible to define a variety of downtime conditions based on triggers coming from one or more machine inputs. As machines operate, all this information is collected and stored. In time, trends from the database appear that determine the leading causes of downtime, when they occurred, why they occurred, what shifts have the highest downtimes, which machines have the lowest downtimes, and why. This information can then be used to determine patterns of failure and appropriate preventive maintenance procedures.
