UC Berkeley

Embedded condition monitoring sensors that eliminate unanticipated failures of critical or high value equipment improve asset utilization while streamlining maintenance and support operations. General Electric and Rolls Royce use embedded sensors on their jet engines that have eliminated failures and allowed maintenance to be performed on an as-needed basis. As a result, airplane utilization increases while enabling the consolidation of maintenance operations. In more traditional industrial settings, condition monitoring of manufacturing and industrial equipment can quickly impact bottom lines through improved productivity and streamlined operations in ways comparable to what is already being realized in the aviation industry.

Wireless sensor nodes provide embedded sensing with little overhead or infrastructure cost as long as appropriate power sources are available to sustain the node over its target lifetime. Energy is the limiting factor for sensor node lifetimes and data streams. Since most manufacturing and industrial equipment have some associated vibration spectrum when operating, transducing the mechanical energy of vibrations to a small amount of electrical energy electromagnetically was explored as a way of powering condition monitoring sensors.

Large pump motors and a machine tool were surveyed to characterize input vibration accelerations associated with manufacturing and industrial operations. Harvestable acceleration peaks occurred below 120 Hz and had magnitudes near or less than 0.1 g. Metal cutting vibrations were characterized and shown to have vibration frequencies proportional to the number of cutting teeth and the spindle RPM. It was also shown that, the 0.4-1.0 g acceleration impulses associated with the rapid axis motion of a machine tool are harvestable.

Simple magnet and coil as well as coreless electromagnetic architectures were pursued using an overall device size constrain of a cube with 2.5 cm sides. That device size was roughly the same as a c-cell battery that is capable of powering a wireless sensor node for five to ten years. The target power for the harvester designs was the time-averaged powers of hundreds of microwatts to single milliwatts required by commercial wireless sensor nodes. Prototype vibration harvesters based on magnet-coil and voice-coil transducer designs were fabricated and evaluated. Both were able to produce about a milliwatt on a vibration platform for an input acceleration amplitude of 0.1 g at frequencies consistent with those characterized on the pump motors and machine tool. The power densities of the unoptomized proof of concept prototypes were comparable to commercial vibration harvester but at less than one seventh the size.

The voice-coil prototype was installed on several 15-30 kW pump motors running support systems for a microfabrication lab, and unrectified powers of 0.2¬-1.5 mW were harvested. Similarly, 0.8-1.8 mW was harvested from metal cutting vibrations while facemilling cast iron and stainless steel, showing that powers comparable to commercial sense node requirements could be harvested from industrial settings.

Coreless motor architectures proved to be best suited for industrial settings because the unconstrained magnetic flux of simple magnet-coil designs interacted with the iron and steel mounting surfaces commonly found on large machines. Simulated coreless magnetic circuit designs showed that gap magnetic flux densities approaching one tesla could be possible but were not implemented.