| dc.description.abstract | Electric motor-driven systems are basic components in most industrial processes. The
four motor mechanical measurements of interest that dictate motor performance are shaft torque,
shaft speed, mechanical power (motor load) and energy conversion efficiency. Torque and speed,
along with the input electric power are needed to estimate load and efficiency. Mechanical
performance monitoring is an ever-increasing trend, present in many industries. Such monitoring
can identify equipment failures, predict system degradation, and monitor overload conditions. It
also provides users with a deeper understanding of the operational demands of their machines,
which is difficult to diagnose through standard vibration analysis. Nevertheless, it is extremely
challenging to measure the in-situ motor shaft torque, and to a lesser extent shaft speed.
A comprehensive literature survey indicates that many estimation methods exist for
motor mechanical measurements. However, almost all of these methods require at least one of
the following: (1) load tests involving measurements of shaft speed and torque at a stable
temperature, (2) no-load tests, i.e. with motor mechanically decoupled from driven load, (3) deenergized
stator resistance measurements, i.e. with motor electrically disconnected. These three
conditions require a level of access to the motor under test not generally acceptable or possible in
an industrial environment, i.e. these are the “intrusive” requirements of the mechanical
measurement estimation methods.
A non-intrusive approach to mechanical measurements is proposed, incorporating the
equivalent circuit model of an operating induction motor. The estimation of the mechanical
measurements is formulated as a non-linear, constrained optimization problem with the variables
to be optimized being the parameters of the equivalent circuit model. The optimization problem
iii
is solved using the Covariance Matrix Adaptation Evolution Strategy (CMA-ES) method. The
required inputs of the proposed method are the three-phase voltage and current phasors and the
motor nameplate information. The resulting parameters of the equivalent circuit model and new
voltage and current phasors are used to estimate the motor mechanical measurements.
Five (5) manufacturer's catalog data sheets, two (2) small motors, 1 HP and 3 HP, and
two (2) larger motors, 100 HP and 200 HP, are used to test the proposed approach against direct
mechanical measurements of torque and speed. The experimental results demonstrate that the
proposed method achieves speed estimates within ±5 RPM of the sensor readings, which is
comparable to existing non-intrusive methods. For torque, load and efficiency, the proposed
method achieves accuracy within ±2-3% for high (>25%) loads, and up to ±5% errors for low
(<25%) loads. The presented shaft torque accuracy is an improvement over existing, nonintrusive
techniques and in the case of load and efficiency estimation is an improvement over
existing, non-intrusive and even some intrusive techniques.
The novel contribution of this research is the estimation of mechanical measurements
based only on motor electrical measurements and motor nameplate information. The need for a
torque transducer, speed sensor and any motor field testing is eliminated. Further testing is
needed to establish the accuracy of the method on a wider motor population. | en |
