Time-optimal rest-to-rest motions employing frequency domain methodologies

Subtitle:contributions towards most time-efficient sensorless stepper motor control

To adapt to the continuously changing market and compete economically, the manufacturing industry requires (production) machines that can be flexibly deployed, operate cost-efficiently through optimised energy consumption or maximised production speed, and are robust enough to withstand changing operating conditions. All these advancements fit a tendency, also referred to as industry 4.0, to enhance and streamline the entire manufacturing process to further optimise the value chain. Simultaneously, in manufacturing machine construction, to simplify control at a system level, machine builders typically utilise a combination of single-axis driven mechanisms, e.g. multiple-rod mechanisms, to compose the machine’s more complex (linear) positioning tasks. Nevertheless, as a consequence, the mechanical properties of the driven motions, e.g., the machine inertia, often vary as a function of the motor position, eventually complicating control at the motor control level. Also, machine builders often opt for stepper motors in positioning applications, e.g., a repetitively moving machine between points A and B, because of their cost-effectiveness and the relatively straightforward open-loop motor control principle. Nonetheless, with the current state-of-the-art, combining these varying mechanical motion properties with the open-loop stepper motor nature is challenging regarding production efficiency, robustness/reliability and sustainability. Therefore, to this end, by employing filtering methods in the frequency domain, this doctoral thesis seeks to contribute towards a sensorless method enabling robust repetitive time-optimal machine motions using stepper motors. First, based on the load angle (a measure of motor efficiency), given a system with varying mechanical properties, a stepper motor drive algorithm is developed, generating motor rotation commands that result in a maximised generated motor torque and, eventually, time-optimal motions. Validation in simulation shows that the generated commands result in a motor torque profile shaped as a one-period asymmetric square wave (also referred to as a bang-bang profile), which is theoretically linked to time optimality. Moreover, lacking a simple method to determine the bang-bang switching point between maximum positive and maximum negative ingoing machine motor torque, a method in the angular domain is designed to calculate this point, solely relying on the position-dependent machine properties. Since small yet difficult-to-encounter influences such as machine temperature depending friction are not considered, a self-learning part is added to the controller to optimise the bang-bang torque profile further and thus guarantee time-optimality. Next, since accurate knowledge of the machine’s position dependant mechanical properties is indispensable for the conceived time-optimal bang-bang angular control strategy, in the third stage, an estimator is developed to directly estimate the inertia profile and the related load torque profile. Fundamentally, the estimator uses a frequency-specific system gain model combined with some SDFT filters. Employing this estimator, the time-optimal bang-bang switching point can be calculated based on estimated position-dependent machine characteristics. Ultimately, this allows a fast establishment of a time-optimal motion without prior application knowledge. Finally, to improve the synchronous machines' sensorless load angle controlled nature, a new load angle estimator with enhanced estimation dynamics is being developed based on an angular sampling strategy and the configurable generalised GDFT. The improved estimator convergence under varying machine loads is an important step towards future sensorless stepper motor-driven time-optimal motions.

Publication year:2023

Keywords:Doctoral thesis

Accessibility:Embargoed