AutoML for Feature Selection and Model Tuning Applied to Fault Severity Diagnosis in Spur Gearboxes

Gearboxes are widely used in industrial processes as mechanical power transmission systems. Then, gearbox failures can affect other parts of the system and produce economic loss. The early detection of the possible failure modes and their severity assessment in such devices is an important field of research. Data-driven approaches usually require an exhaustive development of pipelines including models’ parameter optimization and feature selection. This paper takes advantage of the recent Auto Machine Learning (AutoML) tools to propose proper feature and model selection for three failure modes under different severity levels: broken tooth, pitting and crack. The performance of 64 statistical condition indicators (SCI) extracted from vibration signals under the three failure modes were analyzed by two AutoML systems, namely the H2O Driverless AI platform and TPOT, both of which include feature engineering and feature selection mechanisms. In both cases, the systems converged to different types of decision tree methods, with ensembles of XGBoost models preferred by H2O while TPOT generated different types of stacked models. The models produced by both systems achieved very high, and practically equivalent, performances on all problems. Both AutoML systems converged to pipelines that focus on very similar subsets of features across all problems, indicating that several problems in this domain can be solved by a rather small set of 10 common features, with accuracy up to 90%. This latter result is important in the research of useful feature selection for gearbox fault diagnosis.

Mutual Inductance Calculation of Circular Coils Sandwiched between 3-Layer Magnetic Mediums for Wireless Power Transfer Systems

The mutual inductance between coils directly affects many aspects of performance in wireless power transmission systems. Therefore, a reliable calculation method for the mutual inductance between coils is of great significance to the optimal design of transmission coil structures. In this paper, a mutual inductance calculation for circular coils sandwiched between 3-layer magnetic mediums in a wireless power transmission system is proposed. First, the structure of circular coils sandwiched between 3-layer magnetic mediums is presented, and then a mutual inductance model of the circular coils is established. Accordingly, a corresponding magnetic vector potential analysis method is proposed based on Maxwell equations and the Bessel transform. Finally, the mutual inductance calculation method for circular coils between 3-layer magnetic mediums is obtained. The correctness of the proposed mutual inductance calculation method is verified by comparing the calculated, simulated, and measured mutual inductance data.

Nonlinear Metamaterial Lenses for Inductive Power Transmission Systems Using Duffing-Resonator Based Unit Cells

Metamaterials (MTMs) based on a periodic array of resonant coils have been shown to behave as μ-negative (MNG), enabling the focusing of magnetic flux. The phenomenon has been deployed by designers to boost the efficiency of many inductively coupled systems, such as magnetic resonance imaging, underwater and underground communications, and charging base stations (CBS) for consumer electronics and implanted devices. However, due to their dependency on high-Q unit cells, linear MNG-like MTMs have limited bandwidth, restricting their use in many applications, notably in near-field simultaneous wireless information and power transmission (NF-SWIPT) systems. To improve the tight constraints of the amplitude-bandwidth trade-off of artificial magnetic lenses, this paper presents a theoretical analysis of nonlinear MTMs based on a lattice of Duffing resonators (DRs). Additionally, it introduces a criterium for the quantification and evaluation of the amplitude-bandwidth enhancement. The analytical results are based on a circuit model and further verified by numerical simulations using commercial software. The preliminary findings in this paper open up possibilities for nonlinear MTM lenses and can be applied to enhance the linear amplitude-bandwidth limit.

Analysis of the production process of the forked forging used in the excavator drive system in order to improve the currently implemented technology by the use of numerical modeling

The study concerns a comprehensive analysis of a multistage hot-die forging on hammers, in order to produce a yoke-type forging, used as a component of excavator power transmission systems. The investigations were conducted with the aim to analyze and identify the sensitive areas in the process and then improve the currently implemented forging technology by using finite element (FE) simulation. QuantorForm (the developer of the QForm program) has developed a thermomechanical numerical model for the production of forked forging. The software Computer-Aided Three-Dimensional Interactive Application (CATIA) was used to develop and build Computer-Aided Design (CAD) models of forging tools. As a result of the numerical simulations, the plastic deformations and temperature distributions for the forgings and tools were obtained, and the force courses during the forging process were analyzed. The obtained results enabled a thorough analysis of the forging process, including identification of potential forging defects (laps) as well as those tool areas that are the most loaded and exposed to damage. On this basis, changes were implemented in the production process, which allowed for the improvement of the currently implemented technology and obtaining the corrected forgings.

Optimal Shedding Against Voltage Collapse Based on Genetic Algorithm

The prevalent tendency in power transmission systems is to operate closer and closer to the energy limit, rendering system voltage instability a commonly widespread phenomenon. It is, therefore, necessary that certain remedial corrective controls need be undertaken whenever these systems tend towards failure. In this respect, load shedding stands as a major correction mechanism and such a failure can be prevented and nominal system voltage can be resumed. It is worth noting however that load shedding must be implemented very carefully to ensure the satisfaction of both the customer and the electricity-production company. In this context, our focus of interest is laid on load and machine shedding against voltage collapse as an effective corrective method. It is important to note that such a problem turns out to be commonly defined as an optimization problem under constraints. Using genetic algorithms as resolution methods, the application of the proposed methods was implemented on the 14-node IEEE test network, while considering a number of different case studies.