Polycrystalline Silicon Capacitive MEMS Strain Sensor for Structural Health Monitoring of Wind Turbines
Wind energy is a fast-growing sustainable energy technology and driven by the need for more efficient energy harvesting, size of the wind turbines has increased over the years for both off-shore and land-based installations. Therefore, structural health monitoring and maintenance of such turbine structures have become critical and challenging. In order to keep the number of physical inspections to minimum without increasing the risk of structural failure, a precise and reliable remote monitoring system for damage identification is necessary. Condition-based maintenance which significantly improves safety compared to periodic visual inspections, necessitates a method to determine the condition of machines while in operation and involves the observation of the system by sampling dynamic response measurements from a group of sensors and the analysis of the data to determine the current state of system health. This goal is being pursued in this thesis through the development of reliable sensors, and reliable damage detection algorithms.
Blade strain is the most important quantities to judge the health of wind turbine structure. Sensing high stress fields or early detection of cracks in blades bring safety and saving in rehabilitation costs. Therefore, high performance strain measurement system, consisting of sensors and interface electronics, is highly desirable and the best choice. It has been revealed that the conventional strain gauge techniques exhibit significant errors and uncertainties when applied to composite materials of wind turbine blades. Micro-electro-mechanical system (MEMS) based sensors are very attractive among other sensing techniques owing to high sensitivity, low noise, better scaling characteristics, low cost and higher potential for integration with low power CMOS circuits. MEMS sensors that are fabricated on a chip can be either bonded to the surface of wind turbine blade or embedded into the fiber reinforced composite. Therefore, MEMS technology is selected to fabricate the strain sensor in this work.
Two new sensor structures that can be used for strain measurement are designed. While the proposed sensors focus on high sensitivity, they are based on simple operating principle of comb-drive differential variable capacitances and chevron displacement amplification. Device performances are validated both by analytical solutions and finite element method simulations. The transmission of strain fields in adhesively bonded strain sensors is also studied. In strain sensors that are attached to host structures using adhesive layers such as epoxy, complete strain transfer to the sensor is hindered due to the influence of the adhesive layer on the transfer. An analytical model, validated by finite element method simulation, to provide insight and accurate formulation for strain transfer mechanism for bonded sensors is developed. The model is capable of predicting the strain transmission ratio through a sensor gauge factor, and it clearly establishes the effects of the flexibility, length, and thickness of the adhesive layer and sensor substrate.
Several fabrication steps were required to realize the MEMS capacitive strain sensor in our lab. Polycrystalline silicon is selected as the structural layer and silicon nitride as the sacrificial layer. Polysilicon is deposited using LPCVD and SiN is deposited by PECVD in our lab. A comprehensive material study of silicon nitride and polycrystalline silicon layers is therefore performed. The whole fabrication process involves deposition, etching, and photolithography of five material layers. Although this process is developed to realize the MEMS strain sensors, it is also able to fabricate other designs of surface micromachining structures as well. The fabricated MEMS capacitive strain sensors are tested on a test fixture setup. The measurement setup is created under the probe station by using a cantilever beam fixed on one side and free on other side where a micrometer applies accurate displacement. The displacement creates bending stress on the beam which transfers to the MEMS sensor through the adhesive bond. Measurement results are in a good match with the simulation results.
Finally, a real-time non-destructive health monitoring technique based on multi-sensor data fusion is proposed. The objective is to evaluate the feasibility of the proposed method to identify and localize damages in wind turbine blades. The structural properties of turbine blade before and after damage are investigated and based on the obtained results, it is shown that information from smart sensors, measuring strains and vibrations, distributed over the turbine blades can be used to assist in more accurate damage detection and overall understanding of the health condition of blades. Data fusion technique is proposed to combine the diagnostic tools to improve the detection system with providing a more robust reading and fewer false alarms.
200 University Avenue West
Waterloo, ON N2L 3G1