No abstract

Abstract-In recent years, wireless sensing technologies have provided a much sought-after alternative to expensive cabled monitoring systems. Wireless sensing networks forego the high data transfer rates associated with cabled sensors in exchange for low-cost and low-power communication between a large number of sensing devices, each of which features embedded data processing capabilities. As such, a new paradigm in large-scale data processing has emerged; one where communication bandwidth is somewhat limited but distributed data processing centers are abundant. By taking advantage of this grid of computational resources, data processing tasks once performed independently by a central processing unit can now be parallelized, automated, and carried out within a wireless sensor network. By utilizing the intelligent organization and self-healing properties of many wireless networks, an extremely scalable multiprocessor computational framework can be developed to perform advanced engineering analyses. In this study, a novel parallelization of the simulated annealing stochastic search algorithm is presented and used to update structural models by comparing model predictions to experimental results. The resulting distributed model updating algorithm is validated within a network of wireless sensors by identifying the mass, stiffness, and damping properties of a three-story steel structure subjected to seismic base motion.

Wireless sensing technology has paved the way for the cost-effective deployment of dense networks of sensing transducers within large structural systems. By leveraging the embedded computing power residing within networks of wireless sensors, it has been shown that powerful data analyses can be performed autonomously and in-network, without the need for central data processing. In this study, the power and flexibility of agent-based data processing in the wireless structural monitoring environment is illuminated through the application of market-based techniques to in-network mode shape estimation. Specifically, by drawing on previous wireless sensor work in both decentralized frequency domain decomposition (FDD) and market-based resource allocation, an algorithm derived from free-market principles is developed through which an agent-based wireless sensor network can autonomously and optimally shift emphasis between improving the accuracy of its mode shape calculations and reducing its dependency on any of the traditional limitations of wireless sensor networks: processing time, storage capacity, and power consumption. The developed algorithm is validated by estimating mode shapes using a network of wireless sensors deployed on the mezzanine balcony of Hill Auditorium located at the University of Michigan.

Failure prognostics is the process of predicting the remaining useful life (RUL) of machine components, which is vital for the predictive maintenance of industrial machinery. This paper presents a new deep learning approach for failure prognostics of rolling element bearings based on a Long Short-Term Memory (LSTM) predictor trained simultaneously within a Generative Adversarial Network (GAN) architecture. The LSTM predictor takes the current and past observations of a well-defined health index as an input, uses those to forecast the future degradation trajectory, and then derives the RUL. Our proposed approach has three unique features: (1) Defining the bearing failure threshold by adopting an International Organization of Standardization (ISO) standard, making the approach industryrelevant; (2) Employing a GAN-based data augmentation technique to improve the accuracy and robustness of RUL prediction in cases where the deep learning model has access to only a small amount of training data; (3) Integrating the training process of the LSTM predictor within the GAN architecture. A joint training approach is utilized to ensure that the LSTM predictor model learns both the original and artificially generated data to capture the degradation trajectories. We utilize a publicly available accelerated run-to-failure dataset of rolling element bearings to assess the performance of the proposed approach. Results of a five-fold cross-validation study show that the integration of the LSTM predictor with GAN helps to decrease the average RUL prediction error by 29% over a simple LSTM model without GAN implementation.