Precise time series prediction serves as an important role in constructing a digital twin (DT). The various internal and external interferences result in highly nonlinear and stochastic time series. Although artificial neural networks (ANNs) are often used to forecast time series because of their strong self-learning and nonlinear fitting capabilities, it is a challenging and time-consuming task to obtain the optimal ANN architecture. This paper proposes a hybrid time series prediction model based on an ensemble empirical mode decomposition (EEMD), long short-term memory (LSTM) neural networks, and Bayesian optimization (BO). To improve the predictability of stochastic and nonstationary time series, the EEMD method is implemented to decompose the original time series into several components (each component is a single-frequency and stationary signal) and a residual signal. The decomposed signals are used to train the neural networks, in which the hyperparameters are fine-tuned by the BO algorithm. The following time series data are predicted by summating all the predictions of the decomposed signals based on the trained neural networks. To evaluate the performance of the proposed EEMD-BO-LSTM neural networks, this paper conducts two case studies (the wind speed prediction and the wave height prediction) and implements a comprehensive comparison between the proposed method and other approaches including the persistence model, autoregressive integrated moving average (ARIMA) model, LSTM neural networks, BO-LSTM neural networks, and EEMD-LSTM neural networks. The results show an improved prediction accuracy using the proposed method by multiple accuracy metrics.

A miniature microphone array based on interaural time difference (ITD) is designed. This array contains four microphones with certain arrangement and aims for two-dimensional (azimuth and elevation) direction-of-arrival (DOA) estimation in the whole three-dimensional space. The array can be small because it uses a coupling algorithm that magnifies the time delay between the signals received by every two microphones. The coupling algorithm is built according to a pairwise-coupled multidimensional mechanical model inspired by the ears of the tiny parasitoid fly Ormia ochracea. It was verified that the time-delay magnification can be independent of the incident angle when the parameters in the model satisfy specific relationships. This paper further investigates the multidimensional coupled system and advocates to realize the magnification mechanism in algorithm, where the physical parameters can change according to sound frequency to ensure the time-delay magnification. Moreover, the arrangement of microphones is specially designed to help the array to achieve similar measuring accuracy for all directions in the three-dimensional space. Corresponding signal process procedures are also provided. Simulations that use such an array to estimate the azimuth and elevation angles of sound source are performed via general cross-correlation (GCC) method. Results verify the feasibility of the microphone array and show that the accuracy of the estimation increases after the signals are processed by the coupled system.

Precisions of localization are a function of the size of an array. A kind of parasitoid fly, Ormia ochracea, performs an extraordinary directional hearing ability despite its tiny-scaled auditory organ. In this paper, vibration modes and transfer functions of the Ormia ochracea's ear model were calculated, and the phase difference amplification in responses are analyzed to investigate the directional hearing mechanism. A novel three-element bionic model is proposed for spatial sound source localization for small distance-wavelength ratios. The amplification of the phase difference of this model is verified. In order to realize the bionic localization model, based on electric-mechanic analogy method, a system that consists of a triangular acoustic array and a bionic coupling circuit is designed and tested. Frequency responses of the circuit output, as a means of transfer function of the system, are taken into estimation of the source directions. The result has shown that this circuit design has better performance in estimating the direction of sound sources compared to the uncoupled array with same size.

The Ormia ochracea, a species of parasitic fly, has a remarkable localization ability despite the tiny interaural distance compared with the incoming wavelength. The mechanical coupling between its ears enhances the differences of the two received signals, the main cues to locate the source. Inspired by the coupling mechanism, we present a miniature coupled two-microphone array for estimating sound source horizontal bearing. The coupled array consists of a standard two-microphone array and a two-input, two-output filter which implements the coupling. The relationship between filter parameters and time delay magnification is investigated to provide theoretical support for array design. With appropriate parameters, the time delay of received signals can be linearly magnified. Based on the linear magnification, we present a method for estimating source direction using the coupled array. The influence of time delay magnification on time delay estimation accuracy is explored through the general cross-correlation (GCC) method. Experiments are conducted to verify the coupled array and demonstrate its advantages on improving the resolution of estimation of time delay and accuracy of bearing estimation compared with the standard array with the same element spacing.

The reliable prediction and diagnosis of abnormal events provide much needed guidance for risk management. The traditional Bayesian network (traditional BN) has been used to dynamically predict and diagnose abnormal events. However, its inherent limitation caused by discrete categorization of random variables degrades the assessment reliability. This paper applied a continuous Bayesian network (CBN)-based model to reduce the above-mentioned limitation. To compute complex posterior distributions of CBN, the Markov chain Monte Carlo method (MCMC) was used. A case study was conducted to demonstrate the application of CBN, based on which a comparative analysis of the traditional BN and CBN was presented. This work highlights that the use of CBN can overcome the drawbacks of traditional BN to make dynamic prediction and diagnosis analysis more reliable.

The pathology of the human abdominal aortic aneurysm (AAA) and its relationship to the later complication of intraluminal thrombus (ILT) formation remains unclear. The hemodynamics in the diseased abdominal aorta are hypothesized to be a key contributor to the formation and growth of ILT. The objective of this investigation is to establish a reliable 3D flow visualization method with corresponding validation tests with high confidence in order to provide insight into the basic hemodynamic features for a better understanding of hemodynamics in AAA pathology and seek potential treatment for AAA diseases. A stereoscopic particle image velocity (PIV) experiment was conducted using transparent patient-specific experimental AAA models (with and without ILT) at three axial planes. Results show that before ILT formation, a 3D vortex was generated in the AAA phantom. This geometry-related vortex was not observed after the formation of ILT, indicating its possible role in the subsequent appearance of ILT in this patient. It may indicate that a longer residence time of recirculated blood flow in the aortic lumen due to this vortex caused sufficient shear-induced platelet activation to develop ILT and maintain uniform flow conditions. Additionally, two computational fluid dynamics (CFD) modeling codes (Fluent and an in-house cardiovascular CFD code) were compared with the two-dimensional, three-component velocity stereoscopic PIV data. Results showed that correlation coefficients of the out-of-plane velocity data between PIV and both CFD methods are greater than 0.85, demonstrating good quantitative agreement. The stereoscopic PIV study can be utilized as test case templates for ongoing efforts in cardiovascular CFD solver development. Likewise, it is envisaged that the patient-specific data may provide a benchmark for further studying hemodynamics of actual AAA, ILT, and their convolution effects under physiological conditions for clinical applications.