Drift Model Research Articles

High-precision identification and compensation of nonlinear error in hemispherical resonator gyro

This paper presents a novel nonlinear error identification and compensation method for the Hemispherical Resonator Gyro (HRG) from the signal processing perspective. Firstly, the paper establishes an HRG nonlinear error model that reveals the quantitative correlation between angle measurement error and the amplitude-gap ratio. Subsequently, a dynamic analysis model for gyro drift is developed using nonlinear motion equations. Based on this, a scalar factor nonlinear identification model encompassing assembly and nonlinear errors is proposed. Nonlinear optimization is then used to identify error parameters accurately. Finally, a joint compensation method is devised to mitigate assembly and nonlinear error issues. Empirical findings provide concrete validation for the precision of the identification and compensation method. The experimental results demonstrate that reductions of 97%, 89%, and 40% after the joint compensation are achieved in the fourth harmonic component, scale factor nonlinearity, and bias instability of the gyro, respectively.

Analysis of Periodic and Seasonal Characteristics in Time Series: Taking the Dow Jones Industrial Average as an Example

The objective of this research is to conduct a comprehensive analysis of the Dow Jones Industrial Average (DJIA) with a focus on identifying cyclical and seasonal patterns observed in long-term time series data. In order to accurately predict future market trends, this research employs a combination of fundamental and advanced forecasting techniques, including mean, naive, snaive, drift, Holt, Holt-Winters, and ARIMA models. Furthermore, the research explores the impact of return ratios on market investment opportunities. Customized portfolio optimization strategies are developed based on the risk profiles of different investors, along with quantitative trading strategies that effectively manage market risks using the Capital Asset Pricing Model (CAPM). The research shows that the price of Dow Jones Industrial Index has a gradual upward trend, and its return rate has a certain periodicity and seasonality. The aim of this research is to provide valuable insights into the fundamental framework of the American stock market, enabling economists, investors, and policymakers to make informed decisions.

An improved drift model of dropped containers under wind and current effects

In the face of the increasing risk of Maritime accidents caused by global climate change, it is necessary to improve the accuracy of drift model of the dropped containers at seas for Maritime Search and Rescue (SAR). Sensitivity test on the classic drift model illustrates that immersion ratio (Ir) of container plays a dominate role in the drift trajectory of container compared to the drag coefficients. Inspired from it, a new scheme of variable Ir is proposed to improve the model by considering the water entering process of container in the study. Based on the stability analysis of four floating attitudes of intact empty container under strong wind condition, the variable Ir is estimated according to the orifice inflow theory and the Archimedes' principle of buoyancy. The new scheme is compared with the fixed Ir scheme from the simulation of the drift trajectory of the dropped container at Xiamen Bay of China. The result shows that error arisen from the fixed Ir would be accumulated as the time integral, but the new scheme of variable Ir can obviously improve the model accuracy for the long-time prediction of SAR. In addition, the drag coefficients of wind and water in the drift model of the dropped container are suggested to be 0.8 and 1.2, respectively.

Federated variational generative learning for heterogeneous data in distributed environments

Distributedly training models across diverse clients with heterogeneous data samples can significantly impact the convergence of federated learning. Various novel federated learning methods address these challenges but often require significant communication resources and local computational capacity, leading to reduced global inference accuracy in scenarios with imbalanced label data distribution and quantity skew. To tackle these challenges, we propose FedVGL, a Federated Variational Generative Learning method that directly trains a local generative model to learn the distribution of local features and improve global target model inference accuracy during aggregation, particularly under conditions of severe data heterogeneity. FedVGL facilitates distributed learning by sharing generators and latent vectors with the global server, aiding in global target model training from mapping local data distribution to the variational latent space for feature reconstruction. Additionally, FedVGL implements anonymization and encryption techniques to bolster privacy during generative model transmission and aggregation. In comparison to vanilla federated learning, FedVGL minimizes communication overhead, demonstrating superior accuracy even with minimal communication rounds. It effectively mitigates model drift in scenarios with heterogeneous data, delivering improved target model training outcomes. Empirical results establish FedVGL's superiority over baseline federated learning methods under severe label imbalance and data skew condition. In a Label-based Dirichlet Distribution setting with α=0.01 and 10 clients using the MNIST dataset, FedVGL achieved an exceptional accuracy over 97% with the VGG-9 target model.

Allele frequency dynamics under sex-biased demography and sex-specific inheritance in a pedigreed jay population.

Sex-biased demography, including sex-biased survival or migration, can alter allele frequency changes across the genome. In particular, we can expect different patterns of genetic variation on autosomes and sex chromosomes due to sex-specific differences in life histories, as well as differences in effective population size, transmission modes, and the strength and mode of selection. Here, we demonstrate the role that sex differences in life history played in shaping short-term evolutionary dynamics across the genome. We used a 25-year pedigree and genomic dataset from a long-studied population of Florida Scrub-Jays (Aphelocoma coerulescens) to directly characterize the relative roles of sex-biased demography and inheritance in shaping genome-wide allele frequency trajectories. We used gene dropping simulations to estimate individual genetic contributions to future generations and to model drift and immigration on the known pedigree. We quantified differential expected genetic contributions of males and females over time, showing the impact of sex-biased dispersal in a monogamous system. Due to female-biased dispersal, more autosomal variation is introduced by female immigrants. However, due to male-biased transmission, more Z variation is introduced by male immigrants. Finally, we partitioned the proportion of variance in allele frequency change through time due to male and female contributions. Overall, most allele frequency change is due to variance in survival and births. Males and females make similar contributions to autosomal allele frequency change, but males make higher contributions to allele frequency change on the Z chromosome. Our work shows the importance of understanding sex-specific demographic processes in characterizing genome-wide allele frequency change in wild populations.

Single-beam velocimetry with dual frequency comb absorption spectroscopy

Laser absorption Doppler velocimeters use a crossed-beam configuration to cancel errors due to laser frequency drift and absorption model uncertainty. This configuration complicates the spatial interpretation of the measurement since the two beams sample different volumes of gas. Here, we achieve single-beam velocimetry with a portable dual comb spectrometer (DCS) with high frequency accuracy and stability enabled by GPS-referencing, and a new high-temperature water vapor absorption database. We measure the inlet flow in a supersonic ramjet engine and demonstrate single-beam measurements that are on average within 19 m/s of concurrent crossed-beam measurements. We estimate that the DCS and the new database contribute 1.6 and 13 m/s to this difference respectively.

Remaining useful life prognostics during boiling-induced surface degradation

After sustaining long-term nucleate boiling, any surface is prone to changes over time, which signifies an important operational characteristic for reliable thermal management; however, recent research of this topic is extremely scarce. In this work, we first present long-term pool boiling experiments on samples of different surface topography and morphology in de-ionized water and in an aqueous salt solution. We show the changes induced by vigorous bubble nucleation over the course of several hundred hours of operation, discussing the underlying degradation phenomena. Following this, we present an adaptive data-driven prognostic approach, capable of estimating the remaining useful life (RUL) of a boiling surface in real time, based on change of surface temperature. The method consists of (i) a Kalman filter to identify the degradation drift model and (ii) a Monte Carlo simulation to propagate the drift to the terminal threshold, obtaining a distribution of predicted RUL values. Its practical applicability, validated on experimental results, is reinforced by its advantageous features of (i) prognostic performance not depending on the boiling surface or its degradation mechanisms, (ii) low computational demand and (iii) numerous options for individual tuning and extensions.

Ecological inferences on invasive carp survival using hydrodynamics and egg drift models

Bighead carp (Hypophthalmichthys nobilis), silver carp (H. molitrix), black carp (Mylopharyngodon piceus), and grass carp (Ctenopharyngodon idella), are invasive species in North America. However, they hold significant economic importance as food sources in China. The drifting stage of carp eggs has received great attention because egg survival rate is strongly affected by river hydrodynamics. In this study, we explored egg-drift dynamics using computational fluid dynamics (CFD) models to infer potential egg settling zones based on mechanistic criteria from simulated turbulence in the Lower Missouri River. Using an 8-km reach, we simulated flow characteristics with four different discharges, representing 45–3% daily flow exceedance. The CFD results elucidate the highly heterogeneous spatial distribution of flow velocity, flow depth, turbulence kinetic energy (TKE), and the dissipation rate of TKE. The river hydrodynamics were used to determine potential egg settling zones using criteria based on shear velocity, vertical turbulence intensity, and Rouse number. Importantly, we examined the difference between hydrodynamic-inferred settling zones and settling zones predicted using an egg-drift transport model. The results indicate that hydrodynamic inference is useful in determining the ‘potential’ of egg settling, however, egg drifting paths should be taken into account to improve prediction. Our simulation results also indicate that the river turbulence does not surpass the laboratory-identified threshold to pose a threat to carp eggs.

From the sky and on the beaches: complementary tools to evaluate common dolphin bycatch in the Bay of Biscay

Since 1989, multiple stranding events of common dolphins have been regularly recorded along the French Atlantic coast. Examination of the carcasses revealed that most animals presented evidence of bycatch. Using stranding data to infer bycatch levels reveals the highest levels of bycatch to have been recorded since 2016 (4000 to 9000 bycaught individuals). This approach is directly influenced by drift conditions, which can greatly contribute to or hinder our ability to estimate bycatch at sea. In the winter of 2021/2022, the French stranding network recorded an unusually high number of strandings until mid-February and few records in March. Investigation of drift conditions revealed low probability of stranding in March due to constant east-west winds. Reverse drift modelling of carcasses stranded in January and February resulted in an estimate of 3670 (95% CI [2750; 5170]) bycaught common dolphins. Dedicated aerial surveys were conducted in the same area during this period, designed to assess abundance and distribution of marine megafauna in French waters. A high number of carcasses of small Delphininae were observed in March 2021, and the number of carcasses floating at sea could be estimated using conventional distance sampling methodology. In March 2021, mortality at sea was thus estimated at 3250 (95% CI [1288;10198]) common dolphins. The complementary use of both methodologies resulted in an estimate of 6920 (95%CI [4038;15368]) bycaught individuals during winter 2021/2022. This case study highlights that a decrease in strandings does not imply a decrease in mortality at sea. Trends in strandings need to be considered in the light of scientific evidence to avoid delays in decision making.

Performance of GPT-4 on the American College of Radiology In-training Examination: Evaluating Accuracy, Model Drift, and Fine-tuning

Rationale and ObjectivesIn our study we evaluate GPT-4’s performance on the American College of Radiology (ACR) 2022 Diagnostic Radiology In-Training Examination (DXIT). We perform multiple experiments across time points to assess for model drift, as well as after fine-tuning to assess for differences in accuracy. Materials and MethodsQuestions were sequentially input into GPT-4 with a standardized prompt. Each answer was recorded and overall accuracy was calculated, as was logic-adjusted accuracy, and accuracy on image-based questions. This experiment was repeated several months later to assess for model drift, then again after the performance of fine-tuning to assess for changes in GPT’s performance. ResultsGPT-4 achieved 58.5% overall accuracy, lower than the PGY-3 average (61.9%) but higher than the PGY-2 average (52.8%). Adjusted accuracy was 52.8%. GPT-4 showed significantly higher (p = 0.012) confidence for correct answers (87.1%) compared to incorrect (84.0%). Performance on image-based questions was significantly poorer (p < 0.001) at 45.4% compared to text-only questions (80.0%), with adjusted accuracy for image questions of 36.4%. When the questions were repeated, GPT-4 chose a different answer 25.5% of the time and there was no change in accuracy. Fine-tuning did not improve accuracy. ConclusionGPT-4 performed between PGY-2 and PGY-3 levels on the 2022 DXIT, but significantly poorer on image-based questions, and with large variability in answer choices across time points, but without improvement with fine-tuning. This study underscores the potential and risks of using minimally-prompted general AI models in interpreting radiologic images as a diagnostic tool. Implementers of general AI radiology systems should exercise caution given the possibility of spurious yet confident responses.

Dynamic modeling and optimization of mechanical drift for compliant parallel micromanipulation driven by V-shape linear ultrasonic motors

Designing micro-mechanisms that satisfy large strokes, high precision, and multiple degrees of freedom has consistently been a research focus and challenge in the field of micromanipulation systems. Linear ultrasonic motors (LUMs), with advantages such as millimeter-level travel, nanometer-level motion resolution, and power-off self-locking, significantly contribute to the advancement in the mechanism design of micromanipulation systems. However, the challenge of mechanical drift lacks a comprehensive theoretical analysis to explain its cause and limits its application. This paper presents a mechanical drift modeling method for a compliant parallel micromanipulation (CPM) driven by a LUM. The model primarily focuses on the creep characteristics of the flexure clamp element in the LUM. Based on the stiffness matrix analysis method of the compliant parallel mechanism kinematics, the mechanical drift dynamics model of the CPM is established. The influence of the main structural parameters on the CPM mechanical drift is investigated. The research indicates that the creep of flexure clamps in the CPM is the main factor leading to the mechanical drift. Optimization of mechanical drift is achieved by choosing flexure clamps with higher tangential stiffness for LUM and flexure hinge joints with greater flexibility for CPM. Finally, experimental studies reveal that under equivalent conditions, the mechanical drift of the optimized CPM was significantly reduced than before, proving the rationality of the theoretical model and optimization method.

Research on the mechanism and control methods of mechanical drift in linear ultrasonic motors

Linear ultrasonic motors (LUMs) have advantages such as de-energized self-locking and micro-nano displacement resolution. However, their positioning and control accuracy are negatively affected by mechanical drift, which limits their application in ultra-precision fields. To date, the quantitative mechanism of LUM mechanical drift under power-off conditions remains unreported. To solve the problem, we employ the creep theory to identify the clamp stiffness parameters and consider the internal friction and stick-slip effects of the slider, thereby establishing a non-autonomous dynamic model of the LUM mechanical drift in the power-off state. Subsequently, we utilize this model to investigate how the LUM’s structural parameters influence mechanical drift and explore methods to mitigate this undesirable phenomenon. Finally, we validate the model’s validity through experimental research. Our findings reveal that structural creep is the primary cause of mechanical drift in LUMs. Increasing the tangential stiffness of the clamp component and slider internal friction proves to be an effective approach to reducing mechanical drift. This study holds substantial theoretical and practical significance as it deepens understanding of the mechanisms of mechanical drift in LUMs and offers a pathway to achieve effective mechanical drift control.

Monitoring Ongoing Clinical Trials Under Fractional Brownian Motion With Drift

The standard Brownian motion (Bm) with a linear drift is a convenient statistical structure for monitoring ongoing clinical trials in practice for more than four decades (Lan and DeMets). Under this model, the most current one-point statistic is sufficient. However, in our experience, the sponsor and the data monitoring committee often would like to make decision or recommendation based on the “trend” observed from the history of data, not just a one-point snapshot. In this article, we introduce and advance the fractional Brownian motion (fBm) with drift model to formally accommodate this need. The possible dependence and/or the nonlinear trend (e.g., piecewise linear drift with change-point) of observations in clinical trials may come from uncontrollable factors such as patient entry processes may have seasonal patterns over time, patient survival time may depend on the practices of clinical centers, physicians or censoring time (Lai et al.). The violations of the standard Bm and the need for the fBm are discussed with illustrative examples. The common methods including conditional power and sample size re-estimation used for monitoring clinical trials are derived and implemented in the Dynamic Data Monitoring (DDM) system for practitioners under the fBm.

High-purity germanium semiconductor modeling in the detector response function toolkit

We have extended the detector response function toolkit (DRiFT) to provide modeling capabilities of semiconductor sensors. DRiFT provides realistic nuclear instrumentation response by post-processing Monte-Carlo N-particle (MCNP®) radiation transport outputs. MCNP® is capable of modeling radiation transport in complex environments, but has limited detector physics and readout electronics modeling capabilities. Semiconductor detector response can be calculated with a high-fidelity for a flexible range of environments by utilizing MCNP® to simulate radiation interactions inside of detector volumes, and then using DRiFT to model charge transport and signal formation in the semiconductor, as well as the readout electronics. DRiFT models charge transport in the semiconductor, the preamplifier, shaping amplifier, pulse pile-up, and electronic noise to generate detector response. The semiconductor application in DRiFT can model a range of semiconductor materials, shapes, and sizes; and is demonstrated here for a large volume coaxial high-purity germanium (HPGe) detector. Here, we compare detector response functions of a coaxial HPGe detector with measurement of 60Co, 133Ba, and 137Cs at varying count rates, and we conduct a parameter study to demonstrate the effect of changing parameters in the DRiFT simulation. The HPGe detector response function shows excellent agreement with measurements of difference sources with varying dead times and count rates.

High plateau drift characteristics of spin-stabilized projectile from individual recoilless rifle with rifling

Accurately predicting the drift of spinning projectiles in the highland region has always been a challenging task. The study combines genetic algorithm, modified mass ballistic model, and data from ballistic tests to establish a computational model of projectile plateau drift, which is able to compute the value of projectile drift accurately. The reliability of the model was verified by the tests. The study also explains the reasons for the formation of projectile drift and the characteristics of plateau drift through theory. The results of the study proved that under the same range, the projectile's yaw of repose angle is larger in the plateau region, but its projectile plateau drift is smaller mainly because the flight time of the projectile on the highland is smaller than that on the plains. The results of the study are of great significance for the precision strike of projectiles in the highland area.

Modeling and correction of image drift in dynamic shadowgraphy experiments

The study of phoretic transport phenomena under non-stationary conditions presents several challenges, mostly related to the stability of the experimental apparatus. This is particularly true when investigating with optical means the subtle temperature and concentration fluctuations that arise during diffusion processes, superimposed to the macroscopic state of the system. Under these conditions, the tenuous signal from fluctuations is easily altered by the presence of artifacts. Here, we address an experimental issue frequently reported in the investigation by means of dynamic shadowgraphy of the non-equilibrium fluctuations arising in liquid mixtures under non-stationary conditions, such as those arising after the imposition or removal of a thermal stress, where experiments show systematically the presence of a spurious contribution in the reconstructed structure function of the fluctuations, which depends quadratically from the time delay. We clarify the mechanisms responsible for this artifact, showing that it is caused by the imperfect alignment of the sample cell with respect to gravity, which couples the temporal evolution of the concentration profile within the sample with the optical signal collected by the shadowgraph diagnostics. We propose a data analysis protocol that enables disentangling the spurious contributions and the genuine dynamics of the fluctuations, which can be thus reliably reconstructed.Graphic The imposition of a thermal gradient across a liquid mixture results in a time-dependent refractive index distribution. In the presence of a misalignment of the confining cell with respect to gravity, this leads to a deflection of the optical probe beam used to monitor concentration fluctuations within the sample in quantitative shadowgraphy experiments. If not properly accounted for, this effect can introduce a significant bias in the optical signal.

Physics-informed deep learning with Kalman filter mixture for traffic state prediction

Accurate traffic forecasting is crucial for understanding and managing congestion for efficient transportation planning. However, conventional approaches often neglect epistemic uncertainty, which arises from incomplete knowledge across different spatiotemporal scales. This study addresses this challenge by introducing a novel methodology to establish dynamic spatiotemporal correlations that captures the unobserved heterogeneity in travel time through distinct peaks in probability density functions, guided by physics-based principles. We propose an innovative approach to modifying both prediction and correction steps of the Kalman Filter (KF) algorithm by leveraging established spatiotemporal correlations. Central to our approach is the development of a novel deep learning model called the Physics Informed-Graph Convolutional Gated Recurrent Neural Network (PI-GRNN). Functioning as the state-space model within the KF, the PI-GRNN exploits established correlations to construct dynamic adjacency matrices that utilize the inherent structure and relationships within the transportation network to capture sequential patterns and dependencies over time. Furthermore, our methodology integrates insights gained from correlations into the correction step of the KF algorithm that helps in enhancing its correctional capabilities. This integrated approach proves instrumental in alleviating the inherent model drift associated with data-driven methods, as periodic corrections through update step of KF refine the predictions generated by PI-GRNN. To the best of our knowledge, this study represents a pioneering effort in integrating deep learning and KF algorithms in this unique symbiotic manner. Through extensive experimentation with real-world traffic data, we demonstrate the superior performance of our model compared to the benchmark approaches.

Research on void drift between rod bundle subchannels

Void drift between subchannels in a rod bundle is a crucial phenomenon affecting the calculation accuracy of thermal-hydraulic parameters in SMRs. It holds significant importance in enhancing the precision of safety analysis for SMRs. Existing research on experiment and model of void drift between rod bundle subchannels is relatively rare, and the accuracy of model calculations requires improvement. In this study, experiments on gas-liquid two-phase non-equilibrium flow were conducted to measure the redistribution of two-phase flow induced by void drift in a 1 × 2 rod bundle. The experiment results indicated that in bubby flow regime with void fraction less than 0.3, the void diffusion coefficient showed little variation with changes in void fraction. However, in slug flow and annular flow regimes with void fraction exceeding 0.3, the void diffusion coefficient significantly increased with an increase in void fraction. Furthermore, a new void drift model was developed and validated based on a subchannel code. The overall predicted uncertainty for the outlet void fraction in the rod bundle benchmark was less than 13%.

A method for analyzing discontinuities characterization using the integration of underground drift point cloud with 3DEC software

The geometric statistical characteristics of rock mass discontinuities, such as mean trace length and intensity, cannot be measured directly and require the use of statistical tools (such as sampling lines, sampling windows, and drilling) for analysis. In this paper, we propose a method for extracting discontinuity characteristics using the 3DEC software in three-dimensional (3D) drift models (3DM). Specifically, we introduce the Surface Disk Discontinuity Model (SDDM) as a representation model for planar and linear discontinuities identified in point clouds. By importing the 3DM and SDDM into 3DEC and utilizing a multi-circular window, we calculate the mean trace lengths and two-dimensional (2D) intensity (P21). Additionally, we propose a method to estimate the three-dimensional (3D) intensity (P32), which provides a more comprehensive reflection of the discontinuous characteristics in different spatial orientations. To validate the proposed method, we apply it to the underground drift of the Jiapigou gold mine in Jilin Province to extract the geometric statistical characteristics of discontinuities. The results demonstrate that our method successfully captures the orientation, trace length, and P32 characteristics of four dominant discontinuity groups in the study area. This approach proves to be an effective technique for utilizing 3D laser scanning to obtain accurate discontinuity characteristics of rock mass.

Matrix-based predictive model of residual drift and analytical resilience design approach for self-centering columns

Residual displacement serves as a crucial indicator in evaluating the safety and reparability of structures following seismic events. As a novel self-centering structure, the unbonded prestressed reinforced concrete (UBPRC) column has demonstrated its effectiveness in reducing residual displacement. However, there remains a lack of quantitative research concerning the influence of various structural and material parameters on residual displacement. For the quick and convenient evaluation of the seismic performance and resilience of UBPRC columns, in this paper, the influence of various parameters on the residual drift ratio of UBPRC columns was investigated by establishing a logarithmic linear relationship between the parameters and the residual drift ratio. These parameters include the longitudinal reinforcement and concrete strength (fy and fc), reinforcement ratio of longitudinal and transverse (ρl and ρs), prestressed strand ratio (ρp), aspect ratio (Ar), axial load ratio (αc), and prestressing force ratio (αps). Among the two parameters related to prestressed strands, increasing the prestressing force ratio significantly reduces RDr, while the influence of the prestressed strand ratio on RDr is relatively small. On this basis, a matrix-based probabilistic predictive model for the residual drift ratio was established. According to various physical parameters, the proposed mathematical model accurately predicts the residual drift ratio and describes the influence of each parameter on it. Furthermore, an analytical resilience design approach was proposed, and an analytical expression in matrix form for a multi-parameter resilience design was derived. Lastly, the impact of parameters on the damage state was investigated, and the range of parameter values under investigation was established, according to the analytical resilience design approach.