Stochastic sensitivity analysis method for neural network learning

This paper presents a basic introduction to some popular stochastic analysis methods from an unbiased disciplinary perspective. Examples ranging from fields as diverse as defense analysis, cognitive science, and instruction are illustrated throughout to demonstrate the variety of applications that benefit from such stochastic analysis methods and models. Two applications of longitudinal stochastic analysis methods to collaborative and cognitive training environments are discussed in detail. The first application applies a combination of latent mixed Markov modeling and multidimensional scaling for modeling, analyzing, and supporting the process of online knowledge sharing. In the second application, a combination of iterative nonlinear machine learning algorithms is applied to identify latent classes of problem-solving strategies. The examples illustrated in this paper are instances of an increasing global trend toward interdisciplinary research. As this trend continues to grow, research that takes advantage of the gaps and overlaps in analytical methodologies between disciplines will save time, effort, and research funds.

Multi-scale stochastic dynamic response analysis of offshore risers with lognormal uncertainties

Fatigue crack growth (FCG) in welded joints of orthotropic steel decks (OSDs) is highly stochastic, which complicates the accurate assessment of fatigue reliability. To consider the stochastic factors affecting fatigue performance in welded joints, this study proposed a stochastic FCG analysis method that incorporates machine learning to evaluate the fatigue reliability. Firstly, a probabilistic model for fatigue stress spectrum in welded joints was developed using site-specific traffic data. Subsequently, stochastic FCG analysis method assisted by a random forest model was developed to replace the time-consuming finite element solution. Finally, a fracture limit state function was established to evaluate fatigue reliability with consideration of the stochastic FCG process. Results indicate that a lognormal distribution accurately fits fatigue characteristic of the equivalent stress intensity factor range, with mean values consistently lower at the crack tip compared to the crack edge. Additionally, the deterioration rate of welded joints accelerates significantly at a crack depth of 5 mm, resulting in a shift of the remaining fatigue life distribution from lognormal to exponential. The reliability indices decrease significantly under overloading conditions. A 20% increase in vehicle weight reduces the fatigue life from the expected 100 years to 82 years.

Power system stability has become an increasing concern in renewable power integration because the active power generated by renewable energy varies randomly. Thus, we focus on random outside excitations in this study. Since the power obeys Non‐Gaussian distribution in most case, a novel small disturbance stochastic stability analysis method based on entropy is presented. First, the dynamic model of power system with random outside excitation is established by stochastic differential equation. Then, the values of mean and mean square stability are efficiently derived by transforming the stochastic differential equation into a deterministic mean or variance equation. For non‐Gaussian disturbance, entropy stability is presented by this transformation. At last, the numerical simulations are presented on a WSCC three‐generator nine‐bus power system with a wind farm © 2023 Institute of Electrical Engineer of Japan and Wiley Periodicals LLC.

Dynamic reliability analysis of slopes based on the probability density evolution method

The performance of a railway turnout (switch and crossing) is influenced by a large number of input parameters of the complex train–turnout system. To reach a robust design that performs well for different traffic situations, random distributions (scatter) of these inputs need to be accounted for in the design process. Stochastic analysis methods are integrated with a simulation model of the dynamic interaction between train and turnout. For a given nominal layout of the turnout, using design of experiments methodology and a two-level fractional factorial screening design, four parameters (axle load, wheel–rail friction coefficient, and wheel and rail profiles) are identified to be the most significant. These parameters are further investigated using a three-level full factorial design and stochastic analysis. The random distributions of transverse wheel profile and set of transverse rail profiles along the switch panel are accounted for by the Karhunen–Loève expansion technique. The influence of the random distributions of the input parameters on the statistical outputs of wheel–rail contact forces, wear and rolling contact fatigue is assessed using Latin hypercube sampling to generate a number of stochastic load realizations.

This paper describes the fatigue analysis of Kizomba 'A' FPSO based on the FPSO Fatigue Methodology Specification (FMS) [1] which is developed on the direct stochastic calculation. The FMS provides designers with technical guidance for fatigue control of newly built FPSO and can be characterized as integrated fatigue analysis to ensure the structural safety of all critical FPSO details by means of full stochastic analysis as well as component based one. At the cost of modeling complexity and computer resources, the full stochastic method gives the accurate fatigue damage for FPSO-specific details such as topside, piperack, flare etc., which are governed by hull-wave interactions. On the other hand, the component based one provides the cost-effective tools to screen the most fatigue prone area in topside supports for full stochastic analysis. Besides, the spectrum combination method that is suggested in order to consider both wave-line and wave-hull interactions, is adopted in fatigue damage calculation of the supports for FTL, OOL, riser and mooring. Apart from technical aspect, the interface management between other groups in charge of FPSO-hull, topside, riser and mooring system turns out very important due to tight project schedule. This paper proves the practical possibility and appropriateness of full stochastic fatigue analysis for FPSO and envisions its usefulness for other floating structures such as TLP, semisubmersibles and etc. The considered Kizomba 'A' FPSO to be installed on Angola Block 15 is 285 meters long, 63 meters wide, and 32 meters high, and can store 2.2 million barrels of crude oil in its lower hull facilities [2]. Introduction As the offshore oilfields shift into the deeper waters with little infrastructures, FPSOs have become the more popular solution. Currently one of the most active areas in FPSO installation is the West Africa where the weather is benign and the swell from the south is dominant. Whereas strength of the FPSO in such a calm site is mostly controlled on towing route, the fatigue control is directly dependent on the site-specific environment. It should be ensured that FPSOs could stay for decades without drydocking for maintenance. ExxonMobil has developed FPSO Fatigue Methodology Specification (FMS) with DNV for providing designers with technical guidance for fatigue control of newly built shipshaped FPSOs [3]. The aim of the fatigue control is to ensure that all critical structural details meet the fatigue design requirements. Also, the calculated fatigue lives form the basis of FPSO design and are used as inputs to the inspection plans during fabrication and in-service. The focus of the FMS is a full stochastic fatigue analysis method that gives the accurate fatigue damage for the FPSOspecific details like topside support, piperack, flare tower foundation etc. The structural behaviors of these supports are governed by wave-hull interactions and describable using linear analysis. The wave loads based on the metocean data of installation site are directly calculated and transferred to the structural model. Generally, full-ship structural model is utilized in the analysis. Figure 1 shows the general procedure of full stochastic fatigue analysis.

The aim of the study is to systematize current methods for assessing the economic efficiency of the best available food and agri-biotechnologies in domestic and international science and practice. The research applies methods of comparative analysis, systematization, and critical evaluation of scientific publications focused on the economic and environmental effectiveness of best available technologies (BAT). A classification-based approach was also employed to identify and group methods according to their orientation and application area. The empirical base includes regulatory documents, in particular, methodological guidelines for identifying technologies as BAT, national standards (GOST), and scientific publications on biotechnology, precision agriculture, organic waste processing, alternative protein sources, and more. The study examines methods such as investment feasibility analysis, life cycle environmental impact assessment, life cycle costing, multi-criteria decision analysis (MCDA), stochastic analysis, cluster analysis, among others. The findings demonstrate that integration of several methods enables assessment of both the financial and environmental viability of technology adoption. For example, in the case of cultivated meat, key indicators include both unit production cost and carbon footprint, which necessitates a shift from narrow economic assessment to eco-efficiency evaluation. The use of stochastic frontier analysis and clustering methods reveals that differences in farm efficiency are influenced not only by technology type, but also by the degree of its adoption. Overall, the systematization of methods for evaluating the efficiency of BAT indicates a clear trend toward transitioning from isolated economic indicators to comprehensive, data-driven approaches that combine economic and environmental metrics and validate results through empirical evidence.

A Lattice Structured Algorithm for Recursive Solution of the Extended Yule-Walker Equations

In this paper, we consider the superimposed exponential signals in zero-mean multiplicative and additive noise when all the noise are independently and identically distributed. We use a three-step iterative procedure to estimate the frequencies of the considered model. It is observed that the estimators are consistent and work quite well in terms of biases and mean square errors. Moreover, the convergence rate of the estimators attains O p (N −3/2), which is the best convergence rate in the case of only additive noise and constant amplitude.

A method is proposed to improve the accuracy of the parameter estimates of highly damped sinusoidal signals in additive noise. The method first utilizes singular value decomposition (SVD) of the signal matrix to reduce the noise effect, then based on the noise-reduced signals the weighted sum of squared errors is minimized using an adaptive lattice filter algorithm to estimate the parameters in the backward linear prediction model. In the case of large damping factors, the experimental results show that the proposed method gives more accurate estimates than the eigenvector method. >

How does internal processing contribute to visual pattern perception? By modeling visual search performance, we estimated internal signal and noise relevant to perception of curvature, a basic feature important for encoding of three-dimensional surfaces and objects. We used isolated, sparse, crowded, and face contexts to determine how internal curvature signal and noise depended on image crowding, lateral feature interactions, and level of pattern processing. Observers reported the curvature of a briefly flashed segment, which was presented alone (without lateral interaction) or among multiple straight segments (with lateral interaction). Each segment was presented with no context (engaging low-to-intermediate-level curvature processing), embedded within a face context as the mouth (engaging high-level face processing), or embedded within an inverted-scrambled-face context as a control for crowding. Using a simple, biologically plausible model of curvature perception, we estimated internal curvature signal and noise as the mean and standard deviation, respectively, of the Gaussian-distributed population activity of local curvature-tuned channels that best simulated behavioral curvature responses. Internal noise was increased by crowding but not by face context (irrespective of lateral interactions), suggesting prevention of noise accumulation in high-level pattern processing. In contrast, internal curvature signal was unaffected by crowding but modulated by lateral interactions. Lateral interactions (with straight segments) increased curvature signal when no contextual elements were added, but equivalent interactions reduced curvature signal when each segment was presented within a face. These opposing effects of lateral interactions are consistent with the phenomena of local-feature contrast in low-level processing and global-feature averaging in high-level processing.

Determination of sensitivity gradient is a major prerequisite for structural optimization, reliability assessment, and parameter identification. As the conventional deterministic sensitivity analysis cannot provide complete information, stochastic analysis is needed to tackle the uncertainties in structural parameters. This study focuses on the utility of the stochastic finite-element method for random response sensitivity analysis. The stochastic modeling of a random parameter is based on a commonly used 2D local averaging method generalized for a 3D case. The Choleski decomposition technique is then employed for digital simulation. The Neumann expansion based finite-element simulation method is extended for stochastic sensitivity analysis. This technique leads to a considerable saving of computational time. Example problems are used to compare the accuracy of this method to the direct Monte Carlo simulation and perturbation method in terms of varying stochasticity and efficiency in CPU time.

Objectives. The study aimed to investigate contemporary models, methods, and tools used for analyzing complex social network structures, both on the basis of ready-made solutions in the form of services and software, as well as proprietary applications developed using the Python programming language. Such studies make it possible not only to predict the dynamics of social processes (changes in social attitudes), but also to identify trends in socioeconomic development by monitoring users’ opinions on important economic and social issues, both at the level of individual territorial entities (for example, districts, settlements of small towns, etc.) and wider regions.Methods. Dynamic models and stochastic dynamics analysis methods, which take into account the possibility of self-organization and the presence of memory, are used along with user deanonymization methods and recommendation systems, as well as statistical methods for analyzing profiles in social networks. Numerical modeling methods for analyzing complex networks and processes occurring in them are considered and described in detail. Special attention is paid to data processing in complex network structures using the Python language and its various available libraries.Results. The specifics of the tasks to be solved in the study of complex network structures and their interdisciplinarity associated with the use of methods of system analysis are described in terms of the theory of complex networks, text analytics, and computational linguistics. In particular, the dynamic models of processes observed in complex social network systems, as well as the structural characteristics of such networks and their relationship with the observed dynamic processes including using the theory of constructing dynamic graphs are studied. The use of neural networks to predict the evolution of dynamic processes and structure of complex social systems is investigated. When creating models describing the observed processes, attention is focused on the use of computational linguistics methods to extract knowledge from text messages of users of social networks.Conclusions. Network analysis can be used to structure models of interaction between social units: people, collectives, organizations, etc. Compared with other methods, the network approach has the undeniable advantage of operating with data at different levels of research to ensure its continuity. Since communication in social networks almost entirely consists of text messages and various publications, almost all relevant studies use textual analysis methods in conjunction with machine learning and artificial intelligence technologies. Of these, convolutional neural networks demonstrated the best results. However, the use of support vector and decision tree methods should also be mentioned, since these contributed considerably to accuracy. In addition, statistical methods are used to compile data samples and analyze obtained results.

The independent component analysis and wavelet transform technology are used to separate and identify the internal combustion engine noise signal. According to the basic principle of independent component analysis, FastICA based on negative entropy great with the good stability and convergence speed algorithm is applied to separate the noise signals of six cylinder diesel engine. And the noise signals are decomposed into a series of independent components. The fast Fourier transform and wavelet transform technology are applied on each independent component analysis. Combining with the time-frequency analysis results and the internal combustion engine noise signal spectrum and the structure of the separation of the independent component and getting the corresponding relationship of different internal combustion engine noise sources. The results show that: the independent components correspond to the diesel engine combustion noise sources, the piston knock noise, and fuel injection pump noise and so on.