Structural materials having higher performance in strength, toughness, and fatigue resistance are strongly required. In the conventional materials development, many fatigue tests need to be conducted to validate stati...
详细信息
ISBN:
(纸本)9783319578644;9783319578637
Structural materials having higher performance in strength, toughness, and fatigue resistance are strongly required. In the conventional materials development, many fatigue tests need to be conducted to validate statistical behavior of fatigue failure. Accordingly the evaluation of fatigue properties with shorter time becomes quite essential. based on such background, we are developing fatigue prediction methods for wide range of structural materials by multi-scale finite element analysis (FEA) and machinelearning in the Materials Integration (MI) system. The multi-scale FEA consists of the following procedures: (i) mechanical and thermal properties are estimated by using commercially available software and database;(ii) temperature field, residual stress and distortion generated during a manufacturing process is calculated on the macroscopic model by thermo-mechanical FEA;(iii) macroscopic stress field under cyclic loading condition is calculated with a hardening constitutive model;(iv) the microscopic stress field is derived by finite element model with the polycrystalline structures and the cycles for a fatigue crack initiation is analyzed by strain energy accumulation on the slip plane;(v) the cycles for fatigue crack propagation is analyzed by extended finite element method (X-FEM) and the total number of cycles to the failure is obtained. The second approach is to use machinelearning techniques to obtain empirical prediction formula. The database was prepared from published resources and experiments. Deterministic machinelearning techniques such as multivariate linear regression and artificial neural network provided accurate equations to predict fatigue strength from materials and process parameters. Additionally, the concept of model-based machine learning was adopted to incorporate prior knowledge of microstructures and properties, and to account for uncertainty on fatigue life. The results showed that model-based machine learning was a promising tool fo
The Bayesian approach to machinelearning amounts to computing posterior distributions of random variables from a probabilistic model of how the variables are related (that is, a prior distribution) and a set of obser...
详细信息
The Bayesian approach to machinelearning amounts to computing posterior distributions of random variables from a probabilistic model of how the variables are related (that is, a prior distribution) and a set of observations of variables. There is a trend in machinelearning towards expressing Bayesian models as probabilistic programs. As a foundation for this kind of programming, we propose a core functional calculus with primitives for sampling prior distributions and observing variables. We define measure-transformer combinators inspired by theorems in measure theory, and use these to give a rigorous semantics to our core calculus. The original features of our semantics include its support for discrete, continuous, and hybrid measures, and, in particular, for observations of zero-probability events. We compile our core language to a small imperative language that is processed by an existing inference engine for factor graphs, which are data structures that enable many efficient inference algorithms. This allows efficient approximate inference of posterior marginal distributions, treating thousands of observations per second for large instances of realistic models.
The nonlinear behavior of loudspeakers is of great interest in a number of audio processing algorithms, as it may have a detrimental effect on their performance. These algorithms may be further enhanced when an accura...
详细信息
The nonlinear behavior of loudspeakers is of great interest in a number of audio processing algorithms, as it may have a detrimental effect on their performance. These algorithms may be further enhanced when an accurate model of the loudspeaker's input-output behavior is available. A variety of approaches has been investigated in the past to solve this task via nonlinear adaptive system identification. Their modeling capabilities are often limited due to a mismatch with electroacoustic principles of real loudspeakers. This paper therefore presents a novel approach using recurrent neural networks (RNN) specifically designed to match the dynamical loudspeaker's physical model behavior. By means of the physical model and its corresponding state-space representation, we derive three conceptually different RNN architectures, which are initially trained on synthetic audio data in order to gain insights into the required training procedure and limitations. Further training and evaluation of the preferred architecture on real loudspeaker recordings shows consistent improvements of the mean-squared modeling error compared to a linear system model and to nonlinear baseline algorithms.
Signal processing, communications, and control have traditionally relied on classical statistical modeling techniques. Such model-based methods utilize mathematical formulations that represent the underlying physics, ...
详细信息
Signal processing, communications, and control have traditionally relied on classical statistical modeling techniques. Such model-based methods utilize mathematical formulations that represent the underlying physics, prior information, and additional domain knowledge. Simple classical models are useful but sensitive to inaccuracies and may lead to poor performance when real systems display complex or dynamic behavior. On the other hand, purely data-driven approaches that are model-agnostic are becoming increasingly popular as datasets become abundant and the power of modern deep learning pipelines increases. Deep neural networks (DNNs) use generic architectures that learn to operate from data and demonstrate excellent performance, especially for supervised problems. However, DNNs typically require massive amounts of data and immense computational resources, limiting their applicability for some scenarios. In this article, we present the leading approaches for studying and designing model-based deep learning systems. These are methods that combine principled mathematical models with data-driven systems to benefit from the advantages of both approaches. Such model-based deep learning methods exploit both partial domain knowledge, via mathematical structures designed for specific problems, and learning from limited data. Among the applications detailed in our examples for model-based deep learning are compressed sensing, digital communications, and tracking in state-space models. Our aim is to facilitate the design and study of future systems at the intersection of signal processing and machinelearning that incorporate the advantages of both domains.
We introduce a new general framework for sign recognition from monocular video using limited quantities of annotated data. The novelty of the hybrid framework we describe here is that we exploit state-of-the art learn...
详细信息
ISBN:
(纸本)9791095546009
We introduce a new general framework for sign recognition from monocular video using limited quantities of annotated data. The novelty of the hybrid framework we describe here is that we exploit state-of-the art learning methods while also incorporating features based on what we know about the linguistic composition of lexical signs. In particular, we analyze hand shape, orientation, location, and motion trajectories, and then use CRFs to combine this linguistically significant information for purposes of sign recognition. Our robust modeling and recognition of these sub-components of sign production allow an efficient parameterization of the sign recognition problem as compared with purely data-driven methods. This parameterization enables a scalable and extendable time-series learning approach that advances the state of the art in sign recognition, as shown by the results reported here for recognition of isolated, citation-form, lexical signs from American Sign Language (ASL).
Modern communication systems rely on accurate channel estimation to achieve efficient and reliable transmission of information. As the communication channel response is highly related to the user's location, one c...
详细信息
ISBN:
(纸本)9798350344868;9798350344851
Modern communication systems rely on accurate channel estimation to achieve efficient and reliable transmission of information. As the communication channel response is highly related to the user's location, one can use a neural network to map the user's spatial coordinates to the channel coefficients. However, these latter are rapidly varying as a function of the location, on the order of the wavelength. Classical neural architectures being biased towards learning low frequency functions (spectral bias), such mapping is therefore notably difficult to learn. In order to overcome this limitation, this paper presents a frugal, model-based network that separates the low frequency from the high frequency components of the target mapping function. This yields an hypernetwork architecture where the neural network only learns low frequency sparse coefficients in a dictionary of high frequency components. Simulation results showthat the proposed neural network outperforms standard approaches on realistic synthetic data.
In this paper we attempt to deduce student attrition at a South African higher-education institution with the aim of identifying students who are likely to be in need of academic support so that a focus could be provi...
详细信息
ISBN:
(纸本)9781450388474
In this paper we attempt to deduce student attrition at a South African higher-education institution with the aim of identifying students who are likely to be in need of academic support so that a focus could be provided on improving their academic performance. The significance of this paper is on using computer science and information technology to address learner attrition (an African reality) and thereby impact the low university throughput and retention rates positively. We trained several machinelearning classification models to deduce the student into four risk classes using only Grade 12 marks and background characteristics of the learner. We provide the following contributions: (a) the first known published trained classifier able to calculate the distribution over a students' risk profile for a South African university focused on the conceptual framework;(b) a ranking of employed features according to their entropy to correctly classify the class variable;(c) a comparison of trained classifiers able to calculate the probability of a students' risk profile for a South African higher-education research-intensive university;and (d) an interactive program which is able to calculate the posterior probability over the student's risk profile so that support can be provided to them. The random forest classification model achieves the best performance with a 82% accuracy over these four risk profiles. We argue for introducing predictive tools to enhance student success and student support initiatives in Higher-Education Institutions. This work will benefit academic developers and staff who provide support to students who are at academic risk of completing their undergraduate Science programmes.
This paper explores the usage of artificial neural networks to evaluate forces acting in dynamically loaded finite-length journal bearings. Unlike standard numerical approaches, which require solving a hydrodynamic pr...
详细信息
This paper explores the usage of artificial neural networks to evaluate forces acting in dynamically loaded finite-length journal bearings. Unlike standard numerical approaches, which require solving a hydrodynamic pressure field, the network predicts the forces directly from relative displacements and velocities of a rotating journal to a stationary bearing shell. This practice can significantly accelerate transient simulations of systems supported on such bearings without compromising their nonlinear properties. The proposed method utilises feedforward neural networks, which use a precomputed database of nondimensional forces for training. This database is generated using a finite difference method and supplemented with the corresponding relative displacements and velocities. The performance of the trained networks is also analysed.
Implementing a condition-based maintenance strategy requires an effective condition monitoring (CM) system that can be complicated to develop and even harder to maintain. In this paper, we review the main complexities...
详细信息
Implementing a condition-based maintenance strategy requires an effective condition monitoring (CM) system that can be complicated to develop and even harder to maintain. In this paper, we review the main complexities of developing condition monitoring systems and introduce a four-stage framework that can address some of these difficulties. The framework achieves this by first using process knowledge to create a representation of the process condition. This representation can be broken down into simpler modules, allowing existing monitoring systems to be mapped to their corresponding module. Data-driven models such as machinelearningmodels could then be used to train the modules that do not have existing CM systems. Even though data-driven models tend to not perform well with limited data, which is commonly the case in the early stages of pharmaceutical process development, application of this framework to a pharmaceutical roller compaction unit shows that the machinelearningmodels trained on the simpler modules can make accurate predictions with novel fault detection capabilities. This is attributed to the incorporation of process knowledge to distill the process signals to the most important ones vis-& agrave;-vis the faults under consideration. Furthermore, the framework allows the holistic integration of these modular CM systems, which further extend their individual capabilities by maintaining process visibility during sensor maintenance.
暂无评论