Representing diverse mathematical problems using neural networks in hybrid intelligent systems

In recent years, artificial neural networks have attracted considerable attention as candidates for novel computational systems. Computer scientists and engineers are developing neural networks as representational and computational models for problem solving: neural networks are expected to produce new solutions or alternatives to existing models. This paper demonstrates the flexibility of neural networks for modeling and solving diverse mathematical problems including Taylor series expansion, Weierstrass's first approximation theorem, linear programming with single and multiple objectives, and fuzzy mathematical programming. Neural network representations of such mathematical problems may make it possible to overcome existing limitations, to find new solutions or alternatives to existing models, and to achieve synergistic effects through hybridization.     

An adaptation algorithm for an intelligent natural language tutoring system

The focus of computerised learning has shifted from content delivery towards personalised online learning with Intelligent Tutoring Systems (ITS). Oscar Conversational ITS (CITS) is a sophisticated ITS that uses a natural language interface to enable learners to construct their own knowledge through discussion. Oscar CITS aims to mimic a human tutor by dynamically detecting and adapting to an individual's learning styles whilst directing the conversational tutorial. Oscar CITS is currently live and being successfully used to support learning by university students. The major contribution of this paper is the development of the novel Oscar CITS adaptation algorithm and its application to the Felder–Silverman learning styles model. The generic Oscar CITS adaptation algorithm uniquely combines the strength of an individual's learning style preference with the available adaptive tutoring material for each tutorial question to decide the best fitting adaptation. A case study is described, where Oscar CITS is implemented to deliver an adaptive SQL tutorial. Two experiments are reported which empirically test the Oscar CITS adaptation algorithm with students in a real teaching/learning environment. The results show that learners experiencing a conversational tutorial personalised to their learning styles performed significantly better during the tutorial than those with an unmatched tutorial.     

Intelligent tutoring systems: Formalization as automata and interface design using neural networks

This article proposes a mathematical model of Intelligent Tutoring Systems (ITS), based on observations of the behaviour of these systems. One of the most important problems of pedagogical software is to establish a common language between the knowledge areas involved in their development, basically pedagogical, computing and domain areas. A mathematical model, like the one proposed here, can facilitate the integration of these different areas, as it defines the elements that constitute the system and defines the technological tools to implement it. The article presents an example demonstrating how the formalization was used to design the adaptive mechanism of an ITS to adapt its Interface Module to some student characteristics.     

Computational organization of didactic contents for personalized virtual learning environments

This paper presents an organization model for personalized didactic contents used in individual study environments. For many students the availability of contents in a general form might not be effective. A multilevel structure of concepts is proposed to provide different presentation combinations of the same content. Our work shows that it is possible to personalize the didactic content in order to encourage students, by using proximal learning patterns. These patterns are obtained from the analysis of the actions of students with positive results in the individual content organization. The system uses artificial intelligence techniques to reactively organize and personalize content. Personalization is made possible by means of an artificial neural network that classifies the student's profile and assigns it a proximal learning pattern. Expert rules are used to mediate and adjust the contents reactively. Experimental results indicate that the approach is efficient and provides the student a better use of the content with adaptive and reactive personalized presentation.     

An analysis of COMET and MAGIC using the standard reference model for intelligent multimedia presentation systems

We provide an overview of two multimedia generation systems developed at Columbia University, COMET and MAGIC, and analyze them using the standard reference model for intelligent multimedia presentation systems. COMET (Coordinated Multimedia Explanation Testbed) emphasizes the coordinated generation of text and 3D graphics in a military maintenance and repair documentation domain. It features support for automated determination of what information to present, assignment of information to appropriate media and generation of textual cross-references to the structure and content of generated graphics, including the automated redesign of graphics to make possible better cross-references. MAGIC (Multimedia Abstract Generation for Intensive Care) generates presentations that use temporal media, including speech and animation, to explain the status of hospital patients. It extends the media coordination approach used in COMET to include a multi-stage negotiation process that solves temporal constraints imposed by individual media generators to determine when and for how long actions should occur in each medium.     

Predictive models for emission of hydrogen powered car using various artificial intelligent tools

This paper investigates the use of artificial intelligent models as virtual sensors to predict relevant emissions such as carbon dioxide, carbon monoxide, unburnt hydrocarbons and oxides of nitrogen for a hydrogen powered car. The virtual sensors are developed by means of application of various Artificial Intelligent (AI) models namely; AI software built at the University of Tasmania, back-propagation neural networks with Levenberg–Marquardt algorithm, and adaptive neuro-fuzzy inference systems. These predictions are based on the study of qualitative and quantitative effects of engine process parameters such as mass airflow, engine speed, air-to-fuel ratio, exhaust gas temperature and engine power on the harmful exhaust gas emissions. All AI models show good predictive capability in estimating the emissions. However, excellent accuracy is achieved when using back-propagation neural networks with Levenberg–Marquardt algorithm in estimating emissions for various hydrogen engine operating conditions with the predicted values less than 6% of percentage average root mean square error.     

Improving neural network models of physical systems through dimensional analysis

The paper presents a technique for generating concise neural network models of physical systems. The neural network models are generated through a two-stage process. The first stage uses information embedded in the dimensions or units in which the data is represented. Dimensional analysis techniques are used initially to make this information explicit, and a limited search in the neural network architecture space is then conducted to determine dimensionless representations of variables/parameters that perform well for a given model complexity. The second stage uses information available in the numerical values of the data to search for high-level dimensionless variables/parameters, generated from simple combinations of dimensionless quantities generated in the first stage and which result in concise neural network models with improved performance characteristics. The search for these high-level dimensionless variables/parameters is conducted in an enhanced representation space using functional link networks with flat or near flat architectures. The use and effectiveness of the technique is demonstrated for three applications. The first is the design and analysis of reinforced concrete beams, which is representative of the class of problems associated with the design and analysis of composites. The second is the classical elastica problem, for predicting non-linear post-buckled behaviour of columns and the third, the analysis of a bent bar under a specified combination of loads.     

Unified methodology of neural analysis in decision support systems built for pharmaceutical technology

The objective of this study was to create universal methodology of artificial neural networks (ANNs) application in construction of decision support systems designed for various dosage forms. Two different dosage forms (solid dispersions and microemulsions) were modeled with use of the same methodology, software and hardware environments. Completely different models prepared confirmed their generalization ability both for solid dosage forms (solid dispersions) and liquid dosage forms (microemulsions). ME_expert and SD_expert systems basing on the neural expert committees were created. In the pilot study their application allowed for appropriate choice of qualitative and quantitative composition of particular pharmaceutical formulation. It was also proposed that ME_expert and SD_expert might provide in silico formulation procedures. Unified methodology of neural modeling in pharmaceutical technology was confirmed to be effective in providing valuable tools for pharmaceutical product development.     

Decentralized fault detection and isolation in wireless structural health monitoring systems using analytical redundancy

One of the most critical issues when deploying wireless sensor networks for long-term structural health monitoring (SHM) is the correct and reliable operation of sensors. Sensor faults may reduce the quality of monitoring and, if remaining undetected, might cause significant economic loss due to inaccurate or missing sensor data required for structural assessment and life-cycle management of the monitored structure. This paper presents a fully decentralized approach towards autonomous sensor fault detection and isolation in wireless SHM systems. Instead of physically installing multiple redundant sensors in the monitored structure (“physical redundancy”), which would involve substantial penalties in cost and maintainability, the information inherent in the SHM system is used for fault detection and isolation (“analytical redundancy”). Unlike traditional centralized approaches, the analytical redundancy approach is implemented distributively: Partial models of the wireless SHM system, implemented in terms of artificial neural networks in an object-oriented fashion, are embedded into the wireless sensor nodes deployed for monitoring. In this paper, the design and the prototype implementation of a wireless SHM system capable of autonomously detecting and isolating various types of sensor faults are shown. In laboratory experiments, the prototype SHM system is validated by injecting faults into the wireless sensor nodes while being deployed on a test structure. The paper concludes with a discussion of the results and an outlook on possible future research directions.     

Fuzzy reliability analysis of repairable industrial systems using soft-computing based hybridized techniques

The purpose of the present study is to analyze the fuzzy reliability of a repairable industrial system utilizing historical vague, imprecise and uncertain data which reflects its components’ failure and repair pattern. Soft-computing based two different hybridized techniques named as Genetic Algorithms Based Lambda–Tau (GABLT) and Neural Network and Genetic Algorithms Based Lambda–Tau (NGABLT) along with a traditional Fuzzy Lambda–Tau (FLT) technique are used to evaluate some important reliability indices of the system in the form of fuzzy membership functions. As a case study, all the three techniques are applied to analyse the fuzzy reliability of the washing system in a paper mill and results are compared. Sensitivity analysis has also been performed to analyze the effect of variation of different reliability parameters on system performance. The analysis can help maintenance personnel to understand and plan suitable maintenance strategy to improve the overall performance of the system. Based on results some important suggestions are given for future course of action in maintenance planning.     

Identification using ANFIS with intelligent hybrid stable learning algorithm approaches

This paper suggests novel hybrid learning algorithm with stable learning laws for adaptive network based fuzzy inference system (ANFIS) as a system identifier and studies the stability of this algorithm. The new hybrid learning algorithm is based on particle swarm optimization (PSO) for training the antecedent part and gradient descent (GD) for training the conclusion part. Lyapunov stability theory is used to study the stability of the proposed algorithm. This paper, studies the stability of PSO as an optimizer in training the identifier, for the first time. Stable learning algorithms for the antecedent and consequent parts of fuzzy rules are proposed. Some constraints are obtained and simulation results are given to validate the results. It is shown that instability will not occur for the leaning rate and PSO factors in the presence of constraints. The learning rate can be calculated on-line and will provide an adaptive learning rate for the ANFIS structure. This new learning scheme employs adaptive learning rate that is determined by input–output data.     

Wavelet-based multi-resolution analysis and artificial neural networks for forecasting temperature and thermal power consumption

As part of the OptiEnR research project, the present paper deals with outdoor temperature and thermal power consumption forecasting. This project focuses on optimizing the functioning of a multi-energy district boiler (La Rochelle, west coast of France), adding to the plant a thermal storage unit and implementing a model-based predictive controller. The proposed short-term forecast method is based on the concept of time series and uses both a wavelet-based multi-resolution analysis and multi-layer artificial neural networks. One could speak of “MRA-ANN” methodology. The discrete wavelet transform allows decomposing sequences of past data in subsequences (named coefficients) according to different frequency domains, while preserving their temporal characteristics. From these coefficients, multi-layer Perceptrons are used to estimate future subsequences of 4 h and 30 min. Future values of outdoor temperature and thermal power consumption are then obtained by simply summing up the estimated coefficients. Substituting the prediction task of an original time series of high variability with the estimation of its wavelet coefficients on different levels of lower variability is the main idea of the present work. In addition, the sequences of past data are completed, for each of their components, by both the minute of the day and the day of the year to place the developed model in time. The present paper mainly focuses on the impact on forecast accuracy of various parameters, related with the discrete wavelet transform, such as both the wavelet order and the decomposition level, and the topology of the neural networks used. The number of past sequences to take into account and the chosen time step were also major concerns. The optimal configuration for the tools used leads to very good forecasting results and validates the proposed MRA-ANN methodology.     

Are you a human or a humanoid: Predictive user modelling through behavioural analysis of online gameplay data

Intelligent agents are widely used in robotics, gaming and simulation. A key issue is modelling human behaviours so that intelligent agents can use a human’s behavioural model to imitate them and predict their next moves. In this article, we use Internet-based multiplayer online gaming (MOG) as a case study to present our approach to predictive user modelling through behavioural analysis of online gameplay data. As latency is an inherited bottleneck of the Internet and is likely to remain so into a foreseeable future, a lot of efforts have been made to address the resulting issues. Most of the existing latency handling techniques are based on the assumption that latency is within an acceptable threshold so that they can alleviate or even completely hide its negative impact on players’ quality of experience (QoE) that directly determines consumers’ satisfaction of the provided MOG services. While this assumption is mostly valid, it is worth noting that a player’s Internet connection latency always fluctuates (known as jitter), possibly to the extent of exceeding a MOG’s designated threshold in which case none of the techniques can handle properly but disconnecting the player from the gameplay session. Forcing a player to quit prematurely simply due to a spike of unusual high latency has a significant negative impact both on the gameplay’s fairness and on the player’s QoE. To improve customer satisfaction of a MOG service, we propose a more tolerant approach by temporarily substituting a player with a humanoid bot in the event of latency hikes so that the player always remains in the gameplay session. The challenge in this approach is to create a personalised humanoid bot that can imitate the playing pattern of the individual human player being substituted. Our solution is to first extract key variables that have impact on the human player’s decision-makings through behavioural analysis of the player’s historical gameplay data, then model the relationships among these variables, and finally creates the player’s humanoid bot with the model. In this paper, we use a multiplayer online pong game as a case study to explain behavioural variables, modelling techniques, processes, outcomes, and performance studies.     

FraMePID-3PB software for material parameter identification using fracture tests and inverse analysis

Knowledge regarding the values of fracture-mechanical parameters is critical for the virtual failure modeling of elements and structures made of concrete. A key parameter in nonlinear fracture mechanics modeling is the specific fracture energy of concrete, and its variability. Three-point bending tests on notched-beam specimens are fundamental experiments for the determination of fracture-mechanical parameters. In the present paper, two basic approaches are applied to determine fracture-mechanical parameter values from these tests: (i) the effective crack model/work-of-fracture method, and (ii) inverse analysis using artificial neural networks and stochastic simulations. First, the paper describes suitable methods for the determination of fracture-mechanical parameters. Second, the FraMePID-3PB software tool, which has been developed in order to automate the whole time consuming process of inverse analysis, is described. Finally, the verification of methodology and software is presented using two illustrative examples.     

Developing a robust model predictive control architecture through regional knowledge analysis of artificial neural networks

Chemical processes are nonlinear. Model based control schemes such as model predictive control are highly related to the accuracy of the process model. For a highly nonlinear chemical system, it is clear to implement a nonlinear empirical model, such as artificial neural network model, should be superior to a linear model such as dynamic matrix model. However, unlike linear systems, the accuracy of a nonlinear empirical model strongly depends on its original data or training data based on how the model is built up. A regional-knowledge index is proposed in this study and applied in the analysis of dynamic artificial neural network models in process control. New input patterns that imply extrapolations and thus unreliable prediction by an artificial neural network model can be recognized from a significant decrease in the regional-knowledge index. To tackle the extrapolation problem and assure stability of the control system, we propose to run a neural adaptive controller in parallel with a model predictive control. A coordinator weights the outputs of these two controllers to make the final control decision. The present state of the controlled process and the model fitness to the present input pattern determine the weightings of the controller's output. The proposed analysis method and the modified model predictive control architecture have been applied to a neutralization process and excellent control performance is observed in this highly nonlinear system.     

Human reliability analysis and optimization of manufacturing systems through Bayesian networks and human factors experiments: A case study in a flexible intermediate bulk container manufacturing plant

Human reliability analysis (HRA) and optimization in manufacturing systems are effective to reduce system failure. The purpose of this study is to examine the HRA and optimization through a Bayesian network (BN) model and human factors experiments (HFEs). This study was applied to a flexible intermediate bulk container manufacturing plant. The human physiological and psychological factors consisting of personal abilities of flexibility, coordination, memory, and attention were regarded as the only performance shaping factors in this study. With the BN model, the relationship between human factors and human errors was described qualitatively and the impact of the human factor on system failures was judged quantitatively. Then the workers’ abilities training with HFEs based on the fault diagnosis results was carried out. The total numbers of errors have been decreased by 69.06% and the system failure rate has been reduced significantly after training.     

Nonlinear adaptive control using neural networks and its application to CSTR systems

In this paper, adaptive tracking control is considered for a class of general nonlinear systems using multilayer neural networks (MNNs). Firstly, the existence of an ideal implicit feedback linearization control (IFLC) is established based on implicit function theory. Then, MNNs are introduced to reconstruct this ideal IFLC to approximately realize feedback linearization. The proposed adaptive controller ensures that the system output tracks a given bounded reference signal and the tracking error converges to an -neighborhood of zero with being a small design parameter, while stability of the closed-loop system is guaranteed. The effectiveness of the proposed controller is illustrated through an application to composition control in a continuously stirred tank reactor (CSTR) system.