Iterative Prediction Method Research Articles

An iterative prediction method for designing the moderator used for the boron neutron capture therapy.

To conduct research related to slow neutrons, fast neutrons must be mode-rated and shifted to the desired energy region. In this research, an iterated prediction method, in which the neutron transportation properties of all materials were characterized by a reflection matrix, , and a transmission matrix, , was proposed to bypass a time-consuming Monte Carlo simulation and predict the performance of the moderator, including the epithermal neutron flux and the dose of fast neutrons and gamma rays, used for boron neutron capture therapy (BNCT). To find the optimal solution in the huge parameter space, a genetic algorithm combined with transmission and reflection matrices was utilized. The results showed that a 70-loop iteration was able to find a design for the moderator of BNCT with almost 80 higher epithermal neutron flux per kilowatt than that of the empirically optimized moderator that was previously reported in the literature. Compared with the Monte Carlo method, this method had the advantage of reducing the calculation time and statistical errors. The genetic algorithm with matrices (GAM) method can be used to find an optimal solution in a huge parameter space without brute-force calculations. It could be a promising method for designing the moderator for thermal or epithermal neutronusages.

Adaptive LSSVM based iterative prediction method for NOx concentration prediction in coal-fired power plant considering system delay

Obtaining accurate and real-time value of the pollution concentration is fundamental to effective and energy-saving operation for pollution controlling in coal-fired power plants. However, accurate measurements for NOx concentration cannot be guaranteed, due to the intrinsic hardware and software design in sensors. In this paper, a prediction method, including variables processing and model regression, is proposed for NOx concentration measurement. Specifically, a set of variables is firstly selected adaptively as an input set by using modified transfer entropy (TE), while the relationships among them are guaranteed to be as weak as possible. Then, the input set can cover features without introducing redundant information to the prediction model , and the system delay is reduced based on the TE and sequential displacement. After the variables are processed, a forgetting factor online least square support vector machine (FFOLSSVM) is constructed to predict NOx concentration timely and accurately. The proposed method is the first work that takes the system delay into consideration for NOx prediction model, without mechanism analysis. The simulation indicates that the computational time and prediction accuracy requirements are sufficiently guaranteed by the proposed model.

Computational prediction and experimental validation of revolute joint clearance wear in the low-velocity planar mechanism

Within the multibody dynamics framework, the paper proposed an efficient procedure to calculate revolute joint wear of planar mechanism with low velocity. Firstly, based on the wear test data, a method to calculate the wear coefficient that corresponds to the actual working condition of the mechanism is proposed. Then, an efficient iterative prediction method for joint clearance wear evolving in the multibody dynamic system in is proposed based on the Archard's wear model. Meanwhile, the wear tests of the typical mechanism have verified the method proposed in this paper. The research shows that when the increments of the wear depth of revolute joints are not big, the method can provide high prediction accuracy; however, as the predicted wear depth increases, the prediction error increases as well.

Application of Linear Prediction for Phase and Magnitude Correction in Partially Acquired MRI

Using the boxcar representation in the spatial domain and a signal-space representation of its frequency-weighted k-space, an iterative prediction method is developed to derive an improved low-resolution phase approximation for phase correction. Compared to the homodyne filter, the proposed predictor is found to be more efficient due to its capability of exhibiting an equivalent degree of performance using a lower number of fractional lines. The phase correction performance is illustrated using partially acquired susceptibility weighted images (SWI). An extension of the predictor into higher frequency regions of phase-encodes in conjunction with a signal-space projection in the frequency-weighted partial k-space is shown to provide restoration of fine structural details of sparse magnitude images. The application of subspace projection filtering is demonstrated using magnetic resonance angiogram (MRA).

IMPROVED ITERATIVE PREDICTION FOR MULTIPLE STOP ARRIVAL TIME USING A SUPPORT VECTOR MACHINE

The paper presents an improved iterative prediction method for bus arrival time at multiple downstream stops. A multiple-stop prediction model includes two stages. At the first stage, an iterative prediction model is developed, which includes a single stop prediction model for arrival time at the immediate downstream stop and an average bus speed prediction model on further segments. The two prediction models are constructed with a support vector machine (SVM). At the second stage, a dynamic algorithm based on the Kalman filter is developed to enhance prediction accuracy. The proposed model is assessed with reference to data collected on transit route No 23 in Dalian city, China. The obtained results show that the improved iterative prediction model seems to be a powerful tool for predicting multiple stop arrival time.

Mixed price and load forecasting of electricity markets by a new iterative prediction method

Load and price forecasting are the two key issues for the participants of current electricity markets. However, load and price of electricity markets have complex characteristics such as nonlinearity, non-stationarity and multiple seasonality, to name a few (usually, more volatility is seen in the behavior of electricity price signal). For these reasons, much research has been devoted to load and price forecast, especially in the recent years. However, previous research works in the area separately predict load and price signals. In this paper, a mixed model for load and price forecasting is presented, which can consider interactions of these two forecast processes. The mixed model is based on an iterative neural network based prediction technique. It is shown that the proposed model can present lower forecast errors for both load and price compared with the previous separate frameworks. Another advantage of the mixed model is that all required forecast features (from load or price) are predicted within the model without assuming known values for these features. So, the proposed model can better be adapted to real conditions of an electricity market. The forecast accuracy of the proposed mixed method is evaluated by means of real data from the New York and Spanish electricity markets. The method is also compared with some of the most recent load and price forecast techniques.

Reservoir neural state reconstruction and chaotic time series prediction

ESN(Echo state network) is a new type of recurrent neural network, which is based on the “reservoir”. ESN has been proved to be significantly efficient to deal with some chaotic time series prediction tasks. This paper makes an analysis of the current iterative prediction method based on “reservoir”, and points out some problems in theory and the obstacles in application. And then, a direct prediction method is proposed, which relates the prediction origin and prediction horizon directly. The direct method does not close the loop in the process of prediction, so there are no such problems as instability and error accumulation. The simulation results show how the reservoir property changes when the loop is closed in the iterative prediction, and then demonstrate the feasibility of the direct prediction method in application to the Mackey-Glass benchmark prediction problem.