Papers by Keyword: Prediction Model

Abstract: In abrasive belt grinding, abrasive belt granularity, abrasive belt speed,feeding speed and grinding force have a great influence on the surface roughness. In order to predicate the surface roughness of Ti-6Al-4V,a response surface methodology are used to build the model to predict surface roughness,and the influence of various parameters on surface roughness was analysed. The research shows that with the abrasive belt granularity and abrasive belt speed increasing,the work piece surface roughness decreases;with the grinding force and feeding speed increasing,the work piece surface roughness increases. Through the test,the response surface methodology with high prediction accuracy,provides a theoretical basis for the reasonable selection of abrasive belt grinding parameters.

Abstract: Single crystal silicon has both important application value in the fields of micro-optics and MEMS, and it has been considered as one of the most difficult-to-cut materials because of its hardness and brittleness. Removal mechanism of the silicon was discussed, and the model of undeformed chip thickness was established in this article. According to the data of micro-groove surface roughness from the diamond fly-cutting experiment, the nonlinear relationship curve, between the largest undeformed chip thickness h_max and microgroove surface roughness R_a, were obtained using Gaussian-fitting principle, and the regression equation of the fitting curve was also got. Thus the prediction mathematical model of microgroove surface roughness was derived. The influence laws of the main working parameters on the R_a were obtained based on the result of this experiment and the response surface of the prediction model, and some conclusions were summarized: the surface roughness R_a of microgroove in the single crystal silicon decreases with the decrease of the cutting depth a_p, the feed f and the increase of the spindle speed n under the diamond fly-cutting; the experimental results also showed that feed f affects the value of R_a very much, cutting depth a_p less, and spindle speed n the least.

Abstract: This paper presents an automatic prediction model for ground vibration induced by Taiwan high-speed trains on embankment structures. The prediction model is developed using different field-measured ground vibration data. The main characteristics that affect the overall vibration level are established based on the database of measurement results. The influence factors include train speed, ground condition, measurement distance, and supported structure. Support vector machine (SVM) algorithm, a widely used prediction model, is adopted to predict the vibration level induced by high-speed trains on embankments. The measured and predicted vibration levels are compared to verify the reliability of the prediction model. Analysis results show that the developed SVM model can reasonably predict vibration level with an accuracy rate of 72% to 84% for four types of vibration level, including overall, low, middle, and high frequency ranges. The methodology in developing the automatic prediction system for ground vibration level is also presented in this paper.

Abstract: In the geomagnetic field, stress can induce spontaneous magnetic signals in ferromagnetic materials, the method, named metal magnetic memory testing, can be potentially applied in estimating the fatigue life. In this paper, the normal component of magnetic field, Hp (y), was measured during dynamic tension test on the surfaces of ferromagnetic specimens with stress concentration factor of 5. The results indicated that the gradient of magnetic field intensity, K, was the key parameter to characterize crack initiation life. Then the numerical fitting of K and fatigue cycles were done under three level loads, 568.7MPa, 698.8MPa and 864.4MPa, meanwhile, a simple model was derived.

Abstract: It is commonly known that particles play a critical role in urban stormwater quality because other pollutants can be attached to the particles and transported into receiving waters. Previous research studies have shown a strong relationship between pollutant build-up loads and particle sizes. In this context, accurately estimating the particle amounts in different sizes is extremely important since it can assist in predicting stormwater quality and hence contribute to effective stormwater quality improvement measures. This paper presents a robust model to predict particle size composition on urban road surfaces using heavy-duty vehicle volumes, traffic coefficient and road texture depth by multiple linear regression (MLR) method. The pollutants build-up data was used for model development and was collected on typical urban roads in Shenzhen, China. The relative prediction error and coefficient of variation values were found within the acceptable limits and hence indicated that the developed prediction models are relatively reliable. This developed model can assist in predicting particle size composition on urban road surfaces and thereby contribute to effective stormwater quality assessment and treatment design. Additionally, this developed modelling approach can also provide a guide in terms of particle size composition prediction using more influential factors.

Abstract: Port logistics demand is the basis and foundation of port logistics park planning and construction of the port city. Therefore, it has very important practical significance to research the port logistics demand. By input data preprocessing, determine the structure of the neural network to construct reasonable prediction model of neural network, can make a good prediction of the port logistics demand influenced by many factors, to provide the reference for the port planning and decision making.

Abstract: Experimental studies show that under special conditions, oils in corrosion environment have some inhibiting effect on CO₂ corrosion behavior of gathering pipelines. Oil wetting and corrosion product film are the great difference in existent rate prediction models of sweet corrosion. The progress of CO₂ corrosion rate prediction including empirical model, semi-empirical model, mechanistic model and artificial neural networks model considering the impact of oil in recent years are introduced in detail, the present problems and further research directions are also discussed.

Abstract: An efficient adhesive bonding of exterior ceramic tiles applied on façades is an obvious important factor to ensure the safety and the durability of the façade. The failure of adhesive bonding is a common issue with relevant technical and economic consequences.The aim of this work is to present an evaluation of the performance overtime of adhesives systems in bonded ceramic tiles on façades, based on extensive experimental research works carried out at the Laboratory of Building Physics (LFC).A detailed case study is presented which evaluate the performance of adhesives systems to be used on the façades of a building located near the sea. For this purpose, accelerated ageing tests are performed following two different ageing procedures, allowing the comparison of the performance over time of the systems under analysis).

Abstract: Back Propagation (BP) neural network can learn and store a large number of input-output model nonlinear relationships with simple structure. Niche ant colony algorithm (NACA) combines the ant colony algorithm (ACA) with the niche technology in order to add its local search ability to ACA with preserving the intelligent search ability and robustness of ACA. To optimize predicting model establishment of the dam monitoring data, NACA and BP neural network modeling method are combined to establish a prediction model of horizontal displacement monitoring data. The traditional BP neural network prediction model is established to make a comparison with the NACA. The results show that NACA-BP neural network method can speed up the convergence rate of BP neural network and enhance local search ability and prediction accuracy.

Abstract: This paper describes the study and analysis of different techniques for online solar irradiance prediction algorithms to properly estimate over the 24 hours of the next day in the “Universidad Militar Nueva Granada” (UMNG) campus at Cajicá, Colombia, in order to use predictions for a model predicted control of a DC-micro grid. These models were designed and tested using MATLAB® software. The performance of models were evaluated and compared among them to determine the best forecasting approach for Cajicá. The absence of seasons and the noisy solar irradiance time series caused by cloudy covering as perturbation are the main particularity of the Cajicá’s climate behavior. A meteorological database from 2010 to 2014 was used to estimate or train the model of prediction ARMAX and NNF, NAR, NARX as Artificial Neural Networks (ANNs), which were compared with error criteria such as square and absolute error criteria.