|Fuzzy-Based Sediment Transport Simulation Using Contemporary Modeling Concepts and Measurement Methods as Validation|
|Project manager:||Prof. Dr.-Ing. Silke Wieprecht|
|Research assistants:||Dr.-Ing. Habtamu Tolossa, M.Sc.|
|Duration:||1.2.2009 - 31.1.2012|
|Funding:||International Postgraduate Studies in Water Technologies (IPSWaT) Scholarship|
This project is part of the research area:MMM- Monitoring, Measuring and Modelling
Abstract:An estimate of sediment transport rates in alluvial rivers is important in the context of erosion, sedimentation, flood control, long-term morphological assessment, etc. Extensive research during the last decades has produced a plethora of sediment transport models. Sediment transport is complex and often subject to semi-empirical or empirical treatment. Most of the sediment transport functions are based on simplified assumptions that the rate of sediment transport could be determined by one or two dominant factors, such as water discharge, average flow velocity, energy slope, and shear stress (Yang, 1996). In many practical situations prediction errors of these models are observed to be high.
An alternative approach is to use data driven modelling, which is especially attractive for modelling processes about which adequate knowledge of the physics is limited, like in the case of sediment transport. Over the last decade fuzzy rule-based models have been introduced in engineering as a powerful alternative modelling tool. The fuzzy rule-based approach introduced by Zadeh (1965) is being widely utilized in various fields of engineering. It is a qualitative modelling scheme in which the system behaviour is described using a natural language (Sugeno & Yasukawa, 1993).
This research focuses on the applicability of a data-driven fuzzy rule-based modelling approach in estimating sediment transport rates. It also aims at the comparison of the results of the fuzzy rule-based model with the results of other commonly utilized sediment transport functions.
A number of variables play important roles in determining sediment transport capacity. These variables are: flow depth, particle fall velocity, particle diameter, flow velocity, energy or water surface slope, shear velocity, shear stress, fluid density, sediment density, stream power, unit stream power, and discharge. Additionally; size, shape, and unit weight of bed composition; morphology of bed forms and availability of sediment from source area affect sediment transport capacity.
The most significant factors affecting sediment transport capacity will be identified and used for constructing a fuzzy model. The fuzzy model identification is usually carried out in two steps: (1) determining the number of fuzzy rules and their associated membership functions and (2) optimizing the fuzzy model. The fuzzy logic toolbox in MATLAB will be used for performing the fuzzy modelling.
A general fuzzy system has the components of fuzzification, fuzzy rule base, fuzzy output engine, and defuzzification. Fuzzification converts each piece of input data to degrees of membership by a look-up in one or more several membership functions. Intuition, fuzzy clustering, neural networks, genetic algorithms, and inductive reasoning can be among many ways to assign membership values or functions to fuzzy variables.
In fuzzy rule-based systems, knowledge is represented by if–then rules. Fuzzy rules consist of
two parts: an antecedent part stating conditions on the input variable(s); and a consequent part describing the corresponding values of the output variable(s). In Mamdani–Assilian type models, both antecedent and consequent parts consist of fuzzy statements concerning the value of the variables involved (Mamdani, 1977), whereas in Takagi–Sugeno type models, the consequent part expresses a non-linear relationship between the input variables and the output variable (Takagi and Sugeno, 1985).
The optimisation and validation of the fuzzy model will be done by data-driven tuning of the fuzzy model parameters so that the system output matches the observed output data. Fuzzy system optimization has two main categories: 1) parameter and 2) structure optimization. Parameter optimization is achieved by membership function fine tuning and rule conclusion optimization. Structure optimization includes input variable selection, and rule base reduction. Adaptive Neuro-Fuzzy Inference System (ANFIS), first introduced by Jang (1993), is used for optimizing the fuzzy model. ANFIS provides a fuzzy modeling procedure to learn information about a data set, in order to compute the membership function parameters that best allow the associated fuzzy inference system to track the given input/output data. The hybrid learning algorithm in ANFIS is used to identify parameters of the fuzzy inference system.