Predicting Velocities In Gross Pollutant Trap Environmental Sciences Essay

Rapid urbanization tends to modify the hydrologic, hydraulic and environment characteristics of an area. Urbanization also causes problems of pollution of receiving waterways. Knowing of the causes of pollutants, disposed off in streams, is problematic area of the recent years for the effective control of pollution. Gross pollutant traps are considerable structures to trap the pollutants from the storm water. Velocity of storm water is the significant factor in transporting these gross pollutants. The critical flow where the velocity is minimum will leads the high efficiency of gross pollutant traps. It is important to study the velocity profiles in gross pollutant traps. Present study concentrates on the velocity profiles from the experimental data was compared with the Adaptive Neuro Fuzzy Inference System (ANFIS) approach to predict the velocities in gross pollutant traps. The proposed ANFIS approach produces satisfactory results (R2 = 0.902) compared to the experimental results.

Key words: Pollutants, Gross pollutant traps, ANFIS, Velocity Contours

Introduction

Rapid urbanization tends to modify the hydrologic, hydraulic and environment characteristics of an area. Urbanization also causes problems of pollution of receiving waterways. It is well known that the quality of urban rivers is influenced by many factors among others are the land-use, method of waste disposal (gross pollutants) and sanitation practices. Streams have been used for the disposal of various industrial and municipal wastes since decades. Understanding of mixing of such pollutants in streams is a matter of concern in recent years for the effective control of pollution in the streams. Pollutants carried by urban stormwater runoff are considered a significant contributor to the degradation of receiving waters. Gross pollutants are often targeted first for the removal and many structural measures have been applied with varying results. Urban storm water pollutants include gross pollutants, trace metals and nutrients that are associated with sediments, and dissolved pollutants.

Innumerable pollutants can entry the waterway. The range from gross pollutants (trash, litter and vegetation larger than 5 mm), sediments (fine (<0.062), medium (0.062-0.5mm), course (0.5-5 mm)), attached pollutants (attached to fine sediments specifically nutrients, heavy metals, toxicants and hydro carbons) and dissolved pollutants (typically nutrients , metals and salts)(CISRO, 1999), as well as bacteria, viruses and other organisms, oxygen demanding substances and aquatic weeds. Figure 1 shows the various types of pollutants exists in waterways. The nature of pollutants entering the catchment drainage system is dependent on the land use within the catchment. The excluding gross pollutants is advisable as they are unattractive, disturb physical habitat, degrade water, attract pests and vermin, cause marine animal depths, promote littering and reduce amenity (Allision et al.,1998)

There are now a number of devices (including the conventional and proprietary devices) for trapping of gross pollutants that are based on initially diverting stormwater to a separation and retention chamber in which those pollutants are subjected to mechanism of interception and sedimentation. Gross pollutant traps (GPTs) are the devices that remove solids conveyed by storm water. There is verity of GPTs suitable for use in urban catchments that remove litter and debris greater than 5 millimeters and coarse sediments before they enter the receiving waters. GPTs can operate in isolation to protect immediate downstream receiving waters or as part of a more comprehensive treatment system to prevent overload of downstream infrastructure.

The reduction and removal of urban litter is a complex and difficult problem, particularly for developing countries. Ultimately, the solution depends on each local authority developing an integrated catchment litter management strategy that includes planning controls, source controls, and structural controls (Armitage, 2007). GPTs in Brookvale Creek, in Sydney’s Northern Beaches region, use the outlet approach in which the GPTs act as the single treatment point for the upstream catchments (CSIRO, 1999). This type of GPT is installed in-line and functions via direct screening with a grid or mesh barrier assembly to reduce the quantity of gross pollutants carried by the runoff. The GPT traps the coarse sediment by decreasing the flow velocity. The GPT that was used in this creek has markedly reduced the amount of gross pollutant in the stream (Rawson et al., 2002). Recent studies of GPTs in Brisbane, Australia indicate that weather conditions such as the extent and duration of rainfall can influence the inflow rate to the GPTs. GPT blockage, either partial or full, can change the litter retention characteristics of the structures (Madhani et al., 2009). The gross pollutant trapping efficiency was the highest at the lowest water depth. The efficiency decreased as the velocity of flow increases ( Ab. Ghani et al., 2011).

The Gross pollutants device traps the gross pollutants and coarse sediment by decreasing the velocity of flow. To achieve removal for a range of pollutants a number of treatments are required. Velocity of storm water is the significant factor in transporting these floatable pollutants. The critical flow where the velocity is minimum will leads the high efficiency of gross pollutant traps. It is important to study the velocity profiles in gross pollutant traps. Present study concentrates on the velocity profiles from the experimental data was compared with the Adaptive Neuro Fuzzy Inference System (ANFIS) approach to predict the velocities in gross pollutant traps. The proposed ANFIS approach produces satisfactory results (R2 = 0.902) compared to the experimental results.

Experimental data:

The GPT used in this study was a multi-component system consisting of a sediment trap and gross pollutant trap capable of preventing bed load and removing solid waste from stormwater. Figure 2 shows the structure of the GPT. Basically, it consisted of two compartments: a primary trap and a secondary trap. The primary trap was a sediment trap compartment equipped with a primary trash rack, which was constructed in-line with the channel flow direction to treat stormwater during low flow conditions. The primary trap comprised a uniform channel with an expansion extended from the existing drain and with a drop at the sediment trap to reduce the velocity of the incoming flow. It is essential to reduce the flow velocity to achieve optimal settlement of the sediment. Under low flow conditions, turbulence is expected to be less significant. Denser pollutants will settle out of the water column and onto the bottom of the sediment trap, while other floatables will be trapped at the primary trash rack; the water will continue to filter through the trap. The primary trap was designed to serve an up to a 3-month average recurrent interval (ARI) event, a design requirement set for stormwater quality treatment in Malaysia (Ab. Ghani et al., 2011).

This GPT prototype was constructed based on the Froude Number similarity. In practical settings, pollutants may be found in very large range of sizes, densities, and shapes. However, the most common gross pollutant types found in Malaysian stormwater system (debris and sediment) were used in this laboratory test. The settling velocities and densities of the typical gross pollutants were determined, and similitude laws were then used to identify representative scale particles. Tests were carried out for a variety of flow rates to examine the performance of the GPT in response to both minor and major designed storm events. The 0.3 m depth represents dry-weather flow events, 0.5 m depth represents frequent events and 0.7 m depth represents occasional events (Ab. Ghani et al., 2011).

Each experiment began with the release of water at the main inlet of the GPT. The direction of water flow and cross sections are shown in Figure 4. Flow velocities were measured after the steady flow had achieved at a depth 0.3 m. The velocities were measured using an electromagnetic current meter (Figure 5) at various points of three cross sections. For each depth and each section velocities were measured at (0.2*depth) and (0.8*depth) from the free surface. Velocities were measured at the points having a minimum interval of 150 mm perpendicular to flow of direction. And the measurements were recorded for 0.3 m, 0.5 m and 0.7 m depth.

The ANFIS networks

ANFIS, first introduced by Jang (1993), is a universal imaginative and, as such, is capable of approximating any real continuous function on a compact set to any degree of measurability. Thus, in parameter estimation, where the given data are such that the system associates measurable system variables with an internal system parameter, a functional mapping may be constructed by ANFIS that approximates the process of estimation of the internal system parameter.

The ANFIS is functionally equivalent to fuzzy inference systems. The hybrid learning algorithm, which combines gradient descent and the least-squares method, is introduced, and

the issue of how the equivalent fuzzy inference system can be rapidly calibrated and adapted with this algorithm is discussed herein. Most of the previous works that address artificial neural networks (ANN) applications to water resources have included the feed forward type of the architecture, where there are no back ward connections, which are trained using the error back propagation scheme or the feed forward back propagation (FFBP) configuration. Drawbacks of ANN include that it needs more training time and the difficulties in detecting hidden neurons in hidden layer for better predictions (Azamathulla and Ghani, 2010). Therefore, the present study applies a new soft computing technique-ANFIS. The input in ANFIS is first converted into fuzzy membership functions, which are combined together. After following an averaging process to obtain the output membership functions, the desired output is finally achieved.

Development of ANFIS Model

The network of ANFIS as shown in Figure 6 works as follows: let x and y be the two typical input values fed at the two input nodes, which will then transform those values to the membership functions (say bell- shaped) and give the output as follows: (note in general, w is the output from a node; m is the membership function, and x, y are inputs in Eq. (1))

(1)

where a1, b1, and c1 are changeable premise parameters. Similar computations are carried out for the input of y to obtain µNi(y). The membership functions are then multiplied in the second layer, e.g.

(i=1, 2) (2)

Such products or firing strengths are then averaged:

(i=1, 2) (4)

Nodes of the fourth layer use the above ratio as a weighting factor. Furthermore, using fuzzy if-then rules produces the following output: (An example of an if-then rule is: If x is M1 and y is N1, then f1 = p1x + q1y + r1)

(5)

where p, q, and r are changeable consequent parameters. The final network output f was produced by the node of the fifth layer as a summation of all incoming signals, which is exemplified in the Eq. (5).

A two-step process is used for faster training and to adjust the network parameters to the above network. In the initial step, the premise parameters are kept static, and the information is propagated forward in the network to layer 4. In layer 4, a least-squares estimator identifies the important parameters. In the second step, the backward pass, the chosen parameters are held fixed while the error is propagated. The premise parameters are then modified using gradient descent. Apart from the training patterns, the only user-specified matter required is the number of membership functions for each input. The explanation of the learning algorithm is given in Jang and Sun (1995).

The velocities measured for three depths 0.3 m, 0.5 m and 0.7 m having complete data sets of 162 patterns. Each data set contains depth, distance, and measured velocity as membership functions. Out of 162 data set patterns 0.3 m and 0.7 m (109 data sets, around 67 %) were used for the training, while the remaining patterns (53 data sets, around 33 %) were used for testing, or validating, the ANFIS model. Software program code was developed to perform the analysis. The Figure 7 graph between the data set patterns and distance show the reading of dataset patterns by the ANFIS developed program. The scenarios considered in building the ANFIS model inputs and an output is shown in network (Figure 8). The parameters Distance and depth are given as membership functions. And ANFIS predicted the velocities with 10 rules which is depicted in Figure 8.

Results and discussions

In this study, the average velocities were calculated from the measured velocities. The velocity is zero at the solid boundaries and gradually increases with the distance from the boundary. Figure 9 shows the maximum velocity occurs at certain distance below the free surface. Figure 10 shows the scattered graph plotted for the predicted values by ANFIS, explain that most of the data is under predicted and produce satisfactory results. And Figure 11 (a), (b) & (c) shows the comparison between the measured velocity contour plots and the predicted velocity contours which shows the satisfactory contour plots. From the contour plots, most of the data is under and predicted approximated the testing results of the proposed new ANFIS model are compared with the statistical parameters, and the correlation parameter R2 = 0.902. Such comparison reveals that, the proposed ANFIS model predicts fairly accurate velocity profiles compared with the experimental model. With the advancements in the computer hardware and software, the application of soft tools should not pose problems in even routine applications.

Conclusions

The study investigates the use of ANFIS technique as an alternative to more conventional velocity predicting contours, based on measured laboratory data of Gross Pollutant Trap. ANFIS is less time consuming and more flexible by employing fuzzy rules and membership functions incorporating with real-world systems. Existing predicted velocities are over-predicted the measured velocity values from laboratory data, confirming that ANFIS predicted velocities gave satisfactory performance. The proposed ANFIS model has least root mean square error, the highest coefficient of correlation (R2=0.902) and produces satisfactory results compared to the measured data.