Prediction of Noise Level in Ota Metropolis Using Artificial Neural Network

Capturing noise data is laborious, time-consuming, expensive and dangerous due to the exposure of the investigator to the menace. Also, appropriate software, computational skill and computational time are also required before the captured data could be of any use. In this work, an artificial neural network (ANN) was deployed to learn and train noise data in Ota Metropolis. Data were captured from forty-one (41) locations for the morning, afternoon and evening in Ota Metropolis. ANN with Levenberg Marquardt algorithm and architectural configuration of 2-21-9 (input-hidden neuron-output) was used to predict noise descriptors for Ota Metropolis with 73% accuracy. The two input variables were the latitude and longitude of the location measured in degrees while the nine output variables are the noise descriptors such traffic noise index (TNI), noise pollution level (LNP), and average equivalent noise levels (LAeq) computed for each selected location for morning, afternoon and evening periods. The results could be used in mobile applications, Google Earth and other platforms to guide residence dwellers, travellers, industrialists and technocrats in selecting travelling routes, choice of apartment location and use of appropriate personal protective equipment (PPE) in unavoidable noisy locations.


Introduction
Urbanization and industrialization brought water, land, soil, air, radioactive and noise pollution with them.This paper is on noise pollution.Sound is produced when vibration passes through air or other medium like water.It is termed 'noise' when it is undesired or rather prevented a wanted sound.Noise is a major factor that reduces the quality of life due to its harmful psychological and physiological effects on human beings [1].Hence, adequate information on noise levels is a necessity for informed decisions in industrial, residential, commercial, recreational, construction, and transportation (airway, waterway, railway, highway) areas.However, harvesting such noise data in any area is usually cumbersome.Foremost, the investigator is exposed to the same pollution in the morning, afternoon and evening while taking the noise readings.This requires time, energy and money.Also, the raw data is of little use until hardware, appropriate software and manhours are deployed in the analysis.Oyedepo et al. [2], conducted extensive fieldwork and analysis of noise data for Ota Metropolis in Nigeria.The work specified latitude and longitude for forty-one different locations with their noise pollution data.The aim of this present work is to use an artificial neural network (ANN) to validate, model and forecast noise pollution data for Ota Metropolis.
Much work has been done using ANN.Mohsen et al. [3] used ANN to study composite produced from A356 matrix and B4C powder.Mechanical properties of the composites were predicted and used in developing software code.Babalola et al. [4] applied ANN to predict the electrical and mechanical properties of aluminium and silicon carbide composite produced using the stir casting method.Ultimate tensile strength (MPa), tenacity at fracture (gf/tex), Modulus (N/mm^2), yield strength (MPa), hardness (HV), time at fracture (s), electrical conductivity(MΩ/m), and tensile stress were predicted in the work.Ayoola et al. [5] used response surface methodology (RSM) and ANN to analyse the production of biodiesel from waste groundnut oil.Few researches have been carried out on the application of ANN in noise data analysis.Steinbach and Altinsoy [6] employed ANN to predict the annoyance level of electric vehicles from sound level, roughness, fluctuations in strength, tonality, sharpness and loudness.Bravo-Moncayo et al. [7] used ANN, multiple linear regressions (MLR) and support vector machines (SVM) to predict traffic-noise annoyance levels and found the ANN model to be the best.Hamad et al. [8] on his part modelled roadway traffic noise from the edge of the road, vehicle volume, average speed and roadway temperature using ANN.It was noted that investigators have not linked noise level to the latitude and longitude of the studied location.This is attempted in this work.

Noise level and health issues-WHO standards
Noise pollution is not without its adverse effects.Exposure of the ear to noise levels of 140 dB (A) and above may permanently damage the tympanum (eardrum) in the middle ear (Fig. 1).Noise levels below 140 dB may result in hearing impairment, hearing loss or temporal auditory fatigue.Non-auditory effects include: interference; stress and annoyance; physiological and behavioural effects; occupational hazards and accidents.The allowable noise level stipulated by the World Health Organisation (W.H.O.) for different localities and permissible dosages is shown in Tables 1 and 2 respectively.

Use of ANN to validate, model and forecast
Artificial neural network is a novel mathematical tool motivated by the biological nervous system.It can be engaged to solve myriads of difficult engineering and scientific problems.It is comparable to the natural biological system such that it could acquire information from existing examples and use this information as a guide to obtain existing inherent patterns [12].In this paper, ANN was used to model, validate and forecast noise descriptors generated for Ota Metropolis.Modelling was done with the use of ANN feed-forward networks (nftool) deep learning toolbox 12.1 in MATLAB R2019a by MathWorks, Inc. Levenberg Marquardt backpropagation algorithm and the ANN architectural configuration of the input layer, hidden layer and target (output) layer was adopted for the analysis.Five ANN configurations of 2-10-9 (input-hidden neuron-output),2-15-9, 2-20-9,2-21-9, and 2-30-9 shown respectively in Figures 2-6 were used during the preliminary investigation.However, the ANN topology of 2-21-9 (input-hidden neuron-output) gave the best regression results and was adhered to henceforth.The input had two variables representing the latitude and longitude values of each location.The name with latitude and longitude for all the forty-one locations used in this work is shown in Table 3.There are twenty-one hidden neurons while the output (target) has nine variables (Table 4) representing traffic noise index (TNI-morning, TNI-afternoon, and TNIevening), pollution noise level (LNP-morning, LNP-afternoon, and LNP-evening) and average noise levels (LAV-morning, LAV-afternoon, and LAV-evening) (Fig. 7).ANN was able to predict and generate noise levels for all the forty-one locations in Ota metropolis.The ANNpredicted data are shown in Table 5.

Results
In this study, noise levels of forty-one (41) selected locations in Ota Metropolis, Nigeria were predicted using ANN.ANN used twenty-nine data points representing 70% for training purposes, six data points representing 15% for validation and a further six sets representing 15% for testing purposes.There was convergence after twenty iterations.Regression R Values measured the correlation between outputs and targets.An R-value of 1 means a close relationship, and 0 a random relationship.An overall regression coefficient of 0.73406 (Figure 5b) was obtained in this analysis and ANN-predicted noise data are shown in Table 5.
The study also conducted an F-test on the two-tailed probability that the variances in measured and predicted data are not significantly different.The obtained low value of 0.262397511 shows that variances are generally low. Figure 8 shows the error histogram with significant zero (0) errors.

Conclusion
In conclusion, ANN has predicted accurately the measured noise descriptors at the selected locations in Ota Metropolis.Based on the results of this study, the authors recommended that ANN for noise descriptors in urban areas in developing countries like Nigeria be developed and inculcated in mobile apps, Google Earth and other platforms to guide residence dwellers, travellers, industrialists and technocrats in selecting travelling routes, choice of apartment location and use of appropriate personal protective equipment (PPE) in unavoidable noisy locations.

Figures 9 -
show experimental data and ANN-generated data for traffic noise index (TNImorning, TNI-afternoon, and TNI-evening), pollution noise level (LNP-morning, LNPafternoon, and LNP-evening) and average noise levels (LAeq-morning, LAeq-afternoon, and LAeq-evening).It was noted from the experimental data obtained in the field, that Ota metropolis is adjudged a noisy city and after the deployment of ANN to train, validate and test the experimental data, it was further corroborated without any exception in all the forty-one locations and all nine variables.This is another success story of the versatility of ANN in artificial intelligence (AI).

Figure 9 :Figure 10 :Figure 11 :Figure 12 :Figure 13 :
Figure 9: Measured and ANN Predicted TNI in the Morning for Ota Metropolis

Figure 14 :Figure 15 :Figure 16 :Figure 17 :
Figure 14: Measured and ANN Predicted LNP in the Evening for Ota Metropolis

Table 3 :
Input-Latitude and Longitude Coordinates for the Locations

Table 5 :
ANN Predicted Data-traffic noise index, pollution noise level and average noise levels.