google2c7a87877ffe6fb7.html

A Machine Learning Model for Average Fuel Consumption in Heavy Vehicles

In this paper author is describing concept to predict average fuel consumption in heavy vehicles using Machine Learning Algorithm such as ANN (Artificial Neural Networks). To predict fuel consumption author has extracted 7 predictor features from heavy vehicle dataset such as

num_stops, time_stopped, average_moving_speed, characteristic_acceleration, aerodynamic_speed_squared, change_ in_kinetic_energy, change_in_potential_energy, class

Above seven features are recorded from each vehicle travel up to 100 kilo meters like number of times vehicle stop, total stopped time taken etc. All this values are collected from heavy vehicle and use as dataset to train ANN model. Below are some value from above seven predictor features.

num_stops, time_stopped, average_moving_speed, characteristic_acceleration, aerodynamic_speed_squared, change_ in_kinetic_energy, change_in_potential_energy, class

7.0,7.0,93.0,34,8.4,4,25.6,9

7.0,7.0,91.0,34,8.3,4,25.7,9

8.9,8.9,151.0,26,10.9,6,15.1,12

9.3,9.3,160.0,25,11.3,6,13.7,13

8.4,8.4,158.0,25,11.2,6,13.8,13

All bold names are the dataset column names and all double values are the dataset values for each vehicle. Last column will be consider as class name which represents fuel consumption for that vehicle. Entire dataset will be used to train ANN model and whenever we give new record then ANN algorithm will apply train model on that test record to predict it average fuel consumption.

Below are some test records

num_stops, time_stopped, average_moving_speed, characteristic_acceleration, aerodynamic_speed_squared, change_ in_kinetic_energy, change_in_potential_energy

7.0,7.0,93.0,34,8.4,4,25.6

7.0,7.0,91.0,34,8.3,4,25.7

8.9,8.9,151.0,26,10.9,6,15.1

9.3,9.3,160.0,25,11.3,6,13.7

8.4,8.4,158.0,25,11.2,6,13.8

In above test data class value as fuel consumption is not there and when we applied this test record on ANN model then ANN will predict fuel consumption class value for that test record. Entire train and test data available inside ‘dataset’ folder.

The ANN model is developed by using duty cycle’sdataset collected from asingle truck, with an approximate mass of 8, 700kg exposed to avariety of transients including both urban and highway traffic inthe Indianapolis area. Data was collected using the SAE J1939standard for serial control and communications in heavy dutyvehicle networks.

Abstract

In this paper we used vehicle travel distance rather than the traditional time period whendeveloping individualized machine learning models for fuel consumption.This approach is used in conjunction with seven predictorsderived from vehicle speed and road grade to produce a highlypredictive neural network model for average fuel consumption inheavy vehicles. The proposed model can easily be developed anddeployed for each individual vehicle in a fleet in order to optimizefuel consumption over the entire fleet. The predictors of the modelare aggregated over fixed window sizes of distance travelled. Differentwindow sizes are evaluated and the results show that a 1 kmwindowis able to predict fuel consumptionwith a 0.91 coefficient ofdetermination and mean absolute peak-to-peak percent error lessthan 4% for routes that include both city and highway duty cyclesegments.

ANN Working Procedure

To demonstrate how to build anANN neural network based image classifier, we shall build a 6 layer neural network that will identify and separate one image from other. This network that we shall build is a very small network that we can run on a CPU as well. Traditional neural networks that are very good at doing image classification have many more parameters and take a lot of time if trained on normal CPU. However, our objective is to show how to build a real-world convolutional neural network using TENSORFLOW.

Neural Networks are essentially mathematical models to solve an optimization problem. They are made of neurons, the basic computation unit of neural networks. A neuron takes an input (say x), do some computation on it (say: multiply it with a variable w and adds another variable b) to produce a value (say; z= wx+b). This value is passed to a non-linear function called activation function (f) to produce the final output(activation) of a neuron. There are many kinds of activation functions. One of the popular activation function is Sigmoid. The neuron which uses sigmoid function as an activation function will be called sigmoid neuron. Depending on the activation functions, neurons are named and there are many kinds of them like RELU, TanH.

If you stack neurons in a single line, it’s called a layer; which is the next building block of neural networks. See below image with layers

neural network shown in Tensorflow tutorial

To predict image class multiple layers operate on each other to get best match layer and this process continues till no more improvement left.

Modules Information

This project consists of following modules

Upload Heavy Vehicles Fuel Dataset: Using this module we can upload train dataset to application. Dataset contains comma separated values.

Read Dataset & Generate Model: Using this module we will parse comma separated dataset and then generate train and test model for ANN from that dataset values. Dataset will be divided into 80% and 20% format, 80% will be used to train ANN model and 20% will be used to test ANN model.

Run ANN Algorithm: Using this model we can create ANN object and then feed train and test data to build ANN model.

Predict Average Fuel Consumption: Using this module we will upload new test data and then ANN will apply train model on that test data to predict average fuel consumption for that test records.

Fuel Consumption Graph: Using this module we will plot fuel consumption graph for each test record.