Вы находитесь на странице: 1из 11

E-powertrain Error detection using AI Techniques

E-powertrain error detection using AI Techniques

Abstract

Effective fault-handling in electric vehicle is complex that requires powertrain


controller should have the ability to adapt future coming fault. Fault prediction policies
are responds to possible changes in requirements and divide available resources. This
paper highlight application of AI techniques and predict fault & adaptive to error. Vehicle
should run by implementation of an adaptive fault-tolerance (AFT) using an artificial
intelligence. The chapter also discusses the driver behavior while driving e-car that cause
the system failure and how to avoid it. Vehicle driver behavior monitoring necessary and
train him if his driving behavior is abnormal.

Keyword: Electric Power train adaptive fault, artificial intelligence, sensors, acceleration, assistance
system, error prediction, driver behavior assistance.

Introduction

Error-overcoming technology is a power to do of a knowledge processing


Machine system. in electronic system or network many more component fails so need
to tolerate error and overcome the difficult situation from breakdown of car. With
knowledge of system parameter , it is very easy to predict error as well before coming
permanents failure

The use of artificial intelligence techniques. The following activity has to predict error :
1) Continuously monitor the important parameter

1
E-powertrain Error detection using AI Techniques

2) feed the parameted to algorithm

3) train the algorithm

4) decision of algorithm and warn user whether condition are true or false

The e-powertrain focuses the issue that requires the system should work dynamic and
error free without any internal or external disturbance. In latest tehcnology continental,
bosch did lot of work regarding epowertrain driving assistance system(DAS), off road
speed predition, Maintain speed while approaching hill area , and maintain traction force
during hill downside.

Driving behaviour need to understand. It will be very easy and require less effort to drive
the car. Also need to study the behaviour in extreme cases by using visual camera, snesor
and predict need to adapt steadily fast , slow , distraction , alcoholic, hard braking,
dangerous manuevers, overspeeding, off-road driving.
The main three key parameter that can elevates the use of electric powertrain as
1) Range: It realises the running cost should be very low
2) Power: Sufficient power is possible to store to cover the max or expected distance
3) Efficiency: Enables the maximum use of energy with least loss.
The current status and experience of the available use of resources is possible only when
max use of an artificial intelligence(AI). It help to divide and allocate all resources at max
efficiecny. For max utilisation we need to study all side benefits and initiate the necessity
of using artificial intelligence. Theory

Definition of artificial intelligence:

Artificial intelligence (AI) is an area of computer science is the creation of intelligent


machines that work and react like humans. Driving behavior detection using artificial
intelligence: By applying knowledge gained from recorded data, it’s very easy to

2
E-powertrain Error detection using AI Techniques

understand the behavior of driver. e.g. at the same time if driver applies speed and brake
to vehicle then its very abnormal condition that damages the transmission system very
badly
Types of AI :
.

Artificial intelligence(AI)

Machine learning(ML)

Deep learning

Image 1. AI sub-field

Predictive maintenance:
The data of vibrations, pressures and temperatures of an engine recorded over time is
evaluated with AI techniques. This makes it possible to detect an error that is becoming
established early. A vehicle can be sent to a service shop promptly, or a system can be
designed to be error-tolerant in advance.

Experimental setup for e-powertrain system:

3
E-powertrain Error detection using AI Techniques

Imge2. Onboard with vibration sensor mount on transmission.

Black box architecture

Image 3: data logger

Data logger details:

Data logger is designed using microcontroller with ethernet controller. All channel is
logged via Modbus communication. Datalogger build for the use of monitoring driver
behavior, vehicle speed, vehicle loading, steering angle, acceleration, braking behavior.
Corresponding all channel configured using C programming. Bellow image (no.4) shows
the behavior of recorded data. From the recorded data it is possible to uncover all
required information and understand the behavior of driver. e.g. driver changing gear in

4
E-powertrain Error detection using AI Techniques

dull speed, that’s very harmful to transmission system may cause permanent damage to
system.

Image 4: Data record and logged in CSV format

After viewing the recorded graph of transmission system vibration as bellow that
transmission vibration goes increasing frequently when gear change.

Image 5. Vibration of transmission system

Application of AI technique to detect transmission system.

5
E-powertrain Error detection using AI Techniques

Data analysis step by step: after recording data its very necessity to organize and retrieve
data

Step 1: The recorded data is like heavy truck should be moved to cloud for leering and
load train with the likely behavior with fault data set. For instigation try to reduce specific
dataset and reduce complexity of the dataset. Now data loaded to learning algorithm.
Using large data set is good mindset but it should be processed in the right way

Step 2: Define problem statement with required dataset.: Try to formulate with known
recorded parameter how data set will help very easy? Like engine oil pressure and
temperature are correlated. If pressure down engine cooling will affect mean temperature
will increases automatically. Then need to collect data regarding temperature, vibration
of machine. As shown recorded graph the data set collection.

Step 3: Establish the connection for data recording. I have shown example to record data
for using Arduino and ethernet shield H/W and data logged in .csv format.

Step 4: Set the data range with tolerance. That recorded data feed to algorithm in
continuous mode. For occurrence, data range is 0.0 to 0.5, range should not beyond like
0.6.

Image 4. Recorded data vs peak change value in vibration for 30days

Step 5: Reduce the data set

6
E-powertrain Error detection using AI Techniques

Analyze the data set which help to save time for begins, or with large data set, expert
should not lead to wrong decision. For this there are two method used to reduce data size
a) top specimen, b) grouping. later these methods help to understand sample model result
yields expected result or not.

Step 6: feature selection: While formulating the problem need to take decision like what
type of algorithm should be used. As shown in image AI classification. Those type are
supervised learning, unsupervised learning, deep learning.

Image 5: The scale down data

The classification done with regards to expected answer. For time series data
prediction is possible when previous old value is known. Then take decision from the
available graph that driver is driving properly or not. Is there operating condition of
worsening transmission?

Cluster: find answer with classes and no. of rules. The difference between cluster and
classification is a no. of various group which will part of the problem statement. For this
need to select suitable cluster with characteristic information.

Classification: group classification done like group of machines, car, bicycle. Will get
answer in binary format yes or no.

Data cleaning also important part of Feature selection. Data cleaning is the first and most
important step of your model designing.

7
E-powertrain Error detection using AI Techniques

Regression: It estimate yield the numeric value. It finds the difference in actual result
and expected result.

Ranking: in this method rank decide the importance of result and organize answer
structurally. i.e. google search result show most relevant website first with input search
query.

Feature selection help to reduce the overall error or overfitting, improves accuracy and
reduce training time.

Step 7: data scaling:

It refers to change data variation likely to move averaging reduce min max value and
plot only peak values that have significant values. Also improves the shape of distribution
achieved with change in mean value of each feature. and scaleup or down by it with non-
constant feature by their standard deviation. The scaling functions are normalization,
discretization, categorical selection

Image6. Moving average data value

8
E-powertrain Error detection using AI Techniques

Bearing sustainability: Bearing manufacturing company must pass license of the stress
test. Energy consumption totally proportional to energy transfer transmission.

Analysis of gear transmission fault detection:

Adaptive learning: Recorded data feed to algorithm by online. This data collected
from industrial data loggers’ forces to use of data science and machine learning
techniques. With know data pattern in real time and previous time data with repeated
scenario creates moving forward rule. It means learning machine over time, if data fits
out of the repeated pattern system detect the error and alarm the use called adaptive
system

Image. Transmission Bearing abnormal vibration

9
E-powertrain Error detection using AI Techniques

Results and discussion:

Strong fault hides the weak fault. detecting weak fault those are necessary to control
helps and understand the behavior of vehicle driver. also setup of experiment explained
and how to deal with logged data detailed well.

Acknowledgements
I would like to express my special thanks of gratitude to my teacher Dr. Anil Gupta
(C-DAC Pune) as who gave me the golden opportunity to do this wonderful project on
the important topic of electric powertrain error detection, which also helped me in doing
a lot of research and i came to know about so many new things I am really thankful to
them.

References:

Journal article:
1. J. Liu and H. Peng, ‘‘A systematic design approach for two planetary gear split
hybrid vehicles,’’ Veh. Syst. Dyn., vol. 48, no. 11, pp. 1395–1412,
2. B. Conlon et al., ‘‘The next generation ‘voltec’ extended range EV propulsion
system,’’ SAE Int. J. Alternative Powertrains, vol. 4, no. 2, pp. 248–259, 2015
3. K. Okuda et al., ‘‘Development of multi stage hybrid transmission,’’ SAE Int. J.
Alternative Powertrains, vol. 6, no. 1, pp. 77–83, 03 2017.
4. S. Bai, J. Maguire, and H. Peng, Dynamic Analysis and Control System Design of
Automatic Transmissions, 1st ed. Warrendale, PA, USA: SAE International, 2013

10
E-powertrain Error detection using AI Techniques

5. Wollenberg B. F.: "Feasibility study for an energy management system intelligent


alarm process", IEEE Conference proceedings on Power Industry Computer
Applications, 1985, pp. 249-254.

11

Вам также может понравиться