ArticlePDF Available

Analysis of Methods used to Investigate Engineering Measured Experimental Data

December 2020

December 2020
9(12):41-49

DOI:10.47760/ijcsmc.2020.v09i12.006

Authors:

Mohamad Tariq Barakat

Al-Balqa' Applied University

Obtained results using CFANN

…

Figures - uploaded by Mohamad Tariq Barakat

Content may be subject to copyright.

Content uploaded by Mohamad Tariq Barakat

Content may be subject to copyright.

Content uploaded by Mohamad Tariq Barakat

Content may be subject to copyright.

Mohamad Tariq Mohamad Barakat, International Journal of Computer Science and Mobile Computing, Vol.9 Issue.12, December- 2020, pg. 41-49

Available Online at www.ijcsmc.com

International Journal of Computer Science and Mobile Computing

A Monthly Journal of Computer Science and Information Technology

ISSN 2320–088X

IMPACT FACTOR: 7.056

IJCSMC, Vol. 9, Issue. 12, December 2020, pg.41 – 49

Analysis of Methods used to Investigate

Engineering Measured Experimental Data

Mohamad Tariq Mohamad Barakat

Albalqa Applied University

Faculty of Engineering Technology

Jordan Amman

DOI: 10.47760/ijcsmc.2020.v09i12.006

Abstract: The results of many measured engineering experiments require an analysis and study process in order to determine the

relationships between the independent variables and the dependent variables, and accordingly, the accuracy of the values of the

adopted variables becomes an urgent necessity.

In this research paper we will take a sample of laboratory data and find the necessary relationships between the different variables,

and then we will present some models of artificial neural networks to find solutions to these relationships in order to make the

necessary comparisons to reach some of the necessary recommendations regarding the handling of measured data.

Keywords: Experimental data, regression, ANN, CFANN, FFANN, EANN, MSE.

1- Introduction

Many experiments and studies are carried out in many laboratories and engineering workshops, the results of which are a set of

measurements, which constitute values for a set of values of independent variables and values of a set of approved variables.

The values obtained in the laboratory as a result of the measurement process may be large, which creates difficulties in linking these

values with each other to find the necessary relationships that can be used with high accuracy to find the values of the variables

adopted by knowing the values of the independent variables.

To obtain the relationship between the independent and dependent variables in the measured data we can use regression model [1], [2],

this model can be easily solved using matlab, figure 1 shows an obtained experimentally results with 4 independent variables (x1, x2,

x3, and x4 and dependent variable y), the regressed out can be obtained applying the following code:

Mohamad Tariq Mohamad Barakat, International Journal of Computer Science and Mobile Computing, Vol.9 Issue.12, December- 2020, pg. 41-49

The mean square error between the measured output and the calculated one can calculate using equation1:

Figure 1: Results of measurements and the regressed output

From figure 1 we can see that solving the regression model for the experimental data provide a very small error which can be

accepted, in the following parts we will check whether artificial neural networks [19], [20], [21] gives better results.

Mohamad Tariq Mohamad Barakat, International Journal of Computer Science and Mobile Computing, Vol.9 Issue.12, December- 2020, pg. 41-49

2- Artificial Neural Networks

Artificial neural network (ANN) [8], [9] is a powerful mathematical model which can be used to solve various problems such as

digital images [3], [4], [5] classification, speech recognition [6], [7], function approximation and many other problems.

ANN is asset of fully connected neurons [10], [11] which are arranged in one or more layers [12], each neuron is a computational cell

which as shown in figure 2 performs summation of the products of the weights and inputs [13], then according to the selected

activation function calculates the cell output[14].

Figure 2: Neuron operations

In order to get ANN desired output it must be trained [15], [16], each training cycle is done in two ways as shown in figure 3: the

feedforward phase in which the neurons outputs are calculated, starting from the input layer, and a backward phase to find the error,

and according to obtained error adjust the weights [17], [18]. Figures 4 and 5 show an example of one training cycle calculation.

Figure 3: Training cycle

Mohamad Tariq Mohamad Barakat, International Journal of Computer Science and Mobile Computing, Vol.9 Issue.12, December- 2020, pg. 41-49

Figure 4: Training cycle calculations

Figure 5: Training cycle calculations (continued)

ANN can be used in different variations, mostly and widely used are feedforward (FFANN), cascade-forward (CANN and Elman

(EANN) neural networks [10], [11]. In FFANN (see figure 6) the neurons are organized in layer and each neuron is fully connected to

neurons in the previous layer.

Figure 6: FFANN

Mohamad Tariq Mohamad Barakat, International Journal of Computer Science and Mobile Computing, Vol.9 Issue.12, December- 2020, pg. 41-49

In CFANN the inputs weights are connected to the other layers as shown in figure 7.

Figure 7: CFANN

Figure 8: EANN

In EANN the outputs of the hidden layer feed the input layer and form a context delay nodes.

3- Implementation and Experimental Results

To use ANN as a computation tool we have to follow the following steps [8], [9]:

Select the input dataset.

Normalize the dataset if needed.

Select the target output (outputs).

Create and build ANN architecture by defining the number of layers, defining the number of neurons in each layer, defining

the activation function for each layer [10], [11], selecting ANN type.

Setting all the weights to zeros by initializing the net.

Setting the number of training cycles.

Setting the error to zero.

Training the net.

Run the net.

Checking the error and the calculated outputs, if acceptable save ANN to be used later as a computational tool, else modify

the architecture OF ANN by adding extra hidden layer, or changing the activation function, or by increasing the number of

training cycles and train ANN again.

Mohamad Tariq Mohamad Barakat, International Journal of Computer Science and Mobile Computing, Vol.9 Issue.12, December- 2020, pg. 41-49

The following matlab code was written and it will be used for results analysis:

Mohamad Tariq Mohamad Barakat, International Journal of Computer Science and Mobile Computing, Vol.9 Issue.12, December- 2020, pg. 41-49

The above code was implemented varying ANN type, by replacing newcf to newff to use FFANN and to newelm to use EANN.

Figure 9 shows the obtained results using CFANN; figure 10 shows the obtained results using CFANN, while figure 11 shows the

obtained results using EANN.

Figure 9: Obtained results using CFANN

Mohamad Tariq Mohamad Barakat, International Journal of Computer Science and Mobile Computing, Vol.9 Issue.12, December- 2020, pg. 41-49

Figure 10: Obtained results using FFANN

Figure 11: Figure 10: Obtained results using EANN

From the obtained results we can see that using CFANN gave the best results by minimizing the value of MSE, and the tools sorted

according to the accuracy will be as follows:

- CFANN

- Regression method

- FFANN

- EANN

And here we can highly recommend CFANN to be used as a computational tool to find the relationship between any measured

experimental data.

Mohamad Tariq Mohamad Barakat, International Journal of Computer Science and Mobile Computing, Vol.9 Issue.12, December- 2020, pg. 41-49

Conclusion

A measured experimental data was collected and the relationship between the independent variables and the dependent ones was

obtained using regression model. The results of the regression model were good and the can be acceptable.

An experiments were done to see whether we can enhance the accuracy of the regression model. Different computational tools using

ANN were built and tested. It was shown that using cascade-forward ANN as a computational tool gave an excellent enhancement,

thus it can be highly recommended.

References

[1]. Akram A Moustafa, Ziad A Alqadi, Eyad A Shahroury, Performance evaluation of artificial neural networks for spatial data analysis, WSEAS Transactions

on Computers, vol. 10, issue 11, pp. 115-124, 2011.

[2]. K Matrouk, A Al-Hasanat, H Alasha'ary, Z Al-Qadi, H Al-Shalabi, Speech fingerprint to identify isolated word person, World Applied Sciences Journal, vol.

31, issue 10, pp. 1767-1771, 2014.

[3]. Ziad Alqadi, Bilal Zahran, Jihad Nader, Estimation and Tuning of FIR Lowpass Digital Filter Parameters, International Journal of Advanced Research in

Computer Science and Software Engineering, vol. 7, issue 2, pp. 18-23, 2017.

[4]. Khaled M Matrouk, Haitham A Alasha'ary, Abdullah I Al-Hasanat, Ziad A Al-Qadi, Hasan M Al-Shalabi, Investigation and Analysis of ANN Parameters,

European Journal of Scientific Research, vol. 121, issue 2, pp. 217-225, 2014.

[5]. Haitham Alasha'ary, Abdullah Al-Hasanat, Khaled Matrouk, Ziad Al-Qadi, Hasan Al-Shalabi, A Novel Digital Filter for Enhancing Dark Gray Images,

European Journal of Scientific Research , pp. 99-106, 2014.

[6]. Majed O Al-Dwairi, Ziad A Alqadi, Amjad A Abujazar, Rushdi Abu Zneit, Optimized true-color image processing, World Applied Sciences Journal, vol. 8,

issue 10, pp. 1175-1182, 2010.

[7]. Jamil Al Azzeh, Hussein Alhatamleh, Ziad A Alqadi, Mohammad Khalil Abuzalata, Creating a Color Map to be used to Convert a Gray Image to Color

Image, International Journal of Computer Applications, vol. 153, issue 2, pp. 31-34, 2016.

[8]. Jamil Al-Azzeh, Ziad Alqadi, Mohammed Abuzalata, Performance Analysis of Artificial Neural Networks used for Color Image Recognition and

Retrieving, International Journal of Computer Science and Mobile Computing, vol. 8, issue 2, pp. 20 – 33, 2019.

[9]. Dr. Ghazi. M. Qaryouti, Prof. Ziad A.A. Alqadi, Prof. Mohammed K. Abu Zalata, a Novel Method for Color Image Recognition, IJCSMC, Vol. 5, Issue. 11,

pp.57 – 64, 2016

[10]. Dr. Amjad Hindi Dr. Majed Omar Dwairi Prof. Ziad Alqadi, PROCEDURES FOR SPEECH RECOGNITION USING LPC AND ANN, International

Journal of Engineering Technology Research & Management, vol. 4, issue 2, pp. 48 -55, 2020.

[11]. Amjad Y. Hindi, Majed O. Dwairi, Ziad A. AlQadi, Creating Human Speech Identifier using WPT, International Journal of Computer Science and

Mobile Computing, vol. 9, issue 2, pp. 117 – 123, 2020.

[12]. Ziad Alqadi, Aws Al-Qaisi, Adnan Manasreh, Ahmad Sharadqeh, Digital Color Image Classification Based on Modified Local Binary Pattern

Using Neural Network, IRECAP, vol. 9, issue 6, pp. 403-408, 2019.

[13]. Ziad AlQadi, Yehya Abded Allatif, Musbah J Aqel, A Proposed methodology for image objects recognition using Artificial neural networks, IJCSS, vol. 3,

issue 1, pp. 49-56 , 2011

[14]. Ziad A AlQadi Amjad Y Hindi, O Dwairi Majed, PROCEDURES FOR SPEECH RECOGNITION USING LPC AND ANN, Internationa l Journal

of Engineering Technology Research & Management, vol. 4, issue 2, pp. 48-55, 2020.

[15]. Dr. Amjad Hindi, Dr. Majed Omar Dwairi, Prof. Ziad Alqadi, Analysis of Procedures used to build an Optimal Fingerprint Recognition System,

International Journal of Computer Science and Mobile Computing, vol. 9, issue 2, pp. 21 – 37, 2020.

[16]. Eng. Sameh S. Salma Prof. Ziad A. AlQadi, Eng. Ahmad S. AlOthma, Eng. Mahmoud O. Alleddaw ,Eng. Mohammad Ali Al-Hiar , Eng. Osama T. Ghaza,

Investigation of ANN Used to Solve MLR Problem, IJCSMC, vol. 8, issue 5, pp. 38 – 50, 20, 2019.

[17]. Eng. Sameh S. Salman Prof. Ziad A. AlQadi, Eng. Ahmad S. AlOthman, Eng. Mahmoud O. Alleddawi, Eng. Mohammad AlHiary, Eng. Osama T.

Ghazal, Building Accurate and Efficient Color Image Recognizer, IJCSMC, vol. 8, issue 4, pp. 127 – 135, 2019.

[18]. Jamil Al-Azzeh, Ziad Alqadi, Mohammed Abuzalata, Performance Analysis of Artificial Neural Networks used for Color Image Recognition and

Retrieving, International Journal of Computer Science and Mobile Computing, vol. 8, issue 2, pp. 20-33, 2019.

[19]. Belal Ayyoub, Ahmad Sharadqh, Ziad Alqadi, Jamil Al-azzeh, Simulink based RNN models to solve LPM, International Journal of Research in Advanced

Engineering and Technology, vol. 5, issue 1, pp. 49-55, 2019.

[20]. Abdullah Al-Hasanat, Haitham Alasha'ary, Khaled Matrouk, Ziad Al-Qadi, Hasan Al-Shalabi, Experimental Investigation of Training Algorithms

used in Back propagation Artificial Neural Networks to Apply Curve Fitting, European Journal of Scientific Research, pp. 328-335, 2014.

[21]. Ziad AA Alqadi, Investigation and analysis of mat lab models used to solve second order linear ODE, Internationa l Journal on Numerical and

Analytical Methods in Engineering, vol. 2, issue 1, 2014.

ResearchGate has not been able to resolve any citations for this publication.

Creating Human Speech Identifier using WPT

Article

Full-text available

Feb 2020

Human speech signal is widely used in various vital applications such as computer security system, speech signal usually has big size, which make it difficult to identify the speech by direct matching, sample by sample, so the process of using an efficient and accurate method of creating speech print is an important task in speech identification process. In this research paper we will introduce a wavelet packet decomposition, the generated wavelet packet tree will be analyzed in order to create a unique features (identifier) to each speech signal, we will show how wavelet packet decomposition is flexible in creating speeches identifiers by providing a variety of selections, each features selection will lead to generate a unique and small in size identifier, which can be used later on in any application requiring human speech recognition.

Simulink based RNN models to solve LPM

Article

Full-text available

Jan 2018

Recurrent artificial neural network has shown great success because of its interesting feature, each neuron in this type of artificial neural acts a computational element with memory, providing a neuron with ability to store the calculated output for the next stage of computation. In this paper we will introduce a methodology base on RNN features, this methodology will be used to build and run simulink based RNN models to solve linear programming problems, the introduced simulink models will be implemented in order to find the optimal values of LPM.

Building Accurate and Efficient Color Image Recognizer

Article

Full-text available

Apr 2019

Digital color image recognition is one of the most important tasks used in digital image processing, due to the various applications requiring this task. Color image recognition system deals with huge data, thus minimizing the memory space and recognition time will be an important issue. In this paper a method of color image features extraction will be introduced, the extracted features will be minimized and will be unique for each image. A data set if color images features will be built and passed to artificial neural network to be used for recognition purposes,, deferent ANN architectures will be trained , tested and implemented , the obtained results will be used as a guide for selecting the suitable , efficient and accurate ANN.

Performance Analysis of Artificial Neural Networks used for Color Image Recognition and Retrieving

Article

Full-text available

Feb 2019

Image recognition refers to methods and technologies that identify digital images such as places, logos, people, objects, buildings, and several other variables in images. Users are sharing vast amounts of data through applications, social networks, and websites. The large volume of digital data is being used by companies to deliver better and smarter services to the people accessing it. In this paper we will introduce a methodology of image recognition and retrieval process based on using ANN as a recognizer capable to identify digital image index. ANN will be implemented and tested in order to select an optimal ANN which must give the accurate image index with minimum MSE and maximum PSNR.

Estimation and Tuning of FIR Lowpass Digital Filter Parameters

Article

Full-text available

Mar 2017

Finite impulse response (FIR) digital filters are known to have many distinguishable features such as stability, linear phase characteristic at all frequencies and digital implementation as non-recursive structures. FIR filter design can be considered as an optimization problem. In this paper an estimation method of FIR filter parameters is proposed. The method relies on establishing a relationship between the signal input parameters and the filter parameters. An FIR lowpass filter was implemented and tested using various parameter values. The results showed efficient performance characteristics of FIR lowpass filters. Keywords— FIR lowpass filter, transition bandwidth, sampling frequency, window length, filter order, and stopband attenuation. I. INTRODUCTION Designing digital filters involves the determination of a set of filter coefficients to meet a set of design specifications. Digital filters can be classified in two categories: finite impulse response (FIR) filters and infinite impulse response (IIR) filters. By varying the weight of the coefficients and number of filter taps, virtually any frequency response characteristics can be realized with an FIR filter. FIR filter is an attractive choice because of the ease in design, linear phase shift property and stability [[1], [2]]. Designing an FIR digital filter require specifying passband, stopband, and transition band. In passband, frequencies are needed to be passed unattenuated. In stopband, frequencies needs to be passed attenuated. Transitionband contains frequencies which are lying between the passband and stopband. Therefore, the entire frequency range is split into one or even more passbands, stopbands, and transition bands. In practical, the magnitude is not necessary to be constant in the passband of a filter. A small amount of ripple is usually allowed in the passband. Similarly, the filter response does not to be zero in the stopband. A small, nonzero value is also tolerable in the stopband as shown in Fig 1. The transition band of the filter as shown in Fig.1 is between the passband and the stopband. The frequency ω p denotes the edge of the passband, and the band-edge frequency ωs defines the edge of the stopband. So, the difference of ω s and ω p is the width of the transition band, i.e. ω t = ω s-ω p. The ripple in the passband of the filter is denoted as δp, and the magnitude of the filter varies from 1-δp to 1+ δp. δs are the ripple in the stopband. Usually we use a Logarithmic scale to show the frequency response, hence, the ripple in the passband is 20log10δp dB, and the ripple in the stopband is 20log10δs dB.[[3], [4]] Different methods are used to design a lowpass filter such as: window method, frequency sampling method, and optimization method. In window method, a truncated ideal lowpass filter with a certain bandwidth is generated, and then we use a chosen window to get certain stopband attenuation. The filter order L can be adjusted to meet a specified roll-off rate in the transition band. Any finite-length (order) of the ideal lowpass impulse response may be considered as the product of the infinite-length lowpass impulse response and a window function W, which has a finite number of contiguous nonzero-valued samples [[1]-[4]]

Investigation and analysis of mat lab models used to solve second order linear ODE

Article

Full-text available

May 2014

Ziad A.A Al Qadi

Ordinary differential equations (odes) are used in realistic applications such as applied mathematics, physics, metrology, and engineering. Because of the big efforts needed to solve second order linear ODES, some mat lab methods and functions were investigated , the result of these methods were studied and some judgment were done regarding the results accuracy and implementation time. ODEs are widely used to describe some plants, the behaviors of a plant were described, and the response signal was analyzed in order to improve the system performance by minimizing the absolute error.

Creating a Color Map to be used to Convert a Gray Image to Color Image

Article

Full-text available

Nov 2016

Ziad A.A Al Qadi

We introduce a simple and efficient method for coloring the gray-scale images into the color images with the existing color map or with creating a new one. The general problem of adding chromatic values to a gray-scale image has no exact solution, the current approach attempts to provide a method to minimize the amount of human labor required for this task, rather than choosing RGB colors to individual components, we convert the entire gray-scale image to color by creating a color map then the color map is to be used to color the gray scale image. We show that this simple method can be successfully applied to a variety of gray scale images and it can be applicable in deferent applications

Investigation and Analysis of ANN Parameters

Article

Full-text available

Jan 2014

Different ANN architectures with a range of values for a set of network parameters will be tested and analyzed in order to get the relationship between the performance factors and ANN parameters; it will be shown how to use ANN parameters to evaluate ANN training time and absolute error.

A Novel Digital Filter for Enhancing Dark Gray Images

Article

Full-text available

Jan 2014

Enhancing gray image contrast is the first step of image preprocessing. A novel filter is proposed, implemented and tested using various dark and unclear images. It is seen that the proposed filter enhanced the image contrast by normalizing the image histogram and moving the average pixels value to the middle of the maximum gray level without affecting the clear images.

Speech Fingerprint to Identify Isolated Word-Person

Article

Full-text available

Jan 2014

Analysis of Methods used to Investigate Engineering Measured Experimental Data

Figures

Recommended publications

Building Face Recognition System (FRS)

Building Face Recognition System (FRS)

Simple Procedures to Create HSCS

PROCEDURES FOR SPEECH RECOGNITION USING LPC AND ANN