Microsoft Word - 001.docx


 CHEMICAL ENGINEERING TRANSACTIONS  
 

VOL. 66, 2018 

A publication of 

 
The Italian Association 

of Chemical Engineering 
Online at www.aidic.it/cet 

Guest Editors: Songying Zhao, Yougang Sun, Ye Zhou
Copyright © 2018, AIDIC Servizi S.r.l. 
ISBN 978-88-95608-63-1; ISSN 2283-9216 

Application of Image Processing Algorithm in Chemical 
Production Technology 

Lina Chen 
Shijiazhuang University of Applied Technology, Shijiazhuang 050081, China 
187280350@qq.com 

With the development and application of industrial CT technology, the nondestructive testing of materials and 
the manufacturing errors of materials can be effectively monitored and controlled. The basic principle of 
Industrial CT technology is to adopt the ray bundle of passing through the measured object to get the fault 
projection data, and then with the image reconstruction method the projection data will be scanned to get two-
dimensional grey image. The traditional industrial CT technology is also facing the problem of image quality 
and image segmentation in image processing field. The industrial CT technology has been improved, in order 
to make expanders which can meet the needs of LNG liquefaction factories, and effectively control the quality 
of expanders. First of all, this paper analyzes the expander of LNG production equipment. Secondly, it 
introduces the industrial CT technology in detail. Then, it introduces the algorithm of image processing. Finally, 
combined with the new industrial CT technology, it conducts some experiments to verify that the proposed 
method can effectively detect the equipment like the core components of the expansion machine. 

1. Introduction 
With the development of technology, new materials and new equipment are more and more widely used in the 
development of technology. High strength steel is mainly used in the production of key parts. As a kind of 
clean and high quality fuel, natural gas has attracted more and more attention. However, as global 
environmental problems become increasingly serious, reducing the use of oil, coal and other energy sources 
is becoming the focus of national energy structure adjustment. Accelerating the use of clean energies like 
wind energy, nuclear energy, solar energy, and natural gas is the major energy use policy in China. The 
control of the expander rotor quality can effectively avoid chemical production problems caused by the 
production of expanders. However, any material is not perfect. There a variety of defects in the material. In 
addition to defects, manufacturing errors also lie in the producing process. There are two ways to find defects 
in the material. One is destructive, that is, cutting the material or other preparation methods to observe the 
internal defects. Another method can be used without damage to the material, which is also called non-
destructive testing method. 
Liquefied natural gas (LNG) can make up for the deficiency of pipeline natural gas because of its small size 
and convenient transportation. The volume of natural gas can be reduced by 600 times after being liquefied. 
Liquefied natural gas can play a great role both in large-scale transportation on the sea, and in the blank area 
of the natural gas pipeline on the land. Liquefied natural gas production cannot be separated without air 
separation equipment, whose core is the expander. The precision of the expansion machine have a great 
influence on the service life and the normal operation of the whole chemical production equipment. As the air 
separation equipment manufacturing in the world develops rapidly, the requirements of domestic users and 
large scale equipment have been improved greatly. Therefore, air separation equipment manufacturing 
capacity also has a great development. However, the study of turbo expander is always in the leading position 
in Europe and the United States, and Europe and the United States have mastered the research and 
manufacture of large and oversize turbine technology. In 1883, Tower (1883) finds the dynamic pressure 
phenomenon in the radial sliding bearing test of railway vehicles. The results of Tower are verified by 
Reynolds (1886) in 1886. Some researchers conduct integral transform on Reynolds equation under the 
assumption of infinite bearing, and obtain the analytical solution of simplified Reynolds equation. In 1929, 

                                

 
 

 

 
   

                                                  
DOI: 10.3303/CET1866137 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Please cite this article as: Chen L., 2018, Application of image processing algorithm in chemical production technology, Chemical Engineering 
Transactions, 66, 817-822  DOI:10.3303/CET1866137   

817



Michell (1929) proposes the theory of infinitely short bearing. Air Liquide is the only air separation equipment 
manufacturer in French, which at present has more than 130 subsidiaries in many other countries, and can 
make large air separation equipment with their flow in and above 93440m /h. British Oxygen Company is 
mainly engaged in the design and manufacture of air separation equipment, and it can make air separation 
equipment which can simultaneously extract argon, helium, neon, and krypton. Cryostar Company, a 
subsidiary of Linde Group, is a recognized manufacturer in the field of liquid expander and cryogenic liquid 
pumps in the world (Marcuccilli and Zouaghi, 2007; Josef., 2010; Asutosh, 2010). Dresser-Rand has been one 
of the outstanding suppliers in the field of energy recovery turbine expander. 
CT image processing is widely used in medicine. As a new research field, more and more scholars have paid 
attention to it. In 1895, the famous German physicist Roentgen discovers X –rays (Zheng et al., 2009; Tang 
and Chen, 2016). In 1970s, Computer tomography is born. CT creates a digital imaging precedent, but it is 
different from the ordinary X-ray imaging. CT shows the sectional anatomy images, whose density resolution 
is much higher than that of X-ray images. It is a revolution in medical imaging technology. CT has been widely 
used in the diagnosis of central nervous system, head and neck, lung, heart and large blood vessels and 
organ diseases (Wu, 2003). In 1970s, industrial CT technology is applied to the field of materials science. 
Industrial Computer Tomography (referred to as Industrial CT or ICT) is a kind of technique which uses rays to 
pass through the object and then conducts attenuation and reconstruction. It can obtain the lossless high-
resolution grey image in the detected object. The internal structure, material, defect and density distribution of 
the tested parts can be accurately and visually reflected by industrial CT technology. At present, the industrial 
CT technology has been widely used in the field of non-destructive testing and non-destructive evaluation, and 
it has been recognized as the world's best non-destructive testing means (Legacy, 2008). 
Industrial CT has been widely used in the detection of material defects and parts’ manufacturing error analysis. 
However, with the development of industry, some defects in industrial CT image and the contour segmentation 
cannot meet the needs of industrial production. In this paper, the industrial CT technology has been improved. 
First of all, this paper analyses the expander of LNG production equipment. Secondly, it introduces the 
industrial CT technology in detail. Then, it introduces the algorithm of image processing. Finally, combined 
with the new industrial CT technology, it uses some experiments to verify that the proposed method can 
effectively detect the equipment like the core components of the expansion machine. 

2. Expander technology for LNG production plant 
Long-distance natural gas pipeline transportation generally uses high-pressure transportation, since the high 
pressure natural gas contained in the high-pressure gas transmission line always contains a great deal of 
energy. Because the terminal pressure of users in the downstream is relatively low, generally below 0.4MPa, 
the natural gas receiving stations and gate stations in the cities need to properly make pressure reduction 
according to downstream users’ different requirements for gas pressure (Xiong, 2006). In the process of 
pressure regulating, high pressure natural gas will produce a large pressure drop, and release a lot of energy 
(Shao, 2009). The basic principle of using pressure energy of high pressure natural gas pipeline to produce 
LNG is that the mechanical energy, which is generated by volume expansion of high pressure natural gas 
during depressurization, is used to continue pressurizing high-pressure natural gas. After expanding, the 
temperature of the gas is reduced. Then after repeated expansion, the temperature can reach more than 100 
degrees below zero, and through the heat exchange, LNG will be produced. One end is the cantilever single 
stage centripetal turbine coaxial drive unit. The other end is the cantilever centrifugal turbocharger, and if the 
turbocharger is used to improve the expander inlet pressure, the refrigeration capacity and output work of the 
expander will be increased. This type of unit is called a positive turbo expander. Figure 1 is a schematic 
diagram of using turbo expander refrigeration to produce LNG.  

 

Figure 1: The schematic diagram of using turbo expander refrigeration to produce LNG 

818



2.1 Turbo expander 

Turbo expander is a kind of equipment that uses the expansion of the gas to do work and produce cold 
quantity. It is also an important part of low temperature air separation equipment and gas separation and 
liquefaction equipment. When passing through the turbine expansion machine, high pressure gas’ expansion 
process can be summed up as follows. High pressure gas speeds up in the static nozzle ring, and outflows 
after ripping into the cascade passageway. The change of the inlet and outlet momentum makes the air flow to 
the gate, then the kinetic energy of the air flow is reduced and the stagnation enthalpy is decreased, so as to 
achieve the purpose of external work and refrigeration. The turbine is to achieve energy conversion through 
utilizing the rate change of the working fluid. The working fluid expands in the turbine expander to obtain 
kinetic energy, and the energy is output by external wheel, in order to reduce the energy and temperature of 
the working fluid. This production process is also called differential pressure liquefaction process. High 
pressure natural gas achieves constant entropy expansion and depressurization in the turbo expander, and 
the cooling capacity of this process is:  

1( )in outQ m h h η= −        (1) 
Among them, Q is the refrigeration output, m is the gas mass flow rate, hin and hout are respectively the import 
enthalpy and export enthalpy of expander working substances, η1 is the turbine efficiency. If all of the 
expansion work is used, the process of energy loss is: 

0 0 2 1( )gE T S T m s sΔ = = −  (2) 
Among them, ΔE is the loss of energy, Sg is the entropy, s is the entropy ratio. Figure 2 is the internal structure 
of the expander. 

 

Figure 2: The internal structure of the expander. 

2.2 Analysis of expander rotor system 

The working environment of turbine rotor system is quite complex. There are the expansion flow and 
compressible flow of gas at both ends of the impeller, and different gradients of pressure and temperature are 
formed on the surface of the expansion wheel and boosting wheel, so the rotor system is both affected by and 
temperature field. Since the flow field will affect the boundary conditions of the temperature field, different 
fields will influence each other. In addition, the centrifugal force, caused by the rotation of the rotor, and other 
forces also increase the complexity of the engineering problem. The expansion wheel is the output power unit 
of the rotor system. Gas flows and expands in the expander to drive impeller, whose velocity, temperature and 
pressure will decrease, and then the total energy will also decrease. While the boosting wheel is the rotor 
components in the system, in which the gas is compressed, and the speed, temperature and pressure will be 
larger, then its total energy will increase. As the core component of the expander, the mechanical strength of 
rotors is very important to the production. Figure 3 is the mechanical model of turbo expander rotor. 

819



 

Figure 3: The mechanical model of turbo expander rotor. 

3. Research on industrial CT technology 
The basic principle of Industrial CT technology is to use the ray bundle of passing through the measured 
object to get the fault projection data, and then with the image reconstruction method the projection data will 
be scanned to get two-dimensional grey images. Through the analysis and measurement of the image, we 
can get the information of the detected objects, such as the inside and outside structure, size and density [8-7]. 
It is not only applied to aerospace, pyrotechnics, precision machinery and other important products, but also 
applied to the automotive, petroleum, geology, archaeology and many other fields. 
Computer tomography scan imaging technology is an image contrast technology which combines both 
computer technology and ray scanning technology. It uses an accurate X-ray beam and high sensitivity 
detectors to make continuous tomography through surrounding detected objects. In every process, firstly the 
detector receives the attenuation X-ray information, and the A/D converter inputs the output digital to the 
computer, then the X-ray absorption coefficient of each point in different fault planes will be got after high-the 
speed operation of computer. Finally, these data will be used to make up the matrix of the image. CT image is 
composed of two parts: and pixel size. Among them, pixel intensity reflects the X-ray absorption degree of the 
detected objects, and the pixel size reflects the fine degree of the image, i.e. the spatial resolution of the 
image. CT image is a grey level image. Compared with the X image, the black area is low density because of 
its low density and low X-ray absorption. 
Industrial CT system is generally composed of X-ray source, detector, mechanical scanning system, data 
acquisition system, control system and so on. The combination of X-ray source, detector and mechanical 
scanning system determines the effect of mechanism error on the imaging quality of industrial CT, which is 
very important for the industrial CT system.  

4. Image segmentation algorithm 
Support vector machine (SVM) is a machine learning method based on statistical learning theory, whose 
theoretical basis is the structural risk minimization principle. The VC dimension is an important index to 
describe the function set or machine learning capability.VC dimension reflects the learning ability of the 
function set. In the case of finite dimensional training samples, when the number of samples is fixed, the 
greater the VC dimension of machine learning is, the more complex the learning mechanism of the learning 
machine to the training samples is. The statistical learning theory is based on VC dimension, and deduces the 
relationship between expected risk and empirical risk, which is called the world of generalization. The principle 
of empirical risk minimization is to obtain the empirical risk through the known training data, and use the 
empirical risk to approach the minimum value of the desired risk. In the case of finite samples, the learning 
machine should not adopt the empirical risk minimization principle. Researchers put forward the principle of 
structural risk minimization (SRM). Its basic idea is that firstly, the function set is divided into multiple subsets,. 
Then array these function subsets in ascending order according to the VC dimension, and seek empirical risk 
minimization for each subset. Finally, based on their empirical risk minimization and VC dimension, obtain the 
actual minimum risk of each subset. 
With the introduction of the kernel function, SVM has shown its outstanding advantages in solving nonlinear 
problems. Low dimensional space vector sets are usually hard to partition, and the solution is to map them to 
high dimensional space. But the kernel function smartly solves this problem exactly. In practical problems, the 
kernel function is usually given directly. At present, the most commonly used kernel functions are the following 
three types. 

RBF kernel function: 
2

2( , ) exp( )2
i j

i j

x x
k x x

σ
−

= −  (3) 

820



Polynomial kernel function: ( , ) [1 ( )]di j i jk x x x x= + ⋅  (4) 

Sigmoid kernel function: ( , ) tanh[ ( ) ]i j i jk x x x x cγ= ⋅ +  (5) 

Different kernel functions and their parameters have great influence on the classification performance. 
Therefore, it is important to select the appropriate kernel. 

5. System Improvement and Implementation 
The pressure energy loss of high-pressure natural gas pipeline mainly exists in the process of adjusting 
pressure in the sub transmission station. In order to realize the recovery and utilization of energy, the pressure 
in long natural gas pipeline transportation station can be recycled to produce LNG, which is also known as 
pressure liquefaction production device. In view of the high efficiency of the turbine expander, it can always be 
used as the equipment for recovering the pressure energy.  
The rotor is the only high speed operating component in the turbo expander. It has a great influence on the 
efficiency and stability of the expander. The structural parameters of the rotor have a great effect on the critical 
speed, and the main parameters are the bearing span, the length of the main shaft, the sealing length, the 
diameter of the shaft, the position and the size of the thrust plate. High quality rotor is essential to the 
production of the factory. Advanced industrial CT can be used to detect rotor. Detecting parts’ manufacturing 
errors and internal material defects can avoid operation risk when the factory produces LNG. In this paper, 
optimizing the industrial CT image can effectively improve the efficiency of industrial CT detection. 
In the actual engineering application, the quality of Tomography image obtained by industrial CT scan cannot 
satisfy the application requirements. In this paper, the image is enhanced to improve the image quality, and 
then the contour of the CT image is segmented and extracted. Histogram equalization is firstly used to 
enhance the image. Secondly, the SVM algorithm is used to segment the image. 
(1) Calculate the frequency of each gray level in the histogram of the original image. 
(2) The gray value is calculated by the gray value transform function. 
(3) Update the gray level distribution of the image to complete the gray level histogram equalization. 
(4) The kernel function of SVM algorithm is improved, and the new kernel function is constructed by using 2-2 
weighting the three functions RBF kernel function, Polynomial kernel function and Sigmoid kernel function. 
(5) Optimize the weights to find the best combination of the best weight. 
(6) Use the improved algorithm to segment the CT image contour. 
With regard to the expander rotor of LNG production factories, the optimization algorithm for CT image is put 
forward in this paper. It is not only helpful for rapidly extracting expander rotor image region that needs 
detecting, but also helps to compare and verify CAD drawings and designing drawings, which greatly 
improves the efficiency and accuracy of industrial CT detection. 

6. Experiment and Analysis 
6.1 System Platform 

In order to verify the effect of this algorithm, we choose the expander rotor of a certain LNG production factory 
as CT sample. The improved method is applied in the experiment. Experimental results show that the 
proposed method can quickly and accurately detect the size of the specimen. Our experimental environment 
contains: Hardware environment Intel (R) Core (TM) i5-4300M @ 2.60GHz, 4G memory, and Windows 7 
software environment, with MATLAB as the design language. Fig.4 is the specimen of experiment. 

 

Figure 4: The specimen of experiment. 

821



6.2 System Operating Results and Analysis 

In order to test the accuracy of the industrial CT image, this paper conducts actual measurement on the 
sample. Besides, the vector size value and the actual measurement value are compared to analyze the error 
so as to achieve the purpose of inspection accuracy. 

Table 1: The comparison between the ICT vector size and the actual measured value. 

Size Design size(mm) Measurement(mm) CT size(mm) 
Size 1 220.50 220.41 220.46 
Size 2 150.00 150.10 150.11 
Size 3 10.00 10.03 10.03 
Size 4 55.00 55.17 55.10 
Size 5 35.00 35.01 34.01 
Size 6 50.00 50.05 50.03 

Experimental results show that the proposed method can effectively display the image contour of the sample, 
which has obvious advantages in improving the detection efficiency of industrial CT images. 

7. Conclusion 
In this paper, the industrial CT technology has been improved. First of all, this paper analyses the expander of 
LNG production equipment. Secondly, this paper introduces the industrial CT technology in detail. Then, this 
paper introduces the algorithm of image processing. Finally, combined with the new industrial CT technology, 
some experiments are carried out. Results show that the proposed method can effectively detect the core 
components of the expansion machine and other equipment. 

References 

Asutosh N., 2010, Development of Design Software for Cryogenic Turbo-expander, Rourkela: National 
Institute of Technology, 6-23. 

Josef P., 2010, Cryogenic Turbo-expander, EP1057977B. 
Legacy K., 2008, Industrial CT scanning speedsmold qualification, Plastics Technology, 54(9), 66-70. 
Marcuccilli F., Zouaghi S., 2007, Radial Inflow Turbines for Kalina and Organic Rankine Cycles, Unterhaching: 

Proceedings European Geothermal Congress. 
Michell A.G.M., 1929, Progress of Fluid-Film Lubrication. Transaction of the ASME, 51, 153-163. 
Reynolds O., 1886, On the Theory of Lubrication and its Application to Mr. Beauchamp Tower’s Experiment, 

Including an Experimental Determination of the Viscosity of Olive Oil. Phil Trans, 177(1), 157-234. 
Shao H., 2009, Safety Devices Used in Gas Regulator Station and Their Safety Analysis. Gas & Heat, 29(6), 

B01-B04. 
Tang L., Chen M.J., 2016, Image denoising method using the gradient matching pursuit, Mathematical 

Modelling of Engineering Problems, 3(2), 53-56, DOI: 10.18280/mmep.030201 
Tower B., 1883, First Report on Friction Experiments (Friction on Lubricated Bearings), Proceedings of the 

Institution of Mechanical Engineers, 632-659. 
Wu E.H., 2003, Medical Imaging. Beijing: People's Medical Publishing House, 20-21. 
Zheng G., Gollmer S., Schumann S., 2009, A 2D/3D correspondence building method for reconstruction of a 

patient-specific 3D bone surface model using point distribution models and calibrated X-ray images, 
Medical Image Analysis, 13(6), 883-899, DOI: 10.1016/j.media.2008.12.003 

822