Prediction of corrosion rate and wear rate constant in gas cylinder tubes using neural network

Number of pages: 171 File Format: word File Code: 32195
Year: 2013 University Degree: Master's degree Category: Industrial Engineering
  • Part of the Content
  • Contents & Resources
  • Summary of Prediction of corrosion rate and wear rate constant in gas cylinder tubes using neural network

    Master's Thesis

    Trend of precision instruments and automation in the oil industry

    Abstract

    Prediction of corrosion rate and wear rate constant in gas tube using neural network

     

    Corrosion is a phenomenon that is very complicated due to the influence of various factors and cannot be modeled easily. In order to predict and model corrosion, attention is paid to its physical, chemical and electrochemical reactions and processes, and modeling is done based on it. Despite the success of these models, due to the multitude of influencing factors that are sometimes unknown, there is a need for models that model and predict this phenomenon more accurately.  In addition, the oil and gas industries, especially the upstream industries, have always faced the problem of wear/corrosion phenomena, and in addition to the inherent limitations in the reservoir that affect the production capacity of the gas production wells of each reservoir, the limitation of the fluid velocity in the production line of the well in order to prevent the phenomenon of wear/corrosion is another factor that determines the production capacity of a gas well. A common way to obtain the production rate is to use the relationship suggested by the API RP 14E standard. In this regard, factor C, which is the wear rate constant, has been suggested by the standard in different conditions. Experience has shown that the proposal of this standard is conservative in many cases.

    The purpose of this research is to predict the corrosion rate by neural network as well as the wear rate constant by artificial neural networks, and a suitable numerical proposal for the constant C using field data from gas wells is discussed, so that the phenomenon of wear/corrosion does not occur.

     

    Key words:

    Neural network, metal corrosion, wear constant, brain tube, gas well

     

    Chapter One: Introduction

                                                                                                

    1-1 Introduction of the whole research

         One of the important scientific, technical and economic topics is the issue of corrosion of metals and the protection of metal facilities. Investigating the topic of corrosion is not so simple, and with all the research done, its factors are still not known correctly. Apart from the use of chemical sciences to deal with corrosion, the use of other sciences in controlling and predicting corrosion and using its results in the repair of metal equipment is of particular importance. The problem of corrosion in the oil and gas industry has become more serious than other industries due to the presence of corrosive compounds. The inability to predict the corrosion rate makes it impossible to predict the resulting failure times, which causes problems for the maintenance and repair teams. So far, various methods have been used to deal with this phenomenon. Corrosion modeling can be effective in knowing more and predicting the issues arising from it. In these modelings, more than mechanistic methods have relied on the theoretical background of corrosion and mathematical formulas, but due to the inherent complexity of this phenomenon, these methods have not been very successful.

    Due to the mentioned complexity and the multitude of known and unknown factors affecting this phenomenon, it seems that data-driven methods such as neural networks can be used to predict the corrosion rate, provided that sufficient data is available in this field.

    This research is based on using the capabilities of neural networks to predict the corrosion rate, and for this purpose, the information collected from the gas fields covered by the South Zagros Exploitation Company was used. Another important topic that is of particular importance in the gas industry is the phenomenon of wear/corrosion.The wear phenomenon is very likely in wells with high flow rate or with solid particles suspended in the production fluid. Even in sand-free conditions or clean services where the intensity of sand production is about several pounds per day, the damage caused by wear is very high at high production speeds. In the industry, modified shear stress relations like the one provided by the standard are used to predict the wear rate.

    where

    Ve: fluid wear velocity (ft/s)

    : volume mass of gas and liquid mixture In operating pressure and temperature (pounds per cubic foot)

    : empirical coefficient (no unit)

    The limitations and problems of using equation (1-1) for wells are mostly related to the constant value "". The API RP 14E standard suggests that for wells that do not produce sand and also for wells that use CRA (corrosion resistant alloy) tubing, values ??of 150 to 200 should be considered for the constant. If sand or sand is produced along with the fluid, they consider the number 100.

    Today, after years have passed since the emergence of the API RP 14E standard, its inefficiency has become clear to everyone. The "C" constant in the API RP 14E standard is considered very conservative. In this research, we will predict the wear rate constant (experimental coefficient C). For this purpose, the information collected from the gas fields covered by the South Zagros Exploitation Company has been used. rtl;">     In recent years, the use of artificial intelligence in the field of corrosion process modeling has received attention. Artificial neural network has become one of the most widely used methods in the field of corrosion process modeling. In the following, some researches about corrosion in which neural network is used for modeling are introduced.

         One of the most serious works that has been done to use neural network in corrosion problems and was one of the first in this field, was the prediction of corrosion rate using neural network by Smets and Bogaerts. In their work, they developed a neural network and used it to predict [1] SCC on type 304 stainless steel in the presence of chloride and oxygen compounds and a specified temperature. They found that the neural network method is superior to the traditional fitting method in this regard. ]1[

         In another research, a neural network model was produced to predict the number and depth of pits caused by pitting corrosion. Of course, information about the topology and size of the network and how to train it is not given. The progress in the depth of holes and their number was effectively modeled and showed good results compared to the experimental data. ]2[

         Neural network has been used to predict the type of corrosion from the polarization curve. The inputs of the network are the density and pitting corrosion potential, and the outputs are the risk of each type of general, pitting, and brittle corrosion. [3[

    Nesik and colleagues in an article have mentioned two important problems that cause neural networks to be used less in corrosion topics. The first reason is the lack of familiarity of corrosion engineers with the category of artificial intelligence and neural network and its application in corrosion prediction, and the second reason is the lack of sufficient data in this matter. Of course, in this article, in the first part, there are explanations about the neural network for corrosion engineers, and in the second part, the Monte Carlo method is introduced, and a practical work has been done during it.

  • Contents & References of Prediction of corrosion rate and wear rate constant in gas cylinder tubes using neural network

    List:

    Chapter One: Introduction. 9

    1-1 Introduction of the entire research. 9

    1-2 previous activities and research history. 11

    1-3 research objectives. 17

    Chapter Two: Neural Networks. 18

    2-1 Single neuron modeling 19

    2-2 Activity function. 20

    2-3 neural network architecture. 21

    2-3-1 feed networks 22

    2-3-2 return networks. 22 2-4 learning algorithms. 23

    2-5 MLP neural network. 24

    2-5-1 error back propagation algorithm 25

    2-5-2 error signal 26

    2-5-3 learning rate selection. 26

    2-5-4 training stage. 27

    2-5-5 Generalization ability. 27

    2-5-6 Stop training. 28

    2-6 RBF network. 29

    2-6-1 Radial neural network structure. 30

    2-6-2-1 Determining the location of centers. 35

    2-6-2-2 Determination of standard deviation. 37

    2-6-2-3 Teaching the weight matrix of the output layer. 38

    The third chapter: fuzzy logic. 40

    3-1 introductions to fuzzy systems. 40

    3-2 basic components of fuzzy inference system (FIS) 45

    3-2-1 fuzzy rule base. 45

    3-2-1-1 Characteristics of the set of rules. 45

    3-2-2 Fuzzy inference engine. 47

    3-2-2-1 Inference based on the combination of rules. 47

    3-3 Non-fuzzifier 49

    3-3-1 Non-fuzzifier of center of gravity. 49

    3-3-2 Average center de-fuzzifier. 49

    3-3-3 Maximum non-fuzzifier. 50

    Chapter Four: Adaptive Neural-Fuzzy Inference Systems (ANFIS)). 52

    Chapter Five: Corrosion. 54

    5-1 An introduction to corrosion. 54

    5-1-1 Corrosion costs. 56

    5-1-2 Investigation of types of corrosion. 57

    5-2 Design of organic anti-corrosion systems. 68

    5-3 Corrosion in oil and gas facilities 70

    5-3-1 Corrosion by corrosive carbon dioxide gas. 71

    5-3-2 Corrosion by corrosive liquids of oil tanks. 73

    5-3-3 Corrosion by hydrogen sulphide corrosive gas. 73

    5-4 Corrosion in three-phase systems of wells and gas pipes and its control methods. 77

    5-4-1 Corrosion control methods. 77

    5-4-1-1 Corrosion inhibitors. 78

    5-3-1-2 pH stabilization method. 82

    Sixth chapter: Wear phenomenon in hydrocarbon production systems. 88

    6-1 Wear process in oil and gas production wells 89

    6-2 Wear mechanisms. 90

    6-2-3 Vulnerability of equipment against the phenomenon of wear: 90

    6-3-2-1 Material of equipment. 92

    6-3-2-2 conductive metals and other conventional materials. 92

    6-3-2-3 Special wear-resistant materials. 93

    6-4 Abrasion caused by sand or fine particles. 94

    6-4-1 Sand production and transportation. 94

    6-4-2 Size, shape and hardness of solid particles. 96

    6-5 wear/corrosion. 97

    6-6 Wear caused by hitting liquid drops. 98

    6-7 cavitation. 100

    6-8 Wear due to solid particles in elbows 101

    6-9 Wear of solid particles in closed T-shaped joints. 103

    6-10 methods of monitoring, preventing and managing wear phenomena. 104

    6-10-1 wear management techniques. 105

    6-10-1-1 Reducing production flow. 105

    6-10-1-2 pipeline design. 105

    6-10-1-3 Separation and removal of sand from the flow. 106

    6-10-1-4 instructions and prediction of wear. 107

    6-10-1-5 Wall thickness assessment. 109

    6-11 wear prediction tools and a review of the research done. 110

    6-11-1 Review of the most important standards in pipeline design and wear management. 110

    6-11-2 Wear prediction tools and models. 111

    6-11-2-1 API standard RP 14E. 112

    6-11-2-2 Other wear prediction models. 117

    6-11-3 Comparison of knee wear prediction models 124

    Seventh chapter: research method. 131

    7-1 Prediction of corrosion rate. 134

    7-1-1 Corrosion rate prediction using neural network. 134

    7-1-2 Prediction of corrosion rate using ANFIS. 141

    7-2 constant prediction of wear rate. 151

    Chapter Eight: Conclusion. 158

    Chapter 9: Suggestions. 159

    Resources 160

     

     

    Source:

     

     

     [9] Mustafa Kia, "Neural Networks in MATLAB", Kian Computer Sabz Publications, 1387.

    [10] Nima Jamshidi, Seyed Rasool Moulai and Ali Abuyi

    [10] Nima Jamshidi, Seyyed Rasool Moulai and Ali Abuei Mehrizi, "Applied teaching of advanced electrical engineering topics with MATLAB", Abid Publications, 2016.

    [11] Mehdi Ghazanfari, "Neural Networks (Principles and Functions)", second edition, University of Science and Technology Publications, p. 296.

    [15] Mohammad Baqer Minhaj, "Fundamentals of Networks" Nervous (computational intelligence)", Amirkabir University of Technology Publications, 1377.

    [20] Li Wang, translated by Mohammad Tashnelab, "Fuzzy Systems and Fuzzy Control", Khajeh Nasiruddin Toosi University Publications, 1389.

    [21] March J. Fontana, translated by Ahmad Saatchi, "Corrosion Engineering", Isfahan Industrial Unit Jihad Publications, 1387.

     

     

    List of Latin sources

     

    [1] Smets, H. M. G., Bogaerts, W.F.L.(1992). Neural network prediction of stress corrosion cracking, mater perform,vol.31, PP 64-68

    [2] Urquidi-Macdonald, M. , Eiden, M.N., Macdonald, D.D. (1993). Development of a neural network model for predicting damage function for pitting corrosion in condensing heat exchanger. Modification of passive films, Paris. PP. 336-343. [3] E.M. Rosen and D.C silverman Corrosion prediction from polarization scans Using an artificial neural network integrated with an Expert system. NACE international, corrosion/92, vol. 48, no. 9, pp. 734-745. [4] Nesic, S., Nordsveen, M. Maxwell N., Vrhovac, M. (2001). Probabilistic modeling of CO2 corrosion laboratory data using neural network corrosion science, vol.43, PP. 1373-1392

     

    [5] Trassati, S.P., Gabbetta G. (2006), study of naphthenic acid corrosion by neural network corrosion engineering science, and technology, vol 41, Number 3, PP. 200-211.

    [6] Andrews, P.; Illson, T.F.; Matthews, S.J., "Erosion-Corrosion studies on 13Cr steel in gas well environment", PP. 568-574, December 1999

     

    [7] Esmaeilzedeh, F.,"Future south pars development may include 9 5/8-in.tubing". Oil & Gas Journal/ Sep. 27, 2004, PP. 53-57. [8] J. Kamruzzaman, et. al, "Artificial Neural Networks in Finance and Manufacturing", 2006.

    [12] G. Bortolan and J. L. Willems, "Diagnostic ECG classification based on neural networks", J. Electrocardiol, Vol. 26, pp. 75-79, 1994.

    [13] B. Krose, P. Van der Smagt, “An introduction to Neural Networks”, 1996,

    (http://www.avaye.com/files/articles/nnintro/nn_intro.pdf)

    [17] Y. Kutlu, M. Kuntalp, and D. Kuntalp, “Optimizing the performance of an MLP classifier for the automatic detection of epileptic spikes”, Expe. Syst. App., Vol. 36, pp. 7567–7575, 2009. [18] Hartman. E., Keeler. J. D, and Kowalski. J. M, Layered neural networks with Gaussian hidden units as universal approximations. Neural Computation, vol. 2, no. 2, pp. 210-215, (1990).

    [22] Venkatesh, E.S. Erosion damage in oil and gas wells. Proc. Rocky Mountain Meeting of SPE, Billings, MT, May 19-21, 1986, pp 489-497.

    [23] Haugen, K., Kvernvold, O., Ronold, A. & Sandberg, R. Sand erosion of wear-resistant materials: erosion in choke valves. Wear 186-187, pp 179-188, 1995. [24] Marchino, P. Best practice in sand production prediction. Sand control & Management, London, October 15-16, 2001. [25] Det Norske Vertitas. Recommended practice RP 0501: Erosive Wear in Piping Systems. 1996, Revision 1999. [26] Salama, M.M. & Venkatesh, E.S. Evaluation of API RP14E erosional velocity limitations for offshore gas wells. OTC 4484, OTC Conference, Houston, May 2 – 5 1983, pp371 – 376, 1983.

     

    [27] Shadley, J.R., Shirazi, S.A., Dayalan, E., Ismail, M. & Rybicki, E.F. Erosion-corrosion of a carbon steel elbow in a carbon dioxide environment, Corrosion, Vol 52, No9, September 1996, pp 714 – 723.

    [28] Shinogaya, T., Shinohara, T. & Takemoto, M. Erosion of metals in high speed mist flow evaluation of velocity by acoustic emission system. International Congress on Metallic Corrosion,

Prediction of corrosion rate and wear rate constant in gas cylinder tubes using neural network