Corrosion in Wet Oil Piping: Root Cause Failure Analysis,Mitigation and Neural Networking

This paper discusses the typical corrosion anomalies likely to occur in the wet oil system of a Floating Storage and Offloading Vessel (FPSO). The root causes and operational mitigations are identified. A neural network is proposed for capturing relationships within the large volume and diversity of data and hence permit effective modelling of corrosion and integrity of specific piping sections. Novel mitigations which can minimize severity of wall loss for flowing lines include; corrosion modelling of separator fluids and consequent pressure and temperature adjustments and calculated partition coefficients P derived from corrosion inhibitor (CI) injection and residual lab data for first-pass assessment of the effectiveness of CI injections, hence providing adequate inhibition.     

Root Cause Failure Analysis and Mitigation of Corrosion in Seawater Injection Piping

This paper illustrates the methodology for performing a root cause failure analysis of corrosion anomalies reported in the low pressure section of an FPSO seawater injection piping system. Specific practical actions are recommended for mitigating the identified root causes of failure. The primary root cause of failure is oxygen corrosion due to oxygen ingress at components upstream (u/s) and downstream (d/s) of the sulfate removal (SRM) pumps. Oxygen corrosion was worsened by underdeposit corrosion caused by unavailability of coarse straining of abstracted seawater. The secondary root causes are likely due to: deterioration of graphite-filled gaskets on the flanges; occasional bisulphite overdosing; incorrectly specified oxygen scavenger; non-optimized scavenger injection rates; incorrect risk assessment; and unavailability of hypochlorination of abstracted seawater. This paper recommends that the primary and secondary root causes should be further investigated and mitigated. The less difficult but big-impact mitigations which should reduce significantly the number and severity of reported anomalies include: modeling corrosion (based on dissolved oxygen concentration inputs) with predicted velocities (V) guiding operational adjustments within the target range 2 < V < 10 m/s and the predicted cumulative corrosion rates used to target and optimize inspection requirements; inspecting and testing each component u/s and d/s of the SRM pump, in order to identify locations of oxygen ingress; and checking deaeration performance especially possible leaks along the vacuum pump system. Results of KPIs should comply with the specified targets and be managed as part of the corrosion management system.     

多因素作用下混凝土寿命的BP神经网络预测

影响混凝土及其结构使用寿命的因素主要包括荷载和自然环境的变化,它们对混凝土结构使用寿命的影响过程是错综复杂的,难以用准确的数学式表达.人工神经网络方法具有无需输入变量与输出变量间复杂的相关假设,也无需确定各种计算参数,从而以消除计算参数确定过程中产生的计算误差的特点,使得其在土木结构工程耐久性方面具有广泛的应用.采用动量～自适应学习速率调整算法以及规则化调整对BP神经网络的泛化能力进行了改进,使其误差平方和达到0.000918,提高了BP神经网络的泛化能力;并用改进的BP神经网络对荷载-复合离子-干湿交替作用下混凝土材料的使用寿命进行了预测,避免了在确定计算参数过程中所产生的计算误差,拓宽了多因素作用下结构混凝土寿命预测新方法.     

神经网络在含水率预测中的应用

相分率测量是多相管流研究的重点，γ密度仪是气液两相流和多相流测量中常用的非插入式测量方法，其中含水率的测量度最大。实验测试发现，直接利用高低能计数计算含水率存在较大偏差，本文通过将高低能计数值作为输入参数，利用BP神经网络方法预含水率，大大提高了预测精度，结果表明，BP神经网络方法是改善γ密度仪测量含水率精度的有效方法。     

长北气田某气井油管腐蚀速率增大原因

长北气田油管整体腐蚀速率缓慢,但某气井在多臂井径检测中发现,该油管局部腐蚀速率增大,远远大于整个气田的油管腐蚀速率和该井前期的腐蚀速率.进行了水质、气质组分分析、多臂井径检测和腐蚀挂片等试验,分别对腐蚀挂片的宏观形貌及腐蚀产物进行了分析,并结合油管的腐蚀程度和腐蚀情况,对该井油管腐蚀速率突然加快的原因及腐蚀机理进行了分...     

Pipeline Corrosion,Modeling and Analysis

Degradation of pipelines is the result of the continuous attack by the environment of these conduits like climate change (temperature, rate moisture in the ground, etc.…) that may lead to a corrosive environment. The design and the maintenance of these conduits and pipelines is a challenge for oil industry seen the serious consequences which can occur because of several reasons, example: defects of the cracks types, rust…, etc. Which can cause escapes of the transported matter or ruptures of these conduits with all that involves like economic loss and pollution of the environment.     

BP神经网络对3C钢腐蚀性能的预测分析

利用三层BP神经网络,根据已有的3C钢在不同海水环境参数下的腐蚀速度数据,建立了3C钢在海洋环境中腐蚀速度的人工神经网络模型;并分析预测了海水环境参数与腐蚀速度之间的关系.预测结果表明:在近海条件下,温度和含氧量越低,氧化还原电位越高,腐蚀速度越慢;在弱碱性条件及盐度大于25ppt时,腐蚀速度比较慢.此环境参数下,3C钢材料的腐蚀较小.本文的预测结果能很好地反映出海水环境参数对腐蚀速度的影响.     

Viscosity Modeling and Prediction of Reservoir Fluids: From Natural Gas to Heavy Oils

Viscosity and density are key properties for the evaluation, simulation, and development of petroleum reservoirs. In previous work, the friction theory (f-theory) models have already been shown capable of providing simple but accurate viscosity modeling results of petroleum reservoir fluids with molar masses up to around 200 g · mol^–1. As a base, the f-theory approach requires a compositional characterization procedure to be used in conjunction with a van der Waals type of equation of state (EOS). This is achieved using simple cubic EOS, which are widely used within the oil industry. In this work, the f-theory approach is further extended to the viscosity modeling of heavy reservoir fluids with viscosities up to thousands of mPa · s. Essential to the extended approach presented here is the achievement of accurate pvT results for the EOS characterized fluid. In particular, it has been found that for accurate viscosity modeling of heavy oils, a compressibility correction in the way the EOS is coupled to the viscosity model is required. With the approach presented in this work, the potential of the f-theory for viscosity modeling of reservoir fluids is extended to practically all kind of reservoir fluids, from light ones to heavy ones. Additionally, the approach has been completed with an accurate density modeling scheme.     

神经网络对结构地震反应的预测及试验研究

本文基于神经网络对非线性系统具有辨识和预测功能，并结合具有二阶收敛效应的Levenberg—Marquardt算法，采用一多层前馈网络对建筑结构的地震反应进行了预测。首先以一三层钢筋混凝土结构的振动台试验数据对网络结构进行批量训练，然后用未曾训练的地震波数据对结构进行地震反应预测，并与试验数据进行对比，分析结果表明：Lev—enberg—Marquardt算法能快速收敛，神经网络能准确地预测结构的地震反应。     

A Neural Network Modeling of Stress Behavior in Nonlinear Magnetostrictive Materials

Using accurate magnetostriction simulation models during the various design stages of their related devices can positively contribute to the enhancement of their precision. These models are indispensable to different crucial computational activities such as those dealing with active vibration damping devices and optimum clamping stresses for transformer sheets. In this paper, we present a new contribution for the dynamic hysteresis behavior of magnetostrictive materials. To do this, we have used a neural network to model the relationship between the elongation λ and the magnetization M for different loads σ. The derivative of λ according to M is then calculated numerically and integrated in the Jiles–Atherton model for calculating the hysteresis loops.     

基于BBMNN的多变量非线性软测量建模及应用

基于Boostins思想,提出了一种改进的adaboost算法。在此基础上,提出了一种新的多神经网络构造方法BBMNN。应用于软测量建模,给出了一种新的非线性系统软测量建模方案,并分别针对多变量、非线性典型模型和复杂工业过程,应用实验数据和实际运行数据进行了仿真研究。仿真结果表明,该方案可以较好地解决复杂对象神经网络建模时样本点数量与模型精度之间的矛盾,可同时获得较高的训练精度和预测精度。     

Artificial Neural Network Prediction of Material Removal Rate in Electro Discharge Machining

Thermal erosion theory is widely accepted as an explanation of the erosion process in electro-discharge machining (EDM). Theoretical models are based on the solution of the transient heat conduction equation, which is modeled considering suitable assumptions with appropriate initial and boundary conditions. The closed form solutions result only after considering too many assumptions, which are far from actual machining conditions. The growth of the plasma channel, energy sharing between electrodes, process of vaporization, formation of recast layer, plasma-flushing efficiency, and temperature sensitivity of thermal properties of the work material are a few physical phenomena that render the machining process highly complex and stochastic. The mathematical consideration of all these complex phenomena is very difficult. Therefore, mathematical prediction of material removal rate when compared with the experimental results shows wide variation. In such circumstances, an attempt has been made to develop an artificial feed forward neural network based on the Levenberg-Marquardt back propagation technique of appropriate architecture of the logistic sigmoid activation function to predict the material removal rate. Such a neural network model is expected to perform well under the stochastic environment of actual machining conditions without understanding the complex physical phenomena exhibited in electro-discharge machining. The validity of the neural network model is checked with the experimental data, and we conclude that the artificial neural network model for EDM provides faster and more accurate results.     

人工神经网络在腐蚀与防护领域的应用

介绍了人工神经网络的概念、分类、组成,概括了腐蚀实验数据的特点,即数据量大,来源广泛,存在大量随机数据,由于人工神经网络是一个非线性动力学系统,所以可以轻易地实现函数逼近,数据聚类、模式分类,优化计算等功能,因此是处理腐蚀实验数据的一个有力手段,。人工神经网络迄今为止,已在区分腐蚀类型,确定主要腐蚀因子,解析谱图数据,腐蚀监测,腐蚀预测等方面有了较成功的应用,分析了人工神经网络技术在腐蚀防护领域的应用前景。     

高流速、高盐、高CO2湿气条件下X60管线钢顶部的腐蚀行为

为明确X60管线钢在高流速、高含盐、高含CO2湿气条件下的顶部腐蚀行为和规律,采用高速湿气环路系统对X60管线钢进行湿气顶部腐蚀模拟评价试验.通过腐蚀前后试样的失重量评估X60管线钢湿气顶部腐蚀的情况,并采用扫描电子显微镜(SEM)、能谱分析(EDS)以及X射线衍射仪(XRD)等测试分析手段,研究了不同含水率和不同Cl...     

Stress Corrosion Cracking and Oil Leakage in a Control Oil Piping System

This article describes the results of an investigation concerning with the failure of a pipe which was carrying oil from the control oil unit to the steam turbine control valves servomotor. The failure was in the form of a crack, propagating horizontally along the pipe. The crack initiated on the outside of the pipe. The cause of the failure was investigated by conducting visual examination, detailed macro and microstructural examinations and determining the composition of material from the failed pipe. The composition of the pipe material was analyzed by atomic absorption spectrometer. The failure of the oil pipe was attributed to stress corrosion cracking. The pipe material is A312 TP 304L. Recommendation to minimize such failures includes coating the pipe to prevent contact with chloride from the surrounding marine environment.     

Prediction of Mechanical Properties of Hot Rolled, Low-Carbon Steel Strips Using Artificial Neural Network

Artificial intelligence (AI) has been used in many application areas of engineering. In the present work, an artificial neural network (ANN)-based model is developed to predict the mechanical properties such as yield strength (YS), ultimate tensile strength (UTS), and elongation (EL) of the hot rolled (HR) steel strips/coils. Different network topologies have been investigated to find the appropriate network to simulate the problem. Finally, the best network was chosen as the one with 7-19-3 topology-7 neurons in the input layer, 19 in the hidden layer, and 3 in the output layer. It has been shown that a single network with three output neurons is sufficient to address the problem. The model has been tested with 121 unknown patterns, and the match between the actual values and the simulated ones is found to be very good. The model has been implemented in the hot strip mill (HSM) of Tata Steel, India. This paper describes the methodology adopted to develop the model.     

Stress Corrosion Cracking in Plastic Pipes: Observation and Modeling

Stress corrosion cracking (SCC) in engineering thermoplastics is commonly observed in the form of a microcrack colony within a surface layer of degraded polymer exposed to a combined action of mechanical stresses and chemically aggressive environment. A probabilistic modeling of SCC initiation is briefly discussed. A deterministic modeling of slow stress corrosion (SC) crack growth process is developed using Crack Layer (CL) theory. Numerical solution of SC crack growth equations is discussed. Comparison of the kinetics of cracks driven by SC and by stress only is presented. Conventional plot of SC crack growth rate vs. the stress intensity factor is constructed and analyzed. An algorithm for conservative estimation of lifetime of engineering thermoplastic subject to a combination of mechanical stresses and chemically aggressive environment is discussed.     

气瓶表面裂纹扩展规律与寿命估算的研究

基于Franc3D数值模拟法及子模型技术,对气瓶瓶筒处表面裂纹进行裂纹扩展的动态数值模拟及三维应力强度因子计算,证明该方法的可行性,并用公式迭代法验证该方法的准确性。在此计算结果基础上,对两初始裂纹a_0=3 mm、c_0=6 mm及a_0=3 mm、c_0=9 mm的剩余寿命进行了估算,考虑安全裕度后计算得到两初始裂纹剩余寿命分别为3.5 a和2.6 a。     

Wireless Sensor Network Modeling Using Modified Recurrent Neural Networks: Application to Fault Detection

This paper presents a dynamic model of wireless sensor networks (WSNs) and its application to sensor node fault detection. Recurrent neural networks (NNs) are used to model a sensor node, the node's dynamics, and interconnections with other sensor network nodes. An NN modeling approach is used for sensor node identification and fault detection in WSNs. The input to the NN is chosen to include previous output samples of the modeling sensor node and the current and previous output samples of neighboring sensors. The model is based on a new structure of a backpropagation-type NN. The input to the NN and the topology of the network are based on a general nonlinear sensor model. A simulation example, including a comparison to the Kalman filter method, has demonstrated the effectiveness of the proposed scheme.