Risk Assessment Study Of Drilling System For Deep Well Complex Formation Based On Soft Computing Theory

Since the drilling is a hidden underground engineering, and the distributions of porepressure and in-situ stress have the complex features of great difference and poor regularity indeep formation, there are a lot of complex and uncertain factors. Therefore, there are someunexpected problems which are wellbore instability, blowout, leakage and serious welldeviation. In this study, through some basic theories of system engineering, drillingengineering, finite element analysis and risk assessment, the drilling uncertainty in complexformation for deep well is analyzed, the failure risk of drill-string system and well structure isstudied deeply, the wellbore stability is evaluated using cusp catastrophe model, and finallythe risk assessment model for drilling system is established based on soft-computing theory,and the “Risk Assessment software for Deep-well Drilling system” is developed. It provides ascientific basis to drilling design and risk control for deep well.The main research contents of this paper are as follows:(1) The drilling uncertainty information is boiled down to random information, fuzzyinformation and gray information according to the subjective and objective reasons foruncertainty in drilling process. The basic drilling information database is build based on thetest and field data of core test, logging data and mud logging data. After the analyzed target isdetermined, the original formation information is simplified by reduction process of based onrough set approach. And then the distribution law of uncertainty values for complex formationis obtained by probability method or fuzzy set theory.(2) Taking the drill-string in deep well as an initially straight slender rotating beam withannular cross-section, considering the interaction of bit-rock and the collision ofdrill-string-sidewall, the nonlinear dynamic model of drill-string system for whole well isbuild with axial-bending-torsional coupling vibration based on the extend Hamilton principleand the finite element discretization method. The model uncertainty and parameteruncertainty are modeled using the non-parametric probabilistic method, and the stochasticnonlinear dynamic model of drill-string is build. Taking Von Mises stress, damage cumulatedby failure, stick-slip factor and radial displacement as the integrity limits of the drill-stringsystem, the risk assessment model of drill-string system for deep well is established. The drill-string risk analysis of some well in Tarim Basin is taken by applying the risk assessmentmodel, and the variation of drill-string risk with rotation speed and weight on bit is identified:the drill-string risk increases with the weight on bit increases in the case of same rotationspeed; in the case of different rotation speed, the fluctuation of risk value is small while theweight on bit is small, and the fluctuation increases with the weight on bit increases.(3) The residual strength of casing with casing defects and cement annular defects iscalculated. And then the casing safety factor is calculated. According to the size of casingsafety factor, the failure risk assessment model of well structure is build.(4) Considering the seepage action of formation, the stress and strain of rocks in plasticzone and elastic zone around wellbore are calculated, and the potential function of rock strainenergy is obtained. The cusp catastrophe model of wellbore stability is established byreducing the strength parameters in generalized Hoek-Brown criteria continuously. The safetyfactor of wellbore stability for some well is calculated by applying the established cuspcatastrophe model, and the sensitivity analyses of the factors affecting the wellbore stabilityare taken. The results show that the safety factor of wellbore stability is relative large forrocks with good mechanics under the same conditions, the safety factors increases with thedensity of drilling fluid increases and the safety factors decreases with the well depth.(5) Aiming to the complexity of deep well drilling process, the methods in softcomputing which are neural network, rough set theory and genetic algorithm are applied tothe drilling risk assessment in the uncertain and imprecise environments. The rough neuralnetwork is build by combining the reduction and classification abilities and extractioncapacity of decision rules of rough set, the learning ability of artificial neural network and theoptimize search capability of genetic algorithm. The drilling risk assessment model isestablished by applying the signs parameters and mechanics parameters which cause drillingrisk to the rough neural network, and the “Risk Assessment software for Deep-well Drillingsystem” is developed.