1. Introduction
Due to the exhaustion of conventional petroleum resources, much attention is being paid to some ultra-deep, unconventional and deep-water petroleum resources. To exploit these kinds of petroleum resources, more and more deep wells and ultra-deep wells are utilized [
1,
2]. The initial deep drilling was just made for petroleum exploration and production, but now is being applied to exploit deep geothermal energy and geo-resources, and to conduct international continental scientific drilling [
1]. The initial oil well was only drilled to a depth of 19.8 m, while current oil wells can be drilled to more than 10,000 m, and the deepest well is the SG-3 well that was drilled to a depth of 12,262 m on the Kola Peninsula in Russia in 1984. Currently, most oil wells can be drilled deeper than 6000–8000 m. Deep and ultra-deep drilling usually encounters some challenges, such as the high-temperature and high-pressure (HTHP), wellbore collapse, wellbore fracture, lost circulation, gas kick, and blowout [
1]. The reasons may be the uncertainty of formation pressure, the uncertainty of formation lithology, the uncertainty of the reservoir interface depth, the uncertainty of completion depth, and the interaction of multiple factors. In order to minimize or avoid the above challenges that encountered in deep and ultra-deep drilling, drilling engineers need to keep the wellbore pressure within a safe window. If the safe window is converted to the equivalent density, the safe window can be called a safe mud weight window (SMWW). The design of mud weight should follow the following principle: the safe mud weight should be higher than the lowest safe mud weight and lower than the highest safe mud weight. In general, the lowest safe mud weight depends on the minimum between the pore pressure and collapse pressure, while the highest safe mud weight depends on the fracture pressure [
3,
4,
5]. The important foundation of SMWW should be determined by wellbore stability. The wellbore instability, a classical rock mechanics problem, is one of the most complex problems that encountered during drilling and completion, which cost the drilling industry certainly more than
$100 million per year worldwide [
5,
6]. Wellbore instability is recognized when the hole diameter is markedly different from the bit size and the hole does not maintain its structural integrity [
3]. In general, the mechanical failure occurs when wellbore stress concentrations exceed the rock strength, and the mechanical failure of wellbore can be classified into two main types (
Figure 1): (1) Compressive failure, i.e., “tight hole” or “stuck pipe” incidents, which are time-consuming to solve and therefore expensive, an increased borehole diameter will occur due to brittle failure and the subsequent caving of the wellbore wall in brittle rocks, a reduced borehole diameter which occurs in weak (plastic) shales, sandstones, and salts; and (2) Tensile failure, i.e., “lost circulation” or “mud loss” problems, which are potentially dangerous, thus representing a safety risk that has to be avoided.
In order to investigate the wellbore stability, a large number of analysis methods has been proposed, such as the elastic model, plastic model, elastoplastic model, poro-elastic model, thermo-poro-elastic model, chemo-poro-elastic model and chemo-thermo-poro-elastic model [
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
12,
13]. However, due to the uncertainty of formation lithology, the uncertainty of formation pressure, the uncertainty of mechanical properties of rocks, and the unstable wellbore pressure, the input parameters of wellbore stability analysis never can be known precisely. In other words, the input parameters are often uncertain, which might cause an incorrect result [
14,
15,
16,
17,
18,
19,
20,
21,
22,
23,
24,
25,
26,
27,
28,
29,
30]. In order to quantify the influence of uncertain parameters on wellbore stability and SMWW, it’s necessary to utilize reliability assessment method. Morita [
14] conducted an uncertainty analysis of borehole stability based on a statistical error analysis method. McLellan and Hawkes [
15] indicated that the traditional models fail to account for the inherent variability of rock properties, as well as uncertain values for input parameters, a probabilistic technique was used to evaluate the risks of borehole instability or sand production, and a spreadsheet-based Monte Carlo simulation tool was involved. Ottesen et al. [
16] present a new analysis method of wellbore stability based on the quantitative risk analysis (QRA) principles. De Fontoura [
17] investigated three analytical methods for evaluating the influence of parameter uncertainties on wellbore stability, such as the first order second moment, first order reliability model and statistical error analysis method, and contrasted with the results generated by Monte Carlo method. Moos et al. [
18] and Zoback [
6] presented the use of QRA to formally account for the uncertainty in each input parameter to assess the probability of achieving a desired degree of wellbore stability at a given mud weight. Sheng et al. [
19] indicated that because of uncertainty in some key influential parameters, these can lead to uncertainty of wellbore stability, a Monte Carlo uncertainty analysis technique was combined with a numerical geomechanical modelling method to develop a geostatistical approach to determine the uncertainty of wellbore stability. Al-Ajmi and Al-Harthy [
20] calculated the value of drilling fluid pressure as a probability distribution by utilizing a probabilistic approach captures uncertainty in input variables through running a Monte Carlo simulation. Aadnøy [
21], Aadnoy and Looyeh [
4] also utilized QRA to assess the effects of the errors or uncertainties associated with the key data on the instability analysis. Zhang et al. [
22] deduced a reliability calculation formula for collapse pressure in coal seam drilling by integration of the reliability theory and the Hoek-Brown criterion, and the Weibull distribution and Monte Carlo simulation was utilized in this study. Udegbunam et al. [
23] investigated typical fracture and collapse models with respect to inaccuracies in the input data with a stochastic method. Kinik et al. [
24,
25] presented a mathematical model to estimate the true fracture pressure from the leak-off-test data, and the QRA was also involved to represent the probability-density distribution of fracture pressure. Gholami et al. [
26] applied the QRA to consider the uncertainty of input parameters in determining of the mud weight window for different failure criteria, such as the Mohr-Coulomb (M-C), Hoek-Brown (H-B) and Mogi-Coulomb (MG-C) criterion. Plazas et al. [
27] investigated a stochastic approach along with the conventional wellbore stability analysis enables to take into account uncertainty from input data.
Although the reliability assessment method had been integrated with wellbore stability models, but most of the above studies still have some shortcomings: (1) The required mud weight to prevent wellbore collapse is too high [
28,
29], due to the fact traditional wellbore stability models assume that the critical failure appears at the highest point of stress concentration, in other words, it’s performed to yield no shear failure along the borehole wall [
2,
28]. In real drilling engineering, a tolerable breakout or an appropriate breakout width will not cause an unbearable collapse problem, and it can help lower the required mud weight to prevent collapse; once a tolerable breakout or an appropriate breakout width was involved, the SMWW can be therefore widened, and it’s beneficial to the narrow SMWW formations. However, most of the above studies did not involve a tolerable breakout or an appropriate breakout width, only Morita [
14] and Aadnoy and Looyeh [
4] investigated its influence on wellbore collapse. (2) Most of wellbore stability analysis that involved the reliability assessment method always uses the M-C criterion to determine the shear failure, while the M-C criterion ignored the influence of the intermediate principal stress, which makes the required mud weight to prevent wellbore collapse too conservative. Al-Ajmi and Zimmerman [
7,
8] introduced an analytical solution of collapse pressure in conjunction with a true triaxial criterion, i.e., the MG-C criterion, it overcomes the above defect. (3) Most of the above studies just focus on the uncertainty evaluation of wellbore collapse, and the uncertainty evaluation of SMWW is seldom investigated, and the influences of uncertain basic parameters on SMWW are also seldom investigated. Therefore, this study takes the tolerable breakout and the MG-C criterion into account to determine the required SMWW by using the reliability assessment method. Firstly, the basic principle of reliability assessment theory was introduced briefly. Secondly, the uncertain SMWW model was proposed by involving the tolerable breakout, the MG-C criterion and the reliability assessment method. Thirdly, the uncertainty of input parameters was investigated, and the uncertainty analysis of the equivalent mud weight of collapse pressure (EMWCP), the equivalent mud weight of fracture pressure (EMWFP), the equivalent mud weight of collapse pressure (EMWPP) and the SMWW were investigated. Finally, the field observation of well SC-101X was reported and discussed.
2. Reliability Assessment Theory
Reliability calculations provide a means of evaluating the combined effects of uncertainties, and a means of distinguishing between conditions where uncertainties are particularly high or low [
29]. “Reliability” as it is used in reliability theory is the probability of an event occurring or the probability of a “positive outcome”. Based on the theory of reliability assessment, we can sort the influencing factors into two types: (1) Loads
Q, and (2) Resistances
R [
30]. The SMWW assessment needs to consider the influence of well kick, wellbore collapse, and wellbore fracture:
For well kick, the loads Q denotes the pore pressure, while the resistances R denotes the wellbore pressure; and the basic random variables of loads and resistances of well control can be assumed as , .
For wellbore collapse, the loads Q denotes the collapse pressure that controlled by in-situ stress, pore pressure and rock properties; while the resistances R denotes the wellbore pressure and rock strength; and the basic random variables of loads and resistances of wellbore collapse can be assumed as , .
For wellbore fracture, the loads Q denotes the wellbore pressure; while the resistances R denotes the collapse pressure that controlled by in-situ stress, pore pressure, rock properties and rock strength; and the basic random variables of loads and resistances of wellbore fracture can be assumed as , .
The loads and resistances can be assumed as:
where
Qk and
Rk are the loads and resistances;
k denotes the subscript,
k can value for K, C and F for well kick, wellbore collapse, and wellbore fracture, respectively.
The margin of safety,
M, is the difference between the resistance and the load. The margin of safety for well control, wellbore collapse and wellbore fracture can be expressed as [
30,
31]:
where
Mk is the margin of safety for factor
k.
In general, the loads and resistances are two independent random variables, and the reliability and failure probability can be expressed as [
30,
31]:
where
Prk is the reliability of factor
k;
Pfk is the failure probability of factor
k.
There is a relationship between the reliability and failure probability:
If we assume that the probability density function of
Qk and
Rk are
fQ(
Qk) and
fR(
Rk), respectively.
Figure 2 shows the interference of the probability density function of
Qk and
Rk, we just take the assessment of wellbore collapse as an example, the well control and wellbore fracture is similar. The overlapping zone denotes the probability of failure. The smaller overlapping zone, the wellbore will be more reliable, i.e., the risk of wellbore collapse will be lower; the bigger overlapping zone, the wellbore will be more unreliable, i.e., the risk of wellbore collapse will be higher. The reliability and failure probability can be expressed as [
31]:
where
fQ(
Qk) and
fR(
Rk) are the probability density function of
Qk and
Rk, respectively.