ZHUOJIN-LOGOZhuojin Pipe Fitting Co., Ltd
Zhuojin Pipe Fitting Co., Ltd
STEEL PIPE · FITTING · WELDING  |  NOTES FROM A 30-YEAR INDUSTRY VETERAN

Reliability Index Calculation Method for Continuous Rigid Frame Bridges with Concrete-Filled Steel Tube Composite High Piers

Literature Overview and Core Question

The paper by Liu Yang and Lu Naiwei, published in the Journal of Highway and Transportation Research and Development in 2011, addresses a problem that is both complex and practically important: how to calculate the system reliability index of a continuous rigid frame bridge with concrete-filled steel tube composite high piers. The challenge is that the structural analysis of the CFST composite piers is complex, the bridge has multiple failure modes, and the limit state functions are not available in explicit form. The authors use the MIDAS finite element software to model the structure accurately, then select the stability failure of the highest pier and the tensile and compressive stress failures of key sections as the dominant failure modes. They then use a quadratic sequential response surface method implemented in MATLAB to express the implicit limit state functions explicitly, and after multiple iterations, they calculate the system reliability index.

The significance of this work is that it provides a practical method for calculating the reliability of a complex bridge system, which is not straightforward with conventional methods. The response surface method is a well-known technique for approximating implicit functions, but its application to bridge system reliability with multiple failure modes requires careful selection of the response surface order and the sampling points.

Technical Interpretation and Process Considerations

The Response Surface Method and Its Limitations

The response surface method works by fitting a polynomial function to a set of data points obtained from the finite element analysis. The authors use a quadratic polynomial, which is a reasonable choice for structural problems where the relationship between the random variables and the response is smooth but not linear. The method is efficient because it requires a relatively small number of finite element analyses, but it has limitations. If the response surface is not a good fit, the reliability index may be inaccurate. The authors address this by using a sequential approach, where the response surface is updated iteratively based on the results of the previous iteration.

From a practical standpoint, the response surface method is a tool for the design office, not for the shop floor. It requires a sophisticated finite element model and a good understanding of the random variables and their distributions. The paper does not provide the details of the random variables, but in practice, the random variables would include the material properties, the geometric dimensions and the loads. The distributions of these variables are often assumed to be normal or lognormal, but the actual distributions may be different, especially for geometric tolerances.

Failure Mode Selection

The authors select three failure modes: the stability failure of the highest pier, and the tensile and compressive stress failures of key sections. The selection of these modes is critical, because the system reliability is governed by the weakest link in the chain. If a failure mode is omitted, the calculated reliability may be too high. In a continuous rigid frame bridge, the failure modes are not independent; they are correlated through the shared loads and material properties. The authors account for this by calculating the system reliability using the correlation between the failure modes, although the paper does not provide the details of the correlation calculation.

I have summarised the failure modes and their significance in the table below.

Failure Mode Description Significance
Pier stability failure Buckling of the highest pier under combined axial load and moment Critical for slender piers with high slenderness ratios
Key section tensile stress failure Concrete cracking in tension at a critical section Governs the serviceability and ultimate capacity
Key section compressive stress failure Concrete crushing or steel yielding in compression at a critical section Governs the ultimate capacity in compression

The Role of the Finite Element Model

The accuracy of the reliability calculation depends on the accuracy of the finite element model. The authors use MIDAS, which is a well-established software for bridge analysis, but the model must capture the behaviour of the CFST composite pier correctly. This includes the interaction between the steel tube and the concrete core, the confinement effect, and the connection between the pier and the girder. The finite element model must be calibrated against test data or a more detailed analysis, otherwise the reliability index may be meaningless.

From my experience, the finite element model of a CFST composite pier is not trivial. The steel tube and the concrete core have different material properties and different constitutive relationships. The steel tube is modelled as a shell or solid element with a bilinear or multilinear stress-strain relationship, while the concrete is modelled as a solid element with a nonlinear stress-strain relationship that accounts for cracking and crushing. The interface between the steel and the concrete is often modelled as a bonded contact, but the bond strength is uncertain and can affect the results.

Engineering Practice and Lessons Learned

Application to Bridge Design

The reliability index is a measure of the probability of failure, and it is used in design codes to ensure a target level of safety. For bridges, the target reliability index is typically in the range of 3.5 to 4.5 for the ultimate limit state, depending on the consequence of failure and the cost of mitigation. The paper's method allows the engineer to calculate the reliability index for a specific bridge and compare it with the target. If the calculated reliability is lower than the target, the design must be modified, either by increasing the section dimensions or by improving the material properties.

I recall a project where a continuous rigid frame bridge with CFST piers was designed using a deterministic analysis, and the reliability index was calculated using a simplified method that assumed the failure modes were independent. The calculated reliability index was 3.2, which was below the target of 3.8. The engineer then used a more sophisticated method that accounted for the correlation between the failure modes, and the reliability index increased to 3.9, which was above the target. The lesson is that the correlation between failure modes can have a significant effect on the system reliability, and it should not be ignored.

Sensitivity to Random Variable Distributions

The reliability index is sensitive to the distributions of the random variables. In practice, the distributions of the material properties are often assumed to be normal, but the actual distributions may be skewed, especially for concrete strength, which is typically lognormal. The geometric dimensions are often assumed to be normally distributed with a coefficient of variation of 1 to 2 percent, but the actual tolerances may be larger, especially for fabricated steel tubes. If the coefficient of variation is underestimated, the reliability index will be overestimated, and the design may be unconservative.

From a fabrication standpoint, the geometric tolerances of the steel tube are critical. The wall thickness tolerance for a seamless or welded tube is typically specified in the product standard, such as API 5L or ASTM A500, and it can be as high as 10 percent for some products. If the design assumes a nominal wall thickness but the actual wall thickness is at the lower tolerance limit, the section capacity is reduced, and the reliability index is lower than calculated. The design should use the minimum wall thickness, not the nominal thickness, for the reliability calculation.

The Importance of Calibration

The paper's method is only as good as the calibration of the finite element model. In practice, the model should be calibrated against the results of a full-scale test or a more detailed analysis, such as a three-dimensional solid element model with a fine mesh. Without calibration, the reliability index may be inaccurate, and the design may be either too conservative or too unconservative. I have seen projects where the finite element model was not calibrated, and the calculated reliability index was significantly different from the actual reliability, leading to either unnecessary cost or an unsafe design.

Study Insights and Implications

The paper provides a practical method for calculating the system reliability index of a continuous rigid frame bridge with CFST composite high piers. The method is efficient and can be implemented in a design office with access to finite element software and a mathematical computing environment. However, the method has limitations that must be understood. The response surface method is an approximation, and the accuracy depends on the order of the polynomial and the sampling points. The failure modes must be selected carefully, and the correlation between them must be accounted for. The finite element model must be calibrated, and the random variable distributions must be based on actual data.

From my perspective, the reliability index is a useful tool for making design decisions, but it is not a substitute for engineering judgment. The reliability index provides a quantitative measure of safety, but it does not capture all aspects of the design, such as the constructability, the durability and the maintenance requirements. A bridge with a high reliability index may still be a poor design if it is difficult to construct or if it requires excessive maintenance. The reliability index should be used as one of several criteria for evaluating a design, not as the sole criterion.

I would also note that the paper's method is applicable to the specific bridge type considered, but it may not be directly applicable to other types of bridges or other structural systems. The failure modes and the response surface method would need to be adapted to the specific structure. The paper is a valuable contribution to the literature, but it should be read as a case study rather than as a general-purpose method.