Euro-SiBRAM’2002 Prague, June 24 to 26, 2002, Czech Republic

Session 4

Reference Values for Durability Based Performance Design Criteria

Paul J. Tikalsky, Ph.D., P.E., FACI

Associate Professor of Civil and Environmental Engineering,

The Pennsylvania State University, University Park, PA USA

Introduction

Defining the reference value (RV) for most durability parameters is a difficult task that requires detailed knowledge of the environment of the structural element, material science, and deterioration mechanisms. Determining the probability that a specific structural element will not perform the design function because of a single durability parameter is also difficult, if not improper, when determining the failure of a structural element or system. The multitude of deterioration mechanisms that contribute to potential safety or serviceability related failure is a compounding factor in the definition of reference values for durability based performance design. 

Durability based performance design (DBPD) is the most recent technique to address the long life cycle engineering systems. In the past, designers have focused on load based designs, ensuring that the initial load capacity was not exceeded by one-time physical loadings or the cumulative effects of multiple physical loadings. While this technique has lead to statistically safe engineered systems, with respect to life safety, it has often led to short life structures. DBPD builds upon the current load based probabilistic or semi-probabilistic design techniques, by adding important considerations for longer service lives. 

By its very nature, most durability-based deterioration is time dependent. Since durability related distress is often a function of exposure, time, material properties and acceptable performance, it is necessary to understand the nature of each and its interrelationship with other parameters.  Exposure can be thought of as a load effect, and specific material properties can be considered as resistance effects. The exposure is resisted by specific material properties, and material properties can change over time due to exposure. Therefore the load, i.e. exposure, can be function of resistance, and the resistance can change with time.  For example, corrosion of steel in a structure is a function of access to moisture, air, salts, and galvanic current. This is further complicated in reinforce concrete by the passivating effect of high pH pore water solutions.    

Statistically addressing the potential for exposure and the improvement or degradation of material properties that resist exposure with time is a challenging problem. There are no known holistic models available to design engineers to address the durability requirements for bridges, buildings or infrastructural systems. At this point in time, the need for such a model is very high compared to the need to better understand additional mechanics of materials or load based behavior. The complexity of the overlapping models for deterioration and resistance to deterioration can be addressed using simulation based reliability assessment (SBRA) techniques. The SBRA technique is able to simulate multiple scenarios of exposure as well as the resistance to exposure based deterioration. By simulating the variability of exposure and resistance with time simultaneously, the SBRA technique may be able to provide a probability of failure with regards to durability. This performance distribution can be use to simulate additional scenarios for life cycle cost or life-cycle performance. The end result is a better understanding of the variables that effect deterioration and an engineering tool to better select materials and methods that improve the life-cycle performance and reduce the life cycle cost. 

In addition to defining reference values for any particular deterioration mechanism, there is a need to understand that one deterioration mechanism can accelerate another mechanism, or several combined factors may increase exposure and the probability of failure. For example, the magnesium sulfates in seawater may cause sulfate attack in concrete structures. These exposed structural elements become cracked and the chemistry of the pore water changes from its exposure to seawater, as which time the ingress of chlorides enter the concrete to further decrease the passivity of  the pore water, as well as strengthen the electrolytic solution, by which corrosion is accelerated. The safety related distress is the loss of reinforcing due to corrosion, but it would not have been happened in the given time period had not the sulfate attack accelerated the process. 

With this type of interaction in mind, the technical and design community must consider the various forms for durability related distress and determine the acceptable levels of deterioration.  The acceptable level of deterioration may be different for different structural elements and applications. If may also vary with the desired time in service. Therefore the probability of failure must be considered as part of a life-cycle design philosophy. The designer will have a make a realistic estimate on service life. Short service life buildings (electronics manufacturing facilities, suburban retail space, etc.) may be considered in a different class than long-life structures (highway bridges and government buildings, etc). A third class of structures for ultra-long life might be considered for special structures (nuclear waste repositories, monuments, etc.). 

Discussion

The statistical characterization of permeability, diffusivity, corrosion potential or rate, and sulfate resistance is difficult at best. The good news is that these properties are likely normally distributed around a mean value for any give mixture design. The bad news is that the professional community around the world can not easily decide on a form of measurement of these properties. The AASHTO T277 chloride ion permeability test works well for normal portland cement and portland/pozzolan blended cements, but does not work particularly well for the reduction in permeability from chemical admixtures. In many cases, the volume of data to define a suitable mean and standard deviation is either available, but not compiled, or not collected beyond the initial mixture approval process. The histogram for a fly ash/portland cement concrete mixture is not complicated or expensive to generate. The presence of this data is not enough; the data needs some form of standardization. Since the requirement is often related to long term durability, the 28-day test value is not particularly valuable. This is especially true for pozzolanic blends that act to improve the microstructure with time. A 90-day test is more suitable and should be used to standardize the T277 values for performance based specifications. The next step is to tie the T277 result to the rate at which chlorides ingress concrete. This is a critical step in life cycle prediction.  For a particular mixture design, it is possible to have from 30 to 200 test results to document the T277 property of the mixture.

In a similar manner, the sulfate expansion results could be obtained to generate a histogram of the material behavior of the concrete mixture exposed to sulfate environments. In the absence of the desired volume of data, the maximum allowable intra-laboratory bias could be used to predict the behavior. As the concrete production moves forward, the concrete producer would continue to submit results to adjust the prediction. This could be tied to payment. Meeting the required bias, or demonstrating higher quality control by reduced variation would ensure full payment or payment bonuses if verified by independent laboratories. 

The time dependence of particular exposures also plays a role in the prediction and associated reference value. Rarely are concrete exposures saturated solutions for years, or high concentrations with wet dry solutions. These acceleration techniques have to be used in conjunction with controlled field test. The national bridge inventory in the USA is an attempt to do this at the loading and structural resistance.  The same types of comparisons must be done for chloride diffusion, sulfate exposure, and other forms of durability distress. 

The prediction models need more than raw science. The science must be verified by in-situ performance. This is extremely difficult, because the time required to verified life expectancy is much longer that the time for new generations of materials to be brought to the market. By the time we have the data to find a material combination acceptable, that material may be only a historic record.  As such, the use of prediction models has to be closely tied to mechanisms of durability, not specific materials. The corrosion of reinforcing steel must be tied to diffusion, conductivity, oxidation potential, and other inhibiting mechanisms. Then the resistance to corrosion could be predicted by the materials properties of the concrete. Likewise, this methodology would work for sulfate attack, alkali silica reaction and other forms of environmental distress. 

The reference values used in the design of structures has to be based on scientific data and reasonable statistical methods. As designers, if we desire a design life of 75 years, we must decide, if that is an average of 75 years or a 90 percent chance of reaching 75 years. This formidable discussion has not occurred in the engineering community. It is a necessary discussion for performance and life cycle based designs.    

Summary

A more comprehensive set of criteria is needed to define the desired durability properties for structural elements. For reinforced concrete structures with would include: freeze-thaw resistance, scaling resistance, permeability, abrasion resistance, resistance to alkali aggregate reaction, shrinkage, corrosion potential, autogenous and external thermal properties, and sulfate resistance. Bacterial and fungal resistances have to be added for wood structural elements.  Coating quality and thickness have to be added for steel structures. 

Finally, material properties and laboratory testing must be adequately tied to field performance and deterioration for reference values to have real meaning. While the building codes have spent years perfecting the reference values for strength related issues. Serviceability and durability reference values are much more difficult to define and will require considerable effort to incorporate into probabilistic design. 

An acceptable model and methodology is needed to define the probability of failure for durability related distress. The models must be based on material science and verified by field performance.  The combined materials science and field performance will lead to reliable predictors of high level of performance. Only after these steps are completed can design engineers be expected to complete true life-cycle designs based on probabilistic reality.