We read with great interest the paper by Liu et al. in April’s issue of Geotechnique (see here). The paper examines the determination of representative strength of deep cement-mixed clay from cores extracted from improved ground. We’re in the process of writing up a couple of case studies based around the same topic and we decided to share some of our findings in advance. Straightaway, we’re in agreement that the criterion for examining strength should be the mean strength with consideration of standard deviation. We’ll have to examine the particular proposal put forward by Liu et al. in more detail but, intuitively, we see this as appropriate. Here’s why.
Basic material science tells us that the Normal Distribution (bell curve) for engineered materials like steel and concrete is well-defined and generally quite narrow. Our codes of practice are normally concerned with the 95th percentile or higher. Soil mechanics on the other hand tells us that for natural materials subject to the vagaries of depositional processes, geological time and external influencing factors, such as glaciation or erosion, are inherently variable i.e., the bell curve for any given soil parameter is going to be wide – very wide compared to an engineered material. In ground improvement, we’re concerned with narrowing the bell curve. On the assumption that the innate structure of the soil is more-or-less destroyed by the mixing process, let’s take for example soil density. In an improved ground, we might anticipate a correlation between soil density and increased cube strength, right? Well, what we’ve found over a decent sample size is that the correlation, at best, is weak. In fact, we could say that, for the density of soils that we might expect to be improving, there really isn’t much of a correlation at all.
Equally in examining the strength results statistically, we can observe a wide degree of scatter or deviation from the mean. There is however, a strongly perceptible mean result and, we’ve observed that the deviation from the mean narrows with time as the cement hydrates.
Therefore, we would concur with the central hypothesis of the paper and, given strength is the central design parameter, we might also argue that cement-mixed soil shouldn’t be over-designed. We’ve consistently noticed that target strengths of 1MPa will hit 1.1-1.2MPa at 48 days, i.e. the rate of strength increase carries on well after 28 days and why shouldn’t we factor this into designs and prescribed 48 or 56 day tests to verify?
Kudos to Liu et al. for this very useful paper which marries real data with a solid mathematical approach. We look forward to comparing our own UK-based data to their results from Singapore in more detail soon.