Prediction of the in‐situ secant elasticity modulus (E50) and unconfined compressive strength (UCS) of deep soil mixing (DSM) columns is inevitable for the design process of soil improvement in weak soils. Existing literature commonly rely on deterministic linear correlations between E50 and UCS for DSM columns. The broad range of proposed correlations ( \({E}_{50}\cong 45\sim 450\times \text{UCS}\) ) can significantly influence the estimated settlement of improved ground and may lead to unconservative designs. Furthermore, establishing a single deterministic relationship between E50 and UCS is unlikely due to inherent uncertainties in soil–cement interactions in the field. Hence, to fill the gap, the present study employs a probabilistic approach (the Multiple Stripe Analysis) to correlate E50 and UCS of DSM columns using nearly 7000 UCS tests on field samples. The results revealed substantial discrepancies between measured E50 values and those predicted by commonly used deterministic correlations, such as the FHWA recommendation (E50 = 300 × UCS), highlighting the need for a probabilistic framework. Moreover, the practical probabilistic prediction curves presented in this study demonstrated that, based on an 80% confidence level, achieving 200, 400, and 600 MPa elasticity modulus for DSM columns at the design stage was contingent upon obtaining at least 1.5, 3, and 5 MPa UCS values, respectively. In contrast, the FHWA deterministic recommended equation predicted considerably higher E50 for the mentioned UCS values (i.e., 450, 900, and 1500 MPa, respectively), which may lead to overly optimistic and potentially unsafe design assumptions.
