Long-distance propagation of foam is one key to deep gas mobility control for enhanced oil recovery and CO₂ sequestration. It depends on two processes: convection of bubbles and foam generation at the displacement front. Prior studies with N₂ foam show the existence of a critical threshold for foam generation in terms of a minimum pressure gradient or minimum total interstitial velocity, beyond which strong-foam generation is triggered. The same mechanism controls foam propagation. There are few data for or for CO₂ foam. We extend previous studies to quantify and for CO₂ foam generation, and relate and with factors including injected quality (gas volume fraction in the fluids injected) - f_g, surfactant concentration - C_s, and permeability - K. In each experiment, steady pressure gradient, ∇p, is measured at fixed injection rate and quality, with total interstitial velocity, v_t, increasing-then-decreasing in a series of steps. The trigger for strong-foam generation features an abrupt jump in ∇p upon an increase in v_t. In most cases, the data for ∇p as a function of v_t identify three regimes: coarse foam with low ∇p, an abrupt jump in ∇p, and strong foam with high ∇p. The abrupt jump in ∇p upon foam generation demonstrates the existence of and for CO₂ foam. We further show how and scale with f_g, C_s and K. Conditions that stabilize lamellae reduce the values of the thresholds: both and increase with f_g and decrease with increasing C_s or K. Specifically, scales with f_g as (f_g)² and scales as (f_g)⁴, and both and scale with C_s as (C_s)^−0.4. The effect of K on the thresholds for foam generation is greater than the effects of f_g and C_s. Our data in artificial consolidated cores show that scales with K as K⁻² for CO₂ foam, in comparison to K⁻¹ for N₂ foam in unconsolidated sand/bead packs. More data are needed to verify the confidence of these correlations. It is encouraging that in the cores with K = 270 mD or greater is less than 0.17 bar/m (~ 0.75 psi/ft), 2 to 3 orders of magnitude less than for N₂ foam. Such low can be easily attainable throughout a formation. This suggests that: limited ∇p deep in formations is much less of a restriction for long-distance propagation of CO₂ foam than for N₂ foam. Foam propagation could still be challenging in low-K reservoirs (~ 10 bar/m for K = 27 mD). Nevertheless, formation heterogeneity and alternating slug injection of gas and liquid help foam generation and may well reduce the values of. More research is needed to predict long-distance propagation of foam under those conditions.

Engineering faces many wicked problems: irreducibly interdisciplinary with multiple competing objectives, and of such large scale and complexity that will require processes to deeply rely on human insights and power of computation. The resurgence of machine learning offers the possibility for new forms of human/computer collaboration where each fuels hybrid intelligence in complementary ways. A concept of virtual design assistant (VDA) is developed as a platform to bring the hybrid intelligence in solving complex design challenges. A deep learning-based abstraction process is developed to provide VDA a function to extract structured functional requirements from fragmental design specifications and customer needs.