Experiment
Concentrations, flow, timing, surface capacity, controls and complete sensorgrams.
The physical model remains in control. Machine learning accelerates forward prediction and probabilistic inversion while competing mechanisms, uncertainty and identifiability stay visible.
Concentrations, flow, timing, surface capacity, controls and complete sensorgrams.
Langmuir, Bi-Langmuir, transport-aware and fractional alternatives share one observation model.
Physically valid parameter combinations generate labelled families of complete sensorgrams.
Fast forward prediction and inverse distributions over kinetics, fractional order and nuisance parameters.
Validate identifiability and select the next concentration, flow rate or contact time.
The established 1:1 model remains the reference and limiting case, not a discarded predecessor.
A non-integer order represents effective memory and unresolved timescales; it is not a fractional number of physical states.
Complexity earns its place only through held-out prediction, calibrated uncertainty and identifiable parameters.
The 2016 aligned-macroporous-monolith publication is useful prior literature for templated structures and protein chromatography. It helps frame the broader manufacturing direction.
The external paper is not authored by the notebook author, does not document this code, and must not be presented as its associated publication or as evidence of ownership.
The destination is a transparent SPR oracle: conditions and sensorgrams in; the least complex supported explanation, kinetic states, uncertainty, model adequacy and the next informative experiment out.