Aiming at the problems of over-fitting, local optimum, curve distortion and error amplification in the automatic drawing of raw coal washability curve, this paper proposes an improved method based on BIC model selection, multi-starting point parameter search, monotonicity constraint test and four-curve independent fitting. This method establishes candidate models for the cumulative ash β curve of floats, the λ curve of elementary ash, the cumulative yield θ curve of sediments and the density δ curve, respectively, and automatically selects the best through the BIC criterion. At the same time, monotonicity screening and post-processing constraints are introduced to ensure that the fitting results meet the physical meaning of the optional curve. Based on Python 's open source libraries such as Pandas, NumPy, SciPy, lmfit, scikit-learn and Matplotlib, a full-process automation program from floating and sinking test data reading, curve fitting to standard graphics output was constructed. The verification of 30 sets of measured float-and-sink data shows that this method is superior to the traditional scheme in terms of fitting accuracy, stability and curve shape rationality, which can provide reference for the standardized drawing of optional curve and its engineering application.