Due to its plethora of nitrogen (N), phosphorus (P), potassium (K) in urine, it is bound to trigger phosphorus coprecipitation, thereby adversely affecting K-struvite purity in the coprecipitates. To obtain high pure K-struvite, the present study was to innovatively explore the effect of residual NH4+ concentration, pH and initial Mg2+ concentration on phosphorus coprecipitation in synthetic urine. Importantly, a Back-Propagation Artificial Neural Network (BPANN) model was innovatively proposed to simulate and predict crystal purities in the coprecipitates. It was revealed that K-struvite, struvite, hydroxyapatite and cattiite dominated the coprecipitates. Comparatively, the content of calcium in synthetic urine is far lower than that of phosphorus and potassium, resulting in low hydroxyapatite purity in the coprecipitates. Notably, cattiite purity is highly dependent of Mg2+ concentration, because it was low at the Mg2+ concentration of <10 mmol/L, but increased up to above 50% at the Mg2+ concentration of 50 mmol/L.At 10 mmol/L Mg2+ and pH 10, K-struvite purity in the coprecipitates decreased from 64.6% to 43.3% following the increase of NH4+ concentration from 0 to 300 mg/L. The BPANN model well simulated and predicted the purities of the crystals in the coprecipitates from synthetic urine. At 10 mmol/L Mg2+ and 100 mg/L NH4+, an increase in pH from 8.5 to 10 can facilitate K-struvite crystallization in synthetic urine. The adjustment of pH and initial Mg2+ concentration can significantly mitigate the inhibitory effect of residual NH4+ on K-struvite crystallization. The BPANN model herein can effectively obtain optimized operational parameters for the full-scale implementation of slow-release NPK fertilizers from urine, which can also provide an effective reference for nutrient recovery from various waste streams.
Read full abstract