Machine Learning Algorithm To Predict Urea Crystals In Exhaust Lines

Claims:1. A machine learning algorithm (10) to predict composition and quantity of urea 10 crystals in an exhaust gas mixing section, the machine learning algorithm (10) comprising the steps of: inputting a plurality of parameters including exhaust mass flow rate (12), upstream and downstream NOx emissions (14), selective catalytic reduction NOx conversion efficiency (16), and exhaust temperature and dosing quantity 15 (18) into an algorithm; and estimating (50) the composition and quantity of urea crystals deposited in a mixing section based on discrete time intervals that the urea crystals in the mixing section is subjected to high temperature ranges over finite intervals of time. 2. The machine learning algorithm (10) to predict quantity and composition of urea crystals in an exhaust gas mixing section in accordance with Claim 1 wherein inputting exhaust mass flow rate (12), upstream and downstream NOx emissions (14), selective catalytic reduction NOx conversion efficiency (16), 25 and exhaust temperature and dosing quantity (18) into an algorithm further comprises calculating (20) a theoretical stoichiometric dosing quantity of urea into the mixing section by the algorithm from the input in combination with the actual quantity of NOx that flows into the mixing section from an engine, the theoretical stoichiometric dosing quantity of urea adapted to determine the 30 dosing quantity of urea to be dosed for creating no or minimal urea crystal deposits on a wall of said mixing section. 3. The machine learning algorithm (10) to predict quantity 5 and composition of urea crystals in an exhaust gas mixing section in accordance with Claim 2 further comprising determining (22) if the actual quantity of urea that is dosed into the mixing section is greater than the theoretical stoichiometric quantity of urea dosed into the mixing section. 10 4. The machine learning algorithm (10) to predict quantity and composition of urea crystals in an exhaust gas mixing section in accordance with Claim 3 further comprising estimating (24) by means of a first machine learning algorithm as to the quantity of ammonia that is currently loaded in the selective 15 catalytic reduction system by means of the temperature of exhaust gases, if the actual quantity of urea that is dosed into the mixing section is greater than the theoretical stoichiometric quantity of urea dosed into said mixing section. 5. The machine learning algorithm (10) to predict quantity and composition of 20 urea crystals in an exhaust gas mixing section in accordance with Claim 4 further comprising computing (26) by means of a second machine learning algorithm as to the quantity of ammonia that is loaded or unloaded in the selective catalytic reduction system based on the quantity of ammonia that exists in the selective catalytic reduction system in the span of two consecutive 25 time intervals. 6. The machine learning algorithm (10) to predict quantity and composition of urea crystals in an exhaust gas mixing section in accordance with Claim 5 further comprising calculating (28) the quantity of urea that is deposited in said 30 mixing section by subtracting the quantity of ammonia that is loaded in the selective catalytic reduction system from the actual quantity of urea that is dosed into said mixing section and multiplying the difference with (1 – NOx conversion efficiency) if ammonia is loaded in the selective catalytic 5 reduction system. 7. The machine learning algorithm (10) to predict quantity and composition of urea crystals in an exhaust gas mixing section in accordance with Claim 6 10 further comprising calculating (30) a quantity of ammonia that is deposited in said mixing section by multiplying the actual quantity of urea that is dosed into said mixing section with (1 – NOx conversion efficiency) if ammonia is being consumed from the selective catalytic reduction system. 15 8. The machine learning algorithm (10) to predict quantity and composition of urea crystals in an exhaust gas mixing section in accordance with Claim 7 further comprising (32) a quantity of unreacted urea that exists in said mixing section by subtracting a decomposed quantity of solid urea due to exhaust gas that flows over urea that is deposited into said mixing section at a known 20 temperature, and that occurs over a predetermined time interval from the quantity of urea that is deposited in the mixing section. 9. The machine learning algorithm (10) to predict quantity and composition of urea crystals in an exhaust gas mixing section in accordance with Claim 8 25 further comprising (50) determining the composition of urea deposits in said mixing section based on discrete temperature ranges that the urea deposits in said the mixing section is subjected to over corresponding discrete time intervals as measured over a finite interval of time. , Description:A machine learning algorithm 10 to predict composition and quantity of urea crystals in an exhaust gas mixing section is described. The machine learning algorithm 10 comprises the steps of inputting a plurality of parameters including exhaust mass flow 10 rate12, upstream and downstream NOx emissions 14, selective catalytic reduction NOx conversion efficiency 16, and exhaust temperature 18 into an algorithm. The method further comprises computing a composition of urea crystals deposited in the mixing section based on discrete time intervals that a quantity of urea crystals that exists in the mixing section is subjected to over corresponding discrete temperature 15 ranges over a finite interval of time.