EXHAUST EMISSION PREDICTION SYSTEM AND METHOD

Abstract

An exhaust emission prediction system includes an engine configured to generate a flow of exhaust and a controller configured to determine a first estimation of an amount of an emissions constituent at a first location using an empirical model. The first location is downstream of the engine. The controller is also configured to determine a second estimation of the amount of the emissions constituent at the first location using a physics-based model and determine a third estimation of the amount of the emissions constituent at the first location based on at least one of the first estimation or the second estimation.

Description

TECHNICAL FIELD

The present disclosure relates generally to an exhaust system, and more particularly, to an exhaust emission prediction system and method.

BACKGROUND

Internal combustion engines, including diesel engines, gasoline engines, gaseous fuel-powered engines, and other engines known in the art, may produce a flow of exhaust composed of gaseous and solid compounds, including particulate matter, nitrogen oxides (NOx), and sulfur compounds. Due to heightened environmental concerns, exhaust emission standards have become increasingly stringent. The amount of one or more constituents of the flow of exhaust emitted from the engine may be regulated depending on the type, size, and/or class of engine.

One method that has been implemented by engine manufacturers to comply with the regulation of NOx exhausted to the environment is a strategy called selective catalytic reduction (SCR). SCR is a process by which gaseous or liquid reductant (e.g., a mixture of urea and water) is injected into the flow of exhaust from the engine. The combined flow may form ammonia (NH₃), which may then be absorbed onto an SCR catalyst. The ammonia may react with NOx in the flow of exhaust to form H₂O and N₂, thereby reducing the amount of NOx in the flow of exhaust.

The ability of the SCR catalyst to reduce NOx depends upon many factors, such as catalyst formulation, the size of the SCR catalyst, exhaust gas temperature, and urea dosing rate. With regard to the dosing rate, the NOx reduction efficiency tends to increase linearly until the dosing rate reaches a certain limit. Above the limit, the NOx reduction efficiency may increase at a slower rate because the ammonia may be supplied at a faster rate than the NOx reduction process can consume. The excess ammonia, known as ammonia slip, may be expelled from the SCR catalyst.

The urea dosing rate may be controlled using signals from a NOx sensing device, such as a NOx sensor, placed in the exhaust stream after the SCR catalyst. The NOx sensing device may measure the level of NOx and provide signals to a SCR controller to adjust the urea dosing rate. Although NOx reduction efficiency may be increased using this process, the costs and maintenance associated with NOx sensing devices may make implementing this process unattractive to engine manufacturers.

To minimize the costs associated with physical sensors, some conventional engine control systems may implement virtual sensors. For example, U.S. Pat. No. 6,236,908 issued to Cheng et al. (the '908 patent) describes an engine control module (ECM) including one or more neural networks that act as virtual sensing devices to replace or enhance traditional physical sensors. The ECM receives values associated with various engine operating parameters from a plurality of physical sensors and applies the values to the neural network to produce values for one or more output parameters. For example, the neural network may receive values, such as engine speed, manifold pressure, exhaust gas recirculation, and air/flow ratio values, from physical sensors. Based on the input values, the neural network may determine values of other engine operating parameters, including residual mass fraction, emissions, knock index, peak pressure rise rate, exhaust gas temperature, and exhaust gas oxygen content. The neural network is trained using data produced by a simulation model calibrated with actual engine test data.

Although the '908 patent suggests the ability to use a neural network as a virtual sensing device to replace or enhance traditional physical sensors, the available engine test data used to train the neural network is limited. The '908 patent describes that the simulation model interpolates or extrapolates a more complete set of data. However, the interpolated or extrapolated data may not be accurate, which may cause the neural network to provide inaccurate outputs.

The disclosed system is directed to overcoming one or more of the problems set forth above.

SUMMARY

In one aspect, the present disclosure is directed to an exhaust emission prediction system. The exhaust emission prediction system includes an engine configured to generate a flow of exhaust and a controller configured to determine a first estimation of an amount of an emissions constituent at a first location using an empirical model. The first location is downstream of the engine. The controller is also configured to determine a second estimation of the amount of the emissions constituent at the first location using a physics-based model and determine a third estimation of the amount of the emissions constituent at the first location based on at least one of the first estimation or the second estimation.

In another aspect, the present disclosure is directed to a method of predicting an amount of NOx in a flow of exhaust from an engine using a controller. The method includes determining, using the controller, a first estimation of the amount of NOx at a first location using an empirical model, the first location being downstream of the engine. The method also includes determining, using the controller, a second estimation of the amount of NOx at the first location using a physics-based model, and determining, using the controller, a third estimation of the amount of NOx at the first location based on at least one of the first estimation or the second estimation.

In another aspect, the present disclosure is directed to an engine system including an engine configured to generate a flow of exhaust, an injector configured to inject a reductant into the flow of exhaust, and a catalytic device configured to receive the flow of exhaust after being injected with the reductant. The engine system also includes a processor and a memory module configured to store instructions that, when executed, enable the processor to determine a first estimation of an amount of NOx at a first location using an empirical model. The first location is downstream of the engine and upstream of the catalytic device. The memory module is also configured to store instructions that, when executed, enable the processor to determine a second estimation of the amount of NOx at the first location using a physics-based model, determine a third estimation of the amount of NOx at the first location based on at least one of the first estimation or the second estimation, and adjust an amount of the reductant injected by the injector based on the determined third estimation.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a diagrammatic illustration of an engine and an exhaust emission prediction system, according to an exemplary embodiment;

FIG. 2 is a diagrammatic illustration of a controller for the exhaust emission prediction system of FIG. 1;

FIG. 3 is a diagrammatic illustration of an empirical NOx model for the controller of FIG. 2;

FIG. 4 is a diagrammatic illustration of a physics-based NOx model for the controller of FIG. 2; and

FIG. 5 is a flow chart illustrating an exemplary disclosed method of estimating an amount of an emissions constituent in a flow of exhaust, according to an exemplary embodiment.

DETAILED DESCRIPTION

Reference will now be made in detail to exemplary embodiments, which are illustrated in the accompanying drawings. Wherever possible, the same reference numbers will be used throughout the drawings to refer to the same or like parts.

FIG. 1 is a diagrammatic illustration of a power source, such as an engine 10, of a machine and an exhaust emission prediction system, according to an exemplary embodiment. The disclosed embodiment may be applicable to various types of machines such as, for example, a fixed or mobile machine that performs some type of operation associated with an industry such as mining, construction, farming, transportation, power generation, tree harvesting, forestry, or any other industry known in the art. The engine 10 may be an internal combustion engine, such as, for example, a diesel engine, a gasoline engine, a gaseous fuel-powered engine, or any other engine apparent to one skilled in the art. The engine 10 may alternatively be another source of power such as a furnace or any other suitable source of power for a powered system such as a factory or power plant.

Operation of the engine 10 may produce power and a flow of exhaust. For example, the engine 10 may include a plurality of cylinders 12. Each cylinder 12 may include a combustion chamber that may mix fuel with air and/or recirculated exhaust gas, as described below, and burn the mixture therein to produce the flow of exhaust. The flow of exhaust may contain carbon monoxide, NOx, carbon dioxide, aldehydes, soot, oxygen, nitrogen, water vapor, and/or hydrocarbons.

An exhaust system 14 is provided with the engine 10 such that the flow of exhaust may be fluidly communicated from the engine 10 to the exhaust system 14. The flow of exhaust produced by the engine 10 may be directed from the engine 10 to components of the exhaust system 14 by flow lines. For example, as shown in FIG. 1, the flow lines may include pipes, tubing, conduits, and/or other exhaust-carrying structures known in the art through which the flow of exhaust may be directed through the exhaust system 14 to one or more of a turbine 16 of a turbocharger 18, one or more aftertreatment devices 20, an injector 22, and a catalytic device (e.g., an SCR catalyst 24) in the exhaust system 14. The exhaust system 14 may also include additional components for directing the flow of exhaust out of the engine 10 that are known in the art.

The turbine 16 may be disposed between an exhaust passageway of the engine 10 and the inlet of the exhaust system 14. The turbine 16 may be configured to drive a connected compressor 26 of the turbocharger 18. For example, as the hot exhaust gases exiting the engine 10 expand against blades (not shown) of the turbine 16, the turbine 16 may rotate and drive the compressor 26. The compressor 26 may be located in an air induction system of the engine 10 and may be configured to compress the atmospheric air received by the air induction system to a predetermined pressure level. The air induction system may also include additional components for introducing the compressed air into the cylinders 12 of the engine 10, such as, for example, a filter, a valve, air cleaner, air cooler, waste gate, a venturi, etc., as known in the art.

The aftertreatment device(s) 20 may be configured to remove particulates and other constituents from the flow of exhaust, e.g., a filter for capturing particulates, ash, or other materials from the exhaust gas to prevent their discharge into the surrounding environment, such as a diesel particulate filter (DPF), a system for regenerating the filter by removing the particulate matter trapped by the filter, other catalytic devices, and/or other exhaust gas treatment devices. For example, a diesel oxidation catalyst (DOC) may raise the NO₂/NOx ratio, which may improve the NOx conversion efficiency of the SCR catalyst 24. One or more aftertreatment device(s) may also be located downstream of the SCR catalyst 24, such as an ammonia oxidation (AMOX) catalyst that may oxidize ammonia that slips from the SCR catalyst 24 to form N₂ and H₂O.

The injector 22 may be connected to a reductant supply (not shown) and may inject reductant, such as urea, urea and water, ammonia, and/or other elements or compounds capable of chemically reducing compounds, e.g., NOx, contained within the flow of exhaust in the presence of, for example, catalyst materials. The injector 22 may include a nozzle (not shown) or other flow control device configured to assist in controllably releasing a flow of the reductant into the flow of exhaust from the engine 10. The nozzle may be any type of injector known in the art and may include any device capable of injecting and/or atomizing an injected fluid.

The SCR catalyst 24 may chemically reduce the amount of NOx in the flow of exhaust. The reductant injected into the flow of exhaust by the injector 22 upstream from the SCR catalyst 24 may be absorbed onto the SCR catalyst 24 so that the reductant may react with NOx in the flow of exhaust to form H₂O (water vapor) and N₂ (nitrogen gas). For example, a mixture of urea and water injected by the injector 22 may decompose to ammonia, and the SCR catalyst 24 may facilitate a reaction between the ammonia and NOx in the flow of exhaust to produce water and nitrogen gas, thereby removing NOx from the flow of exhaust. The SCR catalyst 24 may include catalyst materials such as, but not limited to, zeolites (e.g., iron zeolite or copper zeolite) or vanadia.

A portion of the flow of exhaust exiting the SCR catalyst 24 may enter an exhaust gas recirculation (EGR) passageway 28, which may direct a flow of recirculated exhaust to the compressor 26 for subsequent combustion while the remaining portion of the flow of exhaust exiting the SCR catalyst 24 may be output from the exhaust system 14, e.g., released into the surrounding atmosphere, such as through a tail pipe. Alternatively, the EGR passageway 28 may be configured to direct a flow of recirculated exhaust exiting at least one of the aftertreatment device(s) 20 upstream from the injector 22 to the compressor 26 while the remaining portion of the flow of exhaust exiting at least one of the aftertreatment device(s) 20 may be directed to the injector 22 and the SCR catalyst 24 before being output from the exhaust system 14.

The engine 10 may also be provided with an intake manifold 11 and/or an exhaust manifold. The intake manifold 11 may receive the compressed air and/or recirculated exhaust, and allow the compressed air and/or recirculated exhaust to flow to the cylinders 12. The exhaust manifold may receive the flow of exhaust from the cylinders 12 and direct the flow of exhaust to the turbine 16, e.g., via an exhaust passageway.

The exhaust emission prediction system may include a controller 30 connected via communication lines 32 to one or more of the components of the engine 10 and the exhaust system 14. For example, the controller 30 may receive input via the communication lines 32 from a variety of sources including, for example, a timer and/or one or more sensors configured to measure temperature, speed, pressure, fuel quantity consumed, flow rate, amount of reductant injected, and/or other operating characteristics of the engine 10 and/or exhaust system 14. As shown in FIG. 1, the controller 30 may be connected by the communication lines 32 to a NOx sensor 34 and a humidity sensor 36. The sensors 34, 36 may be physical (hardware) sensors. The NOx sensor 34 may be located downstream of the engine 10 and the turbine 16, and upstream of the aftertreatment device(s) 20, the injector 22 and the SCR catalyst 24. Alternatively, the NOx sensor 34 may be disposed in other locations in the exhaust system 14, e.g., downstream from the aftertreatment device(s) 20 and/or the SCR catalyst 24. The humidity sensor 36 may be located in the air induction system, e.g., at a location upstream of an inlet of the compressor 26, the connection of the EGR passageway 28 to the air induction system, and/or one or more filters in the air induction system. Alternatively, the humidity sensor 36 may be disposed at another location that allows the humidity sensor 36 to measure an ambient humidity of the atmospheric air.

The controller 30 may include components required to run an application such as, for example, a computer, a memory module, a secondary storage device (e.g., a database), and a processor or microprocessor, such as a central processing unit, as known in the art. The memory module may be configured to store information used by the processor, e.g., computer programs or code used by the processor to enable the processor to perform functions consistent with disclosed embodiments, e.g., the processes described in detail below. The controller 30 may be communicatively coupled with one or more components of the engine 10 and/or the exhaust system 14 to change the operation thereof. Optionally, the controller 30 may be integrated into the engine 10, e.g., as part of an engine control module (ECM). The controller 30 may use the inputs to form a control signal based on a pre-set control algorithm. The control signal may be transmitted from the controller 30 via the communication lines 32 to various actuation devices, such as one or more components of the engine 10 and/or the exhaust system 14, e.g., the injector 22 to control the timing and amount of injections.

FIG. 2 is a diagrammatic illustration of the controller 30, according to an exemplary embodiment. The controller 30 may include an empirical model and a physics-based model for determining respective estimations of an amount of an emissions constituent at a location in the exhaust system 14. In the exemplary embodiment, the controller 30 includes an empirical NOx model 40 and a physics-based NOx model 60. The empirical NOx model 40 and the physics-based NOx model 60 may each include one or more models, and may be configured to determine respective first and second NOx estimations 42 and 62 of an amount of NOx in the flow of exhaust output from the engine 10. In the exemplary embodiment, the empirical NOx model 40 and the physics-based NOx model 60 are configured to determine the respective first and second NOx estimations 42 and 62 at a location downstream from the turbine 16 and upstream of the aftertreatment device(s) 20. Alternatively, the empirical NOx model 40 and the physics-based NOx model 60 may determine the respective first and second NOx estimations 42 and 62 at other locations downstream from the engine 10.

FIG. 3 is a diagrammatic illustration of the empirical NOx model 40 for determining the first NOx estimation 42 in the flow of exhaust output from the engine 10, according to an exemplary embodiment. The empirical NOx model 40 may include one or more models or maps (e.g., a neural network model and/or map, a curve fitting model and/or map, etc.) that are data-driven, such as a first NOx model 44 and a first map 46. The first NOx model 44 and/or the first map 46 may be created using experimental data from one or more test engines under various operating conditions. For example, the first NOx model 44 and/or the first map 46 may be trained using data for one or more test engines operating under various operating conditions, such as, but not limited to, one or more altitudes (e.g., sea level or a range of altitudes), one or more ambient temperatures, one or more engine speeds, one or more ambient pressures, and/or one or more ambient humidity levels. Other operating conditions that the test engine(s) may be trained under may include one or more application cycles, one or more tasks, one or more aggressive application cycles, one or more non-aggressive application cycles, one or more steady-state operating conditions, one or more transient operating conditions, one or more engine configurations (e.g., engines including certain hardware), one or more fuel injection system calibrations (e.g., fuel injection timing, amount, or pressure), etc. For operating conditions for which data does not exist (e.g., operating conditions for which the first NOx model 44 and/or the first map 46 were not trained), the first NOx model 44 and/or the first map 46 may be used to predict the performance of the engine 10, for example, by interpolation and extrapolation.

In the exemplary embodiment, the first NOx model 44 may receive one or more inputs 48 and determine the first NOx estimation 42 of an amount of NOx in the flow of exhaust output from the engine 10. The inputs 48 may include one or more of a speed 50 of the engine 10 (e.g., in revolutions per minute or RPM), a fuel injection timing 51, a fuel injection amount 52, a fuel injection pressure 53, a pressure of the intake to the engine 10 (e.g., a pressure 54 in the intake manifold 11), a temperature of the intake to the engine 10 (e.g., a temperature 55 in the intake manifold 11), and an EGR amount 56. The inputs 48 may be estimated (e.g., using the controller 30 and/or virtual sensors) or measured (e.g., using physical sensors), as known in the art. Thus, the first NOx model 44 may be trained to determine the first NOx estimation 42 based on the inputs 48 at one or more of the operating conditions described above.

The first NOx model 44 may output the first NOx estimation 42, which may be input into the first map 46. The first map 46 may also receive one or more other inputs, such as an ambient humidity 57 measured by the humidity sensor 36 (FIG. 1). The first map 46 may be used to adjust the first NOx estimation 42 based on the ambient humidity 57. For example, the first map 46 may correlate the ambient humidity 57 and a correction factor for multiplying with the first NOx estimation 42, and the first map 46 may be used to determine the correction factor based on the measured ambient humidity 57. The controller 30 may multiply the first NOx estimation 42 by the correction factor to adjust the first NOx estimation 42. Then, the first NOx estimation 42 may be output from the empirical NOx model 40.

FIG. 4 is a diagrammatic illustration of the physics-based NOx model 60 for determining the second NOx estimation 62 in the flow of exhaust output from the engine 10, according to an exemplary embodiment. The physics-based NOx model 60 may include one or more models, such as a second NOx model 64 and an in-cylinder engine model 66, that may be based on one or more physical and/or chemical equations for determining the performance of the engine 10 (e.g., the second NOx estimation 62 or an estimation of another emissions constituent in the flow of exhaust). For example, the second NOx model 64 and/or the in-cylinder engine model 66 may be created using one or more equations governing the functioning of the engine 10, such as the chemical and physical reactions occurring in the cylinders 12, e.g., the reactions governing the combustion of the mixture of fuel and the compressed air and/or recirculated exhaust, and other reactions relating to the formation of NOx and/or other emissions constituents, etc. In the exemplary embodiment, the second NOx model 64 and/or the in-cylinder engine model 66 are not data-driven and therefore are not trained using experimental data from one or more test engines.

The physics-based NOx model 60 may receive one or more inputs 68. The inputs 68 may include one or more of the inputs 48 to the empirical NOx model 40, such as the engine speed 50, the intake manifold pressure 54, and/or the intake manifold temperature 55. In addition, the inputs 68 may include a crank angle 71 of the engine 10 (e.g., in crank angle degrees or CAD), an air flow rate 72 of air into one or more of the cylinders 12, a fuel flow rate 73 of fuel into one or more of the cylinders 12, and/or a volume percent of EGR (EGR_IVC) 74 in one or more of the cylinders 12 at the closing of the respective intake valve(s). The inputs 68 may be estimated (e.g., using the controller 30 and/or virtual sensors) or measured (e.g., using physical sensors), as known in the art. Based on the inputs 68, the physics-based NOx model 60 may determine the second NOx estimation 62 for the flow of exhaust output from the engine 10.

In the exemplary embodiment, the in-cylinder engine model 66 may receive one or more of the inputs 68 to determine one or more outputs 70, which may include one or more in-cylinder characteristics of one or more of the cylinders 12. The outputs 70 may be determined by the in-cylinder engine model 66 based on one or more equations governing the chemical and physical reactions occurring in one or more of the cylinders 12. For example, the in-cylinder engine model 66 may receive the intake manifold pressure 54, the intake manifold temperature 55, the air flow rate 72, and/or the fuel flow rate 73. The in-cylinder engine model 66 may include the reactions governing the combustion of the mixture of fuel and the compressed air and/or recirculated exhaust. The outputs 70 of the in-cylinder engine model 66 may include one or more in-cylinder characteristics, such as one or more of a gross heat release rate (CylGHRR) 75 of one or more of the cylinders 12 (e.g., in joules per crank angle degrees or J/CAD), a pressure (CylP) 76 of one or more of the cylinders 12 (e.g., one or more values of pressure as a function of the crank angle for a 720-degree CAD cycle of one of the cylinders 12), a bulk gas temperature (CylT) 77 of one or more of the cylinders 12 (e.g., one or more values of bulk gas temperature as a function of the crank angle for a 720-degree CAD cycle of one of the cylinders 12), a bulk gas specific heat (CylCP) 78 of one or more of the cylinders 12 (e.g., one or more values of bulk gas specific heat as a function of the crank angle for a 720-degree CAD cycle of one of the cylinders 12), a bulk gas density (CylRho) 79 of one or more of the cylinders 12 (e.g., one or more values of bulk gas density as a function of the crank angle for a 720-degree CAD cycle of one of the cylinders 12), a bulk gas mass (CylMass) 80 of one or more of the cylinders 12 (e.g., one or more values of bulk gas mass as a function of the crank angle for a 720-degree CAD cycle of one of the cylinders 12), and a total heat transfer (CylQ) 81 to the flow of exhaust out of one or more of the cylinders 12 (e.g., one or more values of total heat transfer to the flow of exhaust out of one of the cylinders 12 as a function of the crank angle for a 720-degree CAD cycle). The outputs 70 from the in-cylinder engine model 66 may be input into the second NOx model 64.

Alternatively, the in-cylinder engine model 66 may be omitted, and one or more of the outputs 70 of the in-cylinder engine model 66 may be measured using one or more physical sensors, such as one or more pressure sensors configured to sense the pressure (e.g., the pressure 76) in one or more of the cylinders 12 and/or one or more temperature sensors configured to sense the temperature (e.g., the bulk gas temperature 77) in one or more of the cylinders 12, to be input into the second NOx model 64. The second NOx model 64 and/or one or more virtual sensors may estimate the gross heat release rate 75, the bulk gas specific heat 78, the bulk gas density 79, the bulk gas mass 80, and/or the total heat transfer 81 using the inputs received from the physical sensors. When the in-cylinder engine model 66 is omitted, the second NOx model 64 may include the reactions governing the combustion of the mixture of fuel and the compressed air and/or recirculated exhaust.

Using either the outputs 70 from the in-cylinder engine model 66 or the measurements from the physical sensors, in addition to one or more of the inputs 68 to the physics-based NOx model 60, the second NOx model 64 may determine the second NOx estimation 62 for the flow of exhaust output from the engine 10. For example, as shown in FIG. 4, the second NOx model 64 may receive the engine speed 50, the crank angle 71, the air flow rate 72, the fuel flow rate 73, and/or the volume percent of EGR 74. The second NOx model 64 may include reactions relating to the formation of NOx and/or other emissions constituents. To determine the second NOx estimation 62, the second NOx model 64 may determine other in-cylinder characteristics, such as an adiabatic flame temperature of a stoichiometric mixture of fuel and air along a burn duration from the start of combustion to the end of combustion and a brake specific fuel consumption (e.g., in grams per kilowatt-hour or g/kW-hr). The brake specific fuel consumption may be determined, for example, based on the air flow rate 72 and the fuel flow rate 73. To determine the second NOx estimation 62, the second NOx model 64 may also include other characteristics (e.g., characteristics of the fuel), such as a stoichiometric air-to-fuel mass ratio of the fuel, a hydrogen-to-carbon ratio of the fuel, or a lower heating value (LHV) of the fuel (e.g., in megajoules per kilogram or MJ/kg).

Referring back to FIG. 2, the empirical NOx model 40 and the physics-based NOx model 60 may output the respective first and second NOx estimations 42 and 62 to a NOx determination module 90 in the controller 30. The NOx determination module 90 may determine a third NOx estimation 92 of the amount of NOx in the flow of exhaust output from the engine 10 (e.g., at a location downstream from the turbine 16 and upstream of the aftertreatment device(s) 20) based on the first NOx estimation 42, the second NOx estimation 62, and the operating condition of the engine 10, as described in further detail below.

The NOx determination module 90 may output the third NOx estimation 92, which may be input into a second map 94. In the exemplary embodiment, the second map 94 may also receive one or more other inputs, such as a measured NOx 95 measured by the NOx sensor 34 (FIG. 1). The second map 94 may be used to adjust the third NOx estimation 92 based on the measured NOx 95 to output a final NOx estimation 96, as described below. Alternatively, the second map 94 and the NOx sensor 34 may be omitted, e.g., if there is no NOx sensor 34 present in the exhaust system 14, and the NOx determination module 90 may determine the third NOx estimation 92, which may be the final NOx estimation, without using input from the NOx sensor 34. The controller 30 may be configured to use the final NOx estimation 96 as feedback for controlling the dosing of the reductant using the injector 22, which may improve the performance of the SCR catalyst 24.

INDUSTRIAL APPLICABILITY

The disclosed exhaust emission prediction system may be applicable to any exhaust system. The exhaust emission prediction system may incorporate both the empirical NOx model 40 and the physics-based NOx model 60 to provide more accurate predictions for emissions characteristics, such as the amount of NOx in the flow of exhaust output from the engine 10.

FIG. 5 shows a flow chart depicting an exemplary embodiment of an algorithm of the software control used in connection with the controller 30. The steps described below may be repeated by the controller 30 periodically.

The controller 30 may determine the first NOx estimation 42 using the empirical NOx model 40 (step 100). As described above, the empirical NOx model 40 includes the first NOx model 44, which may receive one or more of the inputs 48 and may output the first NOx estimation 42. The controller 30 may adjust the first NOx estimation 42 based on the ambient humidity 57 (step 102). As described above, the ambient humidity 57 may be measured using the humidity sensor 36. The controller 30 may use the first map 46 to determine a correction factor based on the ambient humidity 57 and may multiply the first NOx estimation 42 by the correction factor. Because the ambient humidity 57 may affect the rate at which NOx is formed, the controller 30 may provide a more accurate estimate of the amount of NOx in the flow of exhaust output from the engine 10 by taking into account the ambient humidity 57.

The controller 30 may also determine the second NOx estimation 62 using the physics-based NOx model 60 (step 104). As described above, the in-cylinder engine model 66 may receive one or more of the inputs 68, and may determine one or more in-cylinder characteristics (e.g., the gross heat release rate 75, the pressure 76, the bulk gas temperature 77, the bulk gas specific heat 78, the bulk gas density 79, the bulk gas mass 80, and/or the total heat transfer 81), which may be input into the second NOx model 64. The second NOx model 64 may receive the in-cylinder characteristics determined by the in-cylinder engine model 66 and one or more of the inputs 68, and may output the second NOx estimation 62. Alternatively, as described above, the in-cylinder engine model 66 may be omitted, and one or more of the in-cylinder characteristics may be measured using physical sensors (e.g., pressure and temperature in one or more of the cylinders 12) and input into the second NOx model 64. The first and second NOx estimations 42 and 62 may be input into the NOx determination module 90.

The controller 30 (e.g., the NOx determination module 90) may determine whether the engine 10 is operating under the operating conditions for which the empirical NOx model 40 is trained (the trained operating conditions) (step 106). The trained operating conditions may include, e.g., one or a range of altitudes, one or a range of ambient temperatures, one or a range of engine speeds, one or a range of ambient pressures, one or more application cycles, one or more tasks, one or more aggressive application cycles, one or more non-aggressive application cycles, one or more steady-state operating conditions, one or more transient operating conditions, one or more engine configurations (e.g., engines including certain hardware), one or more fuel injection system calibrations (e.g., fuel injection timing, amount, or pressure), etc.

For example, in an exemplary embodiment, the empirical NOx model 40 may be trained at approximately sea level and at engine speeds between approximately 800 RPM and approximately 1,800 RPM. It is understood that the empirical NOx model 40 may be determined to be trained for a range of operating conditions (e.g., engine speeds between approximately 800 RPM and approximately 1,800 RPM) even though the data used to train the empirical NOx model 40 may include a subset of operating conditions within the range.

The NOx determination module 90 may determine that the engine 10 is operating under the trained operating conditions (step 106; yes). In the exemplary embodiment described above, for example, the NOx determination module 90 may determine whether the engine 10 is operating at approximately sea level and at an engine speed between approximately 800 RPM and approximately 1,800 RPM. If so, the NOx determination module 90 may determine the third NOx estimation 92 based at least in part on the first NOx estimation 42 determined by the empirical NOx model 40 (step 108). For example, the NOx determination module 90 may determine that the third NOx estimation 92 equals the first NOx estimation 42.

The controller 30 may receive the measured NOx 95 from the NOx sensor 34 (step 110) and may adjust the third NOx estimation 92 using the measured NOx 95 (step 112). For example, the second map 94 may be used to determine a correction factor based on the measured NOx 95. The correction factor may also depend on the operating condition(s) under which the engine 10 is performing. For example, the correction factor may adjust the third NOx estimation 92 to be closer to the measured NOx 95 when the engine 10 is operating under one or more operation conditions for which there may be relatively more confidence in the accuracy of the NOx sensor 34, e.g., below a certain period of time of use of the NOx sensor 34. The controller 30 may multiply the third NOx estimation 92 by the correction factor to output the final NOx estimation 96. Alternatively, the correction factor may be an offset that is added to (or subtracted from) the third NOx estimation 92 to output the final NOx estimation 96. Thus, the NOx sensor 34 may be used to adjust the third NOx estimation 92 to determine the final NOx estimation 96. Thus, the NOx sensor 34 may measure the amount of NOx using the NOx sensor 34 disposed at or near the same location at which the first, second, and third NOx estimations 42, 62, and 92 are estimating the amount of NOx. Alternatively, if the NOx sensor 34 is located at another location in the exhaust system 14 (e.g., downstream of the aftertreatment device(s) 20), the second map 94 may be configured to take into account any differences in the amount of NOx between the location of the NOx sensor 34 and the location at which the first, second, and third NOx estimations 42, 62, and 92 are estimating the amount of NOx (e.g., due to the aftertreatment device(s) 20).

Optionally, steps 110 and 112 may be omitted, e.g., if there is no NOx sensor 34 present in the exhaust system 14. As another alternative, or in addition, the controller 30 may use the third NOx estimation 92 to diagnose the NOx sensor 34 (e.g., to determine when the NOx sensor 34 fails) or act as a backup virtual NOx sensor if the NOx sensor 34 fails.

Referring back to step 106, the NOx determination module 90 may determine that the engine 10 is not operating under the trained operating conditions (step 106; no). In the exemplary embodiment described above, for example, the NOx determination module 90 may determine that the engine 10 is operating at a relatively high altitude and/or at an engine speed outside the range of approximately 800 RPM and approximately 1,800 RPM. If so, then the NOx determination module 90 may determine the third NOx estimation 92 based at least in part on the second NOx estimation 62 determined by the physics-based NOx model 60 (step 114). For example, the NOx determination module 90 may determine that the third NOx estimation 92 equals the second NOx estimation 62.

Alternatively, the NOx determination module 90 may determine the third NOx estimation 92 based on both the first NOx estimation 42 and the second NOx estimation 62. According to an exemplary embodiment, the NOx determination module 90 may determine a weighing factor that may indicate a relative weight of the empirical NOx model 40 compared to the physics-based NOx model 60. The weighing factor may be determined based on the operating condition(s) under which the engine 10 is performing, and may range from 0 (indicating that the engine 10 is performing under operating condition(s) for which there is more confidence in the accuracy of the empirical NOx model 40) to 1 (indicating that the engine 10 is performing under operating condition(s) for which there is more confidence in the accuracy of the physics-based NOx model 60, e.g., if the engine 10 is operating in the domain to which the physics-based NOx model 60 applies). The third NOx estimation 92 may be determined by applying the weighing factor to the first NOx estimation 42 and the second NOx estimation 62. If the weighing factor is equal to or closer to 0, then the third NOx estimation 92 may be equal to or closer to the first NOx estimation 42 than the second NOx estimation 62. On the other hand, if the weighing factor is equal to or closer to 1, then the third NOx estimation 92 may be equal to or closer to the second NOx estimation 62 than the first NOx estimation 42. The weighing factor may vary from 0 to 1 in order to correspondingly vary the third NOx estimation 92 from the first NOx estimation 42 to the second NOx estimation 62. After determining the third NOx estimation 92, the controller 30 may receive the measured NOx 95 from the NOx sensor 34 (step 110) and may adjust the third NOx estimation 92 using the measured NOx 95 to output the final NOx estimation 96 (step 112), as described above.

The controller 30 may use the final NOx estimation 96 as feedback for controlling the dosing of the reductant using the injector 22. The controller 30 may determine the amount of NOx entering the SCR catalyst 24 based on the final NOx estimation 96 and by taking into account the effect of the other exhaust treatment components (e.g., the aftertreatment device(s) 20) located between the turbine 16 and the SCR catalyst 24 on the composition of the flow of exhaust. The controller 30 may adjust an amount of the reductant injected by the injector 22 based on the determined amount of NOx entering the SCR catalyst 24 to anticipate and mitigate the release of NOx and/or ammonia downstream of the SCR catalyst 24.

The flow chart described above in connection with FIG. 5 depicts an exemplary embodiment of the algorithm and software control. Those skilled in the art will recognize that similar algorithms and software control may be used without deviating from the scope of the present disclosure.

Several advantages over the prior art may be associated with the exhaust emission prediction system. The exhaust emission prediction system may provide a hybrid virtual NOx sensing device that includes both the empirical NOx model 40 and the physics-based NOx model 60. Therefore, the exhaust emission prediction system may provide more accurate and reliable estimations of the amount of NOx in the flow of exhaust output from the engine 10. Optionally, the NOx sensor 34 may be omitted, which may reduce costs, or the NOx sensor 34 may be diagnosed or corrected using the estimations determined by the exhaust emission prediction system.

The exhaust emission prediction system includes both the empirical NOx model 40 and the physics-based NOx model 60, and therefore may provide advantages from both models 40 and 60. The exhaust emission prediction system may determine the final NOx estimation 96 based on whether the engine 10 is operating under trained operating conditions. The empirical NOx model 40 may have higher accuracy than the physics-based NOx model 60 when the engine 10 is operating under trained operating conditions. If the engine 10 is operating under the trained operating conditions, then the final NOx estimation 96 may be determined based at least in part on the output (e.g., the first NOx estimation 42) from the empirical NOx model 40. Therefore, the exhaust emission prediction system may take advantage of the relative accuracy of the empirical NOx model 40 under the trained operating conditions.

Because the empirical NOx model 40 may not be as accurate when the engine 10 is not operating under trained operating conditions, the NOx determination module 90 may switch to the output (e.g., the second NOx estimation 62) from the physics-based NOx model 60 to determine the final NOx estimation 96 or may calibrate the output from the empirical NOx model 40 using the output from the physics-based NOx model 60 to determine the final NOx estimation 96. For example, to calibrate the output from the empirical NOx model 40, the NOx determination module 90 may use the weighing factor to determine how much to weigh the outputs from the empirical NOx model 40 and the physics-based NOx model 60. The weighing factor may favor the empirical NOx model 40 if there is higher confidence in the empirical NOx model 40 (e.g., based on the operating conditions of the engine 10), may favor the physics-based NOx model 60 if there is higher confidence in the physics-based NOx model 60, or may average the two outputs if neither model 40 or 60 is favored. Therefore, when the engine 10 is not operating under trained operating conditions, the exhaust emission prediction system may take advantage of the relative accuracy of the empirical NOx model 40 and/or the physics-based NOx model 60, depending on the operating conditions under which the engine 10 is operating.

It will be apparent to those skilled in the art that various modifications and variations can be made to the disclosed exhaust emission prediction system. Other embodiments will be apparent to those skilled in the art from consideration of the specification and practice of the disclosed exhaust emission prediction system. It is intended that the specification and examples be considered as exemplary only, with a true scope being indicated by the following claims and their equivalents.

Claims

1. An exhaust emission prediction system comprising:

an engine configured to generate a flow of exhaust; and

a controller configured to: determine a first estimation of an amount of an emissions constituent at a first location using an empirical model, the first location being downstream of the engine; determine a second estimation of the amount of the emissions constituent at the first location using a physics-based model; and determine a third estimation of the amount of the emissions constituent at the first location based on at least one of the first estimation or the second estimation.

2. The exhaust emission prediction system of claim 1, wherein:

the empirical model is trained using at least one trained operating condition; and

the controller is further configured to determine an operating condition of the engine and determine the third estimation based on whether the at least one trained operating condition for the empirical model includes the operating condition of the engine.

3. The exhaust emission prediction system of claim 2, wherein:

when the at least one trained operating condition for the empirical model includes the operating condition of the engine, the controller is configured to determine the third estimation based at least in part on the first estimation; and

when the at least one trained operating condition for the empirical model does not include the operating condition of the engine, the controller is configured to determine the third estimation based at least in part on the second estimation.

4. The exhaust emission prediction system of claim 3, wherein, when the at least one trained operating condition for the empirical model includes the operating condition of the engine, the controller is configured to determine that the third estimation is equal to the first estimation.

5. The exhaust emission prediction system of claim 3, wherein, when the at least one trained operating condition for the empirical model does not include the operating condition of the engine, the controller is configured to determine that the third estimation is equal to the second estimation.

6. The exhaust emission prediction system of claim 3, wherein, when the at least one trained operating condition for the empirical model does not include the operating condition of the engine, the controller is configured to determine the third estimation based on the first estimation and the second estimation.

7. The exhaust emission prediction system of claim 2, wherein the operating condition includes at least one of an altitude, an ambient temperature, an engine speed, an ambient pressure, an application cycle, an engine configuration, or a fuel injection system calibration.

8. The exhaust emission prediction system of claim 1, wherein the empirical model includes a neural network model or a curve fitting model.

9. The exhaust emission prediction system of claim 1, wherein, to determine the first estimation, the controller is configured to input into the empirical model at least one of a speed of the engine, a fuel injection timing of the engine, an amount of fuel injected into the engine, a pressure of the fuel injected into the engine, a pressure in an intake manifold of the engine, a temperature in the intake manifold, or an amount of exhaust recirculated into the engine.

10. The exhaust emission prediction system of claim 1, wherein the controller is further configured to determine at least one in-cylinder characteristic of at least one cylinder of the engine using the physics-based model and determine the second estimation using the at least one in-cylinder characteristic.

11. The exhaust emission prediction system of claim 10, wherein the at least one in-cylinder characteristic includes at least one of a pressure, a bulk gas temperature, a bulk gas specific heat, a bulk gas density, a bulk gas mass, or a total heat transfer to the flow of exhaust.

12. The exhaust emission prediction system of claim 1, wherein, to determine the second estimation, the controller is configured to input into the physics-based model at least one of a speed of the engine, a crank angle of the engine, an air flow rate into the engine, a fuel flow rate into the engine, a pressure in the engine, a temperature in the engine, or a volume percent of recirculated exhaust gas in the engine.

13. The exhaust emission prediction system of claim 1, wherein the emissions constituent includes NOx.

14. The exhaust emission prediction system of claim 13, further comprising:

at least one NOx sensor disposed downstream of the engine and configured to output a measured amount of NOx;

wherein the controller is in communication with the at least one NOx sensor and further configured to adjust the third estimation based on the measured amount.

15. A method of predicting an amount of NOx in a flow of exhaust from an engine using a controller, the method comprising:

determining, using the controller, a first estimation of the amount of NOx at a first location using an empirical model, the first location being downstream of the engine;

determining, using the controller, a second estimation of the amount of NOx at the first location using a physics-based model; and

determining, using the controller, a third estimation of the amount of NOx at the first location based on at least one of the first estimation or the second estimation.

16. The method of claim 15, further comprising:

determining an ambient humidity;

wherein the first estimation is further determined, using the controller, based on the ambient humidity.

17. The method of claim 15, further comprising:

determining, using the controller, an operating condition of the engine;

wherein the empirical model is trained using at least one trained operating condition;

wherein the third estimation is further determined, using the controller, based at least in part on the first estimation when the at least one trained operating condition for the empirical model includes the operating condition of the engine; and

wherein the third estimation is further determined, using the controller, based at least in part on the second estimation when the at least one trained operating condition for the empirical model does not include the operating condition of the engine.

18. An engine system comprising:

an engine configured to generate a flow of exhaust;

an injector configured to inject a reductant into the flow of exhaust;

a catalytic device configured to receive the flow of exhaust after being injected with the reductant;

a processor;

a memory module configured to store instructions that, when executed, enable the processor to: determine a first estimation of an amount of NOx at a first location using an empirical model, the first location being downstream of the engine and upstream of the catalytic device; determine a second estimation of the amount of NOx at the first location using a physics-based model; determine a third estimation of the amount of NOx at the first location based on at least one of the first estimation or the second estimation; and adjust an amount of the reductant injected by the injector based on the determined third estimation.

19. The engine system of claim 18, further comprising a turbine upstream of the injector and configured to receive the flow of exhaust, wherein the first location is downstream of the turbine.

20. The engine system of claim 18, wherein:

the memory module is further configured to store instructions that, when executed, enable the processor to:

determine an operating condition of the engine;

determine the third estimation based at least in part on the first estimation when the at least one trained operating condition for the empirical model includes the operating condition of the engine; and

determine the third estimation based at least in part on the second estimation when the at least one trained operating condition for the empirical model does not include the operating condition of the engine.