Methods for health degradation estimation and fault isolation for system health monitoring
By using configurable controllers and probabilistic models in vehicle engine systems, identifying and isolating faults, aging and fault identification problems in the prior art are solved, achieving higher accuracy and system stability.
Patent Information
- Application Number
- CN202210997870.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Priority Date
- 2021-08-19
- Filing Date
- 2022-08-19
- Publication Date
- 2024-11-05
- Estimated Expiration
- 2042-08-19
AI Technical Summary
When existing vehicle engine systems are aging, worn and malfunctioning, the control model does not match the reality, making it difficult to accurately identify degradation and isolation faults, affecting the management and optimization of the system.
Using configurable controllers, including status observers, feature calculators, optimizers, health estimators, and fault isolators, monitor component health status through sensor data, apply probability models to identify faults, and adjust the optimizer's model to adapt to changes.
Improves the accuracy and accuracy of identification of component degradation and failures, ensuring that the system can still be effectively optimized and managed in the event of aging or failure, and extends the uptime of the system.
Smart Images

Figure CN115898681B_ABST
Abstract
Description
Background Art
[0001] Vehicle engine systems have become incredibly complex as new technologies are introduced and new standards for emissions and fuel efficiency are added. Model-based control approaches have been introduced to facilitate the management of complex technologies and requirements. Hardware components may deviate from nominal behavior due to aging, wear and other degradation or failure scenarios, rendering the models used in the control approaches inconsistent with reality. New and alternative solutions are desired to enable systems to identify the occurrence of degradation and isolate faults with greater accuracy and precision. Summary of the invention
[0002] The inventors have recognized that the problems to be solved require, among other things, new and / or alternative approaches to health degradation estimation and fault isolation. A health monitoring system can have as a goal to determine whether a control algorithm relies on a component and / or system model that corresponds to reality. As components age or degrade, the model can be updated or re-adjusted to account for changing performance. If a component fails, the system model can be reconfigured to accommodate continued operation with updated control metrics. The purpose of some examples herein is to identify the occurrence of a change, isolate the source of the change, and then apply mitigation.
[0003] A first illustrative and non-limiting example takes the form of a configurable controller for controlling a physical device, having associated therewith a plurality of actuators for controlling the operation of the physical device and a plurality of sensors for observing a plurality of characteristics of the operation of the physical device, the configurable controller comprising: a state observer configured to capture a current state of the physical device by communicating with the plurality of sensors; a feature calculator configured to receive current state data from the state observer and calculate at least one feature parameter reflecting a health state of at least one component of the physical device; an optimizer configured to optimize the behavior of the physical device using at least the actuator; a health estimator configured to receive at least one feature parameter from the feature calculator and, based on one or more entry conditions, apply a change probability model to at least one feature parameter to determine whether at least one feature parameter has changed to an extent indicating a fault or health-related degradation, and if so, generate a change indicator; and a fault isolator configured to receive the change indicator and identify which component of the physical device is experiencing a fault or health degradation, and generate a fault indicator indicating the identified fault.
[0004] Additionally or alternatively, the configurable controller may further include a mitigator configured to modify the optimizer in response to the fault indicator, wherein: the optimizer is configured to optimize the behavior of the physical device using multiple factors, each factor being associated with a weight, and, before the health estimator generates the change indicator, the optimizer uses a first set of weights; and the mitigator is configured to respond to the fault indicator by modifying or replacing the first set of weights thereby modifying the optimizer.
[0005] Additionally or alternatively, the optimizer is configured to optimize behavior of the physical device using at least the first and second models by: relying on the first model when the fault isolator fails to generate a fault indicator; and relying on the second model when the fault isolator generates a fault indicator.
[0006] Additionally or alternatively, the configurable controller may further include a mitigator configured to modify a state observer in response to a fault indicator, wherein: the state observer is configured to apply one or more adjustment parameters to monitor a current state of the physical device; and the mitigator is configured to respond to the fault indicator by modifying or replacing one or more of the adjustment parameters used by the state observer.
[0007] Another illustrative, non-limiting example takes the form of a system including the above-described configurable controller and physical device, comprising: an engine having an intake manifold and an exhaust manifold with a combustion chamber therebetween, into which a fuel quantity is provided; an air flow system coupled to the engine, the air flow system having a turbocharger and an exhaust gas recirculation (EGR) valve, the turbocharger having a compressor upstream of the intake manifold and a turbine downstream of the exhaust manifold, the exhaust gas recirculation valve being configured to recirculate exhaust gas from the exhaust manifold back to the intake manifold; and an aftertreatment system coupled to the air flow system downstream of the turbine; wherein the at least one characteristic parameter includes: a first characteristic parameter related to mass air flow; a second characteristic parameter related to exhaust manifold air pressure; and a third characteristic parameter related to pressure downstream of the turbine.
[0008] Additionally or alternatively, the fault isolator is configured to determine which component of the physical device is subject to the fault by analyzing which of the at least one characteristic parameter has changed.
[0009] Additionally or alternatively, the fault isolator is configured to: in response to finding that either or both of the first or second characteristic parameters have changed, determine that the fault is related to the EGR valve; or if the third characteristic parameter has changed, determine that the fault is related to aftertreatment.
[0010] Another illustrative, non-limiting example takes the form of a system including the configurable controller and physical device described above, comprising: an engine having an intake manifold and an exhaust manifold with a combustion chamber therebetween, into which an amount of fuel is provided; an air flow system coupled to the engine, the air flow system having a turbocharger with a compressor and a twin-scroll turbine, an exhaust gas recirculation (EGR) valve, a first exhaust air passage from the exhaust manifold to a first scroll of the turbine, a second exhaust air passage from the exhaust manifold to a second scroll of the turbine, and a balancing valve that controllably allows air flow between the first and second exhaust air passages; and an aftertreatment system coupled to the air flow system downstream of the turbine; wherein the at least one characteristic parameter includes: a first characteristic parameter related to mass air flow; a second characteristic parameter related to exhaust manifold air pressure; and a third characteristic parameter related to pressure downstream of the turbine.
[0011] Additionally or alternatively, the fault isolator is configured to determine which component of the physical device is subject to the fault by analyzing which of the at least one characteristic parameter has changed.
[0012] Additionally or alternatively, the fault isolator is configured to: identify an aftertreatment fault if the third characteristic parameter has changed; identify an EGR fault if both the first and second characteristic parameters have changed; or identify a balancing valve fault if the second characteristic parameter has changed but the first characteristic parameter has not changed.
[0013] Another illustrative, non-limiting example takes the form of an engine system, comprising: an engine having an intake manifold and an exhaust manifold, with a combustion chamber between the intake manifold and the exhaust manifold, into which a fuel quantity is provided; an air flow system coupled to the engine, the air flow system having a turbocharger including a compressor and a turbine, and an exhaust gas recirculation (EGR) valve; an aftertreatment system including a lambda sensor; and a configurable controller for controlling the engine, the air flow system, and the aftertreatment system, comprising: a state observer configured to capture current states of the engine, the air flow system, and the aftertreatment system; a feature calculator configured to calculate a first feature parameter related to mass air flow, a second feature parameter related to exhaust manifold air pressure, and a third feature parameter related to exhaust manifold air pressure using data received from the state observer. a first characteristic parameter, a second characteristic parameter, and a third characteristic parameter related to the pressure downstream of the turbine; an optimizer configured to optimize the behavior of the engine, the air flow system, and the aftertreatment system using the current state of the engine and at least one model or set point; a health estimator configured to receive the first, second, and third characteristic parameters, and based on one or more entry conditions, apply a change probability model to the first, second, and third characteristic parameters to determine whether any of the first, second, or third characteristic parameters has changed to an extent indicating a fault, and if so, generate a change indicator; and a fault isolator configured to receive the change indicator and identify the fault by identifying the system component that is subject to the fault as indicated by the change indicator, and generate a fault indicator indicating the identified fault.
[0014] Additionally or alternatively, the fault isolator is configured to: identify an aftertreatment fault if only the third characteristic parameter has changed; or identify an EGR fault if either or both of the first and second characteristic parameters have changed.
[0015] Additionally or alternatively, the optimizer is configured to use each of: a first model associated with aftertreatment air flow; and a second model associated with EGR flow; and the system also includes a health mitigator configured to modify the operation of the optimizer to: replace or reduce reliance on the first model in response to an identified aftertreatment fault; and replace or reduce reliance on the second model in response to an identified EGR fault.
[0016] Additionally or alternatively, the engine system may further include an exhaust manifold pressure sensor, wherein: the optimizer includes one or more models that rely on pressure sensed by the exhaust manifold pressure sensor; and the health mitigator is configured to increase the optimizer's reliance on the one or more models that rely on pressure sensed by the exhaust manifold pressure sensor in response to a fault.
[0017] Additionally or alternatively, the engine system may further include a turbocharger speed sensor, wherein: the optimizer includes one or more models that rely on the speed sensed by the turbocharger speed sensor; and the health mitigator is configured to increase the optimizer's reliance on the one or more models that rely on the speed sensed by the turbocharger speed sensor in response to the fault.
[0018] Another illustrative, non-limiting example takes the form of an engine system comprising: an engine having an intake manifold and an exhaust manifold, with a combustion chamber between the intake manifold and the exhaust manifold, into which a fuel quantity is provided; an air flow system coupled to the engine, the air flow system having a turbocharger with a compressor and a twin-scroll turbine, an exhaust gas recirculation (EGR) valve, a first exhaust air passage from the exhaust manifold to a first scroll of the turbine, a second exhaust air passage from the exhaust manifold to a second scroll of the turbine, and a balancing valve controllably allowing air flow between the first and second exhaust air passages; an aftertreatment system including a lambda sensor; and a configurable controller for controlling the engine, the air flow system, and the aftertreatment system, comprising: a state observer configured to capture current states of the engine, the air flow system, and the aftertreatment system; a feature calculator configured to use a first characteristic parameter related to mass air flow, a second characteristic parameter related to exhaust manifold air pressure, and a third characteristic parameter related to pressure downstream of the turbine are calculated from data obtained from a state observer; an optimizer configured to optimize the behavior of the engine, the air flow system, and the aftertreatment system based on the current state of the engine and at least one model or set point; a health estimator configured to receive the first, second, and third characteristic parameters, and based on one or more entry conditions, apply a change probability model to the first, second, and third characteristic parameters to determine whether any of the first, second, or third characteristic parameters has changed to an extent indicating a fault, and if so, generate a change indicator; and a fault isolator configured to receive the change indicator and identify the fault by identifying the system component that has experienced the fault, and generate a fault indicator indicating the identified fault.
[0019] Additionally or alternatively, the fault isolator is configured to: identify an aftertreatment fault if the third characteristic parameter has changed; identify an EGR fault if both the first and second characteristic parameters have changed; or identify a balancing valve fault if the second characteristic parameter has changed but the second characteristic parameter has not changed.
[0020] Additionally or alternatively, the optimizer is configured to use each of: a first model associated with aftertreatment air flow; a second model associated with EGR flow; and a third model associated with balancing valve flow; and the system also includes a health mitigator configured to modify the operation of the optimizer to: replace or reduce reliance on the first model in response to an identified aftertreatment fault; replace or reduce reliance on the second model in response to an identified EGR fault; replace or reduce reliance on the third model in response to an identified balancing valve fault.
[0021] Additionally or alternatively, the engine system may further include an exhaust manifold pressure sensor, wherein: the optimizer includes one or more models that rely on pressure sensed by the exhaust manifold pressure sensor; and the health mitigator is configured to increase the optimizer's reliance on the one or more models that rely on pressure sensed by the exhaust manifold pressure sensor in response to a fault.
[0022] Additionally or alternatively, the engine system may further include a turbocharger speed sensor, wherein: the optimizer includes one or more models that rely on the speed sensed by the turbocharger speed sensor; and the health mitigator is configured to increase the optimizer's reliance on the one or more models that rely on the speed sensed by the turbocharger speed sensor in response to the fault.
[0023] This summary is intended to introduce the subject matter of this patent application. It is not intended to provide an exclusive or exhaustive explanation. The detailed description is included to provide further information about this patent application. BRIEF DESCRIPTION OF THE DRAWINGS
[0024] In the accompanying drawings, which are not necessarily drawn to scale, like reference numerals may describe similar components in different views. The same reference numerals with different letter suffixes may represent different instances of similar components. The accompanying drawings generally illustrate various embodiments discussed in this document by way of example and not limitation.
[0025] Figure 1 An illustrative engine with a turbocharger is shown;
[0026] Figure 2 An illustrative engine with a twin-scroll turbine is shown;
[0027] Figure 3 An illustrative example of a system responding to a fault is shown;
[0028] Figure 4 Illustrative control systems and physical devices are shown;
[0029] Figure 5 An illustrative health monitor is shown in block diagram form;
[0030] Figure 6A-6B Example health monitors and methods are shown; and
[0031] Figure 7-9 An illustrative expert table for use in a health monitor is shown. DETAILED DESCRIPTION
[0032] Figure 1 and Figure 2 Each shows an illustrative engine system on which the present invention can be operated. These examples are shown to help present subsequent operating steps, calculations, and solutions. Other engine systems with different configurations and components can be used instead as needed.
[0033] Figure 1 An illustrative engine with a turbocharger is shown. System 10 includes an engine 20 having one or more cylinders 22 that receives fuel from a fuel system 24, such as through one or more fuel injectors. Fuel system 24 provides a known or set amount of fuel for each firing of each cylinder 22, thereby forming a determined amount of fuel for each complete firing sequence of the engine cylinders. The speed N of the engine E Indicates how quickly a full firing sequence occurs (whether or not all cylinders are active).
[0034] The system's air flow includes a fresh air intake that passes through an air filter 30 and then enters a turbocharger 40 having a compressor 42 and a turbine 44 linked together by a drive shaft. The air leaving the compressor 42 is considered a "boosted" air flow that has been compressed, and it passes through a charge air cooler 50 to reduce the temperature before reaching a throttle 52. To prevent turbocharger surge (reverse air flow through the compressor 42, which may be caused by a pressure imbalance in response to the throttle 52 closing), a recirculation valve 54 is provided to allow boosted air to be recirculated back to the compressor 42 input.
[0035] After passing through throttle 52, the pressurized fresh air reaches the intake manifold of engine 20 where combustion occurs. Before entering the intake manifold, the pressurized fresh air mixes with recirculated exhaust gas through exhaust gas recirculation (EGR) valve 60, which typically also passes the recirculated gas through EGR cooler 62. The recirculated exhaust gas helps to reduce combustion temperatures in engine 20, thereby reducing certain harmful emissions.
[0036] During combustion, the mass of pressurized air and the mass of fuel m FThe exhaust gas exits the engine at the exhaust manifold. The exhaust gas is then directed to turbine 44, which obtains torque / force from the exhaust gas, which in turn is applied to compressor 42 via a drive shaft. In some examples, the rotational speed of the turbocharger drive shaft is a measured variable, referred to as turbocharger speed.
[0037] Wastegate 46 allows exhaust gas to be discharged without passing through turbine 44 to control turbocharger speed. If desired, a variable geometry turbine (VGT) can be used instead of wastegate 46 to manage the gases entering turbine 44. Gases exiting the system via turbine 44 or wastegate 46 enter exhaust structure 70 where various aftertreatment devices can remove or reduce pollutants. One or more lambda sensors or universal exhaust gas oxygen (UEGO) sensors are disposed in exhaust structure 70. For example, measuring oxygen concentration can be used to determine the air-fuel ratio in engine 20.
[0038] Throughout the figure are shown small boxes representing temperature and pressure at different locations:
[0039] T0 and p0 represent the ambient air temperature and air pressure
[0040] T1 and p1 represent the pressure and temperature at the outlet of the compressor 42;
[0041] T2 and p2 represent the pressure and temperature at the intake manifold;
[0042] T3 and p3 represent the pressure and temperature at the exhaust manifold; and
[0043] T4 and p4 represent the pressure and temperature at the outlet of the turbine 44 .
[0044] In a production system, ambient air temperature and pressure (T0 and p0) and intake manifold pressure and temperature (T2 and p2) may be measured parameters, and the other pressures and temperatures estimated, calculated, and / or inferred using models and other characteristics of the system. Engine speed (N E ) will also be known, as will the mass input via the fuel injectors of the engine 20 and the output of the lambda sensor at the exhaust structure 70. In some examples, p3 may also be measured directly, and / or the turbocharger speed may be measured.
[0045] In a system under test (such as one used to model a system or one tested on a test bench), additional sensors and pressures can be captured throughout the system. The added sensors help develop and calibrate models used in production systems to estimate, calculate, or infer various pressures, temperatures, mass flows, etc. as needed.
[0046] Operation is typically controlled by an engine control unit (ECU) 80. The ECU may include a microcontroller or microprocessor (as required), or other logic / memory, an application specific integrated circuit (ASIC), etc., with associated memory for storing observed characteristics and operating instruction sets in a non-transitory medium such as flash memory or other memory circuits. The ECU 80 will be coupled to various actuators throughout the system and sensors provided to obtain data and issue control signals as required. The ECU may be coupled to other vehicle control systems, such as via a controller area network (CAN) bus or other wired or wireless link.
[0047] Figure 1 An example more or less typical for a diesel engine configuration is shown, with EGR 60 in a "high pressure" position, thereby recirculating exhaust gas from the exhaust manifold of engine 20 to the intake manifold. Other fuels and configurations may be used instead. For example, a gasoline engine may use a low pressure EGR 60, which would link the output of turbine 44 to the input of compressor 42 using a low pressure EGR valve or even using a three-way valve, which would take as input the incoming air from air filter 30 and the filtered and cooled recirculated exhaust gas leaving turbine 44, with the output entering compressor 42. Additional elements may be present, such as by including an electric motor (E-Turbo) that applies additional force to the drive shaft of turbocharger 40 on command, and one or more features may be omitted or exchanged, such as by replacing the RCV with a blow-off valve that blows the boosted air to the atmosphere rather than recirculating it, or any other suitable change.
[0048] Figure 2 An illustrative engine with a twin-scroll turbine is shown. Figure 1 As shown, a known trade-off of a standard turbocharger setup is low-speed response and high-speed efficiency. When the engine is operating at low speeds, the amount of exhaust gas generated can drive the turbocharger turbine itself at a relatively low speed. A command to increase engine speed will cause more exhaust gas to flow, however, the rotational inertia of the turbocharger system components will delay the turbocharger's response. The trade-off is that using a smaller turbocharger to reduce the rotational inertia can improve responsiveness, but this change will reduce efficiency at high speeds. A twin-scroll turbocharger splits the exhaust and applies a portion of the exhaust to a first portion of the turbine and another portion of the exhaust to a different portion of the turbine, thereby improving responsiveness at low speeds without sacrificing efficiency at high speeds.
[0049] exist Figure 2, air flow is shown entering the system at 100 (after exiting the compressor and charge air cooler, as required), combining with the EGR flow and entering the intake manifold of a six cylinder engine 110 (there may be any number of cylinders; typically an even number, such as 4, 6, or 8). The system components shown are controlled by one or more ECUs 102. At the exhaust manifold, the exhaust is split, with some of the cylinder outputs combined together at 114 and routed via a first exhaust path 120, and the remaining cylinder outputs combined together at 116 and routed via a second exhaust path 122. As previously described, the engine receives fuel from a fuel system 124, providing a fuel mass m F As shown, the first exhaust path 120 is directed to a first portion of the blades of the turbine 130, and the second exhaust path 122 is directed to a second portion of the blades of the turbine 130. A balancing valve 150 is disposed between the two exhaust paths 120, 122, and may be opened and / or closed as needed to balance exhaust flow and / or pressure. The exhaust may then exit the turbine 130 to an aftertreatment 160, as shown.
[0050] In this example, an EGR valve (high pressure) is shown at 140, recirculating exhaust gas to mix with the air flow 100. As shown, the EGR valve 140 is linked to the first exhaust path 120. A balancing valve 150 can be used to equalize pressure as needed, meaning that not all recirculated exhaust gas must come from the first exhaust path 120.
[0051] Figure 2 The following variables are noted:
[0052] T0 and p0 represent the ambient air temperature and air pressure;
[0053] T1 and p1 represent the pressure and temperature at the outlet of the compressor (not shown);
[0054] T2 and p2 represent the pressure and temperature at the intake manifold;
[0055] T 3a and p 3a represents the pressure and temperature at the exhaust manifold within the first exhaust path 120; and
[0056] T 3b and p 3b represents the pressure and temperature at the exhaust manifold within the second exhaust path 122; and
[0057] T4 and p4 represent the pressure and temperature at the outlet of the turbine 44 .
[0058] Lambda sensor output and engine speed N EThe twin scroll configuration incorporates an additional mass flow variable (with Figure 1 compared to).
[0059] for Figure 1 and Figure 2 In either configuration, the EGR mass flow can be measured using, for example, a venturi-type sensor or a choke-type flow meter. Instead, the EGR mass flow can be calculated by sensing the pressure at the exhaust manifold of the engine, and knowing the pressure difference between the exhaust manifold and the intake manifold, the mass flow through the EGR can be calculated using a flow model of the EGR itself. Both the sensor type and location (EGR flow sensor / meter or exhaust manifold pressure sensor) increase system expense and complexity, and the added sensor can be subject to failure in hot, dirty environments. Using the exhaust manifold pressure is more complicated due to the inclusion of separate exhaust paths 120, 122 and a balancing valve 150.
[0060] The charge flow (i.e., the mass entering the engine intake manifold) is equal to the air flow (from the compressor) plus the EGR mass flow. In addition, the charge flow can be characterized using the combustion gas fraction BGF, which is a function of the air flow, EGR mass flow, and injected fuel mass. BGF has a strong correlation with engine-out nitrogen oxides (NOx) and is an important control variable.
[0061] exist Figure 1 and Figure 2 In the present invention, the EGR mass flow can be estimated and / or calculated without measuring the mass flow through the EGR. For example, the virtual sensor model can infer the total air flow of the entire system, where the EGR flow is the difference between the air flow entering / leaving the compressor and the boost flow entering the intake manifold. After solving for the boost flow and air flow, a lambda sensor in the exhaust gas stream combined with knowledge of the amount of injected fuel can be used to infer the EGR mass flow. However, in lean burn conditions, the lambda sensor is often not as accurate as desired, and the data provided by the lambda sensor is delayed. When the pressure ratio is very low, the virtual sensor model may not be accurate due to the reliance on the pressure difference from one location in the system to another when calculating the flow rate.
[0062] In some illustrative examples, a weighted combination of a virtual sensor and a lambda sensor is used to determine EGR mass flow. The weighted combination may use weighting coefficients that are dynamically modified in response to engine conditions. For example, under lean burn and / or transient conditions, the weighting coefficients may be selected to favor the virtual sensor model. When the pressure ratio calculated by the virtual sensor is low, the weighting coefficients may be selected to favor the use of the lambda sensor. For example, when one factor or the other is "favorable," the weighting coefficients may be selected so that the favorable factor is the primary input to the resulting calculation (e.g., greater than 50%, 75%, or 90%).
[0063] Figure 2 The mass flow in the illustrative system can be modeled using a finite set of equations. Each variable / model can be formulated using the multivariate polynomial form:
[0064]
[0065] A third or more variables may be included, with extensions to similar structures. Models may also include or use the form of a set of rational functions:
[0066]
[0067] where N(u1,u2) and D(u1,u2) are each polynomials of the form shown in Eq.1. By generating each model as a multivariate and / or rational function, as shown in Eq.1 and Eq.2, the underlying derivations and calculations remain reasonable and can be accurately approximated with infinite sensitivities. The analytical framework allows for rapid evaluation of models and sensitivities. In such models, the error function can be set to detect and handle any cases where the denominator approaches zero.
[0068] like Figure 2 As shown, the mass flow rate of the twin-scroll design can be calculated using the following equation. The aftertreatment (AFT) mass flow rate is shown in Equation 3:
[0069] m T1 (p 3a , p4, T 3a )+m T2 (p 3b , p4, T 3b )=m AFT (p4, p0, T4) {3}
[0070] Where m T1 and m T2 represents the mass flow rate into each vortex of the turbine, and m AFT represents the exhaust / aftertreatment mass flow. The flow balance of vortex 1 is shown in equation 4:
[0071]
[0072] Among them, m EGR is the EGR mass flow rate, m Bal is the mass flow rate through the balancing valve, m Ch is the charge mass air flow, and m F is the fuel mass. The flow balance of vortex 2 is shown in Equation 5:
[0073]
[0074] Pressures p3a, p3b, and p4 can be found by an iterative solver such as Gauss-Newton, Newton-Raphson, or an iterative Kalman filter. In these equations, the exhaust manifold pressure sensor is optional. When an exhaust manifold pressure sensor is present, Equation 6 can be used to mitigate one or more failure modes, as discussed further below:
[0075] p 3a =p 3a_sensor {6}
[0076] Because the system state can be solved without an exhaust manifold sensor, including such a sensor may be used for mitigation as described below.
[0077] The total mass (mA) of the inlet air flow entering the system can be determined using Equation 7:
[0078] m A =m T1 (p 3a , p4, T 3a )+m T2 (p 3b , p4, T 3b )-m F {7}
[0079] Or alternatively, as shown in Equation 8:
[0080] m A =m AFT (p4, p0, T 3a )-m F {8}
[0081] And, the EGR mass flow rate will then be as shown in Equation 9:
[0082] m EGR =m Ch (p2,m F , T2, N e )-m A {9}
[0083] For a single scroll turbine, the equations are slightly simpler. The turbine exit flow balance can be understood from Equation 10:
[0084] m T (p3, p4, T3, u T )=m AFT (p0, p4, T4) {10}
[0085] where u T (now included in the equation) is the turbine flow control actuator position, such as a wastegate or variable geometry turbine (VGT). The mass flow balance can be understood from equation 11:
[0086] m T (p3, p4, T3, u T )+m EGR (p2, p3, T3, u EGR )=m Ch (p2,m F , T2, N e )+m F {11}
[0087] If desired, measured p3 (within the available range) may also be included as one of the parameters to calculate boost quality. Likewise, pressures p3 and p4 may be found by an iterative solver such as Gauss-Newton, Newton-Raphson, or an iterative Kalman filter. In these equations, the exhaust manifold pressure sensor is optional. Including a p3 sensor may be used for fault identification and / or mitigation, using Equation 12:
[0088] p3=p 3_sensor {12}
[0089] The inlet air flow can be calculated for a single scroll turbine using Equation 13 or 14:
[0090] m A =m T (p3, p4, T3, u Trb )-m F {13}
[0091] Or alternatively:
[0092] m A =m AFT (p4, p0, T3)-m F {14}
[0093] Furthermore, the EGR mass flow rate is also the same as the above equation 9.
[0094] Data extraction for the twin scroll turbine model (and particularly with respect to the balancing valve flow) will depend on the accuracy of the turbine flow map, EGR valve modeling, and boost flow model. With so many inputs, the data output may be uncertain. In some examples, exhaust pressure measurements (if available) may be used to directly estimate the mass flow through the balancing valve. In other examples, again assuming that exhaust pressure measurements are available, the turbine mass fraction flow for each scroll may be estimated instead. Either way, using exhaust pressure measurements will reduce model uncertainty, particularly the balancing mass flow m Bal (p 3a , p 3b , T 3a ,u Bal ) uncertainty. This substitution may be particularly useful when a fault is identified. More specifically, if a fault is identified for a balancing valve, the model that relies on the balancing valve becomes unreliable, and mitigation may be to switch to a modified analysis that omits the balancing valve model, on which both Equations 4 and 5 above rely, as discussed below.
[0095] For example, the mass balance equation can be used to solve for the mass flow rate as shown below. The turbine outlet flow balance can be expressed as:
[0096] m T1 (p 3a , p4, T 3a )+m T2 (p 3b , p4, T 3b )-m AFT (p4, p0, T4) = 0 {15}
[0097] The flow balance on vortex 1 can be expressed as:
[0098]
[0099] The flow balance on vortex 2 can be expressed as:
[0100]
[0101] In equations 16 and 17, m Bal is considered as a freely estimated signal rather than being modeled as in Equations 4 and 5 above. The pressure sensor information can be expressed as:
[0102] p 3a =p 3a_sensor {18}
[0103] Then, the augmentation vector can be expressed as p 3a , p 3b ,p4,m Bal .
[0104] As an alternative, the system can start with turbine outlet and aftertreatment (AFT) flow balancing:
[0105] m T (p 3a , p 3b , p4, T 3a , T 3b , T4)-m AFT (p4, p0, T4) = 0 {19}
[0106] The equilibrium on vortex 1 is then:
[0107]
[0108] Where β indicates the fraction of the total exhaust gas that passes through vortex 1. As shown in Equation 20, the mass flow through vortex 1 plus the EGR mass flow equals half of the mass leaving the engine exhaust manifold and is adjusted for the balancing valve flow using β. The vortex 2 flow balance is shown in Equation 21:
[0109]
[0110] The obtained enhancement vector can be expressed as p 3a , p 3b , p4, β. Equations 6 and 7 can again be used to characterize the mass balance; alternatively, if desired, Equation 7 can be combined to reduce the turbine mass flow to m T (p 3a , p 3b , p4, T 3a , T 3b , T4). The EGR mass flow equation can then be modified to use β, as shown in Equation 22:
[0111]
[0112] The set of equations and variables shown above illustrate an air flow model that can be iteratively solved on an ongoing basis to calculate various pressures, temperatures, and air flows in an engine system. The analysis may be referred to in this context as a virtual sensor because several temperatures, pressures, and air flows in the system are calculated rather than sensed or measured, with the output of the calculations acting as "virtual" sensors. Test bench operations and other suitable calibrations may be performed to create and / or update coefficients within the air flow model. For example, as components age or are replaced, the model may be rebuilt at the test station and / or the model may be adjusted over time to accommodate aging. For example, a system health monitor may be used to calculate model changes associated with component aging, such as by using a function of time, a function of usage, or by measuring or virtually monitoring the changed performance / operation of various components.
[0113] The air flow is then calculated using air flow fusion. There are two sources: the air flow calculated from the (filtered) equivalence ratio (which relies on the lambda sensor) and the mass balance from the mass flow solver described above. The lambda sensor is most accurate at steady state and when the exhaust oxygen content is relatively low (e.g., less than 10% oxygen). On the other hand, when p 3a and p 3b When the difference between is small, the mass balance and flow solvers are less accurate.
[0114] The λ relationship is shown in Equation 23:
[0115]
[0116] It can be seen that the output of the lambda or air-fuel ratio (AFR) sensor will be the mass of air flow divided by the mass of fuel multiplied by the inverse of the stoichiometric ratio of the associated fuel. Rearranging equation 23, m A λ Equal to the product of the lambda sensor output, the fuel mass, and the stoichiometric ratio of the associated fuel.
[0117] In some examples, a weighted average may be used to calculate air flow fusion, such as shown in Equation 24, where m A The value of is determined so that the magnitude of the vector using two weighted values is minimized:
[0118]
[0119] The two weights w1 and w2 are each functions of other variables. For example, w1 can be a function of the lambda sensor output, such as by decreasing it from nominal to zero in a linear fashion as the oxygen content changes from 10% to 5% oxygen (or other functions over a similar, wider or narrower range of oxygen values). If the lambda sensor state is false (e.g. due to component failure or transient operation), w1 can be set to zero. Another weight w2 can be p calculated in the pressure model of the system. 3a and p 3b is a function of the difference between 3a and p 3b When the difference between w2 and w3 drops below a preset threshold, w2 can drop from a nominal value to zero within a certain range. In some examples, these constraints on weight values can be omitted.
[0120] Figure 3An illustrative example of a fault mitigation method is shown. The nominal operating mode is shown at 160. At 164, a set of ECU data 162 is provided to a collection of components and subsystems or system models. Component models can be generated for individual components of the system. Examples can be, for example, but not limited to, models of flow characteristics in any system component, such as a compressor, air cooler, EGR valve, wastegate, turbine, and engine, and associated link components between them. Such a model can be determined for each component on an isolated basis, or it can model subsystem and / or system operation. The previous discussion of various flow equations provides a detailed sequence and a set of examples. ECU data from 162 is provided to the model, thereby generating a set of equations that can be solved using a suitable analysis tool, as shown in 166. Any of the solvers and filters mentioned above can be used to provide multivariate analysis and nonlinear (or linear, as the case may be) analysis performed by the ECU and / or dedicated processors and circuits to calculate a set of outputs indicated at 170. Exemplary outputs include EGR valve mass flow 172, mass air flow 174, and burned fuel fraction 176.
[0121] Additional sensors 168 may be used to enhance the analysis as described above, where, for example, the normal operating model 164 and / or solver 166 are able to generate output at 170 without the "additional" sensor. An example from the above discussion is an exhaust manifold pressure sensor, which may be used as an additional sensor 168 that need not be used in the normal analysis. When mitigating a fault, the additional sensor(s) 168 may be used as a replacement for the identified faulty component.
[0122] In the illustrative example, an identified fault is then indicated at 178, and the system switches to using a different analysis, where the newly implemented analysis 180 is specifically configured for fault 178. Faults can be identified using a bottom-up approach, looking directly at a single actuator or component, or looking at parameters affected by a single actuator or component. For example, a stuck or obstructed EGR valve can be identified by determining that the EGR valve fails to actuate in response to a control signal, whether by observing the valve actuator itself, or by observing the failure of a system parameter that changes in response to an attempted EGR valve actuation. A blocked AFT can be identified due to an aftertreatment (AFT) related sensor failing to change value or the sensor output being out of range. Similar to the EGR valve, that is, a blocked or stuck counterbalance valve can be identified by direct observation of the actuator failing to respond to a change in a control signal, or by a system parameter failing to change when the counterbalance valve is actuated.
[0123] In some examples, fault isolation is performed as the second part of a broader process that takes a more top-down approach. First, analysis is performed to determine if a fault has occurred, and if so, fault isolation is performed, as shown below starting with Figure 4 shown in the analysis.
[0124] Mitigation can be performed in a variety of ways. For example, if the identified fault is an obstructed aftertreatment (AFT), the model that relies on AFT sensing can be omitted. Using reference to the numbered equations above, equations 3 (twin scroll turbo) or 10 (single scroll turbo) can be omitted from the set of equations and models used because these equations rely on models of aftertreatment flows that are no longer reliable due to the fault. Instead, additional sensor values (such as exhaust manifold pressure) can be provided to the set of equations.
[0125] If the fault identified is related to EGR obstruction or sticking, the model that relies on EGR can be omitted or replaced. Using reference to the numbered equations above, equations 4 (twin scroll turbo) or 11 (single scroll turbo) can be omitted. Instead, additional sensor values (such as exhaust manifold pressure) can be provided to this set of equations.
[0126] If the balancing valve of a twin scroll turbo implementation is blocked or stuck, an enhanced formulation for calculating mass flow may be implemented. For a twin scroll turbo, Equations 15-18 may be used illustratively and in place of Equations 3-5, with the exhaust manifold pressure (as sensed) being used as p 3a Incorporated into the analysis. 3a The EGR mass flow can be calculated, which in turn resolves the charge air flow entering the engine. Bal Inclusion as a freely estimated variable facilitates the analysis of Eqs. 15-18, which has a set of four equations with four unknowns (p 3a 、p 3b , p4 and m Bal Another method is to use β to represent the fraction of the total exhaust gas that passes through vortex 1, as shown in equations 19-22, resulting in four unknowns (p 3a 、p 3b , p4 and β).
[0127] In another example of a blocked or stuck balancing valve, the exhaust manifold pressure (p 3a ) and the intake manifold pressure p2 can each be used to calculate the air mass flow rate using equation 25:
[0128] m A =m Ch (p2,m F ,T2,N e )-m EGR (p3a ,p2,T3,u EGR ){25}
[0129] Therefore, here, we can use p 3a Direct measurements of p2 and p4 are used to feed the boost emissions model and the EGR valve model flow to characterize the mass flow.
[0130] The above mitigations are for illustration only. Other mitigations may be used instead. In some embodiments, in response to an identified fault, it is sufficient that one or more analytical models used to calculate air flow in the engine system are down-weighted or omitted. Instead of down-weighting or omitted analytical models, sensed parameters that are not relied upon during normal operation are integrated into the analysis. In some examples, "unused" sensors may be used for fault identification, if desired.
[0131] In another alternative approach, instead of an exhaust manifold pressure sensor, the additional sensor 186 may be the turbocharger speed, which provides both turbine and compressor speeds. Using the turbocharger model and the torque balance on the turbocharger shaft, the analysis can determine the charge air mass flow and exhaust mass flow (accounting for the effects of VGT or WG operation), and the iterative solver can again simplify the problem formulation to four equations and four unknowns.
[0132] The newly implemented analysis integrates the "additional" sensor 186 into the analysis, such as by changing the weights applied to multiple models, or by replacing one model with another in the analysis. The ECU data 182 is provided to a modified set of component and system / subsystem models 184, which also use the additional sensor data 186. At 188, the resulting set of equations is submitted to the solver and filter to again provide the same set of outputs 170 as previously described. As a result, the response to the fault 178 provides a set of system data 170 that allows continued operation of the system even after the fault is identified. It may be that the system operation is ultimately limited in some sense, or less optimal than before, but in some examples, the system remains operating within the prescribed limits to avoid the need for a complete shutdown due to, for example, a violation of emissions or other requirements. That is, with the set of output system data 170 still available after the fault 178, the overall control required to continue operation within the operating limits is still possible until the fault 178 can be resolved, for example, in a subsequent maintenance activity.
[0133] The modified analysis can take several forms. In some examples, a set of weighted values is used to determine which of the available analyses dominate the system state calculation. Rather than omitting equations from the analysis, some examples swap new weighted values into the analysis to minimize the impact of one or more models on the resulting analysis. For example, with an iterative Kalman filter (IKF) solver, an adjustment matrix or measurement covariance matrix can be adjusted to apply larger gains to those models deemed unreliable, effectively reducing the impact of such models on the output. For other solvers, the weight values can be reduced instead. For the purpose of using the solver in a particular implementation, the effect may be to reduce the weights or effectively ignore those models that have become unreliable due to an identified fault.
[0134] While the specific manner in which the change is implemented depends on the type of solver used, the goal is to minimize the impact on the analysis output 170 of models that are no longer reliable due to the identified fault 178. In some examples, swapping in a new set of gains or weights in response to the identified fault is used as a mitigation rather than implementing a different solver or loading an entirely new analysis matrix into the solver. Other examples may switch to a different solver, or may reload a different set of equations into the solver, if desired, in place of the equations previously used, as a mitigation for the fault 178. When solvers are switched or equations are swapped, the effect may be to omit those models that have become unreliable due to the identified fault for the purpose of using the solver in a particular implementation.
[0135] If at a later time, it is determined that the fault that caused the model to become unreliable has ceased (ie, a stuck valve becomes unstuck, or an obstruction in a valve or other component is cleared), the system can be restored to the original normal operation described above.
[0136] Figure 4 The system and control configuration are shown. The control system operates on an underlying system 200, which includes a set of actuators 202 operating on a physical device 204. A set of sensors 206 are used to observe the physical device.
[0137] The control device as shown in this example includes a state observer 220, which feeds a set of current state variables x(k) to an optimizer 230. The optimizer 230 calculates a solution for process parameters that can be applied to a set of actuators 202, which in turn control the operation of the physical device 204. The set of actuators 202 can be applied to control, for example, but not limited to, fuel or other injectors, variable nozzle turbine position, throttle, engine braking, aftertreatment (including exhaust), exhaust gas recirculation (EGR), turbochargers, electric motors (such as in electric turbochargers, which can be controlled via pulse width modulation (PWM)), wastegate (WG) actuator position (stroke), recirculation valve actuator position, variable compressor geometry actuator position; and combinations thereof.
[0138] The physical device 204 may be, for example, but not limited to, an internal combustion engine (whether diesel or gasoline) or a subsystem thereof (e.g., a turbocharger, the system air passage as a whole, a catalyst, etc.). For example, the illustrative and non-limiting examples of the physical device 204 and associated actuators 202 and sensors 206 are shown above. Figure 1 and Figure 2 The present invention may also be more generally used in other systems outside of the engine or vehicle context.
[0139] In some examples, sensors 206 may include, for example, but not limited to, sensors that detect manifold absolute pressure (MAP), mass air flow (MAF), EGR flow, turbine speed, NOx, engine speed, fuel quantity, boost pressure, etc. Additional monitoring parameters may include, for example, torque output of the electric motor of the electric turbocharger, wastegate (WG) normalized opening, recirculation valve (RCV) normalized opening, and / or variable geometry compressor configuration. Such sensors may be configured to sample the underlying parameters being sensed and provide the results of such sampling to the state observer 220. The state observer 220 may record the underlying sensed parameters and actuator positions over time to provide a history of system operation.
[0140] The state observer 220 and the optimizer 230 may be implemented, for example, in a microcontroller configured to operate on a set of stored instructions for performing state observation and optimization routines. In another example, an application specific integrated circuit (ASIC) may provide the state observer functionality, which may include the capture or accumulation of data from the actuator 202 and / or the sensor 2066, which in turn may be periodically read by a microcontroller configured with a stored instruction set for performing control and / or optimization calculations using, for example, a model predictive control (MPC) cost function, a linear quadratic regulator (LQR) control, a proportional integral derivative (PID) control, or other control algorithms. The optimizer 230 may be integrated into an on-board diagnostic system (not shown) or provided separately from the on-board diagnostic system (the on-board diagnostic system may be used to record diagnostic variables and present them to a user or store them for later analysis as desired), both of which may additionally be integrated into the overall vehicle processing unit if desired.
[0141] The output of optimizer 230 is used to control actuator 202 to operate the device in a manner that minimizes the distance of the operating parameter from one or more target output values of the controllable output or physical device operating characteristic. For example, the target can be any one of a target turbocharger speed, a target boost pressure, a target pressure difference across the compressor, a target air mass flow rate, or a combination thereof. For example, using the MPC function, the distance of one or more output values (or resulting operating characteristics) from the target value or reference value is minimized, thereby optimizing performance. As an example, a conventional MPC cost function form is shown in Equation 26:
[0142]
[0143] where u d,k corresponds to the desired profile of the manipulated variable, u k represents the manipulated variable, k represents the discrete time instance, and P represents the prediction horizon of the predictive controller. In this example, y r,k represents the output reference value, and y k represents the predicted value provided according to the mathematical model of the physical device to be controlled during the relevant time horizon, while w1 and w2 specify the weighting terms. For simplicity, the k term may be omitted in subsequent equations herein. The conventional MPC cost function is minimized in operation to provide optimal control for the physical device 204. This process may be performed by the optimizer 230.
[0144] In another example, a PID controller may be used to consider each of a proportional difference, an integral difference, and a differential difference from a target operating point. The proportional difference may indicate a current state, the integral difference may identify a process deviation over time, and the differential difference may indicate a direction of change in operation. Using PID control, the proportional difference is minimized while monitoring to ensure that the integral difference and the differential difference do not indicate a change in performance, which may cause the proportional difference to increase after further iterations. For a PID controller, control parameters output to the actuator 202 are adjusted to reduce or minimize the distance of the actual performance from one or more targets on an iterative basis. For example, the optimizer 230 may use PID control instead of MPC. If desired, LQR control may be used instead, thereby applying similar concepts.
[0145] Figure 4 The illustrated system also includes a health monitor 240 that is configured to receive information from the sensors 206, the physical devices 204, and / or the actuators 202, and the state observer 220. For example, the received information may reflect the current state (e.g., whether a valve is open and to what extent), although such information from the actuators may alternatively or additionally be obtained by monitoring control signals from the optimizer 230 to the actuators 202 (as desired).
[0146] Block 240 is shown for illustration purposes, but may be understood to be integrated into optimizer 230. The operations of blocks 220, 230, 240 may be implemented in a microcontroller configured to operate a set of stored instructions for performing state observation, health management, and optimization routines. In another example, each of blocks 220, 230, 240 may be performed by a separate microcontroller or application specific integrated circuit (ASIC).
[0147] Health Monitor 240 Figure 4244 and a health and fault analysis block 244. In one example, the health and fault analysis block 244 uses the received data to determine whether health degradation or component failure is occurring. For the purposes of the following discussion, health degradation and component failure can each be considered a "failure" to the extent that any of them is identified as occurring according to the change analysis block further described below. The fault and health mitigation block 242 responds to the output of the block 244 to apply mitigation to the system. In some examples, mitigation is applied by modifying the operation of the optimizer. For example, the weighting values or model used in the optimizer can be modified, or the data within the model used by the optimizer can be modified, as explained in U.S. patent application Ser. No. 17 / 008,076, filed on Aug. 31, 2020, entitled “HEALTH CONSCIOUS CONTROLLER,” and U.S. patent application Ser. No. 17 / 327,066, filed on May 21, 2021, entitled “ENGINEMASS FLOW OBSERVER WITH FAULT MITIGATION,” the disclosures of which are incorporated herein by reference.
[0148] In some examples, the overall goal is to determine whether the state observer and / or optimizer is utilizing a model that corresponds to reality. If not, the accuracy of the estimated state may be compromised, and the ability of the optimizer model to calculate the optimal solution may be impaired. For example, the state observer may have a health-based reconfiguration function, in which the observer adjustment parameters may be switched (i.e., hard switching, completely replacing one model with another model, or replacing another set of adjustment parameters with a set of adjustment parameters), or continuous readjustment may be performed in response to gradient changes in the health of the underlying components and / or system. In one example, the state observer may rely on, for example, but not limited to, a Kalman filter, and the associated measurement error covariance of the Kalman filter may be readjusted and / or adjusted to account for changes in system health or the presence of faults. The optimizer may also use multiple models and / or weighted values, which may be hard switched or incrementally changed over time in response to changes in component conditions / performance. In another example, the optimizer may be configured to use multiple different models and use a weighting matrix to fuse the outputs; when a fault occurs that makes one of the multiple models unreliable, the fusion or weighting matrix may be adjusted to eliminate or reduce the impact of the unreliable model.
[0149] Figure 5 Shows that it can be used as Figure 4 An illustrative health monitor 310 of element 240 is shown. Figure 4 The health and fault analysis block 244 is expanded herein to include two main components: a change analysis block 320 and a fault identifier 340. In this example, the state observer 300 (which may also be used as Figure 4 The block 220 in FIG. 2 is used to capture the current state of the controlled system, which may be an engine system. The feature calculator 302 generates a plurality of feature indicators 304 I1, I2, and I3 based on the current state of the system obtained from the state observer 300 and / or direct sensor measurements or the stored historical state. The feature indicators 304 are provided to the change analysis block 320. Although three feature indicators are shown in this example, any number (1 to tens or hundreds) of feature indicators may be provided.
[0150] Each characteristic indicator 304 is processed and the processed value is then compared to a corresponding threshold by a change analysis block 320, which is subject to a set of entry conditions 330. A set of change probability models 322 is used to determine whether a change has occurred, thereby triggering a fault identifier 340 and subsequent mitigation 350. In the illustrative example shown, the set of entry conditions 330 is used to enable or disable change analysis for each change probability model and associated characteristic indicator 304. The entry conditions 330 may be provided to ensure that the analysis is performed on useful data. Before explaining the entry conditions 330, a set of characteristic indicators 304 and the associated change analysis 320 will be described in detail.
[0151] In an illustrative example, a set of feature indicators 304 may be as shown in Feature Set 1:
[0152]
[0153]
[0154] {Feature Set 1}
[0155] Among them, for I1, m A,EGRv is the air flow calculated as the boost flow minus the EGR valve flow calculated using the EGR valve model, mx ,λ is the air flow rate calculated using a model based on the lambda sensor output, μ Δ1 is the average difference between the two air flow rates under nominal conditions, and σ Δ1 is the standard deviation of the difference between the two air flow rates under nominal conditions. For I2, p 3a sens is the measured pressure observed by the exhaust manifold pressure sensor, p 3a is the exhaust pressure solved from the balanced flow (for a twin-scroll turbine, in this example), and μ Δ2 is the average difference between the two pressures under nominal conditions, and σ Δ2 is the standard deviation of the difference between the two pressures under nominal conditions. For I3, p 4,nnomis the modeled pressure downstream of the turbine calibrated at nominal healthy conditions, p4 is the downstream turbine pressure solved from the equilibrium flow, μ Δ3 is the average difference between the two pressures under nominal conditions, and σ Δ2 is the standard deviation of the difference between the two pressures under nominal conditions.
[0156] Based on the description of the characteristic parameters in the previous paragraph, the entry conditions may be as follows:
[0157] - For a change probability model applicable to I1, entry conditions may be, for example but not limited to, a set minimum EGR mass flow (eg, greater than 0.3 kg / s, for example but not limited to), a fuel mass greater than zero, and a lambda sensor being ON / activated.
[0158] - For a change probability model applicable to I2, the entry condition can be, for example, p 3a sens and p 3a The absolute value of the difference between φ and φ is lower than a threshold value (such as, for example but not limited to, 20 kPa).
[0159] - For the change probability model applicable to I3, the entry condition can be, for example, p 4,nom The absolute value of the difference between p and p4 is below a threshold value (such as, for example but not limited to, 2.5 kPa).
[0160] The numerical limits described are provided as an example only; other values may apply to different implementations. It can be seen that the entry conditions prevent the change analysis from being performed under abnormal or unreliable conditions. For example, the change conditions when analyzing mass flow may not apply in the absence of a valid lambda sensor output (sensor off) or when there is no fuel flow or EGR flow (because each will make one of the underlying models unreliable). Similarly, highly transient conditions (because they may cause large differences in pressure values) may be excluded from the analysis for I2 and I3. Depending on whether the corresponding entry condition is met, the entry condition box 330 provides an enable or disable signal to each of the individual change analysis boxes 322. The entry conditions shown above are specific to one example, and different thresholds, analyses, or types of entry conditions may be used in other implementations.
[0161] In this example, each change analysis block 322 generates a continuous probability W1, W2, W3 of a change with respect to the corresponding feature input I1, I2, I3. Each output W1, W2, W3 can represent a probability of a selected condition, which can then be used by the fault identifier 340 to determine whether a particular component is in a faulty or unhealthy state. In some examples, these continuous probabilities W1, W2, W3 can be scaled or normalized to an interval between 0 and 1; the scaling or normalization can occur in any suitable manner or can be omitted. In an alternative approach, each of W1, W2, W3 can provide a Boolean indicator (e.g., 0 or 1; true or false) of a particular condition. In one example, using the definitions for I1, I2, and I3 in Feature Set 1, outputs W1, W2, W3 can indicate whether, for W1, the air flow calculated using the EGR valve model is equal to the air flow calculated using the lambda sensor output; W2 can indicate whether the sensed exhaust manifold pressure is equal to the calculated / modeled exhaust manifold pressure, and W3 can indicate whether the modeled pressure downstream of the turbine calibrated under nominal healthy conditions is equal to the downstream turbine pressure solved from the balanced flow (in a twin scroll turbocharger configuration).
[0162] Other features / conditions may be used, as desired, and any suitable number of features may be tested / applied. For example, additional inequality conditions may determine whether the sensed turbocharger speed is equal to the modeled turbocharger speed, whether the air flow sensed using the air flow sensor matches the modeled air flow, and / or whether the pressure or temperature sensed anywhere in the system matches the corresponding modeled pressure or temperature, as desired. Further, while feature set 1 illustrates three features, any number of features may be present, including 1, 2, 3, 4, 5, etc., without limitation. For example, feature set 1 is configured for an implementation of a twin scroll turbocharger with an exhaust manifold pressure sensor available in the system. In another example, the twin scroll feature may be omitted, and the feature set may omit I3 (which relies on a balancing valve model that would not exist in the absence of a twin scroll turbo), and instead I2 would simply involve p3 sens and p3.
[0163] In some examples, the change analysis block 322 is embodied as a software component or a dedicated hardware component that applies a change probability model. A suitable change probability model is a sequential probability ratio test (SPRT). Other suitable change probability models may include, for example, but not limited to, a generalized likelihood ratio algorithm, a scaled sequential probability ratio test, a truncated sequential probability ratio test, and / or an accurate binomial test. In some examples, the fault detection block performs an overall SPRT algorithm that returns a Boolean output of 0 (no change) or 1 (change). When an overall change is identified, the change analysis block 320 generates an output indicating that a change has occurred, and also provides a fault identifier 340 to isolate or identify each SPRT output W1, W2, W3 of the specific fault in question.
[0164] In the specific example of using SPRT, the method can be understood as repeated testing of two alternative hypotheses about the parameter being measured:
[0165] H0: θ=θ0{null hypothesis}
[0166] H1: θ = θ1 {change / error} {Eq. 27}
[0167] As long as the decision favors H0, the test continues. When the decision favors H1, the test may or may not stop. For example, the test may not stop to track a self-healing condition (a valve is temporarily stuck). When the analysis results are used to modify other system parameters or analyses, the modified system parameters or analyses can be restored to their previous state if the decision reverts to H0. The decision rule can be written recursively as shown in Equation 28:
[0168] d=0,if S k+1 ≥0, or
[0169] d=1,if S k+1 =ε{Eq.28}
[0170] The recursive term S k+1 Defined as shown in Equation 29:
[0171] S k+1 =min(ε,S k +s k ), if S k +s k >0, or
[0172] S k+1 =max(0, S k +s k ), if S k +s k ≤0{Eq.29}
[0173] In these two expressions, s k is the log-likelihood ratio, and ε is the threshold alarm value. k Limited to:
[0174]
[0175] where p(·) is the probability density function (PDF). The PDF can be parameterized using a common model, such as a real-valued random variable with a normal distribution. In some examples, the PDF can therefore be parameterized using a mean (μ) and a standard deviation (σ). Then, for the null hypothesis H0, the parameter vector can be expressed as θ0 = [μ, σ]. The alternative hypothesis H1 is parameterized as θ1 = [μ a , σ a ]. The normalized H0 PDF has parameters And given in equation 31:
[0176]
[0177] where x is the vector of realizations and the expression shown is the standard normal distribution. Other distributions can also be used; for illustration purposes, a simplified algorithm is provided here. The normalized H1 PDF has parameters μ a , σ a , and is given by:
[0178]
[0179] The illustrative SPRT method is Figure 6A-6B The process flow is shown in FIG.
[0180] from Fig. 6A Initially, the system relies on a probability model 400 for determining whether two signals are equal, wherein the model is based on input engine data and calibration parameters. As described above, the probability model in this example can be parameterized using a mean value (μ) and a standard deviation (σ). At 402, signal 1 and signal 2 are obtained and compared (in this example by subtraction), and then the difference is compared to the mean difference at 404 (again by subtraction in this example). At 406, the output form 404 is divided by the standard deviation, and the result is provided to the SPRT analysis 410. Next, the two PDFs of the null hypothesis 420 and the alternative hypothesis 422 are compared with the output of box 406. Statistics are output to 424, and then log at 426, thereby implementing equation 30 to provide result A.
[0181] Go to Figure 6B , then A and S k Upper and lower limits apply to A and S kThe difference between . As shown at 440, an output Sk+1 is generated. The value of Sk+1 is compared to a threshold 414 to identify that the signals are not equal. The threshold 414 will be generated during the calibration process, an example of which is shown below. If the threshold 414 is exceeded at block 442, an inequality detected signal is generated, as shown at 444.
[0182] At 450, the value of Sk+1 is also provided to the step delay block for use in subsequent iterations. Block 450 uses Equation 29 for S k --1 generates a new value. Optionally, block 450 may be reset upon fault detection, as indicated by the dashed line in the figure. As indicated at 432, if Sk+1 exceeds the threshold in block 442, the invalid hypothesis is rejected.
[0183] Return to Figure 5 , in block 320, each feature is directed to a separately configured change probability (CP) model. In some examples, each such block 322 has a separately defined and calibrated SPRT analysis. Thus, there is a 1:1 ratio of calibrated SPRT analysis to feature indicators 304, because each feature indicator 304 is analyzed separately. For each calibration of block 322, a separate feature indicator can be analyzed separately.
[0184] For example, calibration can be performed for each SPRT method by starting with generating a nominal error signal for each non-fault condition (null hypothesis). Equation 33 can be used, for example:
[0185] x0=Signal 1|Fault_off -Signal 2|Fault_off {Eq.33}
[0186] For example, using feature set 1, the SPRT for I1 can be calibrated by replacing m A,EGRv Used as signal 1, m A,λ Used as signal 2 to perform, each is determined under non-fault conditions when the entry condition is met. The mean (μ) and standard deviation (σ) of x0 can be calculated separately to normalize x0.
[0187] Next, for each fault analyzed, the error signal of H1 is calculated. For example, assuming three faults A, B, and C, the individual error signals for each fault can be calculated as Equations 34-36:
[0188] x 1,{A} =Signal 1|{A}Fault_On -Signal 2|{A}Fault_On {Eq.34}
[0189] x 1,{B} =Signal 1|{B}Fault_On-Signal 2|{B}Fault_On {Eq.35}
[0190] x 1,{C} =Signal 1|{C}Fault_On -Signal 2|{C}Fault_on {Eq.36}
[0191] In one example, using the air path example provided above, the faults could be A: aftertreatment fault, B: EGR fault, and C: balancing valve fault. Use the mean and standard deviation of x0 to create normalized values for each individual error signal, such as:
[0192]
[0193] Similar normalizations will be calculated for the remaining error signals. For each erroneous realization, the mean and standard deviation can then be calculated, providing two data sets Δ = {μ a{A} , μ a{B} , μ a{C)} and ∑={σ a{A} , σ a{B} , σ a{C} Next, we compute the PDFs for each nominal (null hypothesis) and incorrect realization, and calculate S for each PDF. k+1 Then, refer to the expert table (as shown below Figure 7 ), a threshold alarm value is selected for each SPRT to ensure that the desired logic state can be achieved. Then, a global adjustment for each alternative hypothesis is selected from the set of Δ and ∑ of the fault realizations. In this example, the result is that each alternative hypothesis for each SPRT has a cumulative sum of S k+1 Finally, the expert table logic state is confirmed by evaluating the unselected adjustments.
[0194] Figure 7 An expert table for an air flow system associated with an engine is shown. In the example shown, the expert table 500 is populated with multiple modes at 502, including one non-fault mode and several fault modes. At 510, a set of SPRTs are shown in the remaining columns. In this particular example, each SPRT provides a Boolean 0 or 1 output. In the nominal or non-fault state, each SPRT returns 0. Fault modes are identified by evaluating the combination of SPRTs that return 1, as shown. For example, mode 3 is identified in response to the presence of both SPRT (I) 512 and SPRT (II) 514, while mode 4 requires SPRT (II) 514 instead of SPRT (I) 512, and all do not require the presence of SPRT (III) 516. Therefore, each combination is specific to the selected mode.
[0195] Figure 7The table in can be enhanced with additional lines to consider any and all combinations of SPRT events. If necessary, more than one row in the table can be linked to a specific mode, and / or more than one mode can be linked to a specific combination. In some examples, if all three SPRTs are present, both Mode 2 and Mode 3 may be triggered, assuming that mitigation is available for both faults associated with these modes, and assuming that all mitigation can be performed simultaneously. On the other hand, in other examples, when there are multiple faults, the system may not be able to ensure continued operation within the operating boundaries (whether for safety, performance, environment, emissions or other reasons), resulting in the system issuing an overall error / fault warning or requiring shutdown in response to some combinations of SPRTs. Such determinations may vary from implementation to implementation.
[0196] Figure 7 The examples shown are e.g. Figure 2 The system shown has a twin-scroll turbine and a balancing valve. Figure 1 system (lacking twin-scroll turbine and balancing valve), Figure 7 The expert table of can be simplified by removing the fourth fault mode. In some examples, EGR fault (mode 3) can be identified using a logical "OR" of SPRT (I) and SPRT (II); in other examples, a Boolean 1 for both SPRT (I) or SPRT (II) may be required to identify an EGR fault. As shown in the mode description at 502, the fault can be linked to A, B, and C, as used above.
[0197] Figure 8 Another expert table is shown, this time for both fault identification and health degradation. Figure 5 , there can be any number of CP models and feature inputs. Some feature inputs may be as described above for Figure 7 As described above, it points to the identification of the actual fault. If necessary, a feature input can be monitored by more than one CP model. Figure 8 In the example, I1 of the above feature set 1 is analyzed for changes using multiple different thresholds, where one such threshold is selected for a large divergence between the two mass air flows and indicates a fault. Figure 8 Expert table 550 uses SPRT(I) at 552 to monitor for large changes indicative of EGR faults, as shown in Mode 2. Figure 8 Another CP model, shown as SPRT(II) in FIG. 5 , can be calibrated and adjusted to identify smaller levels of change. In the expert table 550, when SPRT(I) 552 returns a Boolean 1, an EGR fault is identified (if necessary, it can also be considered whether SPRT(II) returns 1 or 0). Conversely, if only SPRT(II) is triggered, an EGR aging flag is identified. Fault / Health Mitigator 350 ( Figure 5 ) can then be performed by requiring reconfiguration of the state observer 300 and / or the optimizer ( Figure 4 230) to adjust how the operation is performed to account for component aging. As the model is updated, SPRT (II) 554 ( Figure 8 ) is reset and further incremental changes are monitored again. In some examples, the state observer can run separate calculations on certain parameters to keep both of the original configuration analyses, which will make Figure 8 The SPRT (I) in can monitor changes relative to the original configuration to identify failures, while the SPRT (II) monitors changes relative to the current configuration to identify aging.
[0198] Fig. 9 shows how probabilistic advantages can be used to track health degradation. This paper provides a Figure 5 340) are resolved, where the correlation between each W factor and the condition it identifies is recorded. Thus, for each of a set of components A, B, C, D 600, the model output {W1, W2 ... W n}602 is assessed as indicating the likelihood of failure and / or current state information. In some examples, the change model output is a continuous variable. Each value of W can be normalized, for example, in the range of 0 to 1, and as shown at 604, the values can be classified into the beginning of life (BOL), full useful life, and end of life (EOL) as needed. Linear normalization does not need to be used. For example, when W1 is within the BOL range, the health of the component represented by W1 has a value assigned and entered into the relevant unit - in this article, row 610 is used for component A, and row 614 is used for component C. In this example, an approach is adopted in which the lowest value assigned to any component is used as the health state of that component. Therefore, due to the value of W1, component A is considered to be within its BOL range, even if other applicable change model outputs (W3 and W n ) has a higher value; other approaches may be used, including blending, averaging, etc. In the example shown, component B would be considered to be at its full useful life, no longer at BOL, but not yet at EOL (assuming no other W factors apply), while component D would be found to be at EOL because the lowest W value of 0.75 indicates EOL. Over time, as the changing model outputs continue to be tracked, it can be expected that one or more W values will increase due to aging of system components AD. For example, when a component reaches an EOL boundary, a system alert can be issued to a user of the system or vehicle and / or can be transmitted to a central database for the fleet. In addition, the mitigations discussed above for managing system operation when a component reaches EOL can be implemented when a component reaches any relevant boundary.
[0199] It may also be noted that the table at 604 may include both aging / health factors as well as fault-related health factors. Thus, for example, one or more W factors may be related to aging and health, while other factors are related to faults. A specific example may be, for example, a factor that considers the ability of a valve to reach its full range of motion when actuated, or something that may be monitored using an SPRT, for example, by observing the limits of motion of a given valve. In addition, with reference to a control signal that may be applied to the valve, the flow through the valve may be compared to the modeled operation of the valve, and a valve fault or obstruction may be identified when the flow does not match the model determined using an SPRT or other change analysis. Thus, individual components may be monitored for each of aging and faults using separate metrics reflected in the expert table.
[0200] In another example, each change model output is a Boolean value, and depending on the use of separate change models corresponding to failure or aging increments, different determinations regarding the meaning of the identified changes can be labeled as shown. That is, for component A, the presence of a Boolean value of 1 for W1 indicates that the component was still operating at the beginning of its life; on the other hand, W n A Boolean value of 1 indicates that component A is at the end of its life or has failed. Boolean methods can also be used by reference Figure 7-8 Come to understand.
[0201] Each of these non-limiting examples may stand alone or may be combined with one or more of the other examples in various permutations or combinations.
[0202] The above detailed description includes reference to the accompanying drawings, which form a part of the detailed description. The accompanying drawings show specific embodiments in an illustrative manner. These embodiments are also referred to as "examples" in this article. Such examples may include elements other than those shown or described. However, the inventors also consider examples in which only those elements shown or described are provided. In addition, the inventors also consider examples using any combination or arrangement of those elements (or one or more aspects thereof) shown or described with respect to a specific example (or one or more aspects thereof), or with respect to other examples (or one or more aspects thereof) shown or described herein.
[0203] In the event of an inconsistency in usage between this document and any document incorporated by reference, the usage in this document controls.
[0204] In this document, as is common in patent documents, the terms "a" or "an" are used to include one or more than one, independent of any other instance or usage of "at least one" or "one or more." Moreover, in the claims, the terms "first," "second," and "third," etc. are used merely as labels and are not intended to impose numerical requirements on their objects.
[0205] The method examples described herein may be implemented at least in part by a machine or computer. Some examples may include a computer-readable medium or machine-readable medium encoded with instructions, which may be operable to configure an electronic device to perform the method described in the above examples. The implementation of this method may include code, such as microcode, assembly language code, high-level language code, etc. This code may include computer-readable instructions for performing various methods. The code may form part of a computer program product. In addition, in one example, the code may be tangibly stored on one or more volatile, non-temporary, or non-volatile tangible computer-readable media, such as during execution or at other times. Examples of these tangible computer-readable media may include, but are not limited to, hard disks, removable disks or optical disks, cassettes, memory cards or memory sticks, random access memories (RAM), read-only memories (ROM), etc.
[0206] The above description is illustrative and not restrictive. For example, the above examples (or one or more aspects thereof) may be used in combination with each other. Other embodiments may be used, for example, by one of ordinary skill in the art after reviewing the above description. The Abstract is provided to comply with 37 CFR § 1.72 (b) to allow the reader to quickly determine the nature of the technical disclosure. It is submitted with the understanding that it is not used to interpret or limit the scope or meaning of the claims.
[0207] In addition, in the above detailed description, various features may be combined together to simplify the present disclosure. This should not be interpreted as meaning that unclaimed disclosed features are essential to any claim. On the contrary, the subject matter of innovation may lie in less than all the features of a particular disclosed embodiment. Therefore, the following claims are thus incorporated into the detailed description as examples or embodiments, wherein each claim exists independently as a separate embodiment, and it is considered that these embodiments can be combined with each other in various combinations or arrangements. The scope of protection should be determined with reference to the attached claims and the full range of equivalents given by these claims.
Claims
1. A configurable controller for controlling a physical device for an engine, having associated therewith a plurality of actuators for controlling the operation of the physical device and a plurality of sensors for observing a plurality of characteristics of the operation of the physical device, the configurable controller comprising: a state observer configured to capture a current state of the physical device by communicating with the plurality of sensors; a characteristic calculator configured to receive current state data from the state observer and calculate at least one characteristic parameter reflecting a health state of at least one component of the physical device; an optimizer configured to optimize a behavior of the physical device using at least the actuator; a health estimator configured to receive the at least one feature parameter from the feature calculator and, based on one or more entry conditions, apply a change probability model for determining whether a change has occurred to the at least one feature parameter to determine whether the at least one feature parameter has changed to an extent indicative of a fault or health-related degradation, and if so, generate a change indicator; as well as A fault isolator is configured to receive the change indicator and identify which component of the physical device is experiencing a fault or health degradation, and generate a fault indicator indicative of the identified fault.
2. The configurable controller of claim 1 , further comprising a mitigator configured to modify the optimizer in response to the fault indicator, wherein: the optimizer being configured to optimize the behavior of the physical device using a plurality of factors, each factor being associated with a weight, and the optimizer using a first set of weights before the health estimator generates the change indicator; and The mitigator is configured to respond to the fault indicator by modifying or replacing the first set of weights thereby modifying the optimizer.
3. The configurable controller according to claim 1, wherein: The optimizer is configured to optimize the behavior of the physical device using at least the first and second models by: When the fault isolator fails to generate the fault indicator, relying on a first model; and When the fault isolator generates the fault indicator, it relies on a second model.
4. The configurable controller of claim 1 , further comprising a mitigator configured to modify the state observer in response to the fault indicator, wherein: The state observer is configured to apply one or more adjustment parameters to monitor a current state of the physical device; as well as The mitigator is configured to respond to the fault indicator by modifying or replacing one or more of the tuning parameters used by the state observer.
5. A system comprising the configurable controller of claim 1 and a physical device, comprising: an engine having an intake manifold and an exhaust manifold, wherein the intake manifold and the exhaust manifold have a combustion chamber therebetween, into which a quantity of fuel is provided; an air flow system coupled to the engine, the air flow system having a turbocharger having a compressor upstream of the intake manifold and a turbine downstream of the exhaust manifold, and an exhaust gas recirculation (EGR) valve configured to recirculate exhaust gas from the exhaust manifold back to the intake manifold; as well as an aftertreatment system coupled to the air flow system downstream of the turbine; The at least one characteristic parameter comprises: a first characteristic parameter related to mass air flow; a second characteristic parameter related to exhaust manifold air pressure; and A third characteristic parameter is related to the pressure downstream of the turbine.
6. The system according to claim 5, wherein: The fault isolator is configured to determine which component of the physical device is subject to a fault by analyzing which of the at least one characteristic parameter has changed.
7. The system of claim 6, wherein the fault isolator is configured to: In response to finding that either or both of the first or second characteristic parameters have changed, determining that the fault is related to the EGR valve; or If the third characteristic parameter has changed, it is determined that the fault is related to post-processing.
8. A system comprising the configurable controller of claim 1 and a physical device, comprising: an engine having an intake manifold and an exhaust manifold, wherein the intake manifold and the exhaust manifold have a combustion chamber therebetween, into which a quantity of fuel is provided; an air flow system coupled to the engine, the air flow system having a turbocharger with a compressor and a twin scroll turbine, an exhaust gas recirculation (EGR) valve, a first exhaust air passage from an exhaust manifold to a first scroll of the turbine, a second exhaust air passage from the exhaust manifold to a second scroll of the turbine, and a balancing valve controllably allowing air flow between the first and second exhaust air passages; as well as an aftertreatment system coupled to the air flow system downstream of the turbine; The at least one characteristic parameter comprises: a first characteristic parameter related to mass air flow; a second characteristic parameter related to exhaust manifold air pressure; and A third characteristic parameter is related to the pressure downstream of the turbine.
9. The system according to claim 8, wherein: The fault isolator is configured to determine which component of the physical device is subject to a fault by analyzing which of the at least one characteristic parameter has changed.
10. The system according to claim 9, wherein: The fault isolator is configured to: If the third characteristic parameter has changed, then identifying a post-processing fault; If both the first and second characteristic parameters have changed, an EGR fault is identified; or If the second characteristic parameter has changed but the first characteristic parameter has not changed, a balancing valve fault is identified.
11. An engine system, comprising: an engine having an intake manifold and an exhaust manifold, wherein the intake manifold and the exhaust manifold have a combustion chamber therebetween, into which a quantity of fuel is provided; an air flow system coupled to the engine, the air flow system having a turbocharger including a compressor and a turbine, and an exhaust gas recirculation (EGR) valve; aftertreatment system, including lambda sensor; as well as A configurable controller for controlling the engine, the air flow system, and the aftertreatment system includes: a state observer configured to capture current states of the engine, air flow system, and aftertreatment system; a characteristic calculator configured to calculate a first characteristic parameter related to mass air flow, a second characteristic parameter related to exhaust manifold air pressure, and a third characteristic parameter related to pressure downstream of the turbine using data received from the state observer; an optimizer configured to optimize behavior of the engine, air flow system, and aftertreatment system using a current state of the engine and at least one model or set point; a health estimator configured to receive the first, second, and third characteristic parameters and, based on one or more entry conditions, apply a change probability model for determining whether a change has occurred to the first, second, and third characteristic parameters to determine whether any of the first, second, or third characteristic parameters has changed to an extent indicating a fault, and if so, generate a change indicator; and A fault isolator is configured to receive the change indicator and identify a fault by identifying a system component experiencing the fault as indicated by the change indicator, and to generate a fault indicator indicative of the identified fault.
12. The engine system according to claim 11, wherein: The fault isolator is configured to: If only the third characteristic parameter changes, a post-processing fault is identified; or If either or both of the first and second characteristic parameters have changed, an EGR fault is identified.
13. The engine system according to claim 12, wherein: The optimizer is configured to use each of the following: a first model associated with aftertreatment air flow; and a second model related to EGR flow; and The system also includes a health mitigator configured to modify the operation of the optimizer to: In response to the identified post-processing fault, replacing or reducing reliance on the first model; as well as In response to the identified EGR fault, reliance on the second model is replaced or reduced.
14. The engine system of claim 13, further comprising an exhaust manifold pressure sensor, wherein: The optimizer includes one or more models that rely on pressure sensed by an exhaust manifold pressure sensor; and The health mitigator is configured to increase the optimizer's reliance on the one or more models that rely on pressure sensed by an exhaust manifold pressure sensor in response to a fault.
15. The engine system of claim 13, further comprising a turbocharger speed sensor, wherein: The optimizer includes one or more models dependent on a speed sensed by a turbocharger speed sensor; as well as The health mitigator is configured to increase the optimizer's reliance on the one or more models that rely on a speed sensed by a turbocharger speed sensor in response to a fault.
16. An engine system, comprising: an engine having an intake manifold and an exhaust manifold, wherein the intake manifold and the exhaust manifold have a combustion chamber therebetween, into which a quantity of fuel is provided; an air flow system coupled to the engine, the air flow system having a turbocharger with a compressor and a twin scroll turbine, an exhaust gas recirculation (EGR) valve, a first exhaust air passage from the exhaust manifold to a first scroll of the turbine, a second exhaust air passage from the exhaust manifold to a second scroll of the turbine, and a balancing valve controllably allowing air flow between the first and second exhaust air passages; aftertreatment system, including lambda sensor; as well as A configurable controller for controlling the engine, air flow system and aftertreatment system includes: a state observer configured to capture current states of the engine, air flow system, and aftertreatment system; a characteristic calculator configured to calculate a first characteristic parameter related to mass air flow, a second characteristic parameter related to exhaust manifold air pressure, and a third characteristic parameter related to pressure downstream of the turbine using data obtained from the state observer; an optimizer configured to optimize behavior of the engine, air flow system, and aftertreatment system based on a current state of the engine and at least one model or set point; a health estimator configured to receive the first, second, and third characteristic parameters and, based on one or more entry conditions, apply a change probability model for determining whether a change has occurred to the first, second, and third characteristic parameters to determine whether any of the first, second, or third characteristic parameters has changed to an extent indicating a fault, and if so, generate a change indicator; and A fault isolator is configured to receive the change indicator and identify a fault by identifying a system component experiencing the fault, and generate a fault indicator indicative of the identified fault.
17. The engine system according to claim 16, wherein: The fault isolator is configured to: If the third characteristic parameter has changed, a post-processing fault is identified; If both the first and second characteristic parameters have changed, an EGR fault is identified; or If the first characteristic parameter has changed but the second characteristic parameter has not changed, a balancing valve fault is identified.
18. The engine system according to claim 17, wherein: The optimizer is configured to use each of the following: a first model associated with aftertreatment air flow; a second model related to EGR flow; and a third model associated with balanced valve flow; and The system also includes a health mitigator configured to modify the operation of the optimizer to: In response to the identified post-processing fault, replacing or reducing reliance on the first model; In response to the identified EGR fault, replacing or reducing reliance on the second model; In response to the identified balancing valve fault, reliance on the third model is replaced or reduced.
19. The engine system of claim 18, further comprising an exhaust manifold pressure sensor, wherein: The optimizer includes one or more models dependent upon a pressure sensed by an exhaust manifold pressure sensor; as well as The health mitigator is configured to increase the optimizer's reliance on the one or more models that rely on pressure sensed by an exhaust manifold pressure sensor in response to a fault.
20. The engine system of claim 18, further comprising a turbocharger speed sensor, wherein: The optimizer includes one or more models dependent on a speed sensed by a turbocharger speed sensor; as well as The health mitigator is configured to increase the optimizer's reliance on the one or more models that rely on a speed sensed by a turbocharger speed sensor in response to a fault.
