System and method for probabilistic on-board diagnostics

Abstract

The present disclosure provides methods and systems for probabilistic on-board diagnostics. In one example, a method can include calculating a probability metric for a sample of a measured operating condition of a vehicle system component, averaging a plurality of probability metrics for a plurality of samples, the plurality of probability metrics including the probability metric, the plurality of samples including the sample, and determining whether the vehicle system component has deteriorated based on the averaged plurality of probability metrics. In this way, functionality of a vehicle system component can be continuously monitored without regard to specific diagnostic entry conditions.

Description

Technical Field

[0001] This description generally relates to methods and systems for on-board diagnostics in vehicle systems. Background Technology

[0002] Modern vehicles typically include controllers with on-board diagnostics (OBD) software to monitor the performance of various vehicle and engine components and alert vehicle operators and / or maintenance personnel when a specific component or system exhibits deteriorated performance. Typical OBD methods require robust detection of deterioration reaching a given threshold under normal driving conditions and on standard emission cycles. That is, conventional OBD methods identify specific entry conditions where a robust gap exists between the nominal system and the threshold system, allowing for accurate detection of deterioration.

[0003] An exemplary method is illustrated by Smith et al. in US9,638,122. In this method, the performance of a selective catalytic reduction (SCR) catalyst is monitored by calculating the SCR conversion efficiency at different catalyst temperatures. When the SCR catalyst temperature is above a threshold, the NOx level in the feed gas is artificially increased, for example, by adjusting the exhaust gas recirculation (EGR) operation, thereby obtaining more reliable readings from the NOx sensor.

[0004] However, the inventors of this paper have recognized the potential problems with such systems. As an example, entry conditions used for diagnostics may restrict classification to operating points that may not be frequently encountered, resulting in a lower Monitoring Usage Performance Ratio (IUMPR) or the use of intrusive strategies, such as proactively altering operating conditions, which degrades vehicle operation. Furthermore, even when results are close to a threshold and therefore ambiguous, degraded or nominal classification decisions may be made. Summary of the Invention

[0005] In one example, the problem described above is addressed using a method for on-board diagnostics. This method includes: using a probabilistic classifier to calculate a probabilistic metric for a sample of measured operating conditions of a vehicle system component; averaging multiple probabilistic metrics across multiple samples, the multiple probabilistic metrics including the probabilistic metrics, the multiple samples including the samples; and determining whether the vehicle system component has deteriorated based on the averaged multiple probabilistic metrics. In this way, vehicle system components, such as SCR catalysts, can be continuously monitored regardless of driving conditions and without actively manipulating operating conditions. Furthermore, this probabilistic method enables the evaluation of vehicle system components, such as SCR catalysts, based on multiple measurements, such as intermediate bed efficiency and exhaust tailpipe efficiency, rather than based on a single measurement typically used in previous methods.

[0006] As an example, a probabilistic classifier can calculate the probability that an SCR catalyst has deteriorated based on the intermediate bed efficiency and exhaust tailpipe efficiency of the SCR catalyst. If the probability is fuzzy or uncertain, it can be discarded or rejected. In this way, the system and method presented in this paper achieve robust detection of degradation of vehicle system components regardless of entry conditions, thereby improving IUMPR.

[0007] It should be understood that the above summary is provided to introduce, in a simplified form, a series of concepts further described in the detailed description. This is not intended to identify key or essential features of the claimed subject matter, the scope of which is uniquely defined by the claims following the detailed description. Furthermore, the claimed subject matter is not limited to implementations that address any of the shortcomings described above or in any part of this disclosure. Attached Figure Description

[0008] Figure 1 A schematic depiction of the vehicle's engine system is shown.

[0009] Figure 2 A high-level flowchart is shown, illustrating an exemplary method for probabilistic on-board diagnostics according to an embodiment.

[0010] Figure 3 A graph is shown illustrating the predicted distribution of the probabilistic classifier for measurements of SCR catalyst efficiency according to an embodiment.

[0011] Figure 4 A set of charts is shown, illustrating exemplary pass / fail scores according to an embodiment.

[0012] Figure 5 A set of charts is shown to illustrate exemplary failure scores according to an embodiment.

[0013] Figure 6 A set of charts is shown to illustrate an exemplary calculation of the average score over time according to a probability metric, based on an embodiment.

[0014] Figure 7 A graph is shown illustrating the threshold for accepting the average score according to an embodiment. Detailed Implementation

[0015] The following description relates to vehicle systems (e.g., Figure 1 Systems and methods for probabilistic on-board diagnostics in (systems). Methods for probabilistic on-board diagnostics, for example, in... Figure 2The method described herein may include: calculating a probability metric for pass / fail under all conditions, also referred to herein as a probability score, and accepting the result only if the probability sufficiently indicates pass or failure, and rejecting the result if they are ambiguous. For example, a probability classifier may be trained to calculate the probability score based on measurements of the efficiency of a selective catalytic reduction (SCR) catalyst, such as... Figure 3 The probability score can indicate the probability that the SCR catalyst will function properly, as described in the diagram. Figure 4 The description can further indicate the probability that the SCR catalyst has deteriorated, such as... Figure 5 The description is as follows. An average score can be calculated based on a given number of samples acquired over time, such as... Figure 6 As shown, this is to further increase the robustness of the probabilistic method. Different thresholds can be used to evaluate the average score, such as... Figure 7 As shown in the image.

[0016] Figure 1 A schematic diagram of one cylinder of a multi-cylinder engine 10 that may be included in the propulsion system of a vehicle is shown. The engine 10 may be controlled at least in part by a control system including a controller 12 and inputs from a vehicle operator 132 via an input device 130. In this example, the input device 130 includes an accelerator pedal and a pedal position sensor 134 for generating a proportional pedal position signal PP. The combustion chamber (i.e., cylinder) 30 of the engine 10 may include a combustion chamber wall 32 and a piston 36 positioned therein. The piston 36 may be coupled to a crankshaft 40 such that the reciprocating motion of the piston is converted into rotational motion of the crankshaft. The crankshaft 40 may be coupled to at least one drive wheel of the vehicle via an intermediate transmission system. Furthermore, a starter motor may be coupled to the crankshaft 40 via a flywheel (not shown) to enable starting operation of the engine 10.

[0017] Combustion chamber 30 can receive intake air from intake manifold 44 via intake duct 42 and can discharge combustion gases via exhaust duct 48. Intake manifold 44 and exhaust duct 48 can be selectively connected to combustion chamber 30 via intake valve 52 and exhaust valve 54, respectively. In some embodiments, combustion chamber 30 may include two or more intake valves and / or two or more exhaust valves.

[0018] exist Figure 1In the example depicted, intake valve 52 and exhaust valve 54 can be controlled by cam actuation via corresponding cam actuation systems 51 and 53. Cam actuation systems 51 and 53 may each include one or more cams and may utilize one or more of a cam profile changing (CPS) system, a variable cam timing (VCT) system, a variable valve timing (VVT) system, and / or a variable valve lift (VVL) system, which controller 12 can operate to change valve operation. The positions of intake valve 52 and exhaust valve 54 can be determined by position sensors 55 and 57, respectively. In alternative embodiments, intake valve 52 and / or exhaust valve 54 can be controlled by electric valve actuation. For example, cylinder 30 may alternatively include an intake valve controlled by electric valve actuation and an exhaust valve controlled by cam actuation including a CPS system and / or a VCT system.

[0019] In some embodiments, each cylinder of engine 10 may be configured with one or more fuel injectors for supplying fuel thereto. As a non-limiting example, cylinder 30 is shown as including a fuel injector 66. Fuel injector 66 is shown as being directly coupled to cylinder 30 to directly inject fuel into said cylinder in proportion to the pulse width of the signal FPW received from controller 12 via electronic actuator 68. In this manner, fuel injector 66 provides what is considered a fuel injector that directly injects fuel (hereinafter also referred to as "DI") into combustion chamber 30. It will also be understood that cylinder 30 may receive fuel from multiple injections during a combustion cycle.

[0020] In one example, engine 10 may be a diesel engine that burns air and diesel fuel via compression ignition. In other non-limiting examples, engine 10 may burn different fuels, including gasoline, biodiesel, or blends containing alcohols (e.g., gasoline with ethanol or gasoline with methanol), via compression ignition and / or spark ignition.

[0021] The intake manifold 42 may include a throttle valve 62 with a throttle plate 64. In this particular example, the position of the throttle plate 64 may be changed by the controller 12 via a signal provided to an electric motor or actuator included with the throttle valve 62; this is a configuration commonly referred to as electronic throttle control (ETC). In this way, the throttle valve 62 can be operated to change the intake air supplied to the combustion chamber 30 in other engine combustion cylinders. The position of the throttle plate 64 may be provided to the controller 12 via a throttle position signal TP. The intake manifold 42 may include a mass airflow sensor 120 and a manifold air pressure sensor 122 for providing corresponding signals MAF and MAP to the controller 12.

[0022] Furthermore, in the disclosed embodiments, the exhaust gas recirculation (EGR) system can deliver a desired portion of the exhaust gas from the exhaust manifold 48 to the intake manifold 44 via the EGR passage 140. The amount of EGR provided can be varied by the controller 12 via the EGR valve 142. By introducing exhaust gas into the engine 10, the amount of available oxygen for combustion is reduced, thereby lowering the combustion flame temperature and (e.g.) reducing NOx formation. As depicted, the EGR system also includes an EGR sensor 144, which can be arranged within the EGR passage 140 and can provide indications of one or more of the exhaust gas pressure, temperature, and concentration. Under certain conditions, the EGR system can be used to regulate the temperature of the air-fuel mixture in the combustion chamber, thereby providing a method for controlling ignition timing during certain combustion modes. Additionally, under certain conditions, a portion of the combustion gases can be retained or trapped in the combustion chamber by controlling the exhaust valve timing, for example, by controlling a variable valve timing mechanism.

[0023] The exhaust system 128 includes an exhaust sensor 126 coupled upstream of the exhaust treatment system 150 to the exhaust duct 48. The exhaust sensor 126 can be any suitable sensor for providing an indication of the exhaust air / fuel ratio, such as a linear oxygen sensor or UEGO (universal or wide-range exhaust oxygen), a two-state oxygen sensor or EGO, HEGO (heated EGO), NOx, HC, or CO sensor. The exhaust treatment system 150 is shown arranged downstream of the exhaust sensor 126 along the exhaust duct 48.

[0024] exist Figure 1 In the example shown, the exhaust treatment system 150 is a urea-based selective catalytic reduction (SCR) system. The SCR system includes at least, for example, an SCR catalyst 152, a urea storage reservoir 154, and a urea injector 156. In other embodiments, the exhaust treatment system 150 may additionally or alternatively include other components, such as particulate filters, lean NOx traps, three-way catalysts, various other emission control devices, or combinations thereof. In the depicted example, the urea injector 156 supplies urea from the urea storage reservoir 154. However, various alternative methods can be used, such as solid urea pellets that generate ammonia vapor, which is then injected or metered onto the SCR catalyst 152. In another example, the lean NOx trap may be positioned upstream of the SCR catalyst 152 to generate NH3 for the SCR catalyst 152 based on the degree or abundance of the air / fuel ratio fed to the lean NOx trap.

[0025] The exhaust treatment system 150 also includes an exhaust tailpipe exhaust sensor 162 located downstream of the SCR catalyst 152. In the depicted embodiment, exhaust sensor 162 may be, for example, a NOx sensor that measures the amount of post-SCR NOx. The exhaust treatment system 150 may also include a feed gas exhaust sensor 158 located upstream of the urea injector 156 and the SCR catalyst 152. In the depicted embodiment, feed gas exhaust sensor 158 may be, for example, a NOx sensor for measuring the amount of pre-SCR NOx received in the exhaust duct for treatment at the SCR catalyst. The exhaust treatment system 150 also includes an intermediate bed exhaust sensor 164 located in the intermediate bed of the SCR catalyst 152. In the depicted embodiment, intermediate bed exhaust sensor 164 may be, for example, a NOx sensor that measures the amount of intermediate SCR NOx.

[0026] In some examples, the efficiency of the SCR system can be determined based on the outputs of one or more of the exhaust tailpipe exhaust sensor 162, the intermediate bed exhaust sensor 164, and the feed gas exhaust sensor 158. For example, the SCR system efficiency can be determined by comparing the NOx level upstream of the SCR catalyst (via sensor 158) with the NOx level downstream of the SCR catalyst (via sensor 162). Specifically, the exhaust tailpipe efficiency of the SCR catalyst 152 can be determined by comparing the NOx level upstream of the SCR catalyst (via feed gas exhaust sensor 158) with the NOx level measured downstream of the SCR catalyst 152 (via exhaust tailpipe exhaust sensor 162). Furthermore, the intermediate bed efficiency of the SCR catalyst 152 can be determined by comparing the NOx level upstream of the SCR catalyst (via feed gas exhaust sensor 158) with the NOx level measured in the intermediate bed of the SCR catalyst 152 (via intermediate bed exhaust sensor 164). As further described herein, the SCR catalyst 152 can be diagnosed by using the exhaust tailpipe efficiency and / or intermediate bed efficiency of the SCR catalyst 152 as features in a probabilistic classifier. In other embodiments that do not include a dedicated NOx sensor 158 located upstream of the SCR catalyst 152, the conversion efficiency can be based on an exhaust sensor 126 located upstream of the SCR system (e.g., when sensor 126 measures NOx).

[0027] The exhaust system 150 also includes a temperature sensor 160, which is positioned only upstream and near the SCR catalyst 152 to measure the temperature of the exhaust gas entering the catalyst. The controller 12 can therefore receive the temperature measurement of the SCR catalyst 152 from the temperature sensor 160. Alternatively, the temperature sensor 160 may be positioned such that it provides an indication of the exhaust manifold temperature.

[0028] Controller 12 in Figure 1 The image shows a microcomputer, which includes a microprocessor unit 102, an input / output port 104, an electronic storage medium for executable programs and calibration values, shown in this particular example as a read-only memory chip 106, a random access memory 108, a non-fail-in memory 110, and a data bus. In addition to the signals previously discussed, the controller 12 can communicate with and receive various signals from sensors coupled to the engine 10, including measurements of intake mass airflow (MAF) from the mass airflow sensor 120; engine coolant temperature (ECT) from the temperature sensor 112 coupled to the cooling sleeve 114; surface ignition sensing signals (PIP) from the Hall effect sensor 118 (or other type) coupled to the crankshaft 40; throttle position (TP) from the throttle position sensor; absolute manifold pressure (MAP) from sensor 122; and exhaust component concentrations from exhaust sensors 126 and 158. An engine speed signal (RPM) can be generated by the controller 12 from the PIP signals.

[0029] The storage medium read-only memory 106 can be programmed with non-transitory computer-readable data, which represents instructions executable by processor 102 to perform the methods described below, as well as other anticipated but not specifically listed variations. (References to...) Figure 2 To describe an exemplary method.

[0030] As described above, Figure 1 Only one cylinder of a multi-cylinder engine is shown, and each cylinder may similarly include its own set of intake / exhaust valves, fuel injectors, spark plugs, etc.

[0031] Figure 2 A high-level flowchart is shown, illustrating an exemplary method 200 for probabilistic on-board diagnostics according to an embodiment. Specifically, method 200 involves determining whether a vehicle component has deteriorated based on a probabilistic metric calculated for a given operating condition. Compared to other diagnostic methods, method 200 does not require entering conditions to measure or sample operating conditions in order to determine the deterioration of vehicle components. (Reference) Figure 1 Method 200 is described using systems and components, but it should be understood that other systems and components may be used to implement the method without departing from the scope of this disclosure. Method 200 may be implemented as executable instructions in non-transitory memory, such as storage medium read-only memory 106, and the method may be executed by a processor such as processor 102.

[0032] Method 200 begins at 202. At 202, Method 200 initializes the number of iterations N by defining N as equal to 1. As further discussed in this paper, Method 200 applies the number of iterations T to a threshold number of samples. N (also referred to in this article as the iteration threshold T) N The calculated probability measures are averaged, and the number of iterations N increases by one for each iteration until the number of iterations N equals the iteration threshold T. N until.

[0033] After initializing the number of iterations N, method 200 proceeds to step 205. At step 205, method 200 samples operating conditions related to the vehicle component. Operating conditions may include, for example, engine speed, engine load, temperature, oxygen concentration, exhaust component concentration, actuator position, airflow, efficiency, etc. Operating conditions may be sampled based on signals generated by one or more sensors, Boolean or numerical values ​​stored in the memory of controller 12, modeling signals generated by controller 12, etc. The specific operating condition sampled by method 200 is related to the vehicle component being diagnosed. For example, to monitor an SCR catalyst, such as SCR catalyst 152, method 200 may sample signals generated by one or more exhaust gas sensors or NOx sensors located near the SCR catalyst, such as feed gas exhaust gas sensor 158, exhaust tailpipe exhaust gas sensor 162, and intermediate bed exhaust gas sensor 164. Furthermore, the operating condition may include the efficiency of the SCR catalyst calculated based on the signals generated by the exhaust gas sensors. For example, the intermediate bed efficiency of the SCR catalyst 152 can be calculated based on the difference between the exhaust gas concentration measured by the feed gas exhaust sensor 158 and the exhaust gas concentration measured by the intermediate bed exhaust sensor 164, and the exhaust tailpipe efficiency of the SCR catalyst 152 can be calculated based on the difference between the exhaust gas concentration measured by the feed gas exhaust sensor 158 or the intermediate bed exhaust sensor 164 and the exhaust gas concentration measured by the exhaust tailpipe exhaust sensor 162. Therefore, in order to monitor the SCR catalyst 152, method 200 can sample the signals generated by the feed gas exhaust sensor 158, the intermediate bed exhaust sensor 164, and the exhaust tailpipe exhaust sensor 162, and can further calculate the intermediate bed efficiency and exhaust tailpipe efficiency of the SCR catalyst 152 based on the sampled signals.

[0034] As further described herein, the sampled operating conditions can be used as features for a probabilistic classifier. These features can be computed over short time windows to increase robustness to time delays between signals. For example, exhaust tailpipe efficiency can be calculated over a time window according to the following formula:

[0035]

[0036] Where η TP It refers to the exhaust tailpipe efficiency, t0 is the time at the beginning of the time window, and t w It is the length of the time window (e.g., 10 seconds), C TP The exhaust concentration is measured by exhaust sensor 162 in the exhaust tailpipe, and C FG The exhaust concentration is measured by the raw material gas exhaust sensor 158.

[0037] Continuing at 210, method 200 calculates a probability metric for the sample, which includes a score P. 通过 And failure score P 失败 The probability metric includes a quantitative measure of how likely it is that a vehicle component being diagnosed is either deteriorated or not. This is expressed through a score P. 通过 Therefore, quantitative measures are taken to assess the likelihood that vehicle components are not deteriorated, while the failure score P... 失败 This includes a quantitative measure of the likelihood that vehicle components have deteriorated. In one example, this is expressed by a score P. 通过 Including probability, and therefore through the score P 通过 The value can range from zero (0) to one (1), where the value is determined by the score P. 通过 This indicates a 100% probability that a vehicle component has not deteriorated. Simultaneously, the failure score P... 失败 It can correspond to a probability, but instead its range can be from zero (0) to negative one (-1), where the failure score P 失败 A negative one indicates a 100% probability that the vehicle component has deteriorated. Therefore, method 200 can calculate the pass probability and assign a pass score P. 通过 Set to equal the pass probability, and in addition, method 200 can calculate the failure probability and assign a failure score P. 失败 Set to a negative failure probability (e.g., P). 失败 = -Probability of failure). In this way, although the probability always ranges between 0 and 1, it is obtained through the score P. 通过 The range is from 0 to 1, and the failure score P 失败 The range is from 0 to -1. It should be understood that in other examples, both the score and the failure score can directly correspond to a probability ranging from 0 to 1. In that case, those skilled in the art will understand that the failure score ranging from 0 to 1 should be considered when adjusting the thresholds and inequalities described herein.

[0038] For example, method 200 can use a supervised learning method of probabilistic classification to compute a probability metric, such as Gaussian process classification, k-nearest neighbor classification, decision tree classification, etc. In some examples, method 200 can use a probabilistic classifier to compute the pass score P of the sample acquired at position 205. 通过And failure score P 失败 In other examples, method 200 can use a probability classifier to compute the pass score P for the sample taken at position 205. 通过 And then, based on the score P, 通过 To calculate the failure score P 失败 For example, method 200 can achieve this by making the score P... 通过 Subtract one to calculate the failure score P 失败 :

[0039] P 失败 =P 通过 -1.

[0040] In this way, through the score P 通过 A value equal to 1 indicates that there is a 100% probability that the vehicle part is a nominal sample, which will have a corresponding failure score P equal to 0. 失败 This indicates that there is a 0% probability that a vehicle component has deteriorated, as indicated by the score P. 通过 A score of 0.7 indicates a 70% probability that a vehicle component is a nominal sample and will have a corresponding failure score of -0.3. 失败 This indicates that there is a 30% probability that a vehicle component has deteriorated (e.g., because failure probability = -P). 失败 =0.3).

[0041] A probabilistic classifier can be trained on multiple training datasets, which include operating conditions of a given vehicle component in nominal or full-life (FUL) condition and in degraded or threshold condition. A nominal or FUL vehicle component includes a vehicle component in its worst-performing but still acceptable condition, while a degraded or threshold vehicle component includes a vehicle component in its best-performing but unacceptable condition.

[0042] As an illustrative and non-limiting example, the vehicle component may include an SCR catalyst. The operating conditions used to train the probabilistic classifier may therefore include the intermediate bed efficiency or the SCR catalyst's NOx conversion efficiency measured at the midpoint of the SCR catalyst, and the exhaust tailpipe efficiency or the SCR catalyst's NOx conversion efficiency measured downstream of the SCR catalyst in the exhaust tailpipe. As discussed above, such efficiencies can be measured based on signals generated by, for example, the feed gas exhaust sensor 158, the intermediate bed exhaust sensor 164, and the exhaust tailpipe exhaust sensor 162. For example, the intermediate bed efficiency of the SCR catalyst can be calculated based on the difference between the exhaust concentration measured by the feed gas exhaust sensor 158 and the exhaust concentration measured by the intermediate bed exhaust sensor 164, while the exhaust tailpipe efficiency of the SCR catalyst can be calculated based on the difference between the exhaust concentration measured by the feed gas exhaust sensor 158 or the intermediate bed exhaust sensor 164 and the exhaust concentration measured by the exhaust tailpipe exhaust sensor 162. The specific operating conditions used to train the probabilistic classifier can therefore include multiple intermediate bed efficiencies and exhaust tailpipe efficiencies measured for one or more FUL SCR catalysts and one or more threshold SCR catalysts under a wide range of other vehicle operating conditions. This article is about Figure 3 Further examples of how a probabilistic classifier can be trained on such data are described. Therefore, the operating condition sampled at 205 could include features used as input to the probabilistic classifier. In the example of diagnosing an SCR catalyst, intermediate bed efficiency and exhaust tailpipe efficiency could include features used as input to the probabilistic classifier.

[0043] In some examples, a single probability classifier can be used to calculate the score P. 通过 In other examples, multiple individual probabilistic classifiers can be trained on individual features, and the results can be combined by averaging the individual results, a process also known in the art as propagation. For example, regarding the SCR catalyst diagnostics described herein, a first probabilistic classifier can be trained on intermediate bed efficiency data, while a second probabilistic classifier can be trained on exhaust tailpipe efficiency data. Therefore, the total pass score P can be obtained by a weighted sum of the individual pass scores output by the individual probabilistic classifiers. 通过 :

[0044] P 通过 =w1P 通过，1 +w2P 通过，2 ,

[0045] Where w1 and w2 are weights, P 通过,1 It is the first pass score output by the first probability classifier, and P 通过,2The second pass score is the output of the second probability classifier. A separate small classifier can be trained faster and results in lower memory requirements than a single classifier trained on a large set of features.

[0046] The calculation at point 210 includes the score P. 通过 And failure score P 失败 After the probability measure, method 200 proceeds to 215. At 215, method 200 determines the score P. 通过 Whether it is greater than a first threshold T1. The first threshold T1 includes a first pass threshold and is configured such that a pass score P greater than the first threshold T1 is considered to be greater than the first threshold T1. 通过 This is considered acceptable for reliably diagnosing vehicle components. If the score is P... 通过 If the score is greater than the first threshold T1 (“Yes”), then the pass score P is accepted. 通过 Method 200 continues to 225, which is further described below.

[0047] However, if scored by P 通过 If the value is less than or equal to the first threshold T1 ("No"), then method 200 continues to 220. At 220, method 200 determines the failure score P. 失败 Whether it is less than a second threshold T2. The second threshold T2 includes a first failure threshold and is configured such that a failure score P less than the second threshold T2 is considered a failure score. 失败 This is considered acceptable for reliably diagnosing vehicle components. Therefore, if the failure score is P... 失败 If the score is not less than the second threshold T2 ("No"), then method 200 has determined that the score P has been passed. 通过 And failure score P 失败 It is ambiguous whether the vehicle component is nominal or deteriorated. Therefore, method 200 proceeds to 217, where method 200 rejects the sample and returns.

[0048] However, referring again to 220, if it fails, the score is P. 失败 If the score is less than the second threshold T2 (“Yes”), then method 200 continues to 225. Therefore, if the score P is... 通过 Greater than the first threshold T1 or if the score P fails 失败 If the score is less than the second threshold T2, then method 200 proceeds to 225. Method 200 therefore calculates the probability measure of pass / fail under all conditions, but accepts the result only if the probability sufficiently indicates pass or failure, and rejects the sample if the score is ambiguous.

[0049] If the score P is passed 通过 The score is greater than the first threshold T1 or the failure score P. 失败If the score is less than the second threshold T2, then method 200 continues to step 225. At step 225, method 200 calculates the average score P by averaging multiple probability measures over multiple samples. avg The plurality of probability measures include a probability measure calculated at 210, and the plurality of samples include samples acquired at 205. In one example, method 200 can be determined by the pass score P for each sample. 通过 With failure score P 失败 The average score P is calculated by summing the differences between them. avg :

[0050]

[0051] Where P 通过,i Including the i-th pass score P 通过 ,|P 失败,i |Including the score P of the i-th failure 失败 The absolute value of , and N is the current iteration number.

[0052] Continuing at 227, method 200 determines whether the number of iterations N equals the iteration threshold T. N The iteration threshold T can be selected or defined. N To enable the calculation of the average score P of a sufficient number of samples. avg If the number of iterations N is not equal to the iteration threshold T N If ("No"), then method 200 continues to 228. At 228, method 200 increases the iteration number N by setting N to be equal to N+1. For example, if the current iteration number N is 1, then method 200 sets the iteration number N to 2, and if the current iteration number N is 10, then method 200 sets the iteration number N to 11, and so on. Method 200 then continues to 205 to sample the operating condition. In this way, method 200 iterates until the iteration number N equals the iteration threshold T. N This makes the sum of the number of samples equal to the iteration threshold T. N The average score P is calculated using the corresponding probability metric. avg .

[0053] Referring again to 227, once the number of iterations N equals the iteration threshold T... N (Yes), Method 200 then proceeds to 230. At 230, Method 200 determines the average score P. avg Whether it is greater than a third threshold T3. The third threshold T3 includes a second pass threshold and is configured such that the average score P greater than the third threshold T3 is greater than the third threshold T3. avg The nominal robustness indicator is considered to be that of a vehicle component. Therefore, if the average score P avgIf the value is greater than the third threshold T3 (“Yes”), then method 200 continues to 232, and therefore method 200 generates an indication that the vehicle component is nominal. This indication that the vehicle component is nominal can be stored in the non-transitory memory of the controller 12. Method 200 then returns.

[0054] However, referring again to 230, if the average score P avg If the score is less than or equal to the third threshold T3, then method 200 continues to step 235. At step 235, method 200 determines the average score P. avg Whether it is less than a fourth threshold T4. The fourth threshold T4 includes a second failure threshold and is configured such that the average score P less than the fourth threshold T4 is less than the average score P. avg This is considered a robust indicator that vehicle components have deteriorated.

[0055] If the average score is not less than the fourth threshold T4 (“No”), then method 200 continues to 237, therefore method 200 rejects the average score P. avg Method 200 then returns. Therefore, if the average score P... avg If the average score P is not greater than the third threshold T3 and not less than the fourth threshold T4, then the average score P is... avg There is ambiguity regarding whether vehicle parts are nominally deteriorated or not, and the average score P is... avg Rejected.

[0056] However, referring again to 235, if the average score P avg If the average score P is less than the fourth threshold T4 (“Yes”), then the average score P avg The method 200 then proceeds to 240, whereby it generates an indication that a vehicle component has deteriorated. For example, this indication may be stored in the non-transitory memory of the controller 12. At 245, the method 200 outputs a warning that a vehicle component has deteriorated. This warning may be displayed to the vehicle operator, for example, via the vehicle's dashboard. The warning may generally indicate that the vehicle requires maintenance, or it may specifically indicate that a vehicle component requires maintenance. Alternatively, the warning may be stored in an OBD database, accessible communicatively to the vehicle's OBD scanner, allowing the technician to retrieve the warning indicating deterioration of a vehicle component via the OBD scanner. The method 200 then returns to its previous state.

[0057] Therefore, a method for probabilistic on-board diagnostics is provided. The method includes: using a probabilistic classifier to calculate a probability metric of a sample of measured operating conditions of a vehicle system component; averaging a plurality of probability metrics over a plurality of samples, the plurality of probability metrics including the probability metrics, the plurality of samples including the samples; and determining, based on the averaged plurality of probability metrics, whether the vehicle system component has deteriorated.

[0058] As an illustrative example of method 200, Figures 3 to 7 An embodiment in which method 200 is configured to diagnose an SCR catalyst, such as SCR catalyst 152, in a vehicle is illustrated graphically. Figure 3 Figure 300 illustrates training data for a probabilistic classifier according to an embodiment. As discussed above, the intermediate bed efficiency and exhaust tailpipe efficiency, calculated based on measurements of exhaust gas concentrations at various points in the flow path of the SCR catalyst (specifically, upstream, intermediate bed, and downstream), can be used as features of the probabilistic classifier. To this end, Figure 300 illustrates multiple measurements of intermediate bed efficiency and exhaust tailpipe efficiency. Specifically, Figure 300 includes multiple measurements 302 of the intermediate bed efficiency and exhaust tailpipe efficiency of the FUL SCR catalyst, depicted as shaded circles in Figure 300, and multiple measurements 304 of the intermediate bed efficiency and exhaust tailpipe efficiency of the threshold SCR catalyst, depicted as unshaded circles in Figure 300. Each circle on Figure 300 represents a specific measurement of intermediate bed efficiency and exhaust tailpipe efficiency collected at the same time. The multiple measurements 302 and 304 are acquired at a wide range of operating points of the vehicle. In other words, the multiple measurements 302 and 304 are not limited to collection during specific entry conditions. As depicted, the measurements of intermediate bed efficiency and exhaust tailpipe efficiency span the entire range of efficiencies for both threshold SCR catalysts and FUL SCR catalysts. However, the measurements 302 for the FUL SCR catalyst are more densely concentrated in the upper right corner of graph 300, while the measurements 304 for the threshold SCR catalyst are also densely concentrated near the upper right corner of graph 300. The method of averaging multiple probability scores calculated over time thus provides a robust approach for diagnosing SCR catalysts based on such measurements 302 and 304. In contrast, threshold-based methods (e.g., when implemented using such measurements 302 and 304), are likely to produce numerous erroneous affirmations and denials, and therefore such threshold-based methods may necessarily be limited by the specific entry conditions used to obtain the measurements.

[0059] A probabilistic classifier, such as a Gaussian process classifier, is trained using a training dataset comprising multiple measurements 302 and 304, where the measurements 302 and 304 are labeled as corresponding to FUL SCR catalyst and threshold SCR catalyst, respectively. During training, the probabilistic classifier infers the probability that the SCR catalyst is nominal or degraded based on the multiple measurements 302 and 304 and further on their associated labels. Figure 300 depicts multiple probability thresholds used to illustrate the various probabilities that the probabilistic classifier can infer. For example, probability threshold 320 corresponds to a 90% probability that the SCR catalyst is nominal. Therefore, if measurements of intermediate bed efficiency and exhaust tailpipe efficiency above probability threshold 320 are input into the probabilistic classifier, the classifier outputs a 90% or greater probability that the SCR catalyst is nominal via a score P. 通过 .

[0060] Similarly, probability threshold 322 corresponds to an 80% probability that the SCR catalyst is nominal, probability threshold 324 corresponds to a 70% probability that the SCR catalyst is nominal, probability threshold 326 corresponds to a 60% probability that the SCR catalyst is nominal, probability threshold 328 corresponds to a 50% probability that the SCR catalyst is nominal, probability threshold 330 corresponds to a 40% probability that the SCR catalyst is nominal, probability threshold 332 corresponds to a 30% probability that the SCR catalyst is nominal, probability threshold 334 corresponds to a 20% probability that the SCR catalyst is nominal, and probability threshold 336 corresponds to a 10% probability that the SCR catalyst is nominal.

[0061] Therefore, if the measured results of the intermediate bed efficiency and exhaust tailpipe efficiency, which are between probability thresholds 320 and 322, are input into the probability classifier, then the probability classifier will output a pass score P between 0.8 and 0.9. 通过 (This corresponds to a probability between 80% and 90% that the SCR catalyst is nominal). Similarly, if the measured results of intermediate bed efficiency and exhaust tailpipe efficiency between probability thresholds 322 and 324 are input into a probability classifier, then the probability classifier outputs a pass score P between 0.7 and 0.8. 通过 (This corresponds to a probability between 70% and 80% that the SCR catalyst is nominal). If the measured results of the intermediate bed efficiency and exhaust tailpipe efficiency, which are between probability thresholds 324 and 326, are input into a probability classifier, then the probability classifier outputs a pass score P between 0.6 and 0.7. 通过(Corresponding to a probability between 60% and 70% that the SCR catalyst is nominal). If the measured results of the intermediate bed efficiency and exhaust tailpipe efficiency between probability thresholds 326 and 328 are input into a probability classifier, then the probability classifier outputs a pass score P between 0.5 and 0.6. 通过 (Corresponding to a probability between 50% and 60% that the SCR catalyst is nominal). If the measured results of the intermediate bed efficiency and exhaust tailpipe efficiency between probability thresholds 328 and 330 are input into a probability classifier, then the probability classifier outputs a pass score P between 0.4 and 0.5. 通过 (Corresponding to a probability between 40% and 50% that the SCR catalyst is nominal). If the measured results of the intermediate bed efficiency and exhaust tailpipe efficiency between probability thresholds 330 and 332 are input into a probability classifier, then the probability classifier outputs a pass score P between 0.3 and 0.4. 通过 (Corresponding to a probability between 30% and 40% that the SCR catalyst is nominal). If the measured results of the intermediate bed efficiency and exhaust tailpipe efficiency between probability thresholds 332 and 334 are input into a probability classifier, then the probability classifier outputs a pass score P between 0.2 and 0.3. 通过 (Corresponding to a probability between 20% and 30% that the SCR catalyst is nominal). If the measured results of the intermediate bed efficiency and exhaust tailpipe efficiency between probability thresholds 334 and 336 are input into a probability classifier, then the probability classifier outputs a pass score P between 0.1 and 0.2. 通过 (This corresponds to a probability between 10% and 20% that the SCR catalyst is nominal). If the measured results of intermediate bed efficiency and exhaust tailpipe efficiency below the probability threshold of 336 are input into a probability classifier, then the probability classifier outputs a pass score P below 0.1. 通过 (This corresponds to a probability of less than 10% that the SCR catalyst is nominal). The corresponding failure score can be calculated based on the pass / fail score, as described above.

[0062] Figure 4 A set of graphs 400 is shown, illustrating exemplary probability metrics according to embodiments. Specifically, the set of graphs 400 illustrates an exemplary pass score 405 for a FUL SCR catalyst and a pass score 410 for a threshold SCR catalyst. The set of graphs 400 further illustrates the above regarding... Figure 2 The pass threshold 415 is referred to as the first threshold T1. As depicted, the pass threshold 415 can be selected or configured such that pass scores above the pass threshold 415 are considered valid, while pass scores below the pass threshold 415 are considered uncertain or ambiguous and are therefore rejected.

[0063] Similarly, Figure 5 A set of graphs 500 is shown, illustrating exemplary probability metrics according to an embodiment. Specifically, the set of graphs 500 illustrates a failure score 505 for a nominal SCR catalyst and a failure score 510 for a threshold SCR catalyst. This can be based on... Figure 4 The diagram depicts the calculation of failure scores 505 and 510 using scores of 405 and 410. The set of charts 500 further illustrates the above regarding… Figure 2 The failure threshold 515 is referred to as the second threshold T2. As depicted, the failure threshold 515 can be selected or configured such that failure scores below the failure threshold 515 are considered valid, while failure scores above the failure threshold 515 are considered uncertain or ambiguous and are therefore rejected.

[0064] Figure 6 A set of graphs 600 is shown illustrating exemplary calculations of an average probability metric according to an embodiment. Specifically, the set of graphs 600 illustrates multiple pass scores 603 (depicted as circles) and corresponding multiple failure scores 605 (depicted as squares) calculated over time for samples of intermediate bed efficiency and exhaust tailpipe efficiency of an SCR catalyst, such as SCR catalyst 152. The set of graphs 600 further illustrates a number of samples 615 suitable for averaging over time. As depicted, the number of samples 615 increases only when individual results are accepted. More specifically, the number of samples 615 increases only when a pass score among the multiple pass scores 603 is higher than a pass threshold 607 or a failure score among the multiple failure scores 605 is lower than a failure threshold 608. As depicted, some pass scores 603 are not higher than the pass threshold 607, and therefore the number of samples 615 does not increase after such samples are collected.

[0065] The set of charts 600 further illustrates the moving average or average score 625 calculated over time. When the number of samples 615 reaches a threshold number of samples, the average score 625 at that time is used to determine whether the SCR catalyst 152 has deteriorated. If the average score 625 is above the pass threshold 627, then the SCR catalyst 152 is considered nominal or functional within normal limits, while if the average score 625 is below the failure threshold 629, then the SCR catalyst 152 is considered deteriorated or faulty. As depicted, the pass threshold 627 for the average score 625 is different from the pass threshold 607 of the probability metric, and similarly, the failure threshold 629 for the average score 625 is different from the failure threshold 608 of the probability metric. In this example, the SCR catalyst 152 is nominal, and therefore the average score 625 is above the pass threshold 627 at both times the threshold number of samples is reached.

[0066] Furthermore, in the illustrated example, the threshold number of samples used for diagnosing the SCR catalyst 152 is set to twelve. However, it should be understood that a number of samples less than or greater than twelve can be used as the threshold number of samples. Moreover, after determining the state of the SCR catalyst 152 when the threshold number of samples is reached, the average score 625 and the number of samples 615 are reset to zero. In this way, the function of the SCR catalyst 152 can be monitored continuously and periodically over time, rather than during specific entry conditions, thereby improving the IUMPR. Furthermore, because the method used for diagnosing the SCR catalyst 152 is not limited to specific entry conditions, the SCR catalyst 152 can be monitored regardless of driving conditions or the state of the emission driving cycle.

[0067] Figure 7 Chart 700 is shown, illustrating multiple average scores according to embodiments. Specifically, chart 700 depicts an average score of 705 for a threshold SCR catalyst and an average score of 715 for a FULSCR catalyst. Chart 700 also depicts a failure threshold of 707, which, as described above, relates to... Figure 2 Designated as the fourth threshold T4, average scores below this failure threshold (e.g., one of the average scores of 705) are considered acceptable. Similarly, Figure 700 depicts the pass threshold 712, which, as described above, relates to... Figure 2 The average score, referred to as the third threshold T3, is considered acceptable if it is higher than the pass threshold (e.g., one of the average scores in 710). The average score between the failure threshold 707 and the pass threshold 712 can be considered uncertain or ambiguous and therefore can be rejected.

[0068] As depicted, some average scores of 705 and 710 were calculated based on a relatively small sample size. For example, for a threshold SCR catalyst with an average score of 705, some average scores of 705 calculated based on fewer than five samples fall within the acceptable range (below the threshold of 707) and the rejection range (between the thresholds of 707 and 712). To obtain an accurate decision on whether an SCR catalyst has deteriorated or failed, the above discussion... Figure 2 The described iteration threshold T N Therefore, it can be selected or configured to calculate the average score based on a sufficient number of samples.

[0069] In this way, probabilistic diagnostics of vehicle system components are enabled when traditional methods fail to assess performance metrics, thereby improving IUMPR. Furthermore, the system and method presented in this paper enable the automatic identification of conditions and measurements with large intervals, thereby improving the robustness of on-board diagnostics.

[0070] The technical advantage of this disclosure is the continuous monitoring and diagnosis of vehicle system components regardless of driving conditions. Another technical advantage is the robust and definitive detection of degradation of vehicle system components. Yet another technical advantage is the generation of warnings indicating deteriorated vehicle system components.

[0071] As one embodiment, a method includes: using a probability classifier to calculate a probability metric of a sample of measured operating conditions of a vehicle system component; averaging a plurality of probability metrics over a plurality of samples, the plurality of probability metrics including the probability metrics, the plurality of samples including the samples; and determining whether the vehicle system component has deteriorated based on the averaged plurality of probability metrics.

[0072] In a first example of the method, the method further includes discarding the sample if the probability metric is within a threshold range. In a second example of the method, optionally including the first example, the vehicle system component includes a selective catalytic reduction (SCR) catalyst, and the measured operating condition includes the efficiency of the SCR catalyst. In a third example of the method, optionally including one or more of the first and second examples, the method further includes sensing exhaust gas concentration at one or more locations upstream of the SCR catalyst, midway between the SCR catalyst and downstream of the SCR catalyst. In a fourth example of the method, optionally including one or more of the first to third examples, the method further includes calculating the intermediate bed efficiency of the SCR catalyst and the exhaust tailpipe efficiency of the SCR catalyst, wherein the efficiency of the SCR catalyst includes the intermediate bed efficiency and the exhaust tailpipe efficiency. In a fifth example of the method, optionally including one or more of the first to fourth examples, the probabilistic classifier includes one of a Gaussian process classifier, a k-nearest neighbor classifier, and a decision tree classifier. In a sixth example of the method, which optionally includes one or more of the first to fifth examples, the probability metric includes a pass score calculated by the probability classifier and a failure score calculated based on the pass score. In a seventh example of the method, which optionally includes one or more of the first to sixth examples, averaging the plurality of probability metrics includes summing the differences between the pass score and the failure score for each of the plurality of samples.

[0073] As another embodiment, a method for on-board diagnostics of a vehicle system component includes: acquiring samples of operating conditions associated with the vehicle system component; calculating a probability metric of the samples; calculating an average score based on a plurality of probability metrics of a plurality of samples, the plurality of probability metrics including the probability metrics, the plurality of samples including the samples; indicating that the vehicle system component is nominal in response to the average score being higher than a first threshold; indicating that the vehicle system component has deteriorated in response to the average score being lower than a second threshold; and discarding the average score in response to the average score being between the first threshold and the second threshold.

[0074] In a first example of the method, the method further includes determining that a second probability metric calculated from a second sample among the plurality of samples falls within a second threshold range, and discarding the second probability metric and the second sample such that the second probability metric and the second sample are not included in the calculation of the average score. In a second example of the method, optionally including the first example, calculating the probability metric of the sample includes inputting the sample into a probability classifier, training the probability classifier using multiple data points of a second vehicle system component in a full-lifetime (FUL) state and a third vehicle system component in a threshold state. In a third example of the method, optionally including one or more of the first and second examples, the probability metric includes a pass score and a failure score, and the method further includes discarding the sample if the pass score is below a third threshold and the failure score is above a fourth threshold. In a fourth example of the method, optionally including one or more of the first to third examples, the method further includes calculating the average score when the number of samples among the plurality of samples reaches a threshold number of samples.

[0075] As another embodiment, a system includes: a vehicle system component; at least one sensor configured to measure the operating condition of the vehicle system component; and a controller communicatively coupled to the at least one sensor and configured with executable instructions in a non-transitory memory, the executable instructions, when executed, causing the controller to: use a probability classifier to compute a probability metric of samples of the operating condition of the vehicle system component measured by the at least one sensor; average a plurality of probability metrics of a plurality of samples, the plurality of probability metrics including the probability metrics, the plurality of samples including the samples; and determine whether the vehicle system component has deteriorated based on the averaged plurality of probability metrics.

[0076] In a first example of the system, the vehicle system component includes a selective catalytic reduction (SCR) catalyst. In a second example of the system optionally including the first example, the at least one sensor includes a first exhaust sensor located upstream of the SCR catalyst, a second exhaust sensor located midway up the SCR catalyst, and a third exhaust sensor located downstream of the SCR catalyst. In a third example of the system optionally including one or more of the first and second examples, the sample of the operating condition includes the intermediate bed efficiency and exhaust tailpipe efficiency of the SCR catalyst measured by the first, second, and third exhaust sensors. In a fourth example of the system optionally including one or more of the first to third examples, the probability metric includes a pass score and a failure score, and the controller is further configured with executable instructions in the non-transitory memory that, when executed, cause the controller to reject a sample from the plurality of samples if the pass score is below a first threshold and the failure score is above a second threshold. In a fifth example of the system, optionally including one or more of the first to fourth examples, the controller is further configured with executable instructions in the non-transitory memory that, when executed, cause the controller to reject the averaged plurality of probability metrics if they are within a threshold range, and to determine that the vehicle system component has deteriorated if the averaged plurality of probability metrics are below the threshold range. In a sixth example of the system, optionally including one or more of the first to fifth examples, the controller is further configured with executable instructions in the non-transitory memory that, when executed, cause the controller to generate and output a warning indicating that the vehicle system component has deteriorated if the averaged plurality of probability metrics are below the threshold range.

[0077] It should be noted that the exemplary control and estimation routines included herein can be used in various engine and / or vehicle system configurations. The control methods and routines disclosed herein can be stored as executable instructions in non-transitory memory and can be executed by a control system comprising a combination of controllers and various sensors, actuators, and other engine hardware. The specific routines described herein can represent one or more of any number of processing strategies, such as event-driven, interrupt-driven, multitasking, multithreading, etc. Therefore, the various actions, operations, and / or functions described may be performed in the illustrated sequence, in parallel, or in some cases omitted. Similarly, the processing order is not necessarily required to achieve the features and advantages of the exemplary embodiments described herein, but is provided for the sake of simplicity of description and illustration. One or more of the illustrated actions, operations, and / or functions can be repeatedly performed according to the specific strategy used. Furthermore, the described actions, operations, and / or functions can clearly represent code to be programmed into a non-transitory memory of a computer-readable storage medium in the engine control system, wherein the described actions are implemented by executing instructions in a system comprising a combination of various engine hardware components and electronic controllers.

[0078] It will be understood that the configurations and routines disclosed herein are exemplary in nature and should not be viewed in a limiting sense, as numerous variations are possible. For example, the above techniques can be applied to V-6, I-4, I-6, V-12, opposed 4-cylinder, and other engine types. The subject matter of this disclosure includes all novel and non-obvious combinations and sub-combinations of the various systems and configurations disclosed herein with other features, functions, and / or properties.

[0079] The appended claims specifically point to particular combinations and sub-combinations considered novel and non-obvious. These claims may refer to an element or a first element or its equivalent. Such claims should be understood to include one or more such elements, neither requiring nor excluding two or more such elements. Other combinations and sub-combinations of the disclosed features, functions, elements, and / or properties may be claimed by amending these claims or by presenting new claims in this or related applications. Such claims, whether broader, narrower, equivalent, or different in scope from the original claims, are also considered to be included within the subject matter of this disclosure.

[0080] According to the present invention, a method includes: using a probability classifier to calculate a probability score for a sample of measured operating conditions of a vehicle system component; averaging a plurality of probability scores for a plurality of samples, the plurality of probability scores including the probability scores, the plurality of samples including the samples; and determining whether the vehicle system component has deteriorated based on the averaged plurality of probability scores.

[0081] According to an embodiment, the present invention is further characterized in that the sample is discarded when the probability score is within a threshold range.

[0082] According to an embodiment, the vehicle system component includes a selective catalytic reduction (SCR) catalyst, and the measured operating conditions include the efficiency of the SCR catalyst.

[0083] According to an embodiment, the invention is further characterized in that exhaust gas concentration is sensed at one or more of the following locations: upstream of the SCR catalyst, midway between the SCR catalyst and downstream of the SCR catalyst.

[0084] According to an embodiment, the present invention is further characterized by calculating the intermediate bed efficiency of the SCR catalyst and the exhaust tailpipe efficiency of the SCR catalyst, wherein the efficiency of the SCR catalyst includes the intermediate bed efficiency and the exhaust tailpipe efficiency.

[0085] According to an embodiment, the probabilistic classifier includes one of a Gaussian process classifier, a k-nearest neighbor classifier, and a decision tree classifier.

[0086] According to an embodiment, the probability score includes a pass score calculated by the probability classifier and a failure score calculated based on the pass score.

[0087] According to an embodiment, averaging the plurality of probability scores includes summing the differences between the pass score and the failure score for each of the plurality of samples.

[0088] According to the present invention, a method for on-board diagnostics of a vehicle system component includes: acquiring a sample of operating conditions related to the vehicle system component; calculating a probability metric of the sample; calculating an average score based on a plurality of probability metrics of a plurality of samples, the plurality of probability metrics including the probability metrics, the plurality of samples including the sample; indicating that the vehicle system component is nominal in response to the average score being higher than the first threshold; indicating that the vehicle system component has deteriorated in response to the average score being lower than the second threshold; and discarding the average score in response to the average score being between the first threshold and the second threshold.

[0089] According to an embodiment, the invention is further characterized in that a second probability measure calculated from a second sample among the plurality of samples is within a second threshold range, and the second probability measure and the second sample are discarded such that the second probability measure and the second sample are not included in the calculation of the average score.

[0090] According to an embodiment, calculating the probability metric of the sample includes inputting the sample into a probability classifier and training the probability classifier using multiple data points of a second vehicle system component in a full-lifetime (FUL) state and a third vehicle system component in a threshold state.

[0091] According to an embodiment, the probability metric includes a pass score and a failure score, and further includes discarding the sample if the pass score is below a third threshold and the failure score is above a fourth threshold.

[0092] According to an embodiment, the invention is further characterized in that the average score is calculated when the number of the plurality of samples reaches a threshold number of samples.

[0093] According to the present invention, a system is provided, the system comprising: a vehicle system component; at least one sensor configured to measure the operating condition of the vehicle system component; and a controller communicatively coupled to the at least one sensor and configured with executable instructions in a non-transitory memory, the executable instructions, when executed, causing the controller to: use a probability classifier to calculate a probability metric of samples of the operating condition of the vehicle system component measured by the at least one sensor; average a plurality of probability metrics of a plurality of samples, the plurality of probability metrics including the probability metrics, the plurality of samples including the samples; and determine whether the vehicle system component has deteriorated based on the averaged plurality of probability metrics.

[0094] According to an embodiment, the vehicle system component includes a selective catalytic reduction (SCR) catalyst.

[0095] According to an embodiment, the at least one sensor includes a first exhaust sensor located near and upstream of the SCR catalyst, a second exhaust sensor located in the middle of the SCR catalyst, and a third exhaust sensor located near and downstream of the SCR catalyst.

[0096] According to an embodiment, the sample of the operating condition includes the intermediate bed efficiency and exhaust tailpipe efficiency of the SCR catalyst measured by the first exhaust sensor, the second exhaust sensor, and the third exhaust sensor.

[0097] According to an embodiment, the probability metric includes a pass score and a failure score, and wherein the controller is further configured with executable instructions in the non-transitory memory, which, when executed, cause the controller to reject a sample from the plurality of samples if the pass score is below a first threshold and the failure score is above a second threshold.

[0098] According to an embodiment, the controller is further configured with executable instructions in the non-transitory memory, which, when executed, cause the controller to reject the averaged plurality of probability measures if the averaged plurality of probability measures are within a threshold range, and to determine that the vehicle system component has deteriorated if the averaged plurality of probability measures are below the threshold range.

[0099] According to an embodiment, the controller is further configured with executable instructions in the non-transitory memory, which, when executed, cause the controller to generate and output a warning indicating that the vehicle system components have deteriorated when the average of a plurality of probability metrics is below the threshold range.

Claims

1. A method for a vehicle, comprising: A probability classifier is used to calculate a probability metric for a sample of measured operating conditions of vehicle system components, wherein the probability metric includes pass scores and failure scores. The average of multiple probability measures for multiple samples is calculated, wherein the multiple probability measures include the probability measures and the multiple samples include the samples; The vehicle system components are determined to be degraded based on the average of multiple probability measures. as well as If the pass score is below a first threshold and the failure score is above a second threshold, the sample is rejected from the plurality of samples.

2. The method of claim 1, wherein the vehicle system component includes a selective catalytic reduction catalyst, i.e., an SCR catalyst, and wherein the measured operating conditions include the efficiency of the SCR catalyst.

3. The method of claim 2, further comprising sensing exhaust gas concentration at one or more of the following locations: upstream of the SCR catalyst, midway between the SCR catalyst and downstream of the SCR catalyst.

4. The method of claim 3, further comprising calculating the intermediate bed efficiency of the SCR catalyst and the exhaust tailpipe efficiency of the SCR catalyst, wherein the efficiency of the SCR catalyst includes the intermediate bed efficiency and the exhaust tailpipe efficiency.

5. The method of claim 1, wherein the probability classifier comprises one of a Gaussian process classifier, a k-nearest neighbor classifier, and a decision tree classifier.

6. The method of claim 1, wherein the pass score is calculated by the probability classifier and the failure score is calculated based on the pass score.

7. The method of claim 6, wherein averaging the plurality of probability measures comprises summing the differences between the pass score and the failure score for each of the plurality of samples.

8. A system for a vehicle, comprising: Vehicle system components; At least one sensor, the at least one sensor being configured to measure the operating conditions of the vehicle system components; A controller, communicatively coupled to the at least one sensor and configured with executable instructions in a non-transitory memory, which, when executed, cause the controller to: A probability classifier is used to calculate a probability metric for samples of the operating conditions of the vehicle system components as measured by the at least one sensor, wherein the probability metric includes pass scores and failure scores. The average of multiple probability measures for multiple samples is calculated, wherein the multiple probability measures include the probability measures and the multiple samples include the samples; The vehicle system components are determined to be degraded based on the average of multiple probability measures. as well as If the pass score is below a first threshold and the failure score is above a second threshold, the sample is rejected from the plurality of samples.

9. The system of claim 8, wherein the vehicle system component includes a selective catalytic reduction catalyst, i.e., an SCR catalyst.

10. The system of claim 9, wherein the at least one sensor comprises a first exhaust sensor located near and upstream of the SCR catalyst, a second exhaust sensor located in the middle of the SCR catalyst, and a third exhaust sensor located near and downstream of the SCR catalyst.

11. The system of claim 10, wherein the sample of the operating condition includes the intermediate bed efficiency and exhaust tailpipe efficiency of the SCR catalyst as measured by the first exhaust sensor, the second exhaust sensor and the third exhaust sensor.

12. The system of claim 8, wherein the controller is further configured with executable instructions in the non-transitory memory, the executable instructions, when executed, causing the controller to reject the averaged plurality of probability measures if the averaged plurality of probability measures are within a threshold range, and to determine that the vehicle system component has deteriorated if the averaged plurality of probability measures are below the threshold range.

13. The system of claim 12, wherein the controller is further configured with executable instructions in the non-transitory memory, the executable instructions, when executed, causing the controller to generate and output a warning indicating that the vehicle system components have deteriorated when the average of the plurality of probability metrics is below the threshold range.

Images