Description:Complete Specification:
The following specification describes and ascertains the nature of this invention and the manner in which it is to be performed

Field of the invention
[0001] The present disclosure relates to the field of air filters used in vehicle systems.

Background of the invention

[0002] Clogged air filter can result in insufficient air flow to the engine. This phenomenon could lead to problems such as-reduced fuel economy and horsepower, high emission due to improper and unwanted combustion, engine misfiring. Improper and unwanted combustion could lead to damage of engine and other components. Further, insufficient air flow results in reduction in torque which in turn reduce the efficiency of the vehicle. Reduction in air mass results increase / decrease in EGR from the required quantity results in engine damage.
[0003] The air filters are replaced at regular kilometer intervals than based on the performance of the filter. This can lead to late replacement of the filters leading to over clogging resulting in above listed issues or early replacement of the filter before complete filter life is used.
[0004] The proposed solution in the present disclosure provides the exact clogging level of the filter suggesting for its optimum replacements. This improves the engine life and reduces the maintenance cost than unmonitored or unconditional replacements of air filter during scheduled intervals.
[0005] A machine learning model is trained with air filters of known clog percentages (levels). The ML model is then built by using engine parameters that are directly correlated with the performance of an air filter when it is clogged. This built ML model is then deployed as an API (application programming interface) in virtual database or a as a processor in the vehicle that takes selected engine parameters as inputs and provides the current clogging level.

Brief description of the accompanying drawings
[0006] An embodiment of the invention is described with reference to the following accompanying drawings:
[0007] Figure 1 depicts a system to detect clogging of current air filter of an internal combustion engine (ICE).
[0008] Figure 2 depicts a flow chart for the method to detect clogging of an air filter of an internal combustion engine (ICE).

Detailed description of the drawings

[0009] At the core of the system to detect the clogging in an air filter, is a Machine learning (ML) model which is trained using different types of air filter with different clog levels which are pre-labeled with exact clog percentage. For instance, five air filters with pre-labelled clog percentages (levels) of 0%, 20%, 40%, 60% and 100% respectively can be used. Using these air filters of known clog levels, the engine parameters are run and the parameters correlated or affected with clogging of air filters are selected. Therefore, classification model is built based on the engine parameters from the vehicle which are indicators of the air filter clog level through the performance variation of the engine which gets reflected through engine parameters.
[0010] In an example, these selected engine parameters can be- Engine RPM (this parameter is related to air filter as a clogged air filter leads to insufficient air flow which in turn reduces torque), Intake manifold absolute pressure (Reduction in pressure due to clogging), Air flow rate from mass air flow sensor(Reduction in air flow due to clogging), Control module voltage (Idle region gives additional load alternator with clogging).
[0011] The significant vehicle parameters are selected by using various feature selection techniques and each selected variable has a correlation to the air filter’s behavior. It is to be noted that any engine parameter correlated to the air filter’s behavior can be selected and this is not be taken as limiting the scope of the present disclosure.
[0012] The above techniques are further described with reference to
Figure 1 and Figure 2. It should be noted that the description and the figures merely illustrate the principles of the present subject matter along with examples described herein and should not be construed as a limitation to the present subject matter. It is thus understood that various arrangements may be devised that, although not explicitly described or shown herein, embody the principles of the present subject matter. Moreover, all statements herein reciting principles, aspects, and implementations of the present subject matter, as well as specific examples thereof, are intended to encompass equivalents thereof.

[0013] Figure 1 depicts a system to detect clogging of current air filter of an internal combustion engine (ICE). Only the important components of the system are disclosed in this document as all other components are commonly known. The system consists of a processor (2) with an associated memory that stores the ML model (3) and the associated database (4) to detect clogging of current air filter of an internal combustion engine of a vehicle (5) from which (the vehicle (5)), it receives input.
[0014] The present disclosure may be implemented as a set of software instructions, combination of software and hardware or any combination of the same. The inputs to the processor may come over a bus or wirelessly or through any other communication. The output interface may comprise a display or a bus.
[0015] The output from the processor may be displayed on a display or sent through the bus which can be read by other devices.
[0016] The processor (2) may be deployed as an Application programming interface (API) (1) in cloud which takes the selected parameters as input through a telematic device connected in the vehicle(5) and provides the actual clog status or class of the air filter in the particular vehicle.
[0017] Alternatively, the processor (2) can be deployed as an edge use case operating from control unit in the vehicle(5) instead of from cloud(1) which based on selected control unit parameters provides the actual clogging status of the air filter and trigger for its optimum replacement.
[0018] The processor (2) is configured to train the ML model with plurality of air filters (10, 20, 30) each of the plurality of air filter having a known clog level. It then selects plurality of engine parameters correlated with the air filter(s)’s (10, 20, 30) clog level and builds the ML model based on the selected plurality of engine parameters. The model so trained and built is termed as a ML model (3).
[0019] The ML model (3) is Machine learning model based on Extra tree classifier. ExtraTrees is a known ensemble machine learning approach. Ensemble methods are techniques that aim at improving the accuracy of results in models by combining multiple models instead of using a single model. The combined models increase the accuracy of the results significantly.
[0020] Under the extra tree classification numerous decision trees are trained and the results from the group of decision trees are aggregated to output a prediction.
[0021] The ML model is trained using filters of different clog levels to build a virtual sensor model for the actual air filter clog detection. The model can be fine tuned for the best accuracies and robustness. Since there can be multiple regions (while running the engine parameters) where the behavior can change, this extra tree classifier algorithm creates a large number of unpruned decision trees and the predictions are made by averaging the prediction of the decision trees and uses majority voting to arrive to a particular class.
[0022] In run time, the raw data from the idle condition of the vehicle will be analyzed by the trained extra-tree classifier model and the result of the model goes to a post- processing algorithm which rectifies sudden spikes/variations due to some anomalies caused by sudden environmental changes or sensor errors and detects the current clog level on the air filter.
[0023] The post-processing algorithm considers the historical clog behavior stored in the database (4) from the vehicle to decide on the current clogging level.
[0024] In an example, three air filters with pre-labelled clog percentages (levels) of 0%, 60% and 100% respectively can be used. Using these air filters of known clog levels, the engine parameters are run and the parameters correlated or affected with clogging of air filters are selected. Therefore, a classification model is built done based on the engine parameters from the vehicle which are indicators of the air filter clog level through the performance variation of the engine which gets reflected through engine parameters.
[0025] In an example, these selected engine parameters can be- Engine RPM (this parameter is related to air filter as a clogged air filter leads to insufficient air flow which in turn reduces torque), Intake manifold absolute pressure (Reduction in pressure due to clogging), Air flow rate from mass air flow sensor(Reduction in air flow due to clogging), Control module voltage (Idle region gives additional load alternator with clogging).

[0026] The processor (2) then takes selected engine parameters received as inputs from the vehicle (5) and detects the unknown clog level of the air filter in real time.

[0027] Figure 2 depicts a flow chart for the method to detect clogging of an air filter of an internal combustion engine (ICE) through a ML model.

[0028] Depicted in Figure 2 is a flow chart for the method to detect clogging of an air filter of an Internal combustion engine (ICE) of a vehicle through a ML model. In the first step, a processor trains the ML model with plurality of air filters, each of the plurality of air filters having a known clog level (100). A Classification model is built by the processor using a machine learning algorithm based on the selected plurality of engine parameters (200). Once the model is built, the model can be deployed to detect unknown clog level of an air filter(300). The detection is done based on selected engine parameters received as inputs the vehicle (400), the said vehicle in communication with the processor.

[0029] The processor may have an associated memory that stores the ML model and the associated database to detect clogging of current air filter of an internal combustion engine of a vehicle from which it receives input.

[0030] The processor may be deployed as an Application programming interface (API) in cloud which takes the selected parameters as input through a telematic device connected in the vehicle and provides the actual clog status or class of the air filter in the particular vehicle.
[0031] Alternatively, the processor can be deployed as an edge use case operating from control unit in the vehicle which based on selected control unit parameters provides the actual clogging status of the air filter and trigger for its optimum replacement.
[0032] The ML model is Machine learning model based on Extra tree classifier. ExtraTrees is a known ensemble machine learning approach. Ensemble methods are techniques that aim at improving the accuracy of results in models by combining multiple models instead of using a single model. The combined models increase the accuracy of the results significantly.
[0033] Under the extra tree classification numerous decision trees are trained and the results from the group of decision trees are aggregated to output a prediction.
[0034] In run time, the raw data from the idle condition of the vehicle will be analyzed by the trained extra-tree classifier model. Optionally, the result of the model can go to a post- processing algorithm which rectifies sudden spikes/variations due to some anomalies caused by sudden environmental changes or sensor errors and detects the current clog level on the air filter.
[0035] The post-processing algorithm can consider the historical clog behavior stored in the database from the vehicle to decide on the current clogging level.
[0036] The disclosure advantageously provides the exact clogging level of the filter suggesting for its optimum replacements. This improves the engine life and reduces the maintenance cost than unmonitored or unconditional replacements of air filter during scheduled intervals.
, Claims:We Claim:
1. A method to detect clogging of an air filter of an Internal combustion engine (ICE) of a vehicle through a machine learning (ML) model, characterized in that method the steps of :
- training the ML model, by a processor, wherein, the ML model is trained with plurality of air filters, each of the plurality of air filters having a known clog level (100);
-selecting, by the processor, plurality of engine parameters correlated with the air filter’s clog level(200);
-building the ML model, by the processor, based on the selected plurality of engine parameters (300); and
- detecting, by the processor, the unknown clog level of the air filter in real time (400).

2. The method to detect clogging of an air filter as claimed in claim 1, wherein, the unknown clog level of the air filter is detected based upon the selected engine parameters received as inputs from the vehicle in communication with the processor.

3. The method to detect clogging of an air filter as claimed in claim 1, wherein, said processor in communication with plurality of vehicle-
is deployed in a virtual database; and
receives the selected plurality of engine parameters as inputs from plurality of vehicles.

4. The method to detect clogging of an air filter as claimed in Claim 1, wherein, said processor deployed in the vehicle and receives the selected engine parameters as inputs from the vehicle.

5. The method to detect clogging of an air filter as claimed in claim 1, wherein, the ML model is trained based on an ensemble learning technique.

6. A processor(2) to detect clogging of current air filter of an internal combustion engine of a vehicle through a ML model, the processor configured to:
-Train the ML model with plurality of air filters (10, 20, 30), each of the plurality of air filter having a known clog level;
-select plurality of engine parameters correlated with the air filter’s clog level;
-build the ML model(3) based on the selected plurality of engine parameters;
-take selected engine parameters received as inputs from the vehicle (5); and
- detect the unknown clog level of the air filter in real time.

7. The processor to detect clogging of current air filter as claimed in Claim 6 wherein, the ML model is trained based on an ensemble learning technique.