KR20210124785A - Method for diagnosing diesel particulate filter using machine learning - Google Patents

Abstract

A diagnosis method of an exhaust gas post-processing apparatus according to the present invention first trains a diagnosis model about the exhaust gas post-processing apparatus using machine learning. And then, the diagnosis method measures temperature of exhaust gas, flow amount of the exhaust gas, and a pressure reduction amount between a front end and a rear end of the exhaust gas post-processing apparatus and diagnose a status of the exhaust gas post-processing apparatus by applying the measured temperature, flow amount, and pressure reduction amount on the trained diagnosis model. The present invention diagnoses installation status and necessity of regeneration of the exhaust gas post-processing apparatus.

Description

Method for diagnosing exhaust gas after-treatment device using machine learning

The present invention relates to a method for diagnosing an exhaust gas post-treatment device using machine learning. In more detail, the present invention learns a diagnostic model for estimating the mass of particulate matter (PM) collected in the exhaust gas post-treatment device and whether the exhaust gas post-treatment device is installed using machine learning, and the learned diagnosis It relates to a method for accurately diagnosing the state of an exhaust gas aftertreatment device using a model.

A diesel engine can burn fuel at a high compression ratio by adopting a compression ignition combustion method, but has a disadvantage in that it discharges a large amount of pollutants such as nitrogen oxides (NOx) and dust. Accordingly, it is compulsory by law to install an exhaust gas aftertreatment device (Diesel Particulate Filter, DPF) in vehicles to which a diesel engine is applied.

The DPF is a device for collecting particulate matter (PM) contained in exhaust gas, and in order to maintain its performance, a regeneration process of burning and removing the captured PM is essential. This regeneration process may be performed periodically or may be performed when the PM is captured above a reference value.

Meanwhile, for regeneration, the DPF must be heated to a high temperature, and in this process, fuel efficiency and output of the vehicle are simultaneously reduced. In addition, there is an inconvenience in that the driver has to wait without turning off the engine until the regeneration process is finished. For this reason, some drivers arbitrarily remove the DPF. However, when the DPF is removed, pollutants such as PM are discharged to the outside as they are, which may adversely affect the environment. For this reason, related laws and regulations mandate that the presence or absence of DPF be diagnosed as essential.

In order to satisfy such a requirement, in the related art, exhaust gas pressures at the front and rear ends of the DPF were measured, and the presence or absence of the DPF installation was indirectly estimated using the pressure drop value. For example, when the DPF is normally installed, since the pressure drops more than when the DPF is removed, it is determined that the DPF has been removed when the pressure drop is less than a certain value. However, in the region where the exhaust gas flow rate is small, the difference in pressure drop according to the presence or absence of the DPF is not large. Therefore, in order to determine whether the DPF is installed, only a region having a large exhaust gas flow rate has to be used, and there is a problem in that the time required for diagnosis increases.

Korean Patent No. 10-1775018 (registered on August 30, 2017)

An object of the present invention is to learn a diagnostic model capable of diagnosing an exhaust gas after-treatment device (DPF) using machine learning, and whether to install and regenerate an exhaust gas after-treatment device (DPF) using the learned diagnostic model The goal is to provide a method for accurately diagnosing a need.

However, the problems to be solved by the present invention are not limited to the above problems, and may be variously expanded within the scope without departing from the spirit and scope of the present invention.

In order to achieve the above object of the present invention, a method for diagnosing an exhaust gas after-treatment apparatus according to exemplary embodiments includes: learning a diagnostic model for the exhaust gas after-treatment apparatus using machine learning; measuring the temperature of the exhaust gas, the flow rate of the exhaust gas, and the amount of pressure drop between the front end and the rear end of the exhaust gas aftertreatment device; and diagnosing the state of the exhaust gas post-processing apparatus by applying the measured temperature, flow rate, and pressure drop data to the learned diagnostic model.

In this case, the learning of the diagnostic model may include: measuring a temperature of exhaust gas, a flow rate of exhaust gas, and an amount of pressure drop between the front and rear ends of the exhaust gas post-processing apparatus; estimating the amount of particulate matter collected in the exhaust gas post-treatment device under the measured conditions using a Gaussian process model; and estimating whether the exhaust gas post-treatment device is installed by comparing the estimated amount of particulate matter collected with a preset first reference value.

In addition, the learning of the diagnostic model may further include measuring the mass of the particulate matter collected in the exhaust gas post-processing device, and verifying the estimation of the particulate matter collection amount.

In addition, the learning of the diagnostic model may include: verifying an estimate of whether the exhaust gas post-treatment device is installed by checking whether the exhaust gas post-treatment device is actually installed; and resetting the first reference value when the estimation of whether the verification result is installed or not is incorrect.

In addition, diagnosing the state of the exhaust gas post-treatment device may include diagnosing whether the exhaust gas post-treatment device is actually installed. If it is determined that the exhaust gas after-treatment device is not installed, a warning signal may be output.

Alternatively, the learning of the diagnostic model may include: measuring a temperature of exhaust gas, a flow rate of exhaust gas, and an amount of pressure drop between the front and rear ends of the exhaust gas post-treatment device; estimating the amount of particulate matter collected in the exhaust gas post-treatment device under the measured conditions using a Gaussian process model; and verifying an estimate of the amount of particulate matter collected by confirming an actual collection amount of the particulate matter collected in the exhaust gas post-treatment device.

In this case, the step of diagnosing the state of the exhaust gas post-treatment device may include determining the need for regeneration of the exhaust gas post-treatment device by comparing the estimated amount of particulate matter collected with a preset second reference value. can

A method for diagnosing an exhaust gas post-treatment device according to another embodiment of the present invention includes: learning a diagnostic model for determining whether the exhaust gas post-treatment device is installed through machine learning; and diagnosing whether the exhaust gas post-treatment device is installed using the learned diagnostic model.

In this case, the learning of the diagnostic model may include: measuring a temperature of exhaust gas, a flow rate of exhaust gas, and an amount of pressure drop between the front and rear ends of the exhaust gas post-processing apparatus; estimating the amount of particulate matter collected in the exhaust gas post-treatment device under the measured conditions using a Gaussian process model; estimating whether the exhaust gas post-treatment device is installed by comparing the estimated amount of particulate matter collected with a preset first reference value; verifying an estimate of whether the exhaust gas post-treatment device is installed by checking whether the exhaust gas post-treatment device is actually installed; and resetting the first reference value when the estimation of whether the verification result is installed or not is incorrect.

The determining whether the exhaust gas post-treatment device is installed may include: measuring the temperature of the exhaust gas, the flow rate of the exhaust gas, and the amount of pressure drop between the front end and the rear end of the exhaust gas post-treatment device; and substituting the measured temperature, flow rate, and pressure drop data into the learned diagnostic model to diagnose whether the exhaust gas post-processing device is installed.

The diagnostic method of the exhaust gas treatment apparatus according to the exemplary embodiments of the present invention can derive a highly reliable diagnostic model through machine learning, and use it to accurately determine the state of the DPF, for example, whether the DPF is installed or the need for regeneration. can judge In particular, since various data including a condition of low exhaust gas flow rate are used in the machine learning process, the state of the DPF can be accurately diagnosed regardless of the exhaust gas flow rate, and the time required for diagnosis can be greatly reduced.

1 is a view showing a conventional exhaust gas after-treatment system.
2 is a graph showing the relationship between the exhaust gas flow rate and the pressure drop.
3 is a flowchart for explaining steps of a method for diagnosing an exhaust gas post-treatment apparatus according to the present invention.

With respect to the embodiments of the present invention disclosed in the text, specific structural or functional descriptions are only exemplified for the purpose of describing the embodiments of the present invention, and the embodiments of the present invention may be embodied in various forms. It should not be construed as being limited to the embodiments described in .

Since the present invention can have various changes and can have various forms, specific embodiments are illustrated in the drawings and described in detail in the text. However, this is not intended to limit the present invention to the specific disclosed form, it should be understood to include all modifications, equivalents and substitutes included in the spirit and scope of the present invention.

Terms such as first, second, etc. may be used to describe various elements, but the elements should not be limited by the terms. The above terms may be used for the purpose of distinguishing one component from another. For example, without departing from the scope of the present invention, a first component may be referred to as a second component, and similarly, a second component may also be referred to as a first component.

When a component is referred to as being “connected” or “connected” to another component, it is understood that the other component may be directly connected or connected to the other component, but other components may exist in between. it should be On the other hand, when it is said that a certain element is "directly connected" or "directly connected" to another element, it should be understood that no other element is present in the middle. Other expressions describing the relationship between elements, such as "between" and "immediately between" or "neighboring to" and "directly adjacent to", should be interpreted similarly.

The terms used in the present application are only used to describe specific embodiments, and are not intended to limit the present invention. The singular expression includes the plural expression unless the context clearly dictates otherwise. In the present application, terms such as “comprise” or “have” are intended to designate that the described feature, number, step, operation, component, part, or combination thereof exists, and includes one or more other features or numbers. , it is to be understood that it does not preclude the possibility of the presence or addition of steps, operations, components, parts, or combinations thereof.

Unless defined otherwise, all terms used herein, including technical and scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. Terms such as those defined in commonly used dictionaries should be interpreted as meanings consistent with the context of the related art, and unless explicitly defined in the present application, they are not to be interpreted in an ideal or excessively formal meaning. .

Hereinafter, preferred embodiments of the present invention will be described in more detail with reference to the accompanying drawings. The same reference numerals are used for the same components in the drawings, and repeated descriptions of the same components are omitted.

First, the problems of the prior art will be described with reference to FIGS. 1 and 2 . 1 is a diagram illustrating a conventional exhaust gas after-treatment system, and FIG. 2 is a graph illustrating a relationship between an exhaust gas flow rate and a pressure drop.

Referring to FIG. 1 , a conventional exhaust gas post-treatment device (hereinafter referred to as 'DPF') includes a Diesel Oxidation Catalyst (DOC) 12 and a Diesel Particulate Filter (13) sequentially disposed inside a can 11 . includes Based on the flow direction of the exhaust gas, the DOC 12 may be installed upstream, and the DPF 13 may be installed downstream. The DOC 12 may oxidize and remove nitrogen monoxide included in the exhaust gas, and the DPF 13 may capture and remove particulate matter (PM) included in the exhaust gas.

Temperature sensors 14a and 14b are respectively installed at the front and rear ends of the DOC 12 to measure the temperature of the exhaust gas. The measured temperature information may be transmitted to an ECU (Electric Control Unit, 18). In addition, a differential pressure sensor 16 for measuring a pressure drop is installed between the front and rear ends of the can 11 or between the front and rear ends of the DPF 13, and the measured pressure drop information can be transmitted to the ECU 18. have.

The ECU 18 uses the received pressure drop information and/or temperature information to determine the mass of particulate matter (PM) collected in the DPF 13, the regeneration timing of the DPF 13, whether the DPF 13 is installed, etc. can judge For example, if the pressure drop amount is equal to or greater than a preset first reference value, it may be determined that the PM is collected in the DPF 13 or greater than the reference value, and the DPF 13 regeneration mode may be forcibly performed. In addition, when the pressure drop is less than or equal to a preset second reference value, it may be determined that the DPF 13 has been removed, and a warning signal may be output through an On-Board Diagnostics (OBD) system or operation of the vehicle may be restricted.

However, it is difficult to perform an accurate diagnosis in all driving areas with such a conventional diagnosis method. This will be described in more detail with reference to FIG. 2 .

As shown in FIG. 2 , when the exhaust gas flow rate increases, the pressure drop ΔP between the front and rear ends of the DPF 13 also increases, and the increase may vary depending on whether the DPF 13 is installed.

For example, in Case 1 in which the DPF 13 is normally installed, as compared to Case 2 in which the DPF 13 is removed, the pressure drop ΔP according to the increase in the exhaust gas flow rate increases significantly. This is because particulate matter (PM) is collected inside the carrier of the DPF (13), and the collected particulate matter (PM) obstructs the flow of exhaust gas. Accordingly, at point B, where the flow rate is relatively large, whether or not the DPF 13 is installed can be determined relatively accurately even using only the pressure drop amount ΔP information. However, since the difference in pressure drop amount (ΔP) between the two cases (Case 1, Case 2) is not large at point A, where the flow rate of exhaust gas is relatively small, it is difficult to make an accurate judgment using only the pressure drop amount (ΔP) information. will be.

The present invention has been devised to solve this problem, and it is possible to accurately determine whether the DPF 13 is installed and the mass of particulate matter (PM) collected in the DPF 13 even when the exhaust gas flow rate is small through machine learning. It is possible to learn a diagnostic model that exists, and to use it to diagnose whether the DPF 13 is installed, and to determine the collected mass of particulate matter (PM). Hereinafter, the present invention will be described in more detail with reference to FIG. 3 .

Referring to FIG. 3 , first, a DPF diagnosis model is trained using machine learning ( S100 ).

In this case, the machine learning may be supervised learning to find an output suitable for a given input using data having an input value and an output value corresponding thereto. For example, the input value may be the temperature and flow rate of exhaust gas passing through the DPF 13, the amount of pressure drop ΔP before and after the DPF 13, and the output value is whether the DPF 13 is installed or not. It may be the mass of particulate matter (PM) collected in the DPF 13 , or the like.

In an embodiment, learning the DPF diagnostic model using machine learning may include learning the DPF diagnostic model using a Gaussian process model.

Specifically, first, the temperature and flow rate of exhaust gas passing through the DPF 13, and the pressure drop ΔP at the front and rear ends of the DPF 13 are measured, and the DPF 13 is installed at this time, and the DPF Measure the mass of particulate matter (PM) collected in (13). For example, the mass of particulate matter (PM) collected in the DPF 13 can be determined by measuring the mass of the DPF 13 before and after flowing the exhaust gas, respectively, and calculating the difference between the measured masses. have. This measurement is repeated a plurality of times.

Among the measured information, data regarding the temperature and flow rate of exhaust gas, and the amount of pressure drop ΔP before and after the DPF 13 may be used as a training data set. In addition, data regarding whether the DPF 13 is installed and the mass of the particulate matter (PM) collected in the DPF 13 may be used as ground truth.

The Gaussian process may estimate the mass and standard deviation of particulate matter (PM) collected in the DPF 13 using the training data set, and verify this by comparing the estimated value with the ground truth. . By repeating this process, it is possible to more accurately estimate the amount of particulate matter (PM) collected.

In one embodiment, the DPF diagnostic model estimates whether the DPF 13 is actually installed using the training data set and compares the estimated value with the ground truth to determine whether the DPF 13 is installed. A diagnostic model to diagnose can be trained.

For example, the DPF diagnostic model can estimate the mass of particulate matter (PM) collected in the DPF 13 under the measured specific conditions (flow rate of exhaust gas, temperature, and pressure drop between the front and rear ends of the DPF). , when the estimated mass of the particulate matter PM is less than or equal to a preset first reference value, it may be determined that the DPF 13 is not installed. In this case, the first reference value may be verified by comparing the estimated result with the ground truth, and may be continuously modified in the process of repeating learning.

In an embodiment, the DPF diagnostic model may determine the need for regeneration of the DPF 13 using the estimated mass of the particulate matter (PM).

For example, if the mass of the particulate matter PM collected in the DPF 13 is equal to or greater than a preset second reference value, it may be determined that the DPF 13 needs to be regenerated. The DPF diagnostic model can estimate the mass of particulate matter (PM) currently collected in the DPF 13, and compare the estimated collection amount with the second reference value to determine the need to regenerate the DPF 13. it can be

In this case, the estimated value of the particulate matter (PM) collection may be continuously modified and supplemented while learning is repeated. Accordingly, the DPF diagnostic model can more accurately estimate the collection amount and more accurately determine the need for regeneration.

By continuously repeating the above learning, it is possible to further improve the accuracy of the diagnosis of the DPF 13 . Meanwhile, the diagnostic model learning process ( S100 ) as described above may be performed at the vehicle manufacturing stage or at the vehicle driving stage.

When the learning of the DPF diagnostic model is completed, the state of the DPF 13 may be diagnosed by using it.

Specifically, it is determined whether a diagnosis for the DPF 13 is necessary (S110). For example, whether or not the DPF 13 is installed or the need to regenerate the DPF 13 may be periodically diagnosed, or alternatively, the diagnosis may be performed manually. That is, when a diagnosis cycle has arrived or a diagnosis execution command is input, it may be determined that the diagnosis of the DPF 13 is necessary.

If it is determined that diagnosis is necessary, the temperature and flow rate of exhaust gas passing through the DPF 13, and the pressure drop ΔP at the front and rear ends of the DPF 13 are measured (S120). The state of the DPF 13 is diagnosed using the measured data S120 and the pre-trained diagnostic model S100 ( S130 ).

For example, it is possible to diagnose whether the DPF 13 is normally installed in the vehicle using the measured temperature, flow rate, and pressure drop ΔP ( S130 ). As a result of the diagnosis, if the DPF 13 is removed (S140), a warning signal may be output through the On-Board Diagnostics (OBD) system or the operation of the vehicle may be restricted (S150).

In addition, it may be determined whether there is a need to regenerate the DPF 13 using the measured temperature, flow rate, and pressure drop ΔP. This is because the collection amount of particulate matter (PM) can be estimated in the diagnostic model learning process ( S100 ), and the collection amount serving as a reference for regeneration can be set through this. In this case, if it is determined that regeneration is necessary, the DPF 13 forcible regeneration can be performed.

As described above, in the diagnostic method of the exhaust gas aftertreatment device (DPF) according to the present invention, a highly reliable diagnostic model can be derived through machine learning, and the state of the DPF 13, for example, the DPF ( 13) It is possible to accurately determine whether to install or whether to regenerate. In particular, in the machine learning process, since various data including a condition of low exhaust gas flow are used, the state of the DPF 13 can be accurately diagnosed regardless of the exhaust gas flow rate, and the time required for diagnosis can be greatly reduced. .

Although the above has been described with reference to the embodiments of the present invention, those skilled in the art can variously modify and change the present invention within the scope without departing from the spirit and scope of the present invention described in the claims below. You will understand that you can.

11: Can 12: DOC
13: DPF 14a, 14b: temperature sensor
16: differential pressure sensor 18: ECU

Claims (10)

In the method of diagnosing the state of an exhaust gas after-treatment device (Diesel Particulate Filter) that collects particulate matter (Particulate Matter) contained in exhaust gas,
learning a diagnostic model for the exhaust gas post-treatment device using machine learning;
measuring the temperature of the exhaust gas, the flow rate of the exhaust gas, and the amount of pressure drop between the front end and the rear end of the exhaust gas aftertreatment device; and
diagnosing a state of the exhaust gas post-processing device by applying the measured temperature, flow rate, and pressure drop data to the learned diagnostic model;
A diagnostic method of an exhaust gas after-treatment device comprising a. According to claim 1,
Learning the diagnostic model comprises:
measuring the temperature of the exhaust gas, the flow rate of the exhaust gas, and the amount of pressure drop between the front end and the rear end of the exhaust gas aftertreatment device;
estimating a collection amount of particulate matter collected in an exhaust gas post-treatment device under the measured conditions using a Gaussian process model; and
estimating whether the exhaust gas post-treatment device is installed by comparing the estimated amount of particulate matter collected with a preset first reference value;
A diagnostic method of an exhaust gas post-treatment device comprising a. The method of claim 2, wherein the learning of the diagnostic model comprises:
The method for diagnosing an exhaust gas post-treatment device, further comprising: measuring a mass of particulate matter collected in the exhaust gas post-treatment device, and verifying an estimate of the particulate matter collection amount. The method of claim 2, wherein the learning of the diagnostic model comprises:
verifying an estimate of whether the exhaust gas post-treatment device is installed by checking whether the exhaust gas post-treatment device is actually installed; and
The method for diagnosing an exhaust gas post-treatment device, further comprising resetting the first reference value when the verification result of the installation or not is incorrect. 3. The method of claim 2,
The diagnosing of the state of the exhaust gas post-treatment device includes diagnosing whether the exhaust gas post-treatment device is actually installed. 6. The method of claim 5,
and outputting a warning signal when it is determined that the exhaust gas post-treatment device is not installed. According to claim 1,
Learning the diagnostic model comprises:
measuring the temperature of the exhaust gas, the flow rate of the exhaust gas, and the amount of pressure drop between the front end and the rear end of the exhaust gas aftertreatment device;
estimating a collection amount of particulate matter collected in an exhaust gas post-treatment device under the measured conditions using a Gaussian process model; and
verifying an estimate of the amount of particulate matter collected by checking an actual amount of particulate matter collected in the exhaust gas post-treatment device;
A diagnostic method of an exhaust gas post-treatment device comprising a. 8. The method of claim 7,
The step of diagnosing the state of the exhaust gas post-treatment device may include determining the need for regeneration of the exhaust gas post-treatment device by comparing the estimated amount of particulate matter collected with a preset second reference value. A diagnostic method of an exhaust gas after-treatment device. In the method of diagnosing the state of an exhaust gas after-treatment device (Diesel Particulate Filter) that collects particulate matter (Particulate Matter) contained in exhaust gas,
learning a diagnostic model for determining whether the exhaust gas post-treatment device is installed through machine learning; and
diagnosing whether the exhaust gas post-processing device is installed using the learned diagnostic model;
Learning the diagnostic model comprises:
measuring the temperature of the exhaust gas, the flow rate of the exhaust gas, and the amount of pressure drop between the front end and the rear end of the exhaust gas aftertreatment device;
estimating a collection amount of particulate matter collected in an exhaust gas post-treatment device under the measured conditions using a Gaussian process model;
estimating whether the exhaust gas post-treatment device is installed by comparing the estimated amount of particulate matter collected with a preset first reference value;
verifying an estimate of whether the exhaust gas post-treatment device is installed by checking whether the exhaust gas post-treatment device is actually installed; and
resetting the first reference value when an estimation of whether the verification result is installed or not is incorrect;
A diagnostic method of an exhaust gas after-treatment device comprising a. 10. The method of claim 9,
The step of determining whether the exhaust gas post-treatment device is installed,
measuring the temperature of the exhaust gas, the flow rate of the exhaust gas, and the amount of pressure drop between the front end and the rear end of the exhaust gas aftertreatment device; and
diagnosing whether the exhaust gas post-processing device is installed by substituting the measured temperature, flow rate and pressure drop data into the learned diagnostic model;
A diagnostic method of an exhaust gas post-treatment device comprising a.

Images