KR102510314B1 - Real-time gas type identification device using gas sensor and method thereof - Google Patents

Abstract

The present invention relates to an apparatus and method for determining the type of gas in real time using a gas sensor, and the effect of determining the type of gas in real time by utilizing transient section and stable section data from detection data of a gas sensor there is

Description

Real-time gas type identification device using gas sensor and method thereof}

The present invention relates to an apparatus and method for determining the type of gas in real time using a gas sensor.

As air pollution becomes more serious today, damage to the human body and ecosystem is increasing day by day. Accordingly, efforts are being made to monitor the atmospheric environment, and in particular, various mobile environment monitoring devices based on the Internet of Things (IoT) are being developed.

Due to the characteristics of mobile devices driven by batteries or solar cells, gas sensors must consume less power, be small in size, have high sensitivity to noise, and be mass-produced at an affordable price.

Compared to various gas detection devices (gas chromatography, electrochemical gas sensor, optical gas sensor), semiconductor gas sensors using metal oxides are attracting attention as gas sensors for mobile devices because they meet the above-mentioned requirements well. .

The semiconductor gas sensor uses the change in electrical resistance of the metal oxide sensing material as it is exposed to the target gas. supplies

At this time, the metal oxide semiconductor gas sensor has a limitation called selectivity difficulty. Referring to FIG. 1, since most of the metal oxide semiconductor gas sensors respond to reactive gases, it is difficult to distinguish which gas the signal is from with only a single sensor signal. can't

In order to solve this problem, the most widely used conventional method is to simultaneously measure (two or more types) multiple types of sensor arrays having slightly different responsiveness for each gas and distinguish the type of gas through machine learning-based signal processing.

However, since the semiconductor gas sensor requires time to fully react with the gas, it is difficult to use sensor detection data in the transient region, making it difficult to implement selective gas detection in real time.

Therefore, it is necessary to develop a technology capable of implementing selective gas detection in real time by applying sensor data of a transient section to machine learning.

Registered Patent No. 10-1852074 (2018.04.25.)

An object of the present invention is to provide a real-time gas type discrimination device and method using a gas sensor.

According to the present invention, it is composed of a gas sensor array having different responsiveness for each gas to be measured, and the data before the gas sensor completely reacts with the gas and the data before the gas sensor completely recovers. A sensor unit for measuring "sensing data" including stable interval data in a fully reacted state and a fully recovered state; a diagnosis unit that learns a machine learning model by using the sensing data as learning data and identifies a type of gas in real time using the machine learning model; and a display unit for guiding the type of gas identified by the diagnosis unit. A real-time gas type determination device using a gas sensor including a may be provided.

In addition, a real-time gas type discrimination device using a gas sensor including reaction speed data, recovery speed data, and gradient data may be provided as the transient section data.

In addition, the diagnosis unit may provide a real-time gas type discrimination device using a gas sensor that applies a time window to the detected data and uses it as the learning data.

In addition, the time window is applied to the transient section data and the stable section data, so that, before the gas sensor completely reacts with the gas during the unit time within the time window according to the movement of the time window, the gas sensor A real-time gas type discrimination device using a gas sensor using a reaction pattern of the gas sensor array for gas before recovery and in a state in which the gas sensor completely reacts with the gas and recovers as the learning data can be provided.

In addition, a real-time gas type discrimination device using a gas sensor that sets the length of the time window in various ways corresponding to the type of gas may be provided.

According to the present invention, (a) in the gas sensor array, the data before the gas sensor completely reacts with the gas and the data before the gas sensor completely recovers = transient period data and the stable period where the gas sensor completely reacts with the gas and recovers Measuring “sensing data” containing data; (b) learning a machine learning model using the sensing data as learning data; and (c) identifying the type of gas in real time using the machine learning model; A real-time gas type determination method using a gas sensor comprising a may be provided.

In addition, the step (b) may include a step of applying a time window to the detected data and using it as the learning data. A real-time gas type determination method using a gas sensor may be provided.

In addition, in the step (b), the time window is applied to the transient section data and the stable section data, so that the gas sensor is completely with the gas during the unit time within the time window according to the movement of the time window. Real-time gas type discrimination method using a gas sensor using the response pattern of the gas sensor array to the gas before reacting, before the gas sensor recovers, and in a state in which the gas sensor completely reacts and recovers with the gas as the learning data this can be provided.

In addition, in the step (a), a real-time gas type determination method using a gas sensor in which the transient section data includes reaction speed data, recovery speed data, and gradient data may be provided.

The apparatus and method for determining the type of gas in real time using a gas sensor according to an embodiment of the present invention have an effect of determining the type of gas in real time by utilizing transient section and stable section data from detection data of a gas sensor.

In addition, the concentration of each gas can be predicted in real time, and even if a plurality of gases are simultaneously measured, the type of each gas can be identified.

1 is a diagram showing reaction characteristics of a metal oxide semiconductor gas sensor to gas.
2 is a diagram showing a dynamic response of a semiconductor gas sensor according to gas injection in a real-time gas type discrimination device using a gas sensor according to an embodiment of the present invention.
3 is a diagram illustrating transient section data obtained by applying a time window to detected data in a real-time gas type determination device using a gas sensor according to an embodiment of the present invention.
FIG. 4 is a diagram showing the reactivity of gas sensors for each gas in a gas sensor array having different sensor detection responses for each gas in the real-time gas type discrimination device using a gas sensor according to an embodiment of the present invention.
5 is a diagram illustrating application of a deep neural network to a machine learning model of a diagnostic unit in a real-time gas type discrimination device using a gas sensor according to an embodiment of the present invention.
6 is a diagram showing the identification accuracy of a gas type in the real-time gas type determination device using a gas sensor according to an embodiment of the present invention.
7 is a diagram in which concentration is predicted for each gas type in the real-time gas type determination device using a gas sensor according to an embodiment of the present invention.
8 is a block diagram of a method for determining a type of gas in real time using a gas sensor according to an embodiment of the present invention.

In order to fully understand the present invention, preferred embodiments of the present invention will be described with reference to the accompanying drawings. Embodiments of the present invention may be modified in various forms, and the scope of the present invention should not be construed as being limited to the examples described in detail below. This embodiment is provided to more completely explain the present invention to those skilled in the art. Therefore, the shapes of elements in the drawings may be exaggerated to emphasize a clearer explanation. It should be noted that in each drawing, the same members are sometimes indicated by the same reference numerals. In addition, detailed descriptions of well-known functions and configurations that may unnecessarily obscure the subject matter of the present invention will be omitted.

Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings.

1 is a diagram showing reaction characteristics of a metal oxide semiconductor gas sensor to gas. 2 is a diagram showing a dynamic response of a semiconductor gas sensor according to gas injection in a real-time gas type discrimination device using a gas sensor according to an embodiment of the present invention. 3 is a diagram illustrating transient section data obtained by applying a time window to detected data in a real-time gas type determination device using a gas sensor according to an embodiment of the present invention. FIG. 4 is a diagram showing the reactivity of gas sensors for each gas in a gas sensor array having different sensor detection responses for each gas in the real-time gas type discrimination device using a gas sensor according to an embodiment of the present invention. 5 is a diagram illustrating application of a deep neural network to a machine learning model of a diagnostic unit in a real-time gas type discrimination device using a gas sensor according to an embodiment of the present invention. 6 is a diagram showing the identification accuracy of a gas type in the real-time gas type determination device using a gas sensor according to an embodiment of the present invention. 7 is a diagram in which concentration is predicted for each gas type in the real-time gas type determination device using a gas sensor according to an embodiment of the present invention. 8 is a block diagram of a method for determining a type of gas in real time using a gas sensor according to an embodiment of the present invention.

A real-time gas type discrimination device using a gas sensor according to an embodiment of the present invention is composed of a gas sensor array having different responsiveness for each gas to be measured, and data before the gas sensor completely reacts with the gas and the gas sensor is completely recovered. A sensor unit 10 that measures "sensed data" including transient section data, which is previous data, and stable section data in which the gas sensor completely reacts with gas and completely recovers; a diagnosis unit (20) that learns a machine learning model by using the sensing data as learning data and identifies a type of gas in real time using the machine learning model; and a display unit 30 for guiding the type of gas identified by the diagnosis unit 20; includes

The gas sensor may use a semiconductor type gas sensor using a metal oxide. Semiconductor gas sensors use the change in electrical resistance of a metal oxide sensing material as it is exposed to a target gas.

The gas sensor array is an array of semiconductor gas sensors with slightly different reactivity for each target gas.

The sensor unit 10 is composed of a gas sensor array and measures sensing data for each gas.

The diagnosis unit 20 identifies the type of gas in real time based on the sensing data.

Gases, as shown in FIG. 1, react with the semiconductor gas sensor over time, and the response of the gas sensor is different for each gas according to the gas concentration. And, when the gas sensor detects gas, it does not immediately generate a signal, but requires time to completely react and recover with the gas, which is called a transient region, a state in which the gas sensor completely reacts with the gas and a fully recovered state is referred to as a steady state.

Referring to FIG. 2, even if the concentration of gas is injected and discharged to maintain a constant concentration, the reactivity of the semiconductor gas sensor forms a gradient, that is, a transient section having a change rate.

The transient section shows different degrees of rebound before and after the stable section based on the steady state in which the gas sensor and gas completely react, and forms a reaction section before the stable section and a recovery section after the stable section. And, the response period and the recovery period change while having different gradients, respectively. In addition, the transient section is different for each gas and gas sensor.

That is, the stable period includes a period in which the gas sensor completely reacts with the gas and a period in which the reaction with the gas is completed and the gas sensor completely recovers. Unlike the transition period, the stable period including these two does not have a rate of change.

The diagnosis unit 20 applies a time window to the sensed data and uses it as the learning data.

The diagnosis unit 20 applies a time window to the detected data, so that the data before the gas sensor completely reacts with the gas, the data before complete recovery, the data after the gas sensor completely reacts with the gas, and the data after the gas sensor completely recovers. The machine learning model is trained using the sensing data including as learning data.

The machine learning model uses sensing data measured by the sensor unit 10 as learning data, and uses the sensing data as input data, but applies a time window to the sensing data and learns using it as input data.

Among the sensing data used as learning data, the transient section data includes information obtained in the transient section generated while the gas sensor reacts with the gas, such as reaction speed, recovery speed, and gradient.

The reaction speed is the reaction speed between the gas sensor and the gas sensor in the reaction period before the stable period when the gas sensor and the gas completely reacted, and the recovery speed is the reaction speed between the gas and the gas sensor in the recovery period after the stable period when the gas completely reacts. It's speed. And the gradient is the change rate for each gradient of the reaction rate and the recovery rate.

Sensing data is different for each type of gas and gas sensor, and by learning a machine learning model based on this, it is possible to identify the type of gas that reacts in real time.

The diagnosis unit 20 applies a time window to the sensing data, and according to the motion of the time window, the reaction pattern of the gas sensor array to the gas for a unit time within the time window is displayed as used as learning data.

The diagnosis unit 20 sets a time window, that is, a constant time interval, and applies it to the sensed data. At this time, the length of the time window can be set in various ways, and the moving rate of the time window can be set to be the same as the sampling rate of the data collection equipment.

A time window is applied to the transient section data and the stable section data, 　 during a unit time within the time window according to the movement of the time window 　 before the gas sensor completely reacts with the gas and before the gas sensor completely recovers , and a response pattern of the gas sensor array to the gas in a state in which the gas sensor completely reacts with and recovers from the gas is used as the learning data.

The time window is applied with a set length and a set moving rate to the transient section. The response pattern of the gas sensor array to the gas for a unit time within the set length of the transient section collected while the time window moves at the set moving rate is used as learning data.

Referring to FIG. 3 as well, a time window having an interval of 10 seconds and a moving rate of 1 second is applied to the transient section of the sensed data. Then, within a time window of 10 seconds, a response pattern of the gas sensor array for gas is collected at unit time intervals of 1 second. That is, in the example of FIG. 3 , while the time window at 10-second intervals moves at a moving rate of 1 second, response data is collected at 1-second intervals within the 10-second time window. As the time window moves from the response section of the transient section to the recovery section, the response pattern of the gas sensor array for gas is collected.

Since the response pattern of the gas sensor array may be different for each type of gas, the length of the time window may be variously set according to the type of gas. In addition, the moving rate of the time window and the unit time within the time window can be set in various ways.

A machine learning model is learned using the collected reaction pattern of the gas sensor array for gas in the transient section as learning data. Machine learning models include pattern recognition and Classification methods are available. Since each method is a well-known learning method, a detailed description thereof will be omitted.

In an embodiment of the real-time gas type discrimination device using the gas sensor of the present invention, the detected data was analyzed using a deep neural network. 4 to 7 show gas classification by analyzing sensing data.

The gas sensor array used a semiconductor gas sensor integrated with a micro electro mechanical systems (MEMS)-based micro heater, and nitrogen dioxide gas, carbon monoxide gas, and ammonia were used as gases to be classified. In order to construct a gas sensor array with different sensor detection responses for each gas, four metal oxides, SnO2, In2O3, WO3, and CuO, were used as sensing materials. The type of ideal gas sensor array and the type of metal oxide are for testing the present embodiment, but are not limited thereto.

Referring to FIG. 4 , it can be seen that the response of the gas sensor collected for each sensing material is different for each gas type as gas is injected. In combination with these, the gas species can be identified by analyzing the pattern of the gas sensor response in the sensing data obtained from the sensor array.

As a result of analyzing the detection response data pattern of the gas sensor array collected by applying a 10-second time window to the transient section in the gas detection data through the deep neural network of FIG. 5, as shown in FIG. 6, the type of gas is identified in real time. (Accuracy: 99.3%) Also, as shown in FIG. 7, the concentration of the gas can also be predicted. (Range of error: 19.5%)

In addition, by using the detection data of the gas sensor array for the mixed gas in which various types of gases are mixed, the transient section data is selected as above, and the response pattern of the gas sensor collected by applying the time window is learned by the machine learning model. In this case, it is possible to identify and classify the gas by type even for the mixed gas.

The display unit 30 guides the type of gas identified by the diagnosis unit 20 to an audio device such as a speaker or a visual device such as a display.

A real-time gas type determination method using a gas sensor according to an embodiment of the present invention includes (a) in a gas sensor array, data before the gas sensor completely reacts with gas and data before the gas sensor completely recovers; Measuring sensing data including stable interval data in which the gas sensor completely reacts with and recovers from the gas (S10); (b) learning a machine learning model using the sensing data as learning data (S20); and (c) identifying the type of gas in real time using the machine learning model (S30); includes

Step (a) measures sensing data from the gas sensor array. The gas sensor array constitutes the sensor unit 10 and measures sensing data from an individual gas or a plurality of mixed gases.

In step (b), a time window is applied to the sensing data, and a response pattern of the gas sensor array to the gas for a unit time within the time window according to the movement of the time window is used as the learning data. .

The diagnostic unit 20 applies a time window to the sensing data received from the sensor unit 10 as a previous step for learning a machine learning model to identify a type of gas.

The detection data includes data before the gas sensor completely reacts with the gas, data before complete recovery, transient section data, and stable section data, which is data after the gas sensor completely reacts with the gas and recovers. At this time, the transient section data includes reaction speed data, recovery speed data, and gradient data.

Then, the machine learning model is trained using the learning data. It is to collect the pattern information that the gas appearing in the learning data reacts to the gas sensor array and learn it using a pattern recognition and classification method such as a deep neural network.

Step (c) identifies the type of gas in real time using a machine learning model.

At this time, not only the type of gas can be identified, but also the concentration of the identified gas can be predicted.

Further, a step of guiding the type or concentration of the identified gas to an audio device such as a speaker or a visual device such as a display may be further included.

The apparatus and method for determining the type of gas in real time using the gas sensor of the present invention can determine the type of gas in real time by utilizing the transient section data and the stable section data in the sensing data of the gas sensor. At this time, by using the transient section, which is the section where the complete reaction between the gas sensor and the gas does not occur, as data, the type of gas can be identified in real time by overcoming the delay caused by the time required for the complete reaction between the gas sensor and the gas. will be.

In addition, even if a plurality of gases are simultaneously measured, the type of each gas can be identified, and the concentration of each gas can be predicted in real time. By learning a mixed gas in which a plurality of gases are mixed using a machine learning model, it is possible to classify and identify a plurality of gases at the same time.

The embodiments of the present invention described above are merely exemplary, and those skilled in the art will appreciate that various modifications and equivalent other embodiments are possible therefrom. Therefore, it will be well understood that the present invention is not limited to the forms mentioned in the detailed description above. Therefore, the true technical protection scope of the present invention should be determined by the technical spirit of the appended claims. It is also to be understood that the present invention includes all modifications, equivalents and alternatives within the spirit and scope of the present invention as defined by the appended claims.

Claims (9)

It is composed of gas sensor arrays with different responsiveness for each gas to be measured, and the data before the gas sensor completely reacts with the gas and the data before the gas sensor completely recovers. a sensor unit that measures sensing data including stable interval data in a fully recovered state;
a diagnostic unit for learning a machine learning model by using the sensing data as learning data, and identifying a type of gas in real time using the machine learning model; and
a display unit for guiding the type of gas identified by the diagnosis unit; including,
The transient section data includes reaction speed data, recovery speed data, and gradient data. Real-time gas type discrimination device using a gas sensor.
delete According to claim 1,
The diagnosis unit,
A real-time gas type discrimination device using a gas sensor that applies a time window to the detected data and uses it as the learning data.
According to claim 3,
The time window is applied to the transient section data and the stable section data so that the gas sensor recovers before the gas sensor completely reacts with the gas for a unit time within the time window according to the movement of the time window. A real-time gas type discrimination device using a gas sensor using a reaction pattern of the gas sensor array to gas before and after the gas sensor completely reacts and recovers with the gas as the learning data.
According to claim 3,
The length of the time window is set according to the type of gas. Real-time gas type discrimination device using a gas sensor.
(a) In the gas sensor array, detection that includes the data before the gas sensor completely reacts with the gas, the data before the gas sensor completely recovers, the transient section data, and the gas sensor completely reacts with the gas and recovers the stable section data. measuring data;
(b) learning a machine learning model using the sensing data as learning data; and
(c) identifying the type of gas in real time using the machine learning model; including,
In the step (a), the transient section data includes reaction speed data, recovery speed data, and gradient data. Real-time gas type discrimination method using a gas sensor.
According to claim 6,
In step (b),
A real-time gas type discrimination method using a gas sensor comprising the step of applying a time window to the detected data and using it as the learning data.
According to claim 7,
In step (b),
The time window is applied to the transient section data and the stable section data so that the gas sensor recovers before the gas sensor completely reacts with the gas for a unit time within the time window according to the movement of the time window. A real-time gas type discrimination method using a gas sensor using, as the learning data, a reaction pattern of the gas sensor array to gas before and after the gas sensor completely reacts with and recovers the gas.
delete

Images