Gas identification and concentration detection method and system based on machine learning

Abstract

The present invention discloses a gas identification and concentration detection method and system based on machine learning, which relates to the technical field of gas concentration detection and includes the following steps: obtaining box-related data based on a transparent incubator, and using an open-circuit laser gas detector to simulate light intensity loss to obtain light intensity loss simulation data; based on a multiple sensor model, and using the light intensity loss simulation data to construct a light intensity loss impact model; based on the open-circuit laser gas detector, constructing a gas detection optical path, and dynamically collecting the laser light intensity passing through the transparent incubator to obtain relevant gas detection data; and performing comprehensive gas concentration detection and analysis, and performing gas detection management on the inside of multiple transparent incubators; the present invention is used to solve the problem that the existing gas concentration detection technology cannot, when detecting the gas concentration inside the transparent incubator, detect the gas concentration inside the transparent incubator in real time and accurately without basically affecting the internal environment of the incubator.

Description

Technical Field

[0001] The present invention relates to the technical field of gas concentration detection, and specifically to a gas identification and concentration detection method and system based on machine learning. Background Art

[0002] Gas concentration detection technology refers to the technology of qualitative analysis or quantitative measurement of the concentration of one or more gases in a specific environment through various physical, chemical or biological methods and means. Gas concentration detection technology is widely used in many fields such as industrial production, environmental monitoring, medical care, safety protection, etc., and is of great significance to ensuring production safety, protecting environmental quality and maintaining human health.

[0003] Existing gas concentration detection technologies often use electrochemical sensors or semiconductor gas sensors when detecting gas concentrations inside transparent incubators. In order to ensure the accuracy of detection, electrochemical sensors are often arranged at multiple locations in the incubator. For small incubators, the heat generated may accumulate, causing the temperature inside the incubator to rise slightly, affecting the stability of the culture environment. Semiconductor gas sensors usually need to be heated to a certain temperature to work properly, which will release more heat into the incubator. This will cause large temperature fluctuations in the incubator, affecting the development of the culture inside. Moreover, electrochemical sensors or semiconductor gas sensors will As time goes by, its performance will gradually drift and requires regular calibration and maintenance. For incubators that have been sealed for a long time, calibration and maintenance cannot be performed. For example, the patent application with publication number CN111521647A discloses a gas concentration detection method, system, computer equipment and storage medium. When detecting gas concentration, this solution requires heating the gas-sensitive material and cannot be applied to incubators that require high temperature control accuracy. Therefore, when detecting the gas concentration inside a transparent incubator, the existing gas concentration detection technology cannot detect the gas concentration inside the transparent incubator in real time and accurately without basically affecting the internal environment of the incubator. Summary of the Invention

[0004] The present invention aims to solve one of the technical problems in the prior art to at least a certain extent, by obtaining box-related data based on a transparent incubator and using an open-circuit laser gas detector to simulate light intensity loss to obtain light intensity loss simulation data; using the light intensity loss simulation data to construct a light intensity loss impact model; constructing a gas detection optical path based on the open-circuit laser gas detector, and dynamically collecting the laser light intensity passing through the transparent incubator to obtain relevant gas detection data; and performing comprehensive gas concentration detection and analysis, and performing gas detection management on the inside of multiple transparent incubators, so as to solve the problem that the existing gas concentration detection technology cannot detect the gas concentration inside the transparent incubator in real time and accurately without basically affecting the internal environment of the incubator when detecting the gas concentration inside the transparent incubator.

[0005] To achieve the above objectives, in a first aspect, the present application provides a gas identification and concentration detection method based on machine learning, comprising the following steps:

[0006] Based on the transparent incubator, relevant data of the box is obtained, and the light intensity loss simulation data is obtained by using an open-circuit laser gas detector;

[0007] Based on the multi-sensor model, the light intensity loss impact model is constructed using light intensity loss simulation data;

[0008] A gas detection optical path is constructed based on an open-circuit laser gas detector, and the laser light intensity passing through the transparent incubator is dynamically collected to obtain relevant gas detection data.

[0009] Comprehensive gas concentration detection and analysis is performed based on relevant gas detection data, and gas detection management is carried out inside multiple transparent incubators.

[0010] Furthermore, based on the transparent incubator, relevant data of the incubator is obtained, and light intensity loss simulation is performed using an open-circuit laser gas detector to obtain light intensity loss simulation data, which includes the following sub-steps:

[0011] Place the transparent incubator between the transmitting end and the receiving end of the open-circuit laser gas detector, so that the laser emitted by the transmitting end is perpendicular to the side of the transparent incubator. The two sides of the transparent incubator that the laser passes through are recorded as the incident side and the exit side in the order in which the laser passes.

[0012] The thicknesses of the incident side surface and the exit side surface are obtained respectively, and are recorded as HR and HC in sequence; and the refractive indices of the incident side surface and the exit side surface are obtained respectively, and are recorded as ZR and ZC in sequence.

[0013] Furthermore, obtaining box-related data based on the transparent incubator and performing light intensity loss simulation using an open-circuit laser gas detector to obtain light intensity loss simulation data also includes the following sub-steps:

[0014] The gas to be detected is recorded as the gas to be detected; the absorption wavelength of the gas to be detected is obtained and recorded as the gas absorption wavelength; and the laser wavelength emitted by the open-path laser gas detector is set to be the gas absorption wavelength;

[0015] The laser light intensity emitted by the emitting end of the open-circuit laser gas detector is recorded as the emitted light intensity, and the light intensity received by the receiving end of the open-circuit laser gas detector is recorded as the received light intensity;

[0016] The incident side and the exit side of the transparent incubator are placed separately between the transmitting end and the receiving end of the open-circuit laser gas detector, and the initial emission light intensity is set to A0. The emission light intensity is continuously reduced in sequence with the first light intensity interval, and each time a new emission light intensity is set, the received light intensity and the ambient temperature are recorded at the same time. Different emission intensities, ambient temperatures and corresponding received light intensities are obtained and marked as the corresponding incident side. The simulation data of the exit side is obtained and combined with the corresponding thickness and refractive index and stored as the corresponding light intensity loss simulation data, where the first light intensity interval is a1;

[0017] Repeatedly obtain light intensity loss simulation data corresponding to the incident side and the exit side of multiple transparent incubators.

[0018] Furthermore, based on the multi-sensor model and using the light intensity loss simulation data, a light intensity loss impact model is constructed, which includes the following sub-steps:

[0019] Normalize all light intensity loss simulation data according to data type, scale all data sizes to [0, 1], and obtain normalized loss simulation data after completion;

[0020] Any set of corresponding emission light intensity, thickness, and refractive index in the normalized loss simulation data is combined into a loss feature vector, denoted as M0={m1, m2, m3, m4}, where m1, m2, m3, and m4 represent the emission light intensity, thickness, refractive index, and ambient temperature, respectively; and M0 is combined and stored with the corresponding received light intensity; all normalized loss simulation data are processed repeatedly to obtain loss simulation training data.

[0021] Furthermore, based on the multi-sensor model and using the light intensity loss simulation data to construct the light intensity loss impact model also includes the following sub-steps:

[0022] An original loss influence model is constructed based on the multi-perceptron model. The original loss influence model includes: input layer, hidden layer and output layer. The number of neurons in the input layer is set to b1, the number of hidden layers is set to b2, the number of neurons in each hidden layer is set to b3, and the number of neurons in the output layer is set to b4. The original loss influence model is trained using loss simulation training data, and the light intensity loss influence model is obtained after completion.

[0023] Furthermore, a gas detection optical path is constructed based on an open-circuit laser gas detector, and the laser light intensity passing through the transparent incubator is dynamically collected to obtain relevant gas detection data; the method includes the following sub-steps:

[0024] Arrange multiple transparent incubators in sequence and place them neatly between the transmitting end and the receiving end of the open-circuit laser gas detector, so that the laser emitted by the transmitting end is perpendicular to the sides of all transparent incubators;

[0025] Place a semi-transparent and semi-reflective mirror between any two transparent incubators to split the laser light path between the two transparent incubators into two mutually perpendicular split light paths. The split light path perpendicular to the original laser light path is recorded as the reference light path, and the split light path along the original laser light path is recorded as the incident light path. Set the splitting ratio of the incident light path to the reference light path to be c1:c2.

[0026] For any transparent incubator, designated as the first incubator, the light intensity of the reference optical path corresponding to the incident side and the exit side of the first incubator is acquired at sampling intervals, while the ambient temperature is also acquired, and the acquisition time is recorded. The laser light intensity entering the first incubator at the incident side and the laser light intensity exiting from the exit side of the first incubator are calculated based on the corresponding splitting ratios, and are denoted as RI and CI, respectively. The sampling interval is T0.

[0027] The ambient temperature, RI, CI, and the corresponding HR, HC, ZR, and ZC are normalized and formed into corresponding loss feature vectors, which are input into the light intensity loss influence model. The laser light intensity emitted from the incident side of the first incubator and the laser light intensity on the exit side entering the first incubator are obtained, and are recorded as the gas incident light intensity QR and the gas exit light intensity QC, respectively.

[0028] Furthermore, constructing a gas detection optical path based on an open-circuit laser gas detector and dynamically collecting the laser light intensity passing through the transparent incubator to obtain relevant gas detection data also includes the following sub-steps:

[0029] Calculate the gas light intensity difference QK, QK = QR - QC; obtain the gas light intensity difference QK corresponding to n acquisitions within the most recent first time length, and sort them from recent to far according to the acquisition time as the first difference sequence; the first time length is W1;

[0030] The unit change rate of the difference between any two adjacent gas intensity differences in the first difference sequence is calculated according to the difference change formula to obtain the first change rate sequence. The difference change formula is as follows: , where BK represents the unit change rate of adjacent gas light intensity differences, QK(i) and QK(i+1) represent the i-th and i+1-th gas light intensity differences, respectively; t(i) and t(i+1) represent the acquisition time corresponding to the i-th and i+1-th gas light intensity differences, respectively;

[0031] All the change rates in the first change rate sequence are recorded in order as BK(1)-BK(n-1); the average change rate PK0 is calculated according to the average formula, which is as follows: ;

[0032] The sampling time interval is adjusted to T1 according to the first interval formula; the first interval formula is as follows: , where Tm is the maximum sampling time interval set; PKm is the average change rate threshold set; λ is the set proportional coefficient;

[0033] Repeat the dynamic adjustment of the acquisition time intervals of all transparent incubators, and obtain the gas incident light intensity QR and the gas exit light intensity QC of all transparent incubators, which are recorded as relevant gas detection data.

[0034] Furthermore, comprehensive gas concentration detection and analysis based on relevant gas detection data is performed, and gas detection management inside multiple transparent incubators includes the following sub-steps:

[0035] Obtain the optical path of the laser through the interior of the first incubator. Calculate the concentration of the gas to be measured inside the first incubator according to the Beer-Lambert law for relevant gas detection data corresponding to any collection time of the first incubator, and record it as the gas concentration to be measured. Repeat the calculation to obtain the gas concentration to be measured corresponding to all collection times of the first incubator. The gas concentrations are sorted from the most recent collection time to the most recent collection time, and recorded as the first gas concentration sequence.

[0036] Set the upper and lower concentration thresholds. If the concentration of the gas to be measured at any sampling moment is greater than the upper concentration threshold or less than the lower concentration threshold, an alarm signal will be issued.

[0037] Repeat the test management for the gas concentration to be tested in all transparent incubators.

[0038] Furthermore, comprehensive gas concentration detection and analysis based on relevant gas detection data and gas detection management inside multiple transparent incubators also include the following sub-steps:

[0039] Obtain the first gas concentration sequence of the second time length in the past, recorded as the second gas concentration sequence, and calculate the difference between any two adjacent measured gas concentrations in the second gas concentration sequence according to the difference formula to obtain the first concentration difference sequence. The difference formula is as follows: , where VC(j) represents the difference between two adjacent gas concentrations to be measured, VK(j) and VK(j+1) represent the j-th and j+1-th gas concentrations to be measured, respectively; wherein the second time length is W2;

[0040] Arrange the first concentration difference sequence in ascending order, remove the smallest x1 differences and the largest x1 differences, and obtain the second concentration difference sequence. Calculate the mean and standard deviation of the second concentration difference sequence, and record them as Y1 and Y2 respectively;

[0041] Y1+k1*Y2 is recorded as the dynamic change threshold, and a static change threshold is set. For the measured gas concentration corresponding to any acquisition, the difference between the measured gas concentration corresponding to the previous acquisition is calculated and recorded as the gas concentration change value. If the gas concentration change value is greater than the dynamic change threshold or the static change threshold, an alarm signal is issued;

[0042] Repeat the detection and management of changes in the concentration of the gas to be tested in all transparent incubators.

[0043] In a second aspect, the present application provides a gas identification and concentration detection system based on machine learning, including a data simulation module, a loss model module, a light intensity acquisition module, and a detection management module;

[0044] The data simulation module includes an acquisition unit and a simulation unit. The acquisition unit obtains box-related data based on the transparent incubator. The simulation unit uses an open-circuit laser gas detector to simulate light intensity loss to obtain light intensity loss simulation data.

[0045] The loss model module is based on a multi-sensor model and uses light intensity loss simulation data to construct a light intensity loss impact model;

[0046] The light intensity acquisition module constructs a gas detection optical path based on an open-circuit laser gas detector and dynamically collects the laser light intensity passing through the transparent incubator to obtain relevant gas detection data;

[0047] The detection management module performs comprehensive gas concentration detection and analysis based on relevant gas detection data, and performs gas detection management inside multiple transparent incubators.

[0048] Beneficial effects of the present invention include: obtaining box-related data based on a transparent incubator, and simulating light intensity loss using an open-circuit laser gas detector to obtain light intensity loss simulation data; constructing a light intensity loss impact model based on a multi-sensor model and using the light intensity loss simulation data; constructing a gas detection optical path based on the open-circuit laser gas detector, and dynamically collecting the laser light intensity passing through the transparent incubator to obtain relevant gas detection data; performing comprehensive gas concentration detection and analysis based on the relevant gas detection data, and performing gas detection management inside multiple transparent incubators; when detecting the gas concentration inside the transparent incubator, the gas concentration inside the transparent incubator can be detected in real time and accurately without substantially affecting the internal environment of the incubator;

[0049] The present invention collects the laser light intensity passing through the transparent incubator and dynamically adjusts the sampling time interval according to the light intensity change rate. The advantage is that it can capture changes in gas concentration more timely and accurately, improve the timeliness and accuracy of detection, and optimize resource utilization; by placing semi-transparent and semi-reflective mirrors between the transparent incubators and calculating the light intensity according to the splitting ratio, this design helps to obtain light intensity data more accurately, eliminate external interference, and improve detection accuracy; when detecting and managing the concentration of the gas to be tested inside the transparent incubator, the change of gas concentration is also detected and managed, and abnormal conditions can be discovered in time and alarm signals can be issued, thereby realizing comprehensive monitoring and management of the gases inside multiple transparent incubators. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 is a functional block diagram of the system of the present invention;

[0051] Figure 2 is a flow chart of the steps of the method of the present invention;

[0052] Figure 3 Schematic diagram of the structure of the gas detection optical path of the present invention;

[0053] Figure 4 Schematic diagram of the structure of the electronic device of the present invention. DETAILED DESCRIPTION

[0054] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0055] Example 1, please refer to Figure 1As shown, the present application provides a gas identification and concentration detection system based on machine learning, including a data simulation module, a loss model module, a light intensity acquisition module, and a detection management module;

[0056] The data simulation module includes an acquisition unit and a simulation unit. The acquisition unit obtains box-related data based on the transparent incubator, and the simulation unit uses an open-circuit laser gas detector to simulate light intensity loss and obtain light intensity loss simulation data.

[0057] The acquisition unit is configured with an acquisition strategy, which includes: placing a transparent incubator between the transmitting end and the receiving end of the open-circuit laser gas detector, so that the laser emitted by the transmitting end is perpendicular to the side of the transparent incubator, and recording the two sides of the transparent incubator through which the laser passes as the incident side and the exit side in the order in which the laser passes;

[0058] The thickness of the incident side and the exit side are obtained respectively, and recorded as HR and HC in order; and the refractive index of the incident side and the exit side are obtained respectively, and recorded as ZR and ZC in order; when the laser is incident from air to other transparent media, reflection occurs at the interface; according to the Fresnel formula, the greater the refractive index, the greater the proportion of the reflected light intensity to the incident light intensity, resulting in a relative weakening of the transmitted light intensity; other transparent media have a certain absorption of laser light, and the absorption coefficient is usually a constant value; the longer the distance the laser propagates in the transparent medium, the more light energy is absorbed; that is, the greater the thickness, the smaller the exit light intensity;

[0059] The simulation unit is equipped with a simulation strategy, which includes: recording the gas to be detected as the gas to be detected; obtaining the absorption wavelength of the gas to be detected and recording it as the gas absorption wavelength; setting the wavelength of the laser emitted by the open-path laser gas detector to the gas absorption wavelength; each gas molecule has a unique electronic energy level and vibrational energy level structure; when the laser is irradiated to the gas, only laser light of a specific wavelength can be absorbed by the gas molecules, which is determined by the characteristics of the gas molecules; different gas molecules absorb laser light of different wavelengths, forming their own characteristic absorption spectra, which can be used to identify the gas;

[0060] The laser light intensity emitted by the emitting end of the open-circuit laser gas detector is recorded as the emitted light intensity, and the light intensity received by the receiving end of the open-circuit laser gas detector is recorded as the received light intensity;

[0061] The incident side and the exit side of the transparent incubator are placed separately between the transmitting end and the receiving end of the open-circuit laser gas detector, and the initial emission light intensity is set to A0. The emission light intensity is continuously reduced in sequence with the first light intensity interval, and each time a new emission light intensity is set, the received light intensity and the ambient temperature are recorded at the same time to obtain emission light intensities, ambient temperatures and corresponding received light intensities of different sizes, marked as the corresponding incident side. The simulation data of the exit side is obtained and combined with the corresponding thickness and refractive index for storage, and recorded as the corresponding light intensity loss simulation data, where the first light intensity interval is a1; that is, the exit light intensity under different incident light intensities is obtained, and A0 and a1 can be set according to the actual application scenario;

[0062] Repeatedly obtain light intensity loss simulation data corresponding to the incident side and the exit side of multiple transparent incubators;

[0063] In the specific implementation process, the open-path laser gas detector is usually composed of a central processing unit, a transmitter, a receiver and a transmission optical cable; the central processing unit contains a laser light source, and the generated laser is transmitted to the transmitter by the transmission optical cable. The transmitter emits the laser, and the laser passes through the gas to be measured and is detected by the receiver. The detection signal is then transmitted back to the central processing unit for processing, analysis and display.

[0064] The loss model module is based on the multi-perceptron model and uses light intensity loss simulation data to build a light intensity loss impact model;

[0065] The loss model module configuration consists of a model building strategy. The model building strategy includes: normalizing all light intensity loss simulation data according to data type, scaling all data sizes to [0, 1], and obtaining normalized loss simulation data. Data types include light intensity, temperature, refractive index, and thickness.

[0066] Combine any corresponding set of emission intensity, thickness, and refractive index in the normalized loss simulation data into a loss feature vector, denoted as M0={m1, m2, m3, m4}, where m1, m2, m3, and m4 represent the emission intensity, thickness, refractive index, and ambient temperature, respectively; and merge and store M0 with the corresponding received light intensity; repeatedly process all normalized loss simulation data to obtain loss simulation training data;

[0067] An original loss impact model is constructed based on a multi-perceptron model. The original loss impact model includes an input layer, a hidden layer, and an output layer. The number of neurons in the input layer is set to b1, the number of hidden layers is set to b2, the number of neurons in each hidden layer is set to b3, and the number of neurons in the output layer is set to b4. The original loss impact model is trained using loss simulation training data to obtain a light intensity loss impact model. In this embodiment, b1=4, i.e., inputs m1, m2, m3, and m4; b2=2, i.e., two hidden layers; b3=64; and b4=1, used to output the received light intensity.

[0068] During the specific implementation process, a light intensity loss impact model is constructed, which can obtain accurate laser light intensity when collecting light intensity in the later stage, providing a basis for accurate detection of gas concentration; the ambient temperature is comprehensively considered because changes in ambient temperature will cause thermal expansion or contraction of the side medium of the incubator, which may change its thickness and surface curvature, thereby affecting the optical path length, optical path focusing and divergence characteristics, and indirectly changing the distribution of transmitted light intensity; comprehensive consideration of ambient temperature can more comprehensively consider the impact of various factors on gas concentration detection, reduce errors, and improve detection reliability.

[0069] The light intensity acquisition module constructs a gas detection optical path based on an open-circuit laser gas detector and dynamically collects the laser light intensity passing through the transparent incubator to obtain relevant gas detection data;

[0070] The light intensity acquisition module is equipped with a light intensity acquisition strategy, which includes: Figure 3 As shown, multiple transparent culture boxes are arranged in sequence and neatly placed between the transmitting end and the receiving end of the open-circuit laser gas detector, so that the laser emitted by the transmitting end is perpendicular to the sides of all transparent culture boxes;

[0071] A semi-transparent, semi-reflective mirror is placed between any two transparent incubators to split the laser light path between the two transparent incubators into two mutually perpendicular split light paths. The split light path perpendicular to the original laser light path is recorded as a reference light path, and the split light path along the original laser light path is recorded as an incident light path. The splitting ratio between the incident light path and the reference light path is set to c1:c2. The splitting ratio of the semi-transparent, semi-reflective mirror is mainly determined by its coating material and coating thickness. Different coating materials have different reflection and transmission characteristics for light of different wavelengths. By precisely controlling the coating thickness, a specific splitting ratio can be achieved. In this embodiment, c1:c2 is 9:1. In addition to the semi-transparent, semi-reflective mirror, other optical elements can also be used for splitting, such as a non-polarizing cubic beam splitter and a fiber beam splitter.

[0072] For any transparent incubator, denoted as the first incubator, the light intensity of the reference optical path corresponding to the incident side and the exit side of the first incubator is obtained at a sampling time interval, the ambient temperature is collected at the same time, and the collection time is recorded; the laser light intensity entering the incident side of the first incubator and the laser light intensity emitted from the exit side of the first incubator are calculated based on the corresponding splitting ratio, and are denoted as RI and CI in sequence; wherein the sampling time interval is T0; in this embodiment, the sampling time interval T0 is 1 second; for the incident side of the first transparent incubator in the laser light path, there is no need to place a semi-transparent semi-reflective mirror; the laser intensity emitted by the transmitting end is directly obtained as the corresponding RI; and for the exit side of the last transparent incubator in the laser light path, there is no need to place a semi-transparent semi-reflective mirror; the laser intensity emitted from the exit side is directly obtained as the corresponding CI;

[0073] The ambient temperature, RI, CI, and the corresponding HR, HC, ZR, and ZC are normalized and formed into corresponding loss characteristic vectors, which are input into the light intensity loss influence model; the laser light intensity emitted from the incident side of the first incubator and the laser light intensity on the exit side entering the first incubator are obtained, which are recorded in order as the gas incident light intensity QR and the gas exit light intensity QC; RI, HR, ZR, and the ambient temperature can be directly combined into the corresponding loss characteristic vectors, which are then input into the light intensity loss influence model to obtain the laser light intensity emitted from the incident side of the first incubator; and the laser light intensity on the exit side entering the first incubator needs to be reversely deduced using the light intensity loss influence model, that is, knowing the model output result CI and the three inputs of the model, namely HC, ZC, and the ambient temperature, the remaining input, namely the laser light intensity on the exit side entering the first incubator, is reversely deduced;

[0074] Calculate the gas light intensity difference QK, where QK = QR - QC; obtain the gas light intensity difference QK corresponding to n acquisitions within the most recent first time length, and sort them from recent to far according to the acquisition time as a first difference sequence; the first time length is W1; in this embodiment, the first time length W1 is 5 minutes;

[0075] The unit change rate of the difference between any two adjacent gas intensity differences in the first difference sequence is calculated according to the difference change formula to obtain the first change rate sequence. The difference change formula is as follows: , where BK represents the unit change rate of adjacent gas light intensity differences, QK(i) and QK(i+1) represent the i-th and i+1-th gas light intensity differences, respectively; t(i) and t(i+1) represent the acquisition time corresponding to the i-th and i+1-th gas light intensity differences, respectively;

[0076] All the change rates in the first change rate sequence are recorded in order as BK(1)-BK(n-1); the average change rate PK0 is calculated according to the average formula, which is as follows: That is, weighted averaging is performed. The closer the data is to the time of collection, the greater the weight. The more accurately the data reflects the current trend of gas concentration changes. Giving more weight to recent data can make the results closer to the actual changes in gas concentration in the incubator, and can capture the latest changes in concentration in a timely manner, which helps to more accurately monitor and analyze the real-time changes in the gas environment in the incubator.

[0077] The sampling time interval is adjusted to T1 according to the first interval formula; the first interval formula is as follows: , where Tm is the maximum sampling time interval set; PKm is the average change rate threshold set; λ is the set proportional coefficient; in this embodiment, λ is 2;

[0078] Repeatedly and dynamically adjust the acquisition time intervals of all transparent incubators, and obtain the gas incident light intensity QR and gas exit light intensity QC of all transparent incubators, and record them as relevant gas detection data;

[0079] During the specific implementation process, the difference in gas light intensity corresponds to the gas concentration, and the sampling time interval is dynamically adjusted according to the change in light intensity. This adaptive sampling method can capture changes in gas concentration more promptly and accurately, improving the timeliness and accuracy of detection. For example, by adjusting the sampling time interval according to the average change rate, the sampling frequency can be increased when the gas concentration changes dramatically, and the frequency can be appropriately reduced when the change is stable, thereby optimizing resource utilization.

[0080] The detection management module performs comprehensive gas concentration detection and analysis based on relevant gas detection data, and manages gas detection inside multiple transparent incubators;

[0081] The detection management module is configured with a detection management strategy, which includes: obtaining the optical path of the laser through the interior of the first incubator, i.e., the distance from the inner surface of the incident side surface to the inner surface of the exit side surface of the first incubator; calculating the concentration of the gas to be measured inside the first incubator according to the Beer-Lambert law for relevant gas detection data corresponding to any collection time of the first incubator, and recording it as the measured gas concentration; repeating the calculation to obtain the measured gas concentration corresponding to all collection times of the first incubator; and sorting the gas concentrations from the most recent collection time to the most recent collection time, and recording them as the first gas concentration sequence;

[0082] The Beer-Lambert law describes the relationship between the attenuation of light intensity when light passes through an absorbing medium and the concentration and optical path length of the absorbing medium. Its mathematical expression is: Where I0 is the incident light intensity; I is the outgoing light intensity; ε is the absorption coefficient, which is related to factors such as the properties of the absorbing substance and the wavelength of the incident light, and is a characteristic constant of the substance; b is the optical path length, that is, the distance light travels in the absorbing medium; c is the concentration of the absorbing substance, that is, the concentration of the gas to be measured;

[0083] Set the upper and lower concentration thresholds. If the concentration of the gas to be measured at any sampling moment is greater than the upper concentration threshold or less than the lower concentration threshold, an alarm signal will be issued.

[0084] Repeat the test management of the gas concentration to be tested in all transparent incubators;

[0085] Obtain the first gas concentration sequence of the second time length in the past, recorded as the second gas concentration sequence, and calculate the difference between any two adjacent measured gas concentrations in the second gas concentration sequence according to the difference formula to obtain the first concentration difference sequence. The difference formula is as follows: , where VC(j) represents the difference between two adjacent gas concentrations to be measured, VK(j) and VK(j+1) represent the j-th and j+1-th gas concentrations to be measured, respectively; wherein the second time length is W2; in this embodiment, the second time length W2 is 3 days;

[0086] Arrange the first concentration difference sequence in ascending order, and remove the smallest x1 differences and the largest x1 differences to obtain a second concentration difference sequence. Calculate the mean and standard deviation of the second concentration difference sequence, which are denoted as Y1 and Y2, respectively. In this embodiment, x1 is 5. The smallest and largest differences may contain outliers, which may be caused by measurement errors, equipment failures, or other accidental factors. Removing these outliers can prevent these outliers from having a significant impact on the calculation of the mean and standard deviation, so that the calculation results can better reflect the central tendency and dispersion of the data, thereby improving the stability and reliability of the data.

[0087] Y1+k1*Y2 is recorded as the dynamic change threshold, and a static change threshold is set. For the measured gas concentration corresponding to any acquisition, the difference between the measured gas concentration and the measured gas concentration corresponding to the previous acquisition is calculated and recorded as the gas concentration change value. If the gas concentration change value is greater than the dynamic change threshold or the static change threshold, an alarm signal is issued; the static change threshold is a fixed value that does not change with the dynamic change of data, and is used to determine whether the gas concentration change exceeds the normal range; it can be a constant pre-set according to the general law of gas concentration change in the incubator, experimental requirements, equipment performance and other factors. Under normal circumstances, the static change threshold is greater than the dynamic change threshold;

[0088] Repeat the detection and management of changes in the concentration of the gas to be tested in all transparent incubators;

[0089] During the specific implementation process, while detecting the instantaneous value of the concentration of the gas to be tested inside the transparent incubator, the change in gas concentration is also detected and managed; the trend and law of concentration change can be grasped more accurately; this makes it possible to detect subtle changes in gas concentration in a timely manner, to detect abnormal situations in a timely manner and to issue an alarm signal, rather than just alarming when the concentration reaches a certain fixed threshold, thereby realizing the whole process management of gas concentration changes.

[0090] Example 2, please refer to Figure 2 As shown, this application provides a gas identification and concentration detection method based on machine learning, comprising the following steps:

[0091] Step S1, based on the transparent incubator, obtains box-related data and uses an open-circuit laser gas detector to simulate light intensity loss to obtain light intensity loss simulation data; Step S1 includes the following sub-steps:

[0092] Step S101: Place a transparent incubator between the transmitting end and the receiving end of the open-circuit laser gas detector, so that the laser emitted by the transmitting end is perpendicular to the side of the transparent incubator. The two sides of the transparent incubator that the laser passes through are recorded as the incident side and the exit side in the order in which the laser passes.

[0093] Step S102, respectively obtaining the thickness of the incident side surface and the exit side surface, which are sequentially recorded as HR and HC; and respectively obtaining the refractive index of the incident side surface and the exit side surface, which are sequentially recorded as ZR and ZC;

[0094] Step S103, record the gas to be detected as the gas to be detected; obtain the absorption wavelength of the gas to be detected, record it as the gas absorption wavelength; set the laser wavelength emitted by the open-path laser gas detector to be the gas absorption wavelength;

[0095] Step S104, recording the laser light intensity emitted by the transmitting end of the open-circuit laser gas detector as the transmitted light intensity, and recording the light intensity received by the receiving end of the open-circuit laser gas detector as the received light intensity;

[0096] Step S105: Place the incident side and the exit side of the transparent incubator separately between the transmitting end and the receiving end of the open-circuit laser gas detector, set the initial emission light intensity to A0, and continuously reduce the emission light intensity in sequence at the first light intensity interval. Each time a new emission light intensity is set, record the received light intensity and the ambient temperature, obtain emission light intensities of different sizes, ambient temperatures, and corresponding received light intensities, mark them as the corresponding incident side, obtain simulation data of the exit side, and combine and store them with the corresponding thickness and refractive index as the corresponding light intensity loss simulation data, where the first light intensity interval is a1;

[0097] Step S106 , repeatedly obtaining light intensity loss simulation data corresponding to the incident side and the exit side of multiple transparent incubators.

[0098] Step S2, based on the multi-sensor model and using the light intensity loss simulation data, constructs a light intensity loss impact model; step S2 includes the following sub-steps:

[0099] Step S201: normalize all light intensity loss simulation data according to data type, and scale the size of all data to [0, 1]. After completion, normalized loss simulation data is obtained.

[0100] Step S202: Combine any corresponding set of emission light intensity, thickness, and refractive index in the normalized loss simulation data into a loss feature vector, denoted as M0 = {m1, m2, m3, m4}, where m1, m2, m3, and m4 represent the emission light intensity, thickness, refractive index, and ambient temperature, respectively; and combine M0 with the corresponding received light intensity and store it; repeatedly process all normalized loss simulation data to obtain loss simulation training data;

[0101] Step S203: constructing an original loss impact model based on the multi-perceptron model. The original loss impact model includes: an input layer, a hidden layer, and an output layer. The number of neurons in the input layer is set to b1, the number of hidden layers is set to b2, the number of neurons in each hidden layer is set to b3, and the number of neurons in the output layer is set to b4.

[0102] Step S204 , using the loss simulation training data to perform model training on the original loss impact model, and upon completion, obtaining the light intensity loss impact model.

[0103] Step S3, constructing a gas detection optical path based on the open-circuit laser gas detector, and dynamically collecting the laser light intensity passing through the transparent incubator to obtain relevant gas detection data; Step S3 includes the following sub-steps:

[0104] Step S301, arranging multiple transparent incubators in sequence and neatly placing them between the transmitting end and the receiving end of the open-circuit laser gas detector so that the laser emitted by the transmitting end is perpendicular to the sides of all the transparent incubators;

[0105] Step S302: Place a semi-transparent, semi-reflective mirror between any two transparent incubators to split the laser light path between the two transparent incubators into two mutually perpendicular split light paths. The split light path perpendicular to the original laser light path is recorded as the reference light path, and the split light path along the original laser light path is recorded as the incident light path. The splitting ratio between the incident light path and the reference light path is set to c1:c2.

[0106] Step S303: For any transparent incubator, designated as the first incubator, the light intensity of the reference optical path corresponding to the incident side and the exit side of the first incubator is acquired at sampling intervals, the ambient temperature is simultaneously acquired, and the acquisition time is recorded. The laser light intensity entering the first incubator at the incident side and the laser light intensity exiting from the exit side of the first incubator are calculated based on the corresponding splitting ratios, and are designated as RI and CI, respectively. The sampling interval is T0.

[0107] Step S304: Normalize the ambient temperature, RI, CI, and the corresponding HR, HC, ZR, and ZC to form a corresponding loss feature vector, which is input into the light intensity loss impact model. The laser light intensity emitted from the incident side of the first incubator and the laser light intensity entering the exit side of the first incubator are obtained, which are recorded as the gas incident light intensity QR and the gas exit light intensity QC, respectively.

[0108] Step S305: Calculate the gas light intensity difference QK, where QK = QR - QC; obtain the gas light intensity difference QK corresponding to n acquisitions within the most recent first time length, and sort them from recent to far according to the acquisition time as a first difference sequence; the first time length is W1;

[0109] Step S306, calculating the unit change rate of the difference between any two adjacent gas intensity differences in the first difference sequence according to the difference change formula to obtain a first change rate sequence. The difference change formula is as follows: , where BK represents the unit change rate of adjacent gas light intensity differences, QK(i) and QK(i+1) represent the i-th and i+1-th gas light intensity differences, respectively; t(i) and t(i+1) represent the acquisition time corresponding to the i-th and i+1-th gas light intensity differences, respectively;

[0110] Step S307: record all the change rates in the first change rate sequence in order as BK(1)-BK(n-1); calculate the average change rate PK0 according to the average formula, which is as follows: ;

[0111] In step S308, the sampling time interval is adjusted to T1 according to the first interval formula; the first interval formula is as follows: , where Tm is the maximum sampling time interval set; PKm is the average change rate threshold set; λ is the set proportional coefficient;

[0112] Step S309 , repeatedly and dynamically adjust the collection time intervals of all transparent incubators, and obtain the gas incident light intensity QR and the gas exit light intensity QC of all transparent incubators, which are recorded as relevant gas detection data.

[0113] Step S4, performing comprehensive gas concentration detection and analysis based on relevant gas detection data, and performing gas detection management inside multiple transparent incubators; Step S4 includes the following sub-steps:

[0114] Step S401: Obtain the optical path of the laser through the interior of the first incubator. For relevant gas detection data corresponding to any acquisition time of the first incubator, calculate the concentration of the gas to be measured inside the first incubator according to the Beer-Lambert law, and record it as the gas concentration to be measured. Repeat the calculation to obtain the gas concentration to be measured corresponding to all acquisition times of the first incubator. The gas concentrations are sorted from the most recent acquisition time to the most recent acquisition time, and recorded as the first gas concentration sequence.

[0115] Step S402: Setting an upper concentration threshold and a lower concentration threshold. If the concentration of the gas to be measured corresponding to any collection moment is greater than the upper concentration threshold or less than the lower concentration threshold, an alarm signal is issued.

[0116] Step S403, repeatedly detecting and managing the concentration of the gas to be tested in all transparent incubators;

[0117] Step S404: Obtain a first gas concentration sequence for a second time period in the past, recorded as a second gas concentration sequence, and calculate the difference between any two adjacent gas concentrations in the second gas concentration sequence according to a difference formula to obtain a first concentration difference sequence. The difference formula is as follows: , where VC(j) represents the difference between two adjacent gas concentrations to be measured, VK(j) and VK(j+1) represent the j-th and j+1-th gas concentrations to be measured, respectively; wherein the second time length is W2;

[0118] Step S405: Arrange the first concentration difference sequence in ascending order, remove the smallest x1 differences and the largest x1 differences, and obtain a second concentration difference sequence. Calculate the mean and standard deviation of the second concentration difference sequence, and record them as Y1 and Y2 respectively.

[0119] Step S406: Y1+k1*Y2 is recorded as the dynamic change threshold, and a static change threshold is set. For the measured gas concentration corresponding to any acquisition, the difference between the measured gas concentration corresponding to the previous acquisition is calculated and recorded as the gas concentration change value. If the gas concentration change value is greater than the dynamic change threshold or the static change threshold, an alarm signal is issued;

[0120] Step S407 , repeatedly detecting and managing the changes in the concentration of the gas to be measured in all transparent incubators.

[0121] Example 3, please refer to Figure 4 As shown, Figure 4A schematic diagram of the structure of an electronic device is provided, which may include: a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other via the communication bus. The memory stores computer-readable instructions, and the processor can call the instructions in the memory. When the computer-readable instructions are executed by the processor, the steps of a gas identification and concentration detection method based on machine learning are executed to achieve the following functions: obtaining relevant data of a transparent incubator and simulating light intensity loss using an open-circuit laser gas detector to obtain light intensity loss simulation data; constructing a light intensity loss impact model based on a multi-perceptron model and using the light intensity loss simulation data; constructing a gas detection optical path based on the open-circuit laser gas detector and dynamically collecting the laser light intensity passing through the transparent incubator to obtain relevant gas detection data; performing comprehensive gas concentration detection and analysis based on the relevant gas detection data, and performing gas detection management within multiple transparent incubators.

[0122] In addition, the logical instructions in the above-mentioned memory can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on this understanding, the technical solution of the present application, or the part that contributes to the existing technology, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, and other media that can store program code.

[0123] Example 4. The present application also provides a computer-readable storage medium. The present application provides a storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps in the gas identification and concentration detection method based on machine learning are executed to achieve the following functions: based on the transparent incubator, the box-related data is obtained, and the light intensity loss is simulated using an open-circuit laser gas detector to obtain light intensity loss simulation data; based on the multiple sensor model, a light intensity loss influence model is constructed using the light intensity loss simulation data; based on the open-circuit laser gas detector, a gas detection optical path is constructed, and the laser light intensity passing through the transparent incubator is dynamically collected to obtain relevant gas detection data; based on the relevant gas detection data, a comprehensive gas concentration detection and analysis is performed, and gas detection management is performed inside multiple transparent incubators.

[0124] Through the description of the above embodiments, the embodiments of the present invention can be provided as methods, systems, or computer program products. Based on this understanding, the essence of the above technical solutions or the portion that contributes to the prior art can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for enabling a computer device (such as a personal computer, server, or network device) to execute the methods described in various embodiments or certain portions of the embodiments.

[0125] In the embodiments provided in this application, it should be understood that the disclosed system or method can be implemented in other ways. The embodiments described above are merely illustrative. For example, the division of modules or units is only a logical function division. There may be other division methods in actual implementation. For example, multiple modules or units can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, and the indirect coupling or communication connection of systems, modules and units can be electrical, mechanical or other forms.

[0126] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A gas concentration detection method based on machine learning, characterized in that: The steps include: Based on the transparent incubator, relevant data of the box is obtained, and the light intensity loss simulation data is obtained by using an open-circuit laser gas detector; Based on the multi-sensor model, the light intensity loss impact model is constructed using light intensity loss simulation data; A gas detection optical path is constructed based on an open-circuit laser gas detector, and the laser light intensity passing through the transparent incubator is dynamically collected to obtain relevant gas detection data. Conduct comprehensive gas concentration detection and analysis based on relevant gas detection data, and manage gas detection inside multiple transparent incubators; Obtaining relevant data of the transparent incubator and simulating light intensity loss using an open-circuit laser gas detector to obtain light intensity loss simulation data includes the following sub-steps: Place the transparent incubator between the transmitting end and the receiving end of the open-circuit laser gas detector, so that the laser emitted by the transmitting end is perpendicular to the side of the transparent incubator. The two sides of the transparent incubator that the laser passes through are recorded as the incident side and the exit side in the order in which the laser passes. The thicknesses of the incident side and the exit side are obtained, which are recorded as HR and HC in sequence; and the refractive indices of the incident side and the exit side are obtained, which are recorded as ZR and ZC in sequence; Obtaining box-related data based on the transparent incubator and using an open-circuit laser gas detector to simulate light intensity loss, and obtaining light intensity loss simulation data also includes the following sub-steps: The gas to be detected is recorded as the gas to be detected; the absorption wavelength of the gas to be detected is obtained and recorded as the gas absorption wavelength; and the laser wavelength emitted by the open-path laser gas detector is set to be the gas absorption wavelength; The laser light intensity emitted by the emitting end of the open-circuit laser gas detector is recorded as the emitted light intensity, and the light intensity received by the receiving end of the open-circuit laser gas detector is recorded as the received light intensity; The incident side and the exit side of the transparent incubator are placed separately between the transmitting end and the receiving end of the open-circuit laser gas detector, and the initial emission light intensity is set to A0. The emission light intensity is continuously reduced in sequence with the first light intensity interval, and each time a new emission light intensity is set, the received light intensity and the ambient temperature are recorded at the same time. Different emission intensities, ambient temperatures and corresponding received light intensities are obtained and marked as the corresponding incident side. The simulation data of the exit side is obtained and combined with the corresponding thickness and refractive index and stored as the corresponding light intensity loss simulation data, where the first light intensity interval is a1; Repeatedly obtain light intensity loss simulation data corresponding to the incident side and the exit side of multiple transparent incubators.

2. The gas concentration detection method based on machine learning according to claim 1, characterized in that: Based on the multi-sensor model and using the light intensity loss simulation data, the light intensity loss impact model is constructed, which includes the following sub-steps: Normalize all light intensity loss simulation data according to data type, scale all data sizes to [0, 1], and obtain normalized loss simulation data after completion; Any set of corresponding emission light intensity, thickness, refractive index and ambient temperature in the normalized loss simulation data is combined into a loss feature vector, denoted as M0={m1, m2, m3, m4}, where m1, m2, m3 and m4 represent the emission light intensity, thickness, refractive index and ambient temperature respectively in order; M0 is merged and stored with the corresponding received light intensity; all normalized loss simulation data are processed repeatedly to obtain loss simulation training data.

3. The gas concentration detection method based on machine learning according to claim 2, characterized in that: Building a light intensity loss impact model based on the multi-sensor model and using light intensity loss simulation data also includes the following sub-steps: An original loss influence model is constructed based on the multi-perceptron model. The original loss influence model includes: input layer, hidden layer and output layer. The number of neurons in the input layer is set to b1, the number of hidden layers is set to b2, the number of neurons in each hidden layer is set to b3, and the number of neurons in the output layer is set to b4. The original loss influence model is trained using loss simulation training data, and the light intensity loss influence model is obtained after completion.

4. The gas concentration detection method based on machine learning according to claim 3, characterized in that: A gas detection optical path is constructed based on an open-circuit laser gas detector, and the laser light intensity passing through the transparent incubator is dynamically collected to obtain relevant gas detection data. The process includes the following sub-steps: Arrange multiple transparent incubators in sequence and place them neatly between the transmitting end and the receiving end of the open-circuit laser gas detector, so that the laser emitted by the transmitting end is perpendicular to the sides of all transparent incubators; Place a semi-transparent and semi-reflective mirror between any two transparent incubators to split the laser light path between the two transparent incubators into two mutually perpendicular split light paths. The split light path perpendicular to the original laser light path is recorded as the reference light path, and the split light path along the original laser light path is recorded as the incident light path. Set the splitting ratio of the incident light path to the reference light path to be c1:c2. For any transparent incubator, designated as the first incubator, the light intensity of the reference optical path corresponding to the incident side and the exit side of the first incubator is acquired at sampling intervals, while the ambient temperature is also acquired, and the acquisition time is recorded. The laser light intensity entering the first incubator at the incident side and the laser light intensity exiting from the exit side of the first incubator are calculated based on the corresponding splitting ratios, and are denoted as RI and CI, respectively. The sampling interval is T0. The ambient temperature, RI, CI, and the corresponding HR, HC, ZR, and ZC are normalized and formed into corresponding loss feature vectors, which are input into the light intensity loss influence model. The laser light intensity emitted from the incident side of the first incubator and the laser light intensity on the exit side entering the first incubator are obtained, and are recorded as the gas incident light intensity QR and the gas exit light intensity QC, respectively.

5. The gas concentration detection method based on machine learning according to claim 4, characterized in that: Constructing a gas detection optical path based on an open-circuit laser gas detector and dynamically collecting the laser light intensity passing through the transparent incubator to obtain relevant gas detection data also includes the following sub-steps: Calculate the gas light intensity difference QK, QK = QR - QC; obtain the gas light intensity difference QK corresponding to n acquisitions within the most recent first time length, and sort them from recent to far according to the acquisition time as the first difference sequence; the first time length is W1; The unit change rate of the difference between any two adjacent gas light intensities in the first difference sequence is calculated according to the difference change formula to obtain the first change rate sequence. The difference change formula is as follows: , where BK represents the unit change rate of adjacent gas light intensity differences, QK(i) and QK(i+1) represent the i-th and i+1-th gas light intensity differences, respectively; t(i) and t(i+1) represent the acquisition time corresponding to the i-th and i+1-th gas light intensity differences, respectively; All the change rates in the first change rate sequence are recorded in order as BK(1)-BK(n-1); the average change rate PK0 is calculated according to the average formula, which is as follows: ; The sampling time interval is adjusted to T1 according to the first interval formula; the first interval formula is as follows: , where Tm is the maximum sampling time interval set; PKm is the average change rate threshold set; λ is the set proportional coefficient; Repeat the dynamic adjustment of the acquisition time intervals of all transparent incubators, and obtain the gas incident light intensity QR and the gas exit light intensity QC of all transparent incubators, which are recorded as relevant gas detection data.

6. The gas concentration detection method based on machine learning according to claim 5, characterized in that: Based on the relevant gas detection data, comprehensive gas concentration detection and analysis are performed to manage gas detection inside multiple transparent incubators, including the following sub-steps: Obtain the optical path of the laser through the interior of the first incubator. Calculate the concentration of the gas to be measured inside the first incubator according to the Beer-Lambert law for relevant gas detection data corresponding to any collection time of the first incubator, and record it as the gas concentration to be measured. Repeat the calculation to obtain the gas concentration to be measured corresponding to all collection times of the first incubator. The gas concentrations are sorted from the most recent collection time to the most recent collection time, and recorded as the first gas concentration sequence. Set the upper and lower concentration thresholds. If the concentration of the gas to be measured at any sampling moment is greater than the upper concentration threshold or less than the lower concentration threshold, an alarm signal will be issued. Repeat the test management for the gas concentration to be tested in all transparent incubators.

7. The gas concentration detection method based on machine learning according to claim 6, characterized in that: Based on the relevant gas detection data, comprehensive gas concentration detection and analysis is performed, and gas detection management inside multiple transparent incubators also includes the following sub-steps: Obtain the first gas concentration sequence of the second time length in the past, recorded as the second gas concentration sequence, and calculate the difference between any two adjacent measured gas concentrations in the second gas concentration sequence according to the difference formula to obtain the first concentration difference sequence. The difference formula is as follows: , where VC(j) represents the difference between two adjacent gas concentrations to be measured, VK(j) and VK(j+1) represent the j-th and j+1-th gas concentrations to be measured, respectively; wherein the second time length is W2; Arrange the first concentration difference sequence in ascending order, remove the smallest x1 differences and the largest x1 differences, and obtain a second concentration difference sequence. Calculate the mean and standard deviation of the second concentration difference sequence, and record them as Y1 and Y2 respectively. The dynamic change threshold is obtained through Y1 and Y2, and the static change threshold is set. For the measured gas concentration corresponding to any acquisition, the difference between the measured gas concentration corresponding to the previous acquisition is calculated and recorded as the gas concentration change value. If the gas concentration change value is greater than the dynamic change threshold or the static change threshold, an alarm signal is issued; Repeat the detection and management of changes in the concentration of the gas to be tested in all transparent incubators.

8. A gas concentration detection system based on machine learning, used to implement the gas concentration detection method based on machine learning according to any one of claims 1 to 7, characterized in that: Including data simulation module, loss model module, light intensity acquisition module and detection management module; The data simulation module includes an acquisition unit and a simulation unit. The acquisition unit obtains box-related data based on the transparent incubator. The simulation unit uses an open-circuit laser gas detector to simulate light intensity loss to obtain light intensity loss simulation data. The loss model module is based on a multi-sensor model and uses light intensity loss simulation data to construct a light intensity loss impact model; The light intensity acquisition module constructs a gas detection optical path based on an open-circuit laser gas detector and dynamically collects the laser light intensity passing through the transparent incubator to obtain relevant gas detection data; The detection management module performs comprehensive gas concentration detection and analysis based on relevant gas detection data, and performs gas detection management inside multiple transparent incubators.

Images