A micro-service-based kitchen fire early warning internet platform

Abstract

The application relates to the technical field of fire warning, and discloses a kitchen fire warning internet platform based on micro services, wherein the process of kitchen fire warning comprises the following steps: signal pretreatment and time synchronization processing are performed on multi-source environment signals; airflow signal data are subjected to airflow disturbance feature extraction; the pretreated multi-source environment signals are subjected to dynamic compensation based on the airflow disturbance features, and environment steady-state features of the multi-source environment signals after dynamic compensation are extracted; fire risk assessment is performed on the environment steady-state features by using a fire warning model, fire warning information is generated according to the fire risk assessment result, and the fire warning information is delivered to an edge node. By performing multi-source signal pretreatment, airflow disturbance analysis and dynamic compensation, stable and reliable environment steady-state features are extracted, so that the internet platform has high accuracy, low false alarm rate and low energy consumption in the fire warning under a complex kitchen environment, and fast and reliable fire warning closed-loop control is realized.

Description

Technical Field

[0001] This invention relates to the field of fire early warning, and more particularly to a microservice-based internet platform for early warning of kitchen fires. Background Technology

[0002] With the widespread use of gas stoves, induction cookers, and range hoods in urban households and the catering industry, fire risks in kitchen environments are characterized by multi-factor coupling, rapid changes, and high concealment. Multi-source environmental signals in kitchen spaces, such as temperature, smoke, humidity, gas concentration, and airflow disturbances, exhibit significant transient characteristics and are easily affected by factors such as range hood duct disturbances, personnel movement, and localized heating. This makes traditional single-sensor or fixed-threshold fire detection methods ineffective in describing real fire trends, leading to false alarms or missed alarms. In recent years, fire early warning methods based on multi-source sensing have gradually become a research hotspot, but they still face challenges such as difficulty in extracting steady-state data, large signal fluctuations, and poor model robustness in the highly disturbed kitchen environment.

[0003] In existing research, CN210691528U discloses a kitchen fire early warning system based on thermal imaging and gas sensors. This system uses a fire detection module and a gas detection module to detect temperature and gas concentration in the kitchen, respectively. Combined with a display module and a back-end monitoring center, it displays abnormal information in real time. This solution enables simultaneous monitoring of temperature and gas, reducing manual inspection costs and improving the real-time performance of fire status identification. However, this system still relies on a single or limited number of sensor sources and cannot jointly model multi-source environmental signals; its early warning logic is mainly based on threshold monitoring and cannot handle airflow disturbances, temperature fluctuations, and transient changes in smoke in the kitchen environment.

[0004] To address this issue, this invention proposes a microservice-based internet platform for early warning of kitchen fires. This platform enhances the robustness of fire warning results to dynamic environmental changes and reduces false alarm and missed alarm rates. It has significant research value and practical significance for improving fire safety levels in home and commercial kitchens and promoting the practical application and intelligence of smart fire protection technologies. Summary of the Invention

[0005] This invention provides a microservice-based internet platform for early warning of kitchen fires. Step S1 preprocesses and synchronizes multiple signals (temperature, smoke, humidity, airflow, etc.) from edge nodes, ensuring alignment of these signals on the same timeline. Subsequent calculations are based on microservice unit integration modules within the internet platform, reducing communication interactions and improving efficiency. Step S2 models disturbance sources by extracting airflow signal features, distinguishing between real fire-induced temperature rise or smoke accumulation and false anomalies caused by airflow disturbances. Step S3 performs dynamic compensation based on airflow disturbance features, correcting disturbance offsets in temperature / smoke / humidity signals to obtain stable environmental characteristics that accurately reflect environmental trends. Step S4 issues fire warning information after risk assessment, dynamically adjusting the edge node sampling / upload frequency according to risk, resolving latency issues caused by fixed sampling. This addresses the core technical problems of high false alarm rates, poor model stability, and delayed warning response caused by high airflow disturbances in kitchen environments, significantly improving the accuracy, real-time performance, and energy efficiency of fire warnings.

[0006] To achieve the above objectives, this invention provides a microservice-based internet platform for early warning of kitchen fires, including the following kitchen fire early warning method:

[0007] S1: Edge nodes collect multi-source environmental signals in the kitchen environment in real time, and perform signal preprocessing and time synchronization processing on the multi-source environmental signals to obtain preprocessed multi-source environmental signals;

[0008] S2: Extract airflow signal data from the preprocessed multi-source environmental signals, and extract airflow disturbance features from the airflow signal data;

[0009] S3: Dynamically compensate the preprocessed multi-source environmental signals based on airflow disturbance characteristics, and extract the environmental steady-state characteristics of the dynamically compensated multi-source environmental signals;

[0010] S4: Use the fire early warning model to assess the fire risk of the steady-state characteristics of the environment, generate fire early warning information based on the fire risk assessment results and send it to the edge nodes. The edge nodes adjust the sampling frequency and upload frequency of the multi-source environmental signals synchronously based on the fire early warning information.

[0011] As a further improvement of the present invention:

[0012] Furthermore, multi-source environmental signals from the kitchen environment are collected, and signal preprocessing and time synchronization processing are performed on the multi-source environmental signals, including:

[0013] The multi-source environmental signal data includes temperature signal data, smoke signal data, airflow signal data, and humidity signal data;

[0014] Edge nodes perform signal preprocessing on multi-source environmental signals to obtain preprocessed multi-source environmental signals. The signal preprocessing includes noise filtering, outlier removal, and standardization.

[0015] Time synchronization processing is performed on the multi-source environmental signals after signal preprocessing. The time synchronization processing flow is as follows:

[0016] The sensor acquires the unified reference time built into the edge node. During the process of uploading signal data, the sensor clock is uploaded synchronously, and the deviation of the sensor clock relative to the unified reference time is calculated.

[0017] The acquisition time of data values ​​in multi-source environmental signals is obtained. Based on the deviation of the sensor clock associated with the data value from the unified reference time, the acquisition time is corrected to eliminate the clock offset of the sensor clock and obtain the preprocessed multi-source environmental signal.

[0018] Furthermore, airflow disturbance features are extracted from the preprocessed multi-source environmental signals, including:

[0019] The airflow signal data from the preprocessed multi-source environmental signal is acquired. The airflow signal data is then segmented into multiple airflow signal segments according to a fixed-length time window R. Local feature extraction is performed on each airflow signal segment to obtain disturbance intensity features, local velocity change rate features, and sustained disturbance estimation features. These features are then sorted according to their order within the airflow signal data and used as the airflow disturbance features. The process for local feature extraction of the airflow signal is as follows:

[0020] S21: Calculate the standard deviation of the data values ​​in the airflow signal, perform a fast Fourier transform on the airflow signal, extract the high-frequency energy of the airflow signal in the high-frequency range, and then fused the high-frequency energy with the standard deviation after logarithmic compression to serve as the disturbance intensity feature of the airflow signal.

[0021] S22: The average rate of change of data values ​​in the airflow signal is calculated using the adjacent difference method, which serves as the local velocity change rate characteristic of the airflow signal; the calculation formula for the local velocity change rate characteristic is as follows:

[0022] ;

[0023] in, This represents the local velocity change rate characteristic of the airflow signal in segment h, where H represents the total number of airflow segments. Let L represent the e-th signal value in the h-th segment of the airflow signal, and L represent the number of data values ​​in each segment of the airflow signal.

[0024] S23: By integrating disturbance intensity characteristics, local velocity change rate characteristics, and incorporating the change in the average data value of adjacent airflow signals, a persistent disturbance estimation feature is generated to characterize whether the current airflow signal is persistent; the formula for generating the persistent disturbance estimation feature is as follows:

[0025] ;

[0026] in, This represents the disturbance intensity characteristics of the airflow signal in segment h. This represents the sustained disturbance estimation characteristics of the airflow signal in segment h. This represents the average data value of the airflow signal in segment h. This represents the average data value of the airflow signal in segment h-1. This represents an exponential function with the natural constant as its base. All of these represent weighting coefficients.

[0027] Furthermore, the calculation process for the disturbance intensity characteristics of the airflow signal in step S21 is as follows:

[0028] Performing a Fast Fourier Transform on the airflow signal yields the complex frequency domain signals of the airflow signal at different frequency indices:

[0029] ;

[0030] in, This represents the complex frequency domain signal of the h-th segment of the airflow signal at frequency index u. Represents the imaginary unit. , This represents the e-th signal value in the h-th segment of the airflow signal, where L represents the number of data values ​​in each segment of the airflow signal. Represents an exponential function with the natural constant as its base;

[0031] High-frequency range is set by cutoff frequency. :

[0032] ;

[0033] in, This indicates the sampling frequency of the airflow sensor that collects airflow signal data. Indicates the first cutoff frequency;

[0034] Based on high frequency range Extracting high-frequency energy from airflow signals in the high-frequency range:

[0035] ;

[0036] in, This represents the high-frequency energy of the airflow signal in the h-th segment within the high-frequency range;

[0037] The high-frequency energy is logarithmically compressed and then fused with the standard deviation to obtain the disturbance intensity characteristic of the airflow signal; the calculation formula for the disturbance intensity characteristic is as follows:

[0038] ;

[0039] in, This represents the disturbance intensity characteristics of the airflow signal in segment h. Indicates the fusion weight. This represents the standard deviation of the data values ​​in the h-th segment of the airflow signal.

[0040] It represents the natural logarithm.

[0041] Furthermore, dynamic compensation is performed on the preprocessed multi-source environmental signals based on airflow disturbance characteristics, including:

[0042] The temperature signal data, smoke signal data and humidity signal data in the preprocessed multi-source environmental signal are segmented according to a fixed-length time window R, and multiple segments of temperature signal, smoke signal and humidity signal are obtained in turn.

[0043] Based on the disturbance intensity characteristics, local velocity change rate characteristics, and continuous disturbance estimation characteristics of each airflow signal segment, dynamic compensation processing is performed on the data values ​​of the corresponding segment's temperature signal, smoke signal, and humidity signal. The dynamically compensated temperature signal, smoke signal, and humidity signal are then sequentially spliced ​​together to obtain dynamically compensated temperature signal data, smoke signal data, and humidity signal data. The dynamically compensated temperature signal data, smoke signal data, humidity signal data, and airflow signal data are used as the dynamically compensated multi-source environmental signal.

[0044] Furthermore, the environmental steady-state characteristics of the dynamically compensated multi-source environmental signals are extracted, including:

[0045] The mean and standard deviation of temperature, smoke, humidity and airflow signals in the multi-source environmental signals after dynamic compensation are calculated. The least squares method is used to fit the linear slope of the temperature, smoke, humidity and airflow signals respectively to obtain the linear slope of the temperature, smoke, humidity and airflow signals.

[0046] Based on the mean, standard deviation, and linear slope of temperature, smoke, humidity, and airflow signal data, the environmental steady-state characteristics of the dynamically compensated multi-source environmental signals are constructed. .

[0047] Furthermore, the fire early warning model is a gradient boosting tree model, which consists of multiple weak regression trees. The weak regression trees are connected in sequence and receive environmental steady-state features respectively. The weighted sum of the outputs of the leaf nodes in all weak regression trees is calculated as the fire risk assessment result. Each weak regression tree consists of a root node, multiple internal nodes, and leaf nodes.

[0048] Furthermore, a fire risk assessment is conducted on the steady-state characteristics of the environment using a fire early warning model. Based on the fire risk assessment results, fire early warning information is generated and distributed to edge nodes, including:

[0049] The fire risk assessment result is in numerical form between 0 and 1, where a higher fire risk assessment result indicates a greater probability of a fire occurring in the kitchen. Fire warning information is generated based on the fire risk assessment result, and the method for generating the fire warning information is as follows:

[0050] ;

[0051] in, Indicates the results of the fire risk assessment. Corresponding fire warning information, This indicates the first warning threshold. This indicates the second warning threshold. This indicates the third warning threshold, where , This indicates the first fire warning information. This indicates the second fire warning information. This indicates the third fire warning information.

[0052] This invention also proposes a microservice-based kitchen fire early warning internet platform. The microservice-based kitchen fire early warning internet platform includes a sensor layer, edge nodes, and a microservice unit integration module. The microservice unit integration module contains multiple microservice units. The microservice units are responsible for performing tasks such as extracting airflow disturbance features, dynamically compensating preprocessed multi-source environmental signals based on airflow disturbance features, extracting environmental steady-state features of dynamically compensated multi-source environmental signals, using a fire early warning model to assess fire risk based on environmental steady-state features, and generating fire early warning information based on the fire risk assessment results and sending it to the edge nodes.

[0053] Compared with existing technologies, this invention proposes a microservice-based internet platform for early warning of kitchen fires, which has the following beneficial effects:

[0054] First, this invention utilizes Fast Fourier Transform to obtain the frequency domain representation of airflow signals at different frequency indices, and constructs a high-frequency range through an adaptive cutoff frequency. This allows for the targeted capture of high-frequency energy fluctuations caused by sudden disturbances and instantaneous wind speed changes. By calculating the high-frequency energy and introducing logarithmic compression, this feature maintains its sensitivity to change while suppressing the impact of extreme values ​​on the overall assessment, thus improving the stability and comparability of this feature. Furthermore, this invention integrates the frequency domain high-frequency energy with the time domain standard deviation according to weights, which can simultaneously reflect the time domain fluctuation intensity and frequency domain disturbance structure of the airflow signal. This constructs a composite disturbance intensity feature that combines time and frequency information, exhibiting higher discriminative power and robustness.

[0055] Meanwhile, this invention employs a fixed-length time window to segment airflow signal data, combining three major features: disturbance intensity characteristics (standard deviation + high-frequency energy logarithmic fusion), local velocity change rate characteristics, and persistent disturbance estimation characteristics (multiple features + adjacent mean changes). This captures both the overall fluctuations and high-frequency disturbances of the airflow signal, as well as the local change rate and disturbance persistence, avoiding the one-sidedness of a single feature and achieving a comprehensive characterization of airflow disturbances at different time periods. Specifically, the persistent disturbance estimation feature introduces the mean change and exponential mapping of adjacent airflow signal segments, strengthening the temporal dimension of disturbance correlation, accurately reflecting the dynamic evolution of airflow disturbances, adapting to the physical characteristics of continuous airflow changes, and improving the adaptability of the persistent disturbance estimation feature to complex disturbance scenarios. Attached Figure Description

[0056] Figure 1 A flowchart illustrating a method included in an internet platform for early warning of kitchen fires, provided as an embodiment of the present invention;

[0057] Figure 2 A flowchart illustrating the calculation of disturbance intensity characteristics according to an embodiment of the present invention;

[0058] Figure 3 This is a schematic diagram of a microservice-based internet platform for early warning of kitchen fires, provided as an embodiment of the present invention.

[0059] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0060] The realization of the objectives, functional characteristics, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0061] This invention provides a microservice-based internet platform for early warning of kitchen fires. The executing entity of this microservice-based internet platform includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this invention: a server, a terminal, etc. In other words, the microservice-based internet platform for early warning of kitchen fires can be executed by software or hardware installed on a terminal device or a server device; the software may be a blockchain platform. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster.

[0062] Reference Figure 1 as well as Figure 2 As shown, Embodiment 1 of the present invention is as follows:

[0063] A microservice-based internet platform for early warning of kitchen fires, comprising the following methods for early warning of kitchen fires:

[0064] S1: Edge nodes collect multi-source environmental signals in the kitchen environment in real time, and perform signal preprocessing and time synchronization processing on the multi-source environmental signals to obtain preprocessed multi-source environmental signals.

[0065] Collect multi-source environmental signals from the kitchen environment, and perform signal preprocessing and time synchronization processing on the multi-source environmental signals, including:

[0066] The multi-source environmental signal data includes temperature signal data, smoke signal data, airflow signal data, and humidity signal data;

[0067] Specifically, the temperature sensor, smoke sensor, airflow sensor, and humidity sensor in the sensor layer sequentially collect temperature signal data, smoke signal data, airflow signal data, and humidity signal data, and upload them to the edge node. The temperature signal data, smoke signal data, airflow signal data, and humidity signal data are collected in the same time period, but the length of the signal data may not be the same, depending on the sampling frequency of the signal data. The higher the sampling frequency, the longer the length of the signal data.

[0068] As an embodiment of the present invention, the data in the temperature signal data, smoke signal data, airflow signal data and humidity signal data includes data values ​​and the sampling time of the data values. The data value types of the temperature signal data, smoke signal data, airflow signal data and humidity signal data are, respectively, temperature value, smoke concentration, airflow velocity and humidity percentage.

[0069] Edge nodes perform signal preprocessing on multi-source environmental signals to obtain preprocessed multi-source environmental signals. The signal preprocessing includes noise filtering, outlier removal, and standardization.

[0070] As an embodiment of the present invention, a moving average filtering method is used to perform noise filtering on signal data in multi-source environmental signals, wherein the filtering formula of the moving average filtering method is:

[0071] ;

[0072] in, Indicating the first element in a multi-source environmental signal The first type of signal data One data value, Indicates the first The signal data lengths of the four types of signal data are as follows: the first four types are temperature signal data, smoke signal data, airflow signal data, and humidity signal data, respectively. Represents data value The noise filtering results Indicating the first element in a multi-source environmental signal The first type of signal data One data value, This indicates the range of the sliding window; K is set to 5.

[0073] Anomaly detection is performed on the noise-filtered signal data using a first-order difference constraint. Data values ​​detected as anomalies are either discarded or replaced with data values ​​from the previous sampling time. The anomaly detection method is as follows:

[0074] ;

[0075] in, Represents data value The abnormal data value detection results, if A value of 1 indicates a data value. These are abnormal data values. Indicates the first The maximum allowable variation range of the various signal data can be optionally set as follows: the maximum allowable variation ranges of temperature signal data, smoke signal data, airflow signal data, and humidity signal data are 1 degree Celsius, 10 micrograms per cubic meter, 0.1 meters per second, and 2%, respectively.

[0076] The first to fourth types of signal data are standardized sequentially, with the standardization method being max-min normalization. Specifically, the data values... The standardized processing method is as follows: ,in Indicates the first The preset maximum value of the signal data Indicates the first Optionally, the preset minimum values ​​for various signal data can be set as follows: the preset maximum values ​​for temperature signal data, smoke signal data, airflow signal data, and humidity signal data are 500 degrees Celsius, 4000 micrograms per cubic meter, 5 meters per second, and 100%, respectively; and the preset minimum values ​​for temperature signal data, smoke signal data, airflow signal data, and humidity signal data are 10 degrees Celsius, 0 micrograms per cubic meter, 0 meters per second, and 0%, respectively.

[0077] Time synchronization processing is performed on the multi-source environmental signals after signal preprocessing. The time synchronization processing flow is as follows:

[0078] The sensor acquires the unified reference time built into the edge node. During the process of uploading signal data, the sensor clock is uploaded synchronously, and the deviation of the sensor clock relative to the unified reference time is calculated.

[0079] In one embodiment of the present invention, the unified reference time can be the NTP (Network Time Protocol) synchronization time or a stable clock built into the edge node itself, and the formula for calculating the deviation of the sensor clock relative to the unified reference time is as follows:

[0080] ;

[0081] in, Indicates the collection of the first Sensors of various signal data relative to a unified reference time deviation, Indicates the collection of the first The sensor clock that uploads signal data from the sensor;

[0082] The acquisition time of data values ​​in multi-source environmental signals is obtained. Based on the deviation of the sensor clock associated with the data value from the unified reference time, the acquisition time is corrected to eliminate the clock offset of the sensor clock and obtain the preprocessed multi-source environmental signal.

[0083] Specifically, the correction formula for the acquisition time is:

[0084] ;

[0085] in, Represents data value The corresponding data collection time, Indicates the time of data collection The corrected results.

[0086] S2: Extract airflow signal data from the preprocessed multi-source environmental signals and extract airflow disturbance features from the airflow signal data.

[0087] Airflow disturbance feature extraction is performed on the airflow signal data from the preprocessed multi-source environmental signals, including:

[0088] The airflow signal data from the preprocessed multi-source environmental signal is acquired. The airflow signal data is then segmented into multiple airflow signal segments according to a fixed-length time window R. Local feature extraction is performed on each airflow signal segment to obtain disturbance intensity features, local velocity change rate features, and sustained disturbance estimation features. These features are then sorted according to their order within the airflow signal data and used as the airflow disturbance features. The process for local feature extraction of the airflow signal is as follows:

[0089] S21: Calculate the standard deviation of the data values ​​in the airflow signal, perform a fast Fourier transform on the airflow signal, extract the high-frequency energy of the airflow signal in the high-frequency range, and then fused the high-frequency energy with the standard deviation after logarithmic compression to serve as the disturbance intensity feature of the airflow signal.

[0090] S22: The average rate of change of data values ​​in the airflow signal is calculated using the adjacent difference method, which serves as the local velocity change rate characteristic of the airflow signal; the calculation formula for the local velocity change rate characteristic is as follows:

[0091] ;

[0092] in, This represents the local velocity change rate characteristic of the airflow signal in segment h, where H represents the total number of airflow segments. Let L represent the e-th signal value in the h-th segment of the airflow signal, and L represent the number of data values ​​in each segment of the airflow signal.

[0093] S23: By integrating disturbance intensity characteristics, local velocity change rate characteristics, and incorporating the change in the average data value of adjacent airflow signals, a persistent disturbance estimation feature is generated to characterize whether the current airflow signal is persistent; the formula for generating the persistent disturbance estimation feature is as follows:

[0094] ;

[0095] in, This represents the disturbance intensity characteristics of the airflow signal in segment h. This represents the sustained disturbance estimation characteristics of the airflow signal in segment h. This represents the average data value of the airflow signal in segment h. This represents the average data value of the airflow signal in segment h-1. This represents an exponential function with the natural constant as its base. All represent weighting coefficients, which can be configured empirically. The values ​​are 0.3, 0.3, and 0.4 respectively.

[0096] In one embodiment of the present invention, the length of the time window R is 5 seconds, that is, the airflow signal data is divided into windows of 5 seconds each to obtain multiple airflow signals with a time range of 5 seconds.

[0097] For reference Figure 2 The flowchart shown illustrates the calculation process for the disturbance intensity characteristics. In step S21, the calculation process for the disturbance intensity characteristics of the airflow signal is as follows:

[0098] S211: Perform a Fast Fourier Transform on the airflow signal to obtain the frequency domain complex signals of the airflow signal at different frequency indices:

[0099] ;

[0100] in, This represents the complex frequency domain signal of the h-th segment of the airflow signal at frequency index u. Represents the imaginary unit. , This represents the e-th signal value in the h-th segment of the airflow signal, where L represents the number of data values ​​in each segment of the airflow signal. This represents an exponential function with the natural constant as the base, and H represents the total number of segments in the airflow signal;

[0101] S212: Setting the high-frequency range by cutoff frequency :

[0102] ;

[0103] in, This indicates the sampling frequency of the airflow sensor that collects airflow signal data. Indicates the first cutoff frequency;

[0104] Optionally, a first cutoff frequency is set. The sampling frequency of the sensor represents the number of data values ​​collected by the sensor per second;

[0105] S213: Based on high frequency range Extracting high-frequency energy from airflow signals in the high-frequency range:

[0106] ;

[0107] in, This represents the high-frequency energy of the airflow signal in the h-th segment within the high-frequency range;

[0108] S214: The high-frequency energy is logarithmically compressed and then fused with the standard deviation to obtain the disturbance intensity characteristic of the airflow signal; the calculation formula for the disturbance intensity characteristic is as follows:

[0109] ;

[0110] in, This represents the disturbance intensity characteristics of the airflow signal in segment h. Indicates the fusion weight, which can be set. It is 0.6. This represents the standard deviation of the data values ​​in the h-th segment of the airflow signal. It represents the natural logarithm.

[0111] S3: Based on the airflow disturbance characteristics, the preprocessed multi-source environmental signals are dynamically compensated, and the environmental steady-state characteristics of the dynamically compensated multi-source environmental signals are extracted.

[0112] Dynamic compensation of preprocessed multi-source environmental signals based on airflow disturbance characteristics includes:

[0113] The temperature signal data, smoke signal data and humidity signal data in the preprocessed multi-source environmental signal are segmented according to a fixed-length time window R, and multiple segments of temperature signal, smoke signal and humidity signal are obtained in turn.

[0114] Based on the disturbance intensity characteristics, local velocity change rate characteristics, and continuous disturbance estimation characteristics of each airflow signal segment, dynamic compensation processing is performed on the data values ​​of the corresponding segment's temperature signal, smoke signal, and humidity signal. The dynamically compensated temperature signal, smoke signal, and humidity signal are then sequentially spliced ​​together to obtain dynamically compensated temperature signal data, smoke signal data, and humidity signal data. The dynamically compensated temperature signal data, smoke signal data, humidity signal data, and airflow signal data are used as the dynamically compensated multi-source environmental signal.

[0115] As an embodiment of the present invention, the data values ​​in the h-th segment of the temperature signal The dynamic compensation formula is:

[0116] ;

[0117] in, Represents data value The dynamic compensation results All represent temperature compensation coefficients, set The values ​​are 0.4 and 0.6 respectively. Represents the hyperbolic tangent function. All represent temperature control parameters, settings They are 2 and 1.5 respectively. This represents the mean value of the data in the h-th segment of the temperature signal;

[0118] It should be noted that, Disturbance intensity characteristics are used to compensate for the impact of gas disturbances on heat transfer. The stronger the heat dissipation / heat transfer, the more intense the heat dissipation / heat transfer, resulting in a greater deviation between the perceived temperature and the actual temperature. Nonlinear amplification of local velocity change rate characteristics Due to the contribution of gas flow rate, when there is a sudden change in gas flow rate, the heat exchange rate increases sharply, and the compensation amount needs to be increased accordingly. Used to compensate for temperature deviations under continuous disturbances, the exponential term makes the correction stronger for data values ​​that deviate more from the mean, avoiding the accumulation of distortion in extreme measurements, and making the compensated data values ​​closer to the actual temperature that should be shown.

[0119] The data values ​​in the h-th segment of the smoke signal The dynamic compensation formula is:

[0120] ;

[0121] in, Represents data value The dynamic compensation results All represent the smoke compensation coefficient, set The values ​​are 0.6 and 0.4 respectively. This represents the mean of the data values ​​in the h-th segment of the smoke signal. Indicates the smoke control factor, set It is 1.1;

[0122] It should be noted that, Used to compensate for the dilution effect of airflow on smoke concentration, where It exhibits an S-shaped nonlinearity, with smooth compensation under low disturbances and saturation under high disturbances. This conforms to the physical law that the dilution effect of airflow blowing away smoke has an upper limit. An exponential mapping is used to ensure that the compensation amount increases monotonically with the disturbance intensity, thus restoring the concentration of the diluted smoke. Used to compensate for the coupling bias between smoke diffusion and persistent disturbances. A non-linear approach is used to amplify and compensate for data values ​​deviating from the mean. High-concentration smoke is more affected by airflow and requires stronger compensation. Logarithmic terms are used to avoid compensation overflow. This reflects the real physical phenomenon that smoke diffusion is more intense in scenarios with rapid changes in flow velocity and continuous disturbance;

[0123] The data values ​​in the h-th segment of the humidity signal The dynamic compensation formula is:

[0124] ;

[0125] in, Represents data value The dynamic compensation results All represent humidity compensation coefficients, set The values ​​are 0.3 and 0.7 respectively. Represents a hyperbolic sine function. This represents the mean value of the data in the h-th segment of the humidity signal. Indicates the humidity control factor, set It is 0.6;

[0126] It should be noted that, Used to compensate for the water vapor carry-away effect of continuous airflow disturbance, and through The contribution of nonlinear amplification of disturbance intensity, in which the water vapor loss rate surges under high disturbance, and the compensation amount increases synchronously; The secondary effect reflects the characteristics of the local velocity change rate, among which abrupt velocity changes have a more significant impact on water vapor disturbance in low humidity environments. Avoid overcompensation during high flow rate abrupt changes;

[0127] The dynamic compensation method achieves a nonlinear mapping between characteristics and compensation quantities through hyperbolic functions, exponential functions, power functions, etc., adapting to the complex physical relationship between airflow disturbances and environmental parameter deviations.

[0128] Extracting the environmental steady-state characteristics of the dynamically compensated multi-source environmental signals, including:

[0129] The mean and standard deviation of temperature, smoke, humidity and airflow signals in the multi-source environmental signals after dynamic compensation are calculated. The least squares method is used to fit the linear slope of the temperature, smoke, humidity and airflow signals respectively to obtain the linear slope of the temperature, smoke, humidity and airflow signals.

[0130] Based on the mean, standard deviation, and linear slope of temperature, smoke, humidity, and airflow signal data, the environmental steady-state characteristics of the multi-source environmental signals after dynamic compensation are constructed. :

[0131] ;

[0132] in, The values ​​are represented by [mean, standard deviation, and linear slope] for temperature signal data, smoke signal data, humidity signal data, and airflow signal data, respectively.

[0133] S4: Use the fire early warning model to assess the fire risk of the steady-state characteristics of the environment, generate fire early warning information based on the fire risk assessment results and send it to the edge nodes. The edge nodes adjust the sampling frequency and upload frequency of the multi-source environmental signals synchronously based on the fire early warning information.

[0134] The fire early warning model is a gradient boosting tree model, which consists of multiple weak regression trees. The weak regression trees are connected in sequence and receive environmental steady-state features respectively. The weighted sum of the outputs of the leaf nodes in all weak regression trees is calculated as the fire risk assessment result. Each weak regression tree consists of a root node, multiple internal nodes, and leaf nodes.

[0135] Specifically, each weak regression tree takes environmental steady-state features as input, obtains multiple sets of environmental steady-state features and corresponding real fire risks (0 indicates no fire, 1 indicates fire) as training sets, uses Gini coefficient gain as the splitting index, and generates feature thresholds based on a certain dimension of the environmental steady-state features to divide the tree into root nodes, internal nodes, and leaf nodes. The root node is split for the first time based on a certain dimension of the environmental steady-state features input to the weak regression tree. Internal nodes are further split for the second or multiple times based on other features to generate feature thresholds. Leaf nodes store the fire risk increment value under that branch. Multiple sets of environmental steady-state features different from the training set and corresponding real fire risks (0 indicates no fire, 1 indicates fire) are obtained as validation sets. The training accuracy of each weak regression tree is output based on the validation set and normalized to obtain the training accuracy weights of the weak regression tree.

[0136] For the constructed weak regression tree, by receiving the steady-state features of the environment, and according to the feature partitioning rules and feature thresholds in the weak regression tree, the steady-state features of the environment are controlled to reach the leaf nodes, and the fire risk increment values ​​stored in the leaf nodes are extracted. The fire risk increment values ​​are weighted and superimposed using the training accuracy weights of the weak regression tree itself, so as to obtain the fire risk assessment results with values ​​ranging from 0 to 1.

[0137] A fire risk assessment is conducted on the steady-state characteristics of the environment using a fire early warning model. Based on the fire risk assessment results, fire early warning information is generated and distributed to edge nodes, including:

[0138] The fire risk assessment result is in numerical form between 0 and 1, where a higher fire risk assessment result indicates a greater probability of a fire occurring in the kitchen. Fire warning information is generated based on the fire risk assessment result, and the method for generating the fire warning information is as follows:

[0139] ;

[0140] in, Indicates the results of the fire risk assessment. Corresponding fire warning information, This indicates the first warning threshold. This indicates the second warning threshold. This indicates the third warning threshold, where ,set up The values ​​were 0.3, 0.6, and 0.9, respectively. This indicates the first fire warning information. This indicates the second fire warning information. This indicates the third fire warning information.

[0141] As an embodiment of the present invention, the first fire warning information indicates that the edge node needs to record the first fire warning information generated this time and the generation timestamp of the first fire warning information, while increasing the sampling frequency of each sensor; the second fire warning information indicates that the edge node needs to record the second fire warning information generated this time and the generation timestamp of the second fire warning information, while increasing the sampling frequency of each sensor, the frequency of edge node's own calculation of environmental steady-state characteristics, and the frequency of uploading environmental steady-state characteristics; the third fire warning information indicates that the edge node needs to record the third fire warning information generated this time and the generation timestamp of the third fire warning information, and simultaneously execute emergency actions such as power outage / ventilation control / audio-visual alarm. Example

[0142] For reference Figure 3 The diagram shows a microservice-based internet platform architecture for early warning of kitchen fires, including a sensor layer 101, edge nodes 102, and a microservice unit integration module 103. The microservice unit integration module contains various microservice units.

[0143] Sensor layer 101 is used to sense multi-source environmental signals in the kitchen environment and upload them to the edge nodes;

[0144] Edge node 102 is used to perform signal preprocessing and time synchronization processing on multi-source environmental signals to obtain preprocessed multi-source environmental signals.

[0145] The microservice unit integration module 103 includes a disturbance feature extraction microservice unit, a dynamic compensation microservice unit, an environmental steady-state feature extraction microservice unit, a fire risk assessment microservice unit, and an early warning information generation microservice unit. Each microservice unit can be deployed independently according to the required technical support. The disturbance feature extraction microservice unit is used to extract airflow signal data from the preprocessed multi-source environmental signals and extract airflow disturbance features from the airflow signal data. The dynamic compensation microservice unit performs dynamic compensation on the preprocessed multi-source environmental signals based on the airflow disturbance features. The environmental steady-state feature extraction microservice unit extracts the environmental steady-state features of the dynamically compensated multi-source environmental signals. The fire risk assessment microservice unit uses a fire early warning model to assess the fire risk based on the environmental steady-state features. The early warning information generation microservice unit generates fire early warning information based on the fire risk assessment results and sends it to the edge nodes.

[0146] It should be noted that different microservice units can be scaled up or down independently according to the load. For example, the fire risk assessment microservice unit can increase computing power independently during peak periods without affecting other microservice units. Each microservice unit can be configured with different access permissions and can be updated independently. Multiple algorithm versions can be deployed in parallel for A / B comparison to improve the iteration efficiency of the algorithm.

[0147] Specifically, the sensor layer includes various sensors such as temperature sensors, smoke sensors, airflow sensors, and humidity sensors, which are installed in key locations such as the kitchen stove area, around the fume duct, and the exhaust duct to collect multi-source environmental signals of the kitchen environment. Each sensor collects the corresponding environmental signal at the set sampling frequency and sends the collected environmental signal to the edge node deployed in the kitchen.

[0148] Optionally, a temperature sensor is installed near the heating source above the stove, located at the lower edge of the range hood or near the side wall of the stove, to monitor the local temperature rise around the stove in real time and to quickly detect fires in their early stages. A smoke sensor is placed in the air inlet area inside the range hood or near the flue inlet below it to fully capture abnormal changes in the concentration of cooking fumes. An airflow sensor is installed in the flue or exhaust duct near the exhaust fan inlet to monitor airflow disturbances, eddy current intensity, and changes in airflow velocity during the emission of cooking fumes. By sensing changes in airflow, it can identify possible abnormal exhaust, backflow, or turbulent flow caused by a fire. A humidity sensor is placed on the walls on both sides of the stove or in the center of the kitchen environment to avoid direct impact from water vapor while reflecting the overall humidity trend of the kitchen. In the early stages of a fire, the ambient humidity may decrease due to localized combustion, and this sensor is used to help determine abnormal trends.

[0149] Edge nodes are installed in kitchen electrical control cabinets or equipment cabinets, and have local computing and caching capabilities. They are used to integrate environmental signals collected by various sensors into multi-source environmental signals, and to perform signal preprocessing and time synchronization on the multi-source environmental signals. It should be noted that the edge nodes and the microservice unit integration module establish a secure communication connection through a gateway, and use lightweight message transmission protocols (such as MQTT / CoAP protocol) to achieve reliable transmission of multi-source environmental signals, environmental steady-state characteristics and fire early warning information.

[0150] By deploying the aforementioned sensor layer and edge nodes, the computational burden on the microservice unit integration module is reduced, while ensuring the real-time performance and reliability of kitchen fire early warning.

[0151] It should be understood that the embodiments described are for illustrative purposes only and are not limited to this structure in terms of the scope of the patent invention.

[0152] It should be noted that the sequence numbers of the above embodiments of the present invention are merely for descriptive purposes and do not represent the superiority or inferiority of the embodiments. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, apparatus, article, or method. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0153] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0154] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A microservice-based internet platform for early warning of kitchen fires, characterized in that, The aforementioned kitchen fire early warning internet platform includes the following kitchen fire early warning methods: S1: Edge nodes collect multi-source environmental signals in the kitchen environment in real time, and perform signal preprocessing and time synchronization processing on the multi-source environmental signals to obtain preprocessed multi-source environmental signals; S2: Extract airflow signal data from the preprocessed multi-source environmental signals, and extract airflow disturbance features from the airflow signal data; The airflow signal data from the preprocessed multi-source environmental signal is acquired. The airflow signal data is then segmented into multiple airflow signal segments according to a fixed-length time window R. Local feature extraction is performed on each airflow signal segment to obtain disturbance intensity features, local velocity change rate features, and sustained disturbance estimation features. These features are then sorted according to their order within the airflow signal data and used as the airflow disturbance features. The process for local feature extraction of the airflow signal is as follows: S21: Calculate the standard deviation of the data values ​​in the airflow signal, perform a fast Fourier transform on the airflow signal, extract the high-frequency energy of the airflow signal in the high-frequency range, and then fused the high-frequency energy with the standard deviation after logarithmic compression to serve as the disturbance intensity feature of the airflow signal. S22: The average rate of change of data values ​​in the airflow signal is calculated using the adjacent difference method, which serves as the local velocity change rate characteristic of the airflow signal; the calculation formula for the local velocity change rate characteristic is as follows: ； in, This represents the local velocity change rate characteristic of the airflow signal in segment h, where H represents the total number of airflow segments. Let L represent the e-th signal value in the h-th segment of the airflow signal, and L represent the number of data values ​​in each segment of the airflow signal. S23: By integrating disturbance intensity characteristics, local velocity change rate characteristics, and incorporating the change in the average data value of adjacent airflow signals, a persistent disturbance estimation feature is generated to characterize whether the current airflow signal is persistent; the formula for generating the persistent disturbance estimation feature is as follows: ； in, This represents the disturbance intensity characteristics of the airflow signal in segment h. This represents the sustained disturbance estimation characteristics of the airflow signal in segment h. This represents the average data value of the airflow signal in segment h. This represents the average data value of the airflow signal in segment h-1. This represents an exponential function with the natural constant as its base. All represent weighting coefficients; S3: Dynamically compensate the preprocessed multi-source environmental signals based on airflow disturbance characteristics, and extract the environmental steady-state characteristics of the dynamically compensated multi-source environmental signals; S4: Use the fire early warning model to assess the fire risk of the steady-state characteristics of the environment, generate fire early warning information based on the fire risk assessment results and send it to the edge nodes. The edge nodes adjust the sampling frequency and upload frequency of the multi-source environmental signals synchronously based on the fire early warning information.

2. The microservice-based kitchen fire early warning internet platform as described in claim 1, characterized in that, Collect multi-source environmental signals from the kitchen environment, and perform signal preprocessing and time synchronization processing on the multi-source environmental signals, including: The multi-source environmental signal data includes temperature signal data, smoke signal data, airflow signal data, and humidity signal data; Edge nodes perform signal preprocessing on multi-source environmental signals to obtain preprocessed multi-source environmental signals. The signal preprocessing includes noise filtering, outlier removal, and standardization. Time synchronization processing is performed on the multi-source environmental signals after signal preprocessing. The time synchronization processing flow is as follows: The sensor acquires the unified reference time built into the edge node. During the process of uploading signal data, the sensor clock is uploaded synchronously, and the deviation of the sensor clock relative to the unified reference time is calculated. The acquisition time of data values ​​in multi-source environmental signals is obtained. Based on the deviation of the sensor clock associated with the data value from the unified reference time, the acquisition time is corrected to eliminate the clock offset of the sensor clock and obtain the preprocessed multi-source environmental signal.

3. The microservice-based kitchen fire early warning internet platform as described in claim 1, characterized in that, The calculation process for the disturbance intensity characteristics of the airflow signal in step S21 is as follows: Performing a Fast Fourier Transform on the airflow signal yields the complex frequency domain signals of the airflow signal at different frequency indices: ； in, This represents the complex frequency domain signal of the h-th segment of the airflow signal at frequency index u. Represents the imaginary unit. , This represents the e-th signal value in the h-th segment of the airflow signal, where L represents the number of data values ​​in each segment of the airflow signal. This represents an exponential function with the natural constant as the base, and H represents the total number of segments in the airflow signal; High-frequency range is set by cutoff frequency. : ； in, This indicates the sampling frequency of the airflow sensor that collects airflow signal data. Indicates the first cutoff frequency; Based on high frequency range Extracting high-frequency energy from airflow signals in the high-frequency range: ； in, This represents the high-frequency energy of the airflow signal in the h-th segment within the high-frequency range; The high-frequency energy is logarithmically compressed and then fused with the standard deviation to serve as the disturbance intensity characteristic of the airflow signal.

4. The microservice-based kitchen fire early warning internet platform as described in claim 1, characterized in that, Dynamic compensation of preprocessed multi-source environmental signals based on airflow disturbance characteristics includes: The temperature signal data, smoke signal data and humidity signal data in the preprocessed multi-source environmental signal are segmented according to a fixed-length time window R, and multiple segments of temperature signal, smoke signal and humidity signal are obtained in turn. Based on the disturbance intensity characteristics, local velocity change rate characteristics, and continuous disturbance estimation characteristics of each airflow signal segment, dynamic compensation processing is performed on the data values ​​of the corresponding segment's temperature signal, smoke signal, and humidity signal. The dynamically compensated temperature signal, smoke signal, and humidity signal are then sequentially spliced ​​together to obtain dynamically compensated temperature signal data, smoke signal data, and humidity signal data. The dynamically compensated temperature signal data, smoke signal data, humidity signal data, and airflow signal data are used as the dynamically compensated multi-source environmental signal.

5. A microservice-based kitchen fire early warning internet platform as described in claim 4, characterized in that, Extracting the environmental steady-state characteristics of the dynamically compensated multi-source environmental signals, including: The mean and standard deviation of temperature, smoke, humidity and airflow signals in the multi-source environmental signals after dynamic compensation are calculated. The least squares method is used to fit the linear slope of the temperature, smoke, humidity and airflow signals respectively to obtain the linear slope of the temperature, smoke, humidity and airflow signals. Based on the mean, standard deviation, and linear slope of temperature, smoke, humidity, and airflow signal data, the environmental steady-state characteristics of the multi-source environmental signals after dynamic compensation are constructed. .

6. A microservice-based kitchen fire early warning internet platform as described in claim 1, characterized in that, The fire early warning model is a gradient boosting tree model, which consists of multiple weak regression trees. The weak regression trees are connected in sequence and receive environmental steady-state features respectively. The weighted sum of the outputs of the leaf nodes in all weak regression trees is calculated as the fire risk assessment result. Each weak regression tree consists of a root node, multiple internal nodes, and leaf nodes.

7. A microservice-based kitchen fire early warning internet platform as described in claim 6, characterized in that, A fire risk assessment is conducted on the steady-state characteristics of the environment using a fire early warning model. Based on the fire risk assessment results, fire early warning information is generated and distributed to edge nodes, including: The fire risk assessment result is in numerical form between 0 and 1, where a higher fire risk assessment result indicates a greater probability of a fire occurring in the kitchen. Fire warning information is generated based on the fire risk assessment result, and the method for generating the fire warning information is as follows: ； in, Indicates the results of the fire risk assessment. Corresponding fire warning information, This indicates the first warning threshold. This indicates the second warning threshold. This indicates the third warning threshold, where , This indicates the first fire warning information. This indicates the second fire warning information. This indicates the third fire warning information.

Images