Intelligent fire risk assessment and prevention system
Through intelligent fire risk assessment and prevention systems, combined with environmental and radio spectrum data, using deep belief networks to conduct fire risk assessment and development path prediction, the problem that traditional systems cannot predict the location and fire in advance is solved, and the early stage of fire monitoring is achieved Detect and precise positioning, alert in advance and optimize emergency response.
Patent Information
- Application Number
- CN202410934521.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-12
- Publication Date
- 2024-11-15
AI Technical Summary
Traditional fire monitoring systems cannot achieve accurate positioning of fire sources and early prediction of fire situations, resulting in delayed alarms and unable to effectively reduce the losses of fires to personnel and property.
Adopt intelligent fire risk assessment and prevention system, combined with environmental monitoring module, radio spectrum monitoring and intervention module, data analysis module, fire source positioning and isolation module, and early warning generation module, to collect temperature, humidity, smoke concentration and air flow data in real time , analyze abnormal fluctuations in the radio spectrum, use deep belief networks to predict fire risks and development paths, and automatically initiate fire prevention measures.
It realizes early detection and precise positioning of fires, early alarm time, dynamic adjustment of evacuation routes and fire extinguishing strategies, improves emergency response efficiency, and reduces fire losses and impacts.
Smart Images

Figure CN118968694A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of fire prevention, and in particular to an intelligent fire risk assessment and prevention system. Background Art
[0002] In modern building environments, timely detection and early warning of fires are crucial and can significantly reduce casualties and property losses. Studies have shown that when a fire occurs, specific radio frequencies will be interfered with due to the influence of temperature, smoke and flames, while traditional fire monitoring systems mainly rely on smoke detectors and temperature sensors to alarm fires.
[0003] At present, traditional fire monitoring systems can usually sound an alarm within a few minutes after a fire occurs. However, due to the complexity of building space and the variability of fire sources, these traditional methods often fail to accurately locate the fire source and predict the fire early. Summary of the invention
[0004] Based on the above objectives, the present invention provides an intelligent fire risk assessment and prevention system.
[0005] The intelligent fire risk assessment and prevention system includes an environmental monitoring module, a radio spectrum monitoring and intervention module, a data analysis module, a fire source location and isolation module, and an early warning generation module, among which;
[0006] The environmental monitoring module collects environmental data in real time, including temperature, humidity, smoke concentration and air flow;
[0007] The radio spectrum monitoring and intervention module analyzes abnormal fluctuations in the radio spectrum, including signal attenuation caused by high temperature and smoke, for early fire detection;
[0008] The data analysis module is connected with the environment monitoring module and the radio spectrum monitoring and intervention module, and uses the collected environment data and radio frequency data to analyze fire risks and predict the fire development path;
[0009] The fire source location and isolation module locates the fire source according to the radio spectrum data and the environmental data. After locating the fire source, the fire source location and isolation module automatically initiates fire prevention measures, including closing the automatic fire door or activating the local water spray device;
[0010] The warning generation module generates warning information according to the risk assessment of the data analysis module and the fire source location result of the fire source location and isolation module.
[0011] Furthermore, the environmental monitoring module specifically includes:
[0012] Temperature sensor: Thermistor sensor is used to measure the temperature of the air;
[0013] Humidity sensor: A resistive humidity sensor is used to measure the moisture content in the air. The resistive humidity sensor can provide real-time data on ambient humidity, which is particularly important for predicting fire risks.
[0014] Photoelectric smoke sensor: Photoelectric smoke sensor detects smoke particles in the air through scattering or absorption of light beams;
[0015] Hot wire anemometer: Use hot wire anemometer to monitor air flow.
[0016] Furthermore, the radio spectrum monitoring and intervention module specifically includes:
[0017] Radio frequency monitoring: The radio spectrum monitoring and intervention module is equipped with a radio frequency scanner to monitor radio signals within the frequency range. The radio frequency scanner scans the radio waves in the environment and captures radio frequency data.
[0018] Signal Analysis and Anomaly Detection: The Radio Spectrum Monitoring and Intervention module uses signal processing algorithms to analyze the collected radio frequency data, including frequency analysis, signal attenuation, and noise level monitoring.
[0019] Furthermore, the radio spectrum monitoring and intervention module in the signal analysis and anomaly detection uses statistical models and anomaly detection algorithms to analyze the collected radio frequency data, specifically including:
[0020] Establish a baseline model of the signal: Collect radio frequency data under normal conditions and establish a baseline model of the radio signal, including calculating the average power spectrum density, average signal strength and frequency distribution of the radio signal under no-fire conditions;
[0021] Real-time signal monitoring and data collection: The radio spectrum monitoring and intervention module collects radio frequency data in the environment, including measuring the real-time power spectrum density, signal attenuation and noise level of radio signals;
[0022] Anomaly Detection Algorithm: A Z-score detection algorithm is used to identify anomalies in radio signals, calculating the number of standard deviations of each measurement from the baseline model.
[0023] Furthermore, the establishment of the baseline model of the signal specifically includes:
[0024] Data collection: Radio frequency data under normal conditions without fire is collected through radio spectrum scanners, covering a variety of environmental conditions and time periods to ensure the comprehensiveness of radio frequency data;
[0025] Data preprocessing: The collected radio frequency data are preprocessed, including filtering out noise, filling in missing values, and calibrating differences between devices;
[0026] Calculate statistical indicators: Establishing the baseline model includes calculating statistical indicators of average power spectral density, average signal strength and frequency distribution;
[0027] Establish a baseline model: The calculated average power spectral density, average signal strength and frequency distribution are integrated to form a baseline model, which reflects the behavior of radio signals under no-fire conditions.
[0028] Furthermore, the anomaly detection algorithm specifically includes:
[0029] Calculate the baseline model mean and standard deviation, mean: Among them, x i is the i-th data point in the baseline dataset, and n is the total number of data points;
[0030] Standard Deviation: Among them, x i is the i-th data point in the baseline dataset, and μ is the mean of these data points;
[0031] Calculate the Z-score: Calculate the Z-score of the new measurement X to see how it deviates from the baseline model;
[0032] The Z-score detection algorithm is calculated as:
[0033] Where X is the new measurement, μ is the mean of the baseline model, and σ is the standard deviation of the baseline model.
[0034] Furthermore, the data analysis module specifically includes:
[0035] Feature extraction: Extract variables that identify fire characteristics from environmental data and radio frequency data, including temperature rise rate and smoke concentration changes in environmental data, and signal attenuation and spectrum changes in radio frequency data;
[0036] Statistical modeling: Building predictive models using machine learning methods based on selected features;
[0037] Risk assessment: Use the established prediction model to conduct real-time assessment of current environmental data and radio frequency data to calculate the risk level of fire. The prediction model outputs risk ratings, including low, medium, and high levels;
[0038] Fire Growth Prediction: After detecting initial signs of a fire, use predictive models to predict the path and speed of the fire's spread.
[0039] Furthermore, the statistical modeling uses a deep belief network algorithm to establish a prediction model. The deep belief network is composed of multiple layers of restricted Boltzmann machines. The deep belief network specifically includes:
[0040] Build and train a deep belief network: The structure of a deep belief network consists of multiple layers of restricted Boltzmann machines. Each restricted Boltzmann machine is trained layer by layer using contrastive divergence, allowing each restricted Boltzmann machine to learn and reconstruct the distribution of its input data. After all restricted Boltzmann machine layers are pre-trained by contrastive divergence, the entire network is fine-tuned through back-propagation to optimize the final prediction performance.
[0041] The basic formula of the restricted Boltzmann machine includes the energy function and the activation probability, where;
[0042] The energy function is calculated as: E(v, h) = -∑ i a i v i -∑ j b j h j -∑ i ∑ j v i w ij h j , where v is the visible layer unit, h is the hidden layer unit, and a i and b j are the biases of the visible layer and the hidden layer, respectively, and w ij is the connection weight;
[0043] The activation probability is calculated as: p(h j =1|v)=σ(b j +∑ i v i w ij );
[0044] p(v i =1|h)=σ(a i +∑ j h j w ij ),in, is the sigmoid function;
[0045] Extracting features from a deep belief network: The deep belief network automatically learns features of the input data through its multi-layer structure. Features include patterns of temperature changes, abnormal increases in smoke concentration, and changes in radio frequency signals.
[0046] Risk assessment and development path prediction: Add a classification layer on top of the deep belief network. The classification layer is a logistic regression layer. The classification layer uses the features learned from the bottom layer of the deep belief network to predict fire and development paths.
[0047] The classification layer (logistic regression layer) is calculated as: z = W T x+b;
[0048] Where W is the weight vector, x is the input feature vector, and b is the bias term;
[0049] Sigmoid function:
[0050] p(y=1|x) represents the probability of fire occurrence;
[0051] Calculation of risk level: According to the probability of fire occurrence, it is converted into risk level;
[0052] The fire development prediction adopts a method combining physical model and prediction model to predict the expansion path and speed of the fire, specifically including:
[0053] Physical model: Uses a physical model based on fire dynamics, flame propagation equations and heat conduction equations, Where T is temperature, α is the thermal diffusivity, and Q is the heat source term;
[0054] The combination of physical model and prediction model is expressed as:
[0055]
[0056] Here, f is the regression function learned from the deep belief network and β is the tuning parameter.
[0057] Furthermore, the fire source location and isolation module specifically includes:
[0058] Data fusion: Use Kalman filter to integrate environmental data and radio spectrum data to form an environmental perception model;
[0059] The Kalman filter calculation is:
[0060] in, is the state estimate, K k is the Kalman gain, z k is the observed value, H is the observation model;
[0061] Fire source location algorithm: Use triangulation to determine the specific location of the fire source, and estimate the distance from the fire source to each sensor by analyzing the sensor's temperature, smoke concentration, and radio signal attenuation data;
[0062] There are three sensors, located at (x A ,y A ),(x B ,y B ),(x C ,y C ), the fire source position is (x, y), and the measured distances are d A ,dB ,d C , establish the equation system:
[0063]
[0064] Solve the system of equations to obtain the location of the fire source (x,y).
[0065] Furthermore, the warning generation module specifically includes:
[0066] Data input: receiving fire risk assessment results from the data analysis module, including the probability and risk level of fire occurrence, and fire source location and fire development information from the fire source location and isolation module;
[0067] Generate early warning information: Determine the level of early warning based on the fire risk assessment results and fire source location information;
[0068] Early warning information release: Select channels to release early warning information to ensure that the information can be quickly transmitted to relevant personnel and systems. Channels include SMS, email, broadcast system and fire system interface.
[0069] This invention combines environmental monitoring data and radio frequency data, and uses a deep belief network to perform in-depth data analysis and feature extraction, so that the fire monitoring system can accurately detect potential fire sources in the early stages of fire formation. Compared with traditional systems that only rely on smoke or temperature sensors, this method can identify fire warning signals more quickly, thereby significantly advancing the alarm time and buying valuable time for taking emergency response measures:
[0070] The present invention uses data from multiple sensors to accurately calculate the location of the fire source through algorithms and monitor the dynamic changes of the fire. This precise positioning and real-time tracking capability not only enhances the emergency response team's spatial cognition of the fire, but also dynamically adjusts the evacuation route and fire-fighting strategy according to the specific situation of the fire development, effectively reducing the threat of fire to personnel safety and building structures.
[0071] The present invention, through real-time data analysis and prediction models, can predict the expansion path and speed of the fire, thereby allowing the emergency management department to deploy fire prevention, rescue and evacuation resources in advance. This optimized resource allocation and targeted emergency response strategy not only improves the efficiency of fire fighting and rescue, but also greatly reduces the losses and impacts that may be caused by the fire. BRIEF DESCRIPTION OF THE DRAWINGS
[0072] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0073] Figure 1 A schematic diagram of a system flow of an embodiment of the present invention;
[0074] Figure 2 The figure is a flow chart of a radio spectrum monitoring and intervention module according to an embodiment of the present invention. DETAILED DESCRIPTION
[0075] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments.
[0076] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.
[0077] like Figure 1-Figure 2 As shown, the intelligent fire risk assessment and prevention system includes an environmental monitoring module, a radio spectrum monitoring and intervention module, a data analysis module, a fire source location and isolation module, and an early warning generation module, wherein;
[0078] The environmental monitoring module collects environmental data in real time, including temperature, humidity, smoke concentration and air flow;
[0079] The Radio Spectrum Monitoring and Intervention module analyzes abnormal fluctuations in the radio spectrum, including signal attenuation caused by heat and smoke, for early fire detection;
[0080] The data analysis module is connected with the environmental monitoring module and the radio spectrum monitoring and intervention module to analyze fire risks and predict fire development paths using the collected environmental data and radio frequency data;
[0081] The fire source location and isolation module locates the fire source according to the radio spectrum data and the environmental data. After locating the fire source, the fire source location and isolation module automatically initiates fire prevention measures, including closing the automatic fire door or activating the local water spray device;
[0082] The warning generation module generates warning information based on the risk assessment of the data analysis module and the fire source location results of the fire source location and isolation module. This module is responsible for issuing warnings through various communication means to ensure timely communication and wide coverage of information.
[0083] The environmental monitoring module specifically includes:
[0084] Temperature sensor: Thermistor sensor is used to measure the temperature of the air;
[0085] Humidity sensor: A resistive humidity sensor is used to measure the moisture content in the air. The resistive humidity sensor can provide real-time data on ambient humidity, which is particularly important for predicting fire risks.
[0086] Photoelectric smoke sensor: Photoelectric smoke sensor detects smoke particles in the air through scattering or absorption of light beams;
[0087] Hot wire anemometer: Hot wire anemometer is used to monitor air flow. Hot wire anemometer measures the speed and direction of air flow, which is critical for analyzing and predicting the path of fire spread.
[0088] The radio spectrum monitoring and intervention module specifically includes:
[0089] Radio frequency monitoring: The radio spectrum monitoring and intervention module is equipped with a radio frequency scanner to monitor radio signals within the frequency range. The radio frequency scanner scans the radio waves in the environment and captures radio frequency data.
[0090] Signal Analysis and Anomaly Detection: The Radio Spectrum Monitoring and Intervention module uses signal processing algorithms to analyze the collected radio frequency data, including frequency analysis, signal attenuation, and noise level monitoring;
[0091] When a fire occurs, the release of smoke, dust and gas, as well as the increase in temperature, causes changes in the propagation characteristics of radio waves, such as signal attenuation or enhanced multipath effects. The radio spectrum monitoring and intervention module identifies these changes by comparing real-time data with historical data or a preset baseline.
[0092] The Radio Spectrum Monitoring and Intervention module in Signal Analysis and Anomaly Detection uses statistical models and anomaly detection algorithms to analyze the collected radio frequency data, including:
[0093] Establish a baseline model of the signal: Collect radio frequency data under normal conditions and establish a baseline model of the radio signal, including calculating the average power spectrum density, average signal strength and frequency distribution of the radio signal under no-fire conditions;
[0094] Real-time signal monitoring and data collection: The radio spectrum monitoring and intervention module collects radio frequency data in the environment, including measuring the real-time power spectrum density, signal attenuation and noise level of radio signals;
[0095] Anomaly Detection Algorithm: A Z-score detection algorithm is used to identify anomalies in radio signals, calculating the number of standard deviations of each measurement from the baseline model.
[0096] Establishing a baseline model for a signal specifically includes:
[0097] Data collection: Radio frequency data under normal conditions without fire is collected through radio spectrum scanners, covering a variety of environmental conditions and time periods to ensure the comprehensiveness of radio frequency data;
[0098] Data preprocessing: The collected radio frequency data are preprocessed, including filtering out noise, filling in missing values, and calibrating differences between devices;
[0099] Calculate statistical indicators: Establishing the baseline model includes calculating statistical indicators of average power spectral density, average signal strength and frequency distribution;
[0100] The average power spectral density describes how the power of the signal is distributed in the frequency domain. The signal is converted from the time domain to the frequency domain through Fourier transform, and then the power of each frequency component is calculated and its average value is taken. Fourier transform: In practical applications, fast Fourier transform (FFT) is usually used to process digital signals and calculate power: Where |X(f| is the modulus of the FFT result at frequency f, N is the number of samples, and the average power spectral density is: Where M is the number of measurement windows, P i (f) is the power of the ith window at frequency f;
[0101] Average signal strength: Average signal strength is the average energy or power level of a signal over a period of time. It is usually obtained by calculating the square of the average amplitude of the signal. Average signal strength: Where |x(n)| is the signal amplitude at time n;
[0102] Frequency distribution: Frequency distribution indicates the relative frequency of each frequency component in the signal, which involves calculating the number of occurrences of each frequency component and normalizing it. The statistics of frequency components are: Among them, fi is the frequency of the i-th sample, 1 is the indicator function, when f i =f takes the value of 1, otherwise it takes the value of 0. Frequency distribution: Where N is the total number of samples;
[0103] Establish a baseline model: Integrate the calculated average power spectrum density, average signal strength and frequency distribution to form a baseline model. The baseline model reflects the behavior of radio signals under non-fire conditions. The expression of the baseline model is:
[0104] Baseline Model = {PSD(f), S avg ,F(f)};
[0105] Where PSD(f) is the average power spectral density at different frequencies f, S avg is the average signal strength, and F(f) is the relative frequency of occurrence of frequency f.
[0106] The anomaly detection algorithms specifically include;
[0107] Calculate the baseline model mean and standard deviation, mean: Among them, x i is the i-th data point in the baseline dataset, and n is the total number of data points;
[0108] Standard Deviation: Among them, x i is the i-th data point in the baseline dataset, and μ is the mean of these data points;
[0109] Calculate the Z-score: Calculate the Z-score of the new measurement X to see how it deviates from the baseline model;
[0110] The Z-score detection algorithm is calculated as:
[0111] Where X is the new measurement, μ is the mean of the baseline model, and σ is the standard deviation of the baseline model;
[0112] Example:
[0113] Assuming that in the absence of a fire, the average radio signal strength μ of a specific frequency is -85 dBm, and the standard deviation σ is 3 dBm. If the monitoring system records a reading of signal strength X = -95 dBm, the Z-score can be calculated as follows:
[0114]
[0115] The calculated Z-score Z = -3.33 is much smaller than -2, which indicates that the signal strength is significantly lower than the normal range, which may indicate a fire.
[0116] The data analysis module specifically includes:
[0117] Feature extraction: Extract variables that identify fire characteristics from environmental data and radio frequency data, including temperature rise rate and smoke concentration changes in environmental data, and signal attenuation and spectrum changes in radio frequency data;
[0118] Statistical modeling: Building predictive models using machine learning methods based on selected features;
[0119] Risk assessment: Use the established prediction model to conduct real-time assessment of current environmental data and radio frequency data to calculate the risk level of fire. The prediction model outputs risk ratings, including low, medium, and high levels;
[0120] Fire development prediction: After detecting the initial signs of a fire, use predictive models to predict the path and speed of the fire’s spread;
[0121] Through such a comprehensive analysis and prediction process, the data analysis module can effectively utilize the data collected from relevant monitoring modules to not only identify fire risks in advance, but also predict the development trend of fires, thereby providing more effective early warning and response measures, which is crucial for protecting the safety of life and property, optimizing resource allocation and improving emergency response efficiency.
[0122] Statistical modeling uses a deep belief network algorithm to build a prediction model. The deep belief network is composed of multiple layers of restricted Boltzmann machines. In the context of fire detection and prediction, the deep belief network can learn complex representations from environmental data and radio frequency data to help identify potential fire risks. The deep belief network specifically includes:
[0123] Build and train a deep belief network: The structure of a deep belief network consists of multiple layers of restricted Boltzmann machines. Each restricted Boltzmann machine is trained layer by layer using contrastive divergence, allowing each restricted Boltzmann machine to learn and reconstruct the distribution of its input data. After all restricted Boltzmann machine layers are pre-trained by contrastive divergence, the entire network is fine-tuned through back-propagation to optimize the final prediction performance.
[0124] The basic formula of the restricted Boltzmann machine includes the energy function and the activation probability, where;
[0125] The energy function is calculated as: E(v, h) = -∑ i a i v i -∑ j b j h j -∑ i ∑ j v i w ijh j , where v is the visible layer unit, h is the hidden layer unit, and a i and b j are the biases of the visible layer and the hidden layer, respectively, and w ij is the connection weight;
[0126] The activation probability is calculated as: p(h j =1|v)=σ(b j +∑ i v i w ij );
[0127] p(v i =1|h)=σ(a i +∑ j h j w ij ),in, is the sigmoid function;
[0128] Contrastive divergence includes forward propagation, reconstruction of visible layers, back propagation, and updating weights, where;
[0129] The forward propagation calculation is: p(h j =1|v)=σ(b j +∑ i v i w ij );
[0130] Use this probability to sample h j ;
[0131] The reconstructed visible layer is calculated as: p(v i ′=1|h)=σ(a i +∑ j h j w ij );
[0132] Use this probability to sample the reconstructed visible layer v i ′;
[0133] The back propagation calculation is: p(h j ′=1|v′)=σ(b j +∑ i v i ′w ij );
[0134] Use this probability to sample the reconstructed hidden layer h j ′;
[0135] The updated weight is calculated as: Δw ij =∈( <v i hj > data - <v i h j > recon );
[0136] Where ∈ is the learning rate, <·> data and recon ;
[0137] Extracting features from a deep belief network: The deep belief network automatically learns features of the input data through its multi-layer structure. Features include patterns of temperature changes, abnormal increases in smoke concentration, and changes in radio frequency signals.
[0138] Risk assessment and development path prediction: Add a classification layer on top of the deep belief network. The classification layer is a logistic regression layer. The classification layer uses the features learned from the bottom layer of the deep belief network to predict fire and development paths.
[0139] The classification layer (logistic regression layer) is calculated as: z = W T x+b;
[0140] Where W is the weight vector, x is the input feature vector, and b is the bias term;
[0141] Sigmoid function:
[0142] p(y=1|x) represents the probability of fire occurrence;
[0143] Calculation of risk level: According to the probability of fire occurrence, it is converted into risk level;
[0144] For example: Low risk: if p(y=1|x)<0.3;
[0145] Medium risk: if 0.3≤p(y=1|x)<0.7;
[0146] High risk: if p(y=1|x)≥0.7;
[0147] Fire development prediction uses a combination of physical models and predictive models to predict the path and speed of fire expansion, including:
[0148] Physical model: Uses a physical model based on fire dynamics, flame propagation equations and heat conduction equations, Where T is temperature, α is the thermal diffusivity, and Q is the heat source term;
[0149] The combination of physical model and prediction model is expressed as:
[0150]
[0151] Among them, f is the regression function learned from the deep belief network, and β is the adjustment parameter;
[0152] Through these steps and formulas, the deep belief network can effectively learn and extract useful features from environmental data and radio frequency data, thereby realizing the analysis of fire risks and the prediction of development paths.
[0153] The fire source location and isolation module specifically includes:
[0154] Data fusion: Use Kalman filter to integrate environmental data and radio spectrum data to form an environmental perception model;
[0155] The Kalman filter calculation is:
[0156] in, is the state estimate, K k is the Kalman gain, z k is the observed value, H is the observation model;
[0157] Fire source location algorithm: Use triangulation to determine the specific location of the fire source, and estimate the distance from the fire source to each sensor by analyzing the sensor's temperature, smoke concentration, and radio signal attenuation data;
[0158] There are three sensors, located at (x A ,y A ),(x B ,y B ),(x C ,y C ), the fire source position is (x, y), and the measured distances are d A ,d B ,d C , establish the equation system:
[0159]
[0160] Solve the system of equations to obtain the location of the fire source (x,y).
[0161] The warning generation module specifically includes:
[0162] Data input: receiving fire risk assessment results from the data analysis module, including the probability and risk level of fire occurrence, and fire source location and fire development information from the fire source location and isolation module;
[0163] Generate early warning information: Determine the level of early warning based on the fire risk assessment results and fire source location information;
[0164] Example of rule:
[0165] Low risk: The probability of fire is low and the fire source is far away from important areas;
[0166] Medium risk: The probability of fire is medium, the fire source is close to important areas or the fire has a tendency to spread;
[0167] High risk: The probability of fire is high, the fire source is in a critical area or the fire spreads rapidly;
[0168] Early warning information release: Select channels to release early warning information to ensure that the information can be quickly transmitted to relevant personnel and systems. Channels include SMS, email, broadcast system and fire system interface.
[0169] Those skilled in the art should understand that the discussion of any of the above embodiments is merely illustrative and is not intended to imply that the scope of the present invention is limited to these examples. Under the concept of the present invention, the technical features in the above embodiments or different embodiments may be combined, the steps may be implemented in any order, and there are many other variations of the different aspects of the present invention as described above, which are not provided in detail for the sake of simplicity.
[0170] The present invention is intended to cover all such substitutions, modifications and variations that fall within the broad scope of the claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.
Claims
1. Intelligent fire risk assessment and prevention system, It is characterized in that It includes an environmental monitoring module, a radio spectrum monitoring and intervention module, a data analysis module, a fire source location and isolation module, and an early warning generation module, among which; The environmental monitoring module collects environmental data in real time, including temperature, humidity, smoke concentration and air flow; The radio spectrum monitoring and intervention module analyzes abnormal fluctuations in the radio spectrum, including signal attenuation caused by high temperature and smoke, for early fire detection; The data analysis module is connected with the environment monitoring module and the radio spectrum monitoring and intervention module, and uses the collected environment data and radio frequency data to analyze fire risks and predict the fire development path; The fire source location and isolation module locates the fire source according to the radio spectrum data and the environmental data. After locating the fire source, the fire source location and isolation module automatically initiates fire prevention measures, including closing the automatic fire door or activating the local water spray device; The warning generation module generates warning information according to the risk assessment of the data analysis module and the fire source location result of the fire source location and isolation module.
2. The intelligent fire risk assessment and prevention system according to claim 1, It is characterized in that The environmental monitoring module specifically includes: Temperature sensor: Thermistor sensor is used to measure the temperature of the air; Humidity sensor: A resistive humidity sensor is used to measure the moisture content in the air; Photoelectric smoke sensor: Photoelectric smoke sensor detects smoke particles in the air through scattering or absorption of light beams; Hot wire anemometer: Use hot wire anemometer to monitor air flow.
3. The intelligent fire risk assessment and prevention system according to claim 1, It is characterized in that The radio spectrum monitoring and intervention module specifically includes: Radio frequency monitoring: The radio spectrum monitoring and intervention module is equipped with a radio frequency scanner to monitor radio signals within the frequency range. The radio frequency scanner scans the radio waves in the environment and captures radio frequency data. Signal Analysis and Anomaly Detection: The Radio Spectrum Monitoring and Intervention module uses signal processing algorithms to analyze the collected radio frequency data, including frequency analysis, signal attenuation, and noise level monitoring.
4. The intelligent fire risk assessment and prevention system according to claim 3, It is characterized in that The radio spectrum monitoring and intervention module in the signal analysis and anomaly detection uses statistical models and anomaly detection algorithms to analyze the collected radio frequency data, specifically including: Establish a baseline model of the signal: Collect radio frequency data under normal conditions and establish a baseline model of the radio signal, including calculating the average power spectrum density, average signal strength and frequency distribution of the radio signal under no-fire conditions; Real-time signal monitoring and data collection: The radio spectrum monitoring and intervention module collects radio frequency data in the environment, including measuring the real-time power spectrum density, signal attenuation and noise level of radio signals; Anomaly Detection Algorithm: A Z-score detection algorithm is used to identify anomalies in radio signals, calculating the number of standard deviations of each measurement from the baseline model.
5. The intelligent fire risk assessment and prevention system according to claim 4, It is characterized in that The signal baseline model is established by: Data collection: Radio frequency data under normal conditions without fire is collected through radio spectrum scanners, covering a variety of environmental conditions and time periods to ensure the comprehensiveness of radio frequency data; Data preprocessing: The collected radio frequency data are preprocessed, including filtering out noise, filling in missing values, and calibrating differences between devices; Calculate statistical indicators: Establishing the baseline model includes calculating statistical indicators of average power spectral density, average signal strength and frequency distribution; Establish a baseline model: The calculated average power spectral density, average signal strength and frequency distribution are integrated to form a baseline model, which reflects the behavior of radio signals under no-fire conditions.
6. The intelligent fire risk assessment and prevention system according to claim 5, It is characterized in that The anomaly detection algorithm specifically includes: Calculate the baseline model mean and standard deviation, mean: Among them, x i is the i-th data point in the baseline dataset, and n is the total number of data points; Standard Deviation: Among them, x i is the i-th data point in the baseline dataset, and μ is the mean of these data points; Calculate the Z-score: Calculate the Z-score of the new measurement X to see how it deviates from the baseline model; The Z-score detection algorithm is calculated as: Where X is the new measurement, μ is the mean of the baseline model, and σ is the standard deviation of the baseline model.
7. The intelligent fire risk assessment and prevention system according to claim 1, It is characterized in that The data analysis module specifically includes: Feature extraction: Extract variables that identify fire characteristics from environmental data and radio frequency data, including temperature rise rate and smoke concentration changes in environmental data, and signal attenuation and spectrum changes in radio frequency data; Statistical modeling: Building predictive models using machine learning methods based on selected features; Risk assessment: Use the established prediction model to conduct real-time assessment of current environmental data and radio frequency data to calculate the risk level of fire. The prediction model outputs risk ratings, including low, medium, and high levels; Fire Growth Prediction: After detecting initial signs of a fire, use predictive models to predict the path and speed of the fire's spread.
8. The intelligent fire risk assessment and prevention system according to claim 7, It is characterized in that The statistical modeling uses a deep belief network algorithm to establish a prediction model. The deep belief network is composed of multiple layers of restricted Boltzmann machines. The deep belief network specifically includes: Build and train a deep belief network: The structure of a deep belief network consists of multiple layers of restricted Boltzmann machines. Each restricted Boltzmann machine is trained layer by layer using contrastive divergence, allowing each restricted Boltzmann machine to learn and reconstruct the distribution of its input data. After all restricted Boltzmann machine layers are pre-trained by contrastive divergence, the entire network is fine-tuned through back-propagation to optimize the final prediction performance. The basic formula of the restricted Boltzmann machine includes the energy function and the activation probability, where; The energy function is calculated as: E(v, h) = -∑ i a i v i -∑ j b j h j -∑ i ∑ j v i w ij h j , where v is the visible layer unit, h is the hidden layer unit, and a i and b j are the biases of the visible layer and the hidden layer, respectively, and w ij is the connection weight; The activation probability is calculated as: p(h j =1|v)=σ(b j +∑ i v i w ij ); p(v i =1|h)=σ(a i +∑ j h j w ij ),in, is the sigmoid function; Extracting features from a deep belief network: The deep belief network automatically learns features of the input data through its multi-layer structure. Features include patterns of temperature changes, abnormal increases in smoke concentration, and changes in radio frequency signals. Risk assessment and development path prediction: Add a classification layer on top of the deep belief network. The classification layer is a logistic regression layer. The classification layer uses the features learned from the bottom layer of the deep belief network to predict fire and development paths. The classification layer (logistic regression layer) is calculated as: z = W T x+b; Where W is the weight vector, x is the input feature vector, and b is the bias term; Sigmoid function: p(y=1|x) represents the probability of fire occurrence; Calculation of risk level: According to the probability of fire occurrence, it is converted into risk level; The fire development prediction adopts a method combining physical model and prediction model to predict the expansion path and speed of the fire, specifically including: Physical model: Uses a physical model based on fire dynamics, flame propagation equations and heat conduction equations, Where T is temperature, α is the thermal diffusivity, and Q is the heat source term; The combination of physical model and prediction model is expressed as: Among them, f is the regression function learned from the deep belief network and β is the tuning parameter.
9. The intelligent fire risk assessment and prevention system according to claim 1, It is characterized in that The fire source location and isolation module specifically includes: Data fusion: Use Kalman filter to integrate environmental data and radio spectrum data to form an environmental perception model; Fire source location algorithm: Use triangulation to determine the specific location of the fire source, and estimate the distance from the fire source to each sensor by analyzing the sensor's temperature, smoke concentration, and radio signal attenuation data; There are three sensors, located at (x A ,y A ),(x B ,y B ),(x C ,y C ), the fire source position is (x, y), and the measured distances are d A ,d B ,d C , establish the equation system: Solve the system of equations to obtain the location of the fire source (x,y).
10. The intelligent fire risk assessment and prevention system according to claim 1, It is characterized in that The warning generation module specifically includes: Data input: receiving fire risk assessment results from the data analysis module, including the probability and risk level of fire occurrence, and fire source location and fire development information from the fire source location and isolation module; Generate early warning information: Determine the level of early warning based on the fire risk assessment results and fire source location information; Early warning information release: Select channels to release early warning information to ensure that the information can be quickly transmitted to relevant personnel and systems. Channels include SMS, email, broadcast system and fire system interface.
Citation Information
Cited By
Multi-source fire data fusion and fire analysis and prediction system
CN119323857A
Control method for intelligent fire prevention of flame-retardant material for power transmission and distribution complete switch
CN119541738A
Power distribution room remote monitoring and fire-fighting inspection system based on cloud platform
CN119656519A
Cost-efficient multi-mode fusion building and industrial structure real-time sensing monitoring method
CN120197023A
Forestry fire prevention early warning method and system based on Internet of Things
CN120388444A
