Fire monitoring and early warning system

Abstract

A fire monitoring and early warning system belongs to the field of fire fighting and comprises a pollution-resistant sensor, a central control system, a data filtering and processing module, a real-time monitoring platform, an intelligent alarm module and an environment monitoring module. The data filtering and processing module uses a machine learning model to identify and eliminate abnormal data, and filters out false signals caused by dust; the real-time monitoring platform transmits data of the pollution-resistant sensor to the central control system through a wireless network; the intelligent alarm module is provided with a threshold value and an alarm mechanism, and when the temperature or the smoke concentration exceeds a normal range, the intelligent alarm module automatically gives an alarm and notifies related personnel; and the environment monitoring module is used for monitoring the temperature, humidity and dust concentration of the factory environment so as to adjust an early warning threshold value in real time. According to the comprehensive fire monitoring and early warning system provided by the scheme, abnormal signal identification and dynamic threshold adjustment are carried out by combining various environmental monitoring data and utilizing a machine learning model.

Description

Technical Field

[0001] The present invention relates to the field of fire fighting, and more particularly to a fire monitoring and early warning system. Background Art

[0002] With the acceleration of industrialization, many factories are facing fire risks caused by high temperature and high dust environments. Existing fire monitoring technologies mostly rely on single sensor monitoring, such as temperature or smoke sensors. Although they can detect abnormal signals, the accuracy and reliability of these technologies are significantly affected in environments with high dust concentrations, which can easily lead to false alarms or missed alarms, posing hidden dangers to production safety. In addition, existing systems usually lack the ability to dynamically adapt to environmental changes and fail to adjust alarm thresholds in a timely manner, further reducing response efficiency. Summary of the invention

[0003] To address this problem, this solution proposes a comprehensive fire monitoring and early warning system that combines a variety of environmental monitoring data (such as temperature, humidity, and dust concentration) and uses machine learning models to identify abnormal signals and dynamically adjust thresholds. This method can not only effectively filter out false signals caused by dust, but also automatically adjust the early warning threshold according to real-time environmental changes, thereby improving the system's monitoring accuracy and response capabilities to ensure safe factory operation. Technical solution: A fire monitoring and early warning system, including pollution-resistant sensors, a central control system, a data filtering and processing module, a real-time monitoring platform, an intelligent alarm module, and an environmental monitoring module; The data filtering and processing module uses a machine learning model to identify and exclude abnormal data and filter out false signals caused by dust; The real-time monitoring platform transmits the pollution resistance sensor data to the central control system via a wireless network; The intelligent alarm module sets a threshold and an alarm mechanism. When the temperature or smoke concentration exceeds the normal range, the intelligent alarm module automatically sounds an alarm and notifies relevant personnel; Data collection: Get temperature and smoke concentration data from sensors in real time.

[0004] Data update frequency: Set the frequency of data collection and update (such as once per second).

[0005] Monitoring mechanism: Compare the real-time data with the set threshold every time the data is updated.

[0006] Conditional judgment: Set the conditions: If the temperature or smoke concentration exceeds the upper threshold, the alarm mechanism is triggered.

[0007] Alarm signal generation: Once the conditions are triggered, an alarm signal is generated.

[0008] Information content: Notification information should include the alarm type, occurrence time, specific values ​​(such as current temperature and smoke concentration) and the area where it is located.

[0009] Notification channel: Determine the notification channel, such as mobile app, email, SMS, etc.

[0010] Alarm log: record alarm events in the database, including alarm time, trigger conditions, responding personnel and other information.

[0011] Feedback mechanism: ensure that relevant personnel can respond quickly after receiving notification and provide feedback after processing.

[0012] Regular testing: Regularly test the functionality of the alarm module to ensure it can work properly in an emergency.

[0013] Threshold adjustment: Regularly review and adjust thresholds based on environmental changes and historical alarm data to improve the accuracy of the alarm system.

[0014] The environmental monitoring module adds monitoring of the factory environment (such as temperature, humidity, and dust concentration) to adjust the warning threshold in real time.

[0015] The pollution-resistant sensor includes an infrared temperature sensor, a smoke sensor, and a humidity sensor that are dust-resistant and high-temperature-resistant.

[0016] The working steps of the data filtering and processing module are as follows: S1, collect real-time data from sensors, including temperature, smoke concentration, humidity and dust concentration; S2. Data preprocessing: remove missing values ​​and obvious erroneous data, and extract useful features from the raw data; S3. Labeled data: Use historical data to label normal and abnormal samples, including false signals caused by dust, for model training; S4. Model training: Use random forest labeled data to train the model, adjust hyperparameters to optimize performance, and use cross-validation to ensure the robustness of the model; S5. Anomaly detection: Use the trained model to predict real-time data, identify abnormal signals, set thresholds, and use probability distribution to determine whether the signal is normal or not; S6. Filter false signals: further analyze the detected abnormal signals to confirm whether they are caused by dust, and perform secondary verification based on dust concentration; S7. Feedback and optimization: Regularly evaluate the accuracy of the model, update and retrain based on new data, integrate feedback mechanisms into the system, and continuously improve model performance.

[0017] The S6 comprises the following steps: S6.1. Data association analysis: Calculate the correlation coefficients between abnormal signals including temperature, smoke concentration and dust concentration. A high correlation indicates that the abnormal signal is related to dust. Analyze the time series data of abnormal signals and dust concentration to observe whether they fluctuate in similar time periods. S6.2, Feature construction: Combine the abnormal signal and dust concentration data to obtain the ratio of the signal change rate to the dust concentration; classify the dust concentration into multiple levels (low, medium, and high), and combine it with the abnormal signal to observe the performance of the abnormal signal at different levels; S6.3, Secondary validation of the model: Use the machine learning model to take the abnormal signal and dust concentration as input features, predict whether the abnormal signal is related to dust, prepare a labeled data set to indicate which abnormal signals are caused by dust, and train the model for classification; S6.4, Threshold setting: Determine the critical value of dust concentration based on historical data. When the concentration exceeds this value, give priority to the dust correlation of abnormal signals, set a specific threshold for abnormal signals, and conduct further verification when the threshold is exceeded; S6.5, Multi-sensor fusion: Combined with other sensor data, multi-dimensional analysis is performed to confirm whether the abnormal signal is caused by dust. Data fusion algorithms such as Kalman filtering are used to comprehensively consider the data of each sensor to improve the accuracy of abnormal signal identification; S6.6. Feed the results of the secondary verification back to the monitoring system, adjust the alarm strategy, reduce false alarms caused by dust, update the model based on new data, and improve the adaptability and accuracy of the model in dusty environments.

[0018] The environmental monitoring module combines real-time monitoring data and the learning results of the machine learning model to establish a threshold adjustment rule: when the dust concentration is higher than the dust concentration threshold, the warning threshold of the smoke concentration is increased accordingly; the new threshold is calculated using the weighted moving average method, and the warning threshold is dynamically adjusted according to the fluctuation of the real-time environmental data.

[0019] When the real-time dust concentration >Dust concentration threshold hour: Adjusted smoke density threshold: in, is the adjusted smoke density threshold, is the smoke concentration threshold before adjustment, k is the adjustment coefficient set according to historical data and model learning results; The formula for the weighted moving average method of smoke concentration data is: in, is the weighted moving average at the nth moment, is the real-time monitoring data of smoke concentration at the i-th moment; is the weight corresponding to the i-th moment, and the closer to the current moment, the greater the weight; The formula for the adjusted smoke concentration threshold value calculated by combining the weighted moving average is as follows: .

[0020] Compared with the prior art, the advantages of the present invention are: by integrating multiple environmental monitoring data, it can significantly reduce false alarms and missed alarms caused by dust, improve the accuracy of the fire monitoring system, adjust the warning threshold in real time, enable the system to adapt to environmental changes, and ensure effective monitoring under different working conditions. Combining real-time data and machine learning models, it can quickly identify abnormal signals and issue alarms, thereby shortening response time and reducing fire risks. This solution has the ability to continuously learn and optimize. As data accumulates, the monitoring effect will continue to improve and adapt to more complex environmental conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 The figure is a schematic diagram of the overall system of a fire monitoring and early warning system of the present invention. DETAILED DESCRIPTION

[0022] For examples, see Figure 1 , a fire monitoring and early warning system, including a pollution-resistant sensor, a central control system, a data filtering and processing module, a real-time monitoring platform, an intelligent alarm module, and an environmental monitoring module; The data filtering and processing module uses a machine learning model to identify and exclude abnormal data and filter out false signals caused by dust; The real-time monitoring platform transmits the pollution resistance sensor data to the central control system via a wireless network; The intelligent alarm module sets a threshold and an alarm mechanism. When the temperature or smoke concentration exceeds the normal range, the intelligent alarm module automatically sounds an alarm and notifies relevant personnel; The environmental monitoring module adds monitoring of factory environment temperature, humidity, and dust concentration to adjust the warning threshold in real time.

[0023] The pollution-resistant sensor includes an infrared temperature sensor, a smoke sensor, and a humidity sensor that are dust-resistant and high-temperature-resistant.

[0024] The working steps of the data filtering and processing module are as follows: S1, collect real-time data from sensors, including temperature, smoke concentration, humidity and dust concentration; S2. Data preprocessing: remove missing values ​​and obvious erroneous data, and extract useful features from the raw data; S3. Labeled data: Use historical data to label normal and abnormal samples, including false signals caused by dust, for model training; S4. Model training: Use random forest labeled data to train the model, adjust hyperparameters to optimize performance, and use cross-validation to ensure the robustness of the model; Dataset division: The collected labeled data is divided into training set and test set, generally divided into 70% and 30%.

[0025] Select important features: Select features that have a greater impact on model performance based on feature importance assessment (for example, using the feature importance score provided by random forest).

[0026] Build Random Forest Model: Build a random forest model using the training set.

[0027] Define hyperparameter ranges: Identify the hyperparameters that need to be optimized, for example: n_estimators (number of trees) max_depth (maximum depth of the tree) min_samples_split (minimum number of samples required for a node to split) min_samples_leaf (minimum number of samples required for a leaf node) Grid Search or Random Search: Use grid search (GridSearchCV) or random search (RandomizedSearchCV) to perform a systematic search within a defined range of hyperparameters to find the best combination.

[0028] Set up cross validation: Use K-fold cross validation (for example, K=5 or K=10) to further divide the training set into K subsets.

[0029] Training and validation: In each training process, K-1 subsets are used for training and 1 subset is used for validation, and this is repeated K times to obtain the average performance of the model.

[0030] Evaluation metrics: Select appropriate evaluation metrics (such as accuracy, precision, recall, F1 score, etc.) and calculate the average score of the model in cross-validation.

[0031] Compare models: Compare with the baseline model to ensure that the new model’s performance improves.

[0032] Train the model with the best hyperparameters: Select the best hyperparameters based on the cross-validation results and train the final model again on the entire training set.

[0033] Evaluate on the test set: Use unseen test set data to evaluate the performance of the final model to ensure the generalization ability of the model.

[0034] Record model performance: Save the performance indicators of the model and record the optimal parameter configuration for subsequent use and adjustment.

[0035] Model interpretability: Visualize feature importance to understand the rationale behind model decisions.

[0036] S5. Anomaly detection: Use the trained model to predict real-time data, identify abnormal signals, set thresholds, and use probability distribution to determine whether the signal is normal or not; S6. Filter false signals: further analyze the detected abnormal signals to confirm whether they are caused by dust, and perform secondary verification based on dust concentration; S7. Feedback and optimization: Regularly evaluate the accuracy of the model, update and retrain based on new data, integrate feedback mechanisms into the system, and continuously improve model performance.

[0037] The S6 comprises the following steps: S6.1. Data correlation analysis: Calculate the correlation coefficients between abnormal signals including temperature, smoke concentration and dust concentration. A high correlation may indicate that the abnormal signal is related to dust. Analyze the time series data of abnormal signals and dust concentration to observe whether they fluctuate in similar time periods. S6.2, Feature construction: Combine the abnormal signal and dust concentration data to obtain the ratio of the signal change rate to the dust concentration; classify the dust concentration into multiple levels (low, medium, and high), and combine it with the abnormal signal to observe the performance of the abnormal signal at different levels; S6.3, Secondary validation of the model: Use the machine learning model to take the abnormal signal and dust concentration as input features, predict whether the abnormal signal is related to dust, prepare a labeled data set to indicate which abnormal signals are caused by dust, and train the model for classification; S6.4, Threshold setting: Determine the critical value of dust concentration based on historical data. When the concentration exceeds this value, give priority to the dust correlation of abnormal signals, set a specific threshold for abnormal signals, and conduct further verification when the threshold is exceeded; S6.5, Multi-sensor fusion: Combined with other sensor data, multi-dimensional analysis is performed to confirm whether the abnormal signal is caused by dust. Data fusion algorithms such as Kalman filtering are used to comprehensively consider the data of each sensor to improve the accuracy of abnormal signal identification; S6.6. Feed the results of the secondary verification back to the monitoring system, adjust the alarm strategy, reduce false alarms caused by dust, update the model based on new data, and improve the adaptability and accuracy of the model in dusty environments.

[0038] The environmental monitoring module combines real-time monitoring data and the learning results of the machine learning model to establish threshold adjustment rules: when the dust concentration is higher than a certain level, the warning threshold of the smoke concentration is increased accordingly; the new threshold is calculated using the weighted moving average method, and the warning threshold is dynamically adjusted according to the fluctuation of the real-time environmental data.

[0039] When the real-time dust concentration >Dust concentration threshold hour: Adjusted smoke density threshold: in, is the adjusted smoke density threshold, is the smoke concentration threshold before adjustment, k is the adjustment coefficient set according to historical data and model learning results; The formula for the weighted moving average method of smoke concentration data is: in, is the weighted moving average at the nth moment, is the real-time monitoring data of smoke concentration at the i-th moment; is the weight corresponding to the i-th moment, and the closer to the current moment, the greater the weight; The formula for the adjusted smoke concentration threshold value calculated by combining the weighted moving average is as follows: .

[0040] The basic principles, main features and advantages of the present invention are shown and described above; those skilled in the art should understand that the present invention is not limited to the above embodiments, and the above embodiments and descriptions are only preferred examples of the present invention and are not used to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, and these changes and improvements all fall within the scope of the present invention to be protected; the scope of protection claimed by the present invention is determined by the attached claims and their effects.

Claims

1. A fire monitoring and early warning system, It is characterized in that It includes pollution-resistant sensors, central control systems, data filtering and processing modules, real-time monitoring platforms, intelligent alarm modules, and environmental monitoring modules; The data filtering and processing module uses a machine learning model to identify and exclude abnormal data and filter out false signals caused by dust; The real-time monitoring platform transmits the data detected by the pollution resistance sensor to the central control system via a wireless network; When the temperature and smoke concentration exceed the normal range, the intelligent alarm module automatically sounds an alarm and notifies relevant personnel; The environmental monitoring module monitors the factory environment temperature, humidity, and dust concentration, and adjusts the warning threshold in real time.

2. A fire monitoring and early warning system according to claim 1, It is characterized in that The pollution-resistant sensors include infrared temperature sensors, smoke sensors and humidity sensors that are dust and high temperature resistant.

3. According to the fire monitoring and early warning system of claim 2, It is characterized in that The working steps of the data filtering and processing module are as follows: S1, collect real-time data from pollution resistance sensors and environmental monitoring modules, including temperature, smoke concentration, humidity and dust concentration data; S2. Data preprocessing: remove missing values ​​and obvious erroneous data, and extract useful features from the raw data; S3. Labeled data: Use historical data to label normal and abnormal samples, including false signals caused by dust, for model training; S4. Model training: Use random forest labeled data to train the model, adjust hyperparameters to optimize performance, and use cross-validation to ensure the robustness of the model; S5, anomaly detection: use the model trained in S4 to predict real-time data, identify abnormal signals, set thresholds and use probability distribution to determine whether the signal is normal or not; S6. Filter false signals: further analyze the detected abnormal signals to confirm whether they are caused by dust, and perform secondary verification based on dust concentration; S7. Feedback and optimization: Regularly evaluate the accuracy of the model, update and retrain based on new data, and integrate feedback mechanisms into the system to improve model performance.

4. A fire monitoring and early warning system according to claim 3, It is characterized in that The S6 comprises the following steps: S6.

1. Data association analysis: Calculate the correlation coefficients between abnormal signals including temperature, smoke concentration and dust concentration. A high correlation indicates that the abnormal signal is related to dust. Analyze the time series data of abnormal signals and dust concentration to observe whether they fluctuate in similar time periods. S6.2, Feature construction: Combine the abnormal signal and dust concentration data to obtain the ratio of the signal change rate to the dust concentration; classify the dust concentration into multiple levels, namely low, medium and high, and combine it with the abnormal signal to observe the performance of the abnormal signal at different levels; S6.3, Secondary validation of the model: Use the machine learning model to take the abnormal signal and dust concentration as input features, predict whether the abnormal signal is related to dust, prepare a labeled data set, mark the abnormal signal caused by dust, and train the model for classification; S6.4, Threshold setting: Determine the critical value of dust concentration based on historical data. When the concentration exceeds this value, give priority to the dust correlation of abnormal signals, set a specific threshold for abnormal signals, and conduct further verification when the threshold is exceeded; S6.5, Multi-sensor fusion: Combined with other sensor data, multi-dimensional analysis is performed to confirm whether the abnormal signal is caused by dust. The Kalman filter data fusion algorithm is used to comprehensively consider the data of each sensor to improve the accuracy of abnormal signal identification; S6.

6. Feed the results of the secondary verification back to the monitoring system, adjust the alarm strategy, reduce false alarms caused by dust, update the model based on new data, and improve the adaptability and accuracy of the model in dusty environments.

5. A fire monitoring and early warning system according to claim 4, It is characterized in that The environmental monitoring module combines real-time monitoring data and the learning results of the machine learning model to establish a threshold adjustment rule: when the dust concentration is higher than the dust concentration threshold, the warning threshold of the smoke concentration is increased accordingly; the new threshold is calculated using the weighted moving average method, and the warning threshold is dynamically adjusted according to the fluctuation of the real-time environmental data.

6. A fire monitoring and early warning system according to claim 5, It is characterized in that When the real-time dust concentration >Dust concentration threshold hour: Adjusted smoke density threshold: in, is the adjusted smoke density threshold, is the smoke concentration threshold before adjustment, k is the adjustment coefficient set according to historical data and model learning results; The formula for the weighted moving average method of smoke concentration data is: in, is the weighted moving average at the nth moment, is the real-time monitoring data of smoke concentration at the i-th moment; is the weight corresponding to the i-th moment, and the closer to the current moment, the greater the weight; The formula for the adjusted smoke concentration threshold value calculated by combining the weighted moving average is as follows: .

Images