Precise environment monitoring and early warning system

Abstract

The invention discloses a precise environment monitoring and early warning system, and belongs to the technical field of environment monitoring. Aiming at the problems that an existing system is fixed in threshold value, poor in adaptability and complex in calibration, the system integrates a temperature sensor, a humidity sensor, a vibration strength sensor and a structural stress sensor through a multi-parameter acquisition module, and environment parameter data are obtained in real time; the central processing module operates a dynamic threshold adjustment algorithm to automatically update the dynamic threshold of each parameter, triggers a multi-sensor calibration mechanism when detecting the sudden change of the parameter, and automatically corrects the reference value of the corresponding parameter after calibration; the early warning execution module responds immediately when any parameter exceeds a dynamic threshold value. According to the system, through the dynamic threshold optimization and automatic calibration technology, the environmental parameter monitoring precision and the system adaptability are remarkably improved, and the system can be widely applied to real-time safety monitoring and early warning in the fields of prefabricated building hoisting operation, building construction, ultra-deep foundation pit supporting and the like.

Description

Technical Field

[0001] The present invention relates to the field of precise environmental monitoring. More specifically, the present invention relates to a precise environmental monitoring and early warning system. Background Art

[0002] In the field of construction environment monitoring, the limitations of traditional monitoring and early warning systems are as follows: First, the fixed threshold judgment mechanism is difficult to adapt to the non-linear change characteristics of parameters. This is because the building structure bears dynamic stress changes during construction, and there are complex interferences such as mechanical vibration and temperature and humidity fluctuations at the construction site. Taking the construction of high-rise building concrete structures as an example, during the pouring process of the core tube shear wall, the structural stress change rate can reach 0.5 MPa / h. The traditional fixed threshold cannot match this non-linear growth trend, resulting in a lag in early warning response; during the cantilever casting of the main girder of a long-span cable-stayed bridge, the structural stress change gradient during the cable force tensioning process can reach 2 MPa / m. The traditional fixed threshold cannot match the dynamic characteristics of the construction progress, resulting in a lag in stress over-limit early warning and affecting the linear control accuracy of the closure section; Second, in the construction scenario, there are special challenges in the calibration of sensors. Vibration sensors are prone to zero drift due to mechanical shock, the calibration period of temperature and humidity sensors is significantly shortened in high-dust environments, and the sensitivity attenuation rate of stress sensors accelerates under long-term loads. In the monitoring of ultra-deep foundation pit support, vibration sensors are easily affected by the high-frequency impact of pile drivers, and the zero drift can reach ±1.5 g. Manual calibration requires interrupting the monitoring for 8 - 12 hours, which is difficult to meet the real-time monitoring requirements of foundation pit deformation; during the hoisting operation of prefabricated buildings, environmental temperature and humidity fluctuations (the daily temperature difference can reach 15°C, and the humidity change rate is 30% / h) will cause signal drift of stress sensors. The traditional system cannot automatically compensate for environmental interference, resulting in large monitoring data deviations; during the construction of deep-water bridge piers, vibration sensors are interfered by the high-frequency vibration of pile driving boats, and the zero drift exceeds ±2 g. Manual calibration requires interrupting the monitoring for 12 - 24 hours, which is in sharp contradiction with the continuous operation requirements of bridge foundation construction; when hoisting steel box girders, the thermal expansion and contraction effect caused by the solar radiation temperature difference (locally up to 20°C) causes non-linear drift of stress sensors. The traditional system does not set an environmental compensation mechanism, resulting in high stress monitoring deviations; during the erection of the main cable of a suspension bridge, electromagnetic radiation generated by construction machinery (such as the radiation intensity reaches 200 μT when the electric welding machine is working) often causes sudden changes in sensor data. The existing system cannot automatically identify the sudden change signals, and the false alarm rate is high. More seriously, under the action of long-term loads (such as the prestress tensioning of long-span structures), the sensitivity attenuation rate of stress sensors can reach 0.3% / month. The traditional manual calibration period (usually 6 months) is difficult to meet the accuracy maintenance requirements, and ultimately leads to the accumulation of safety assessment errors.

[0003] The limitations of traditional monitoring and early warning systems jointly restrict the timeliness and reliability of construction safety monitoring. It is urgent to break through the inherent limitations of traditional monitoring modes through dynamic threshold optimization and automatic calibration technologies. Summary of the Invention

[0004] An object of the present invention is to solve the problems that the traditional environmental monitoring system cannot adapt to the dynamic changes of environmental parameters in real time, the fixed threshold leads to false alarms or missed alarms, and the calibration of sensors depends on manual operation with low efficiency.

[0005] Another object of the present invention is to solve the problem that the existing environmental monitoring system lacks a scientific determination of parameter mutations, resulting in misjudgment or missed judgment of the abnormal state of sensors.

[0006] Another object of the present invention is to solve the problem that the parameter adjustment mechanism of the existing environmental monitoring system does not consider historical false alarm data and cannot dynamically optimize the sensitivity of monitoring.

[0007] Another object of the present invention is to solve the problem that the coupling relationship between multiple parameters of the existing environmental monitoring system is not fully utilized, and the threshold calculation model lacks environmental adaptability.

[0008] Another object of the present invention is to solve the problem that the calibration deviation correction of the existing environmental monitoring system is not related to the collaborative influence of multiple parameters, and the adjustment of the threshold benchmark is one-sided.

[0009] Another object of the present invention is to solve the problem that the early warning execution method of the existing environmental monitoring system is single and does not realize device linkage control, resulting in insufficient emergency response ability.

[0010] Another object of the present invention is to provide a precise environmental monitoring and early warning system, which effectively reduces the false alarm rate, improves the system adaptability and data reliability through a dynamic threshold adjustment algorithm and multi-sensor automatic calibration technology, and is applicable to safety early warning scenarios such as building construction and water conservancy projects.

[0011] To achieve these and other advantages according to the present invention, a precise environmental monitoring and early warning system is provided, including: A multi-parameter acquisition module, which includes a temperature sensor, a humidity sensor, a vibration intensity sensor, and a structural stress sensor to monitor the parameter data in real time; A central processing module, which runs a dynamic threshold adjustment algorithm to obtain the dynamic thresholds of each parameter, and the multi-parameter acquisition module is connected to the central processing module; An early warning execution module, which is triggered when any parameter data monitored in real time by the multi-parameter acquisition module exceeds its dynamic threshold; Wherein, when mutation data appears in any parameter collected by the multi-parameter acquisition module, the central processing module sends a calibration instruction to the multi-parameter acquisition module to perform zero calibration of the temperature sensor, saturated salt solution calibration of the humidity sensor, standard vibration table calibration of the vibration intensity sensor, and weight loading calibration of the structural stress sensor. After the multi-parameter acquisition module completes the calibration, the central processing module automatically corrects the reference value of the corresponding parameter dynamic threshold according to the calibration deviation of any sensor.

[0012] Preferably, any of the following situations of any parameter is mutation data: A. Within a continuous time window T, the absolute value of the first derivative of any parameter data exceeds K times the average derivative of the same historical time period, and the change direction of this data is opposite to the change direction of its associated parameter; B. When the data of a single parameter exceeds its dynamic threshold, its associated parameter does not change co - ordinately; C. Based on the ARIMA model to predict the predicted parameter values at the next N time points, if the real - time monitored parameter value exceeds the prediction interval more than M times, the prediction interval is the predicted parameter value ± 10%; D. When the noise level measured by any sensor through the self - check circuit exceeds 150% of the nominal value; Wherein, T is 30 - 90s; K takes 2 - 4; N takes 5 - 8, M takes 3 - 5, and M < N; the temperature - humidity is a group of associated parameters, and the vibration intensity - structural stress is a group of associated parameters.

[0013] Preferably, the central processing module dynamically adjusts the values of T, K, N, and M according to the false alarm rate of historical mutation data. For every 10% increase in the false alarm rate of the mutation data, the value of T increases by 30s, and the values of K, N, and M increase by 1 synchronously.

[0014] Preferably, the central processing module obtains the dynamic thresholds of each parameter through a dynamic threshold adjustment algorithm, and the implementation includes the following steps: S1. Establish a coupling relationship model of temperature T, humidity H, vibration intensity V, and structural stress S based on historical data: ; S2. According to the current parameter values and the coupling relationship model in step S1, calculate the upper and lower limits of the dynamic thresholds of each parameter: ; S3. Recalculate the dynamic thresholds of each parameter every 30 - 90s; Wherein, △S, △V, △T, and △H are the change amounts of structural stress, vibration intensity, temperature, and humidity respectively; A, b, c, d, and e are the weight coefficients of the multiple regression fitting; μ t is the mean value of the current time period parameter, σ t is the standard deviation, and △P is the co-variation of the correlation parameter; f is the correlation strength coefficient, with a value range of 0.1 - 0.5, and g is the safety factor, with a value range of 1.5 - 3.

[0015] Preferably, the value of the safety factor g is dynamically adjusted through the LSTM neural network, specifically: After calculating the health index j of the sensor through the LSTM neural network, the safety factor g is adjusted according to the health index j of the sensor: When the parameters are a set of correlation parameters of vibration intensity - structural stress, g = 1.5 + 1.5j; When the parameters are a set of correlation parameters of temperature - humidity, g = 1.2 + 1.2j; where the value of j ranges from 0 to 1, and 1 indicates that the sensor is in the best state.

[0016] Preferably, the rule for the central processing module to automatically correct the reference value of the corresponding parameter dynamic threshold according to the calibration deviation of any sensor is as follows: ; where △Q is the correction amount of the parameter dynamic threshold reference value; △c is the calibration deviation value of the parameter corresponding sensor; h i is the weight of the correlation parameter, the weight of temperature - humidity is 0.3, and the weight of vibration intensity - structural stress is 0.7; △ci is the mean value of the historical calibration deviation of the parameter corresponding sensor.

[0017] Preferably, the influence of the fuzzy logic compensation term needs to be considered for the correction amount of the parameter dynamic threshold reference value: ; where ρ is the fuzzy compensation coefficient, and its value range is 0.8 - 1.5.

[0018] Preferably, the value rule of the fuzzy compensation coefficient ρ is as follows: When the relative deviation of the sensor < ±2%, ρ takes a value range of 0.8 - 1.1; When ±2% ≤ the relative deviation of the sensor ≤ ±5%, ρ takes a value range of 1.2 - 1.3; When the relative deviation of the sensor > ±5%, ρ takes a value range of 1.4 - 1.5.

[0019] Preferably, the warning execution module includes an acoustic-optic alarm, a GSM short message sending unit, and a relay control unit. The central processing unit sends a pulse signal with a frequency of 2 to 3 kHz to the acoustic-optic alarm, and at the same time sends a text message containing the parameter over-standard value to the GSM short message sending unit, and sends a signal to cut off the power supply of the high-risk construction equipment to the relay control unit.

[0020] Preferably, when the acoustic-optic alarm receives a pulse signal of 2 kHz to 3 kHz, it drives the buzzer and the LED flashing module with a duty cycle of 1:3. The sound pressure level of the buzzer is ≥90 dB@10 m, and the LED flashing frequency is 5 Hz±0.5 Hz; the text message generated by the GSM short message sending module includes the over-standard parameter name, real-time value, GPS geographical coordinates, and the associated construction equipment code. The positioning accuracy of the geographical coordinates is ±1 m, and the equipment code is bound to the equipment topology map pre-stored in the cloud server; the relay control unit is built-in with a buffer circuit. When cutting off the power supply of the high-risk equipment, the control current drop slope is ≤30 A / s, and an RC absorption circuit is connected in parallel at both ends of the relay contact. The resistance value of the RC absorption circuit is 15 Ω±5%, and the capacitance value is 0.2 μF±10%.

[0021] The present invention has at least the following beneficial effects: First, the precise environmental monitoring and warning system provided by the present invention adopts a dynamic threshold adjustment algorithm combined with a multi-sensor automatic calibration technology, enabling the system to adapt to the non-linear changes of environmental parameters in real time. Compared with the traditional fixed threshold system, the false alarm rate is extremely reduced, the warning response time is effectively compressed, and the timeliness of environmental monitoring is significantly improved. The automatic calibration mechanism makes the sensor calibration period more accurate and the calibration accuracy is significantly improved; Second, the determination of the mutation data of the precise environmental monitoring and warning system provided by the present invention establishes a scientific abnormal discrimination system through the collaborative analysis of associated parameters, effectively reducing the false alarms caused by mechanical vibration interference, and greatly improving the accuracy of sensor abnormal identification; Third, the dynamic parameter adjustment mechanism based on the false alarm rate of the precise environmental monitoring and warning system provided by the present invention enables the system to have an adaptive optimization ability. When the complexity of the construction environment increases and the false alarm rate rises, the system can automatically extend the detection window T and synchronously increase the K / N / M parameter values to maintain the detection sensitivity stable; Fourth, the multi-parameter coupling relationship model of the precise environmental monitoring and warning system provided by the present invention introduces the cross-influence coefficients (a, b, c, d, and e) of vibration-stress and temperature-humidity, enabling the threshold calculation to comprehensively consider the environmental synergistic effect; the measured data shows that the stress threshold deviation of this model under the condition of large temperature difference (ΔT>20 °C) is reduced by 42% compared with the traditional single-parameter model, and the dynamic response ability is increased by 3 times; Fifthly, the precision environmental monitoring and early warning system provided by the present invention dynamically adjusts the safety factor g based on the health index of the LSTM neural network, realizing precise compensation for the aging state of sensors. During the long-term monitoring of bridges, when the sensor health index j decays from 1.0, the safety factor g is automatically adjusted to ensure that the risk assessment confidence level is always maintained above 95%; Sixthly, the associated parameter calibration deviation correction mechanism of the precision environmental monitoring and early warning system provided by the present invention makes the adjustment of the reference value more in line with the actual working conditions by introducing a weight coefficient (0.3 / 0.7). In a super high-rise project, this correction method reduces the reference value adjustment error of the stress sensor from ±5% to ±1.2%, and reduces the long-term monitoring data drift by 78%; Seventhly, the introduction of the fuzzy logic compensation term ρ of the precision environmental monitoring and early warning system provided by the present invention effectively solves the non-linear problem of calibration deviation; the adjustment accuracy of the reference value after compensation reaches ±0.8%; Eighthly, the multi-modal early warning execution system of the precision environmental monitoring and early warning system provided by the present invention realizes the full-chain response from on-site warning to equipment control, and extremely shortens the accident response time.

[0022] Other advantages, objectives and features of the present invention will be partially reflected by the following description, and partially will also be understood by those skilled in the art through the research and practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is a schematic diagram of the early warning process of the precision environmental monitoring and early warning system described in a technical solution of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0024] The following will further elaborate on the present invention with specific embodiments, so that those skilled in the art can implement it with reference to the description in the specification.

[0025] It should be understood that the terms such as "having", "comprising" and "including" used herein do not exclude the presence or addition of one or more other elements or their combinations.

[0026] As Figure 1 shown, the present invention provides a precision environmental monitoring and early warning system, including: A multi-parameter acquisition module, which includes a temperature sensor, a humidity sensor, a vibration intensity sensor and a structural stress sensor to real-time monitor the parameter data; A central processing module, which runs a dynamic threshold adjustment algorithm to obtain the dynamic thresholds of each parameter, and the multi-parameter acquisition module is connected to the central processing module; An early warning execution module, which is triggered when any parameter data real-time monitored by the multi-parameter acquisition module exceeds its dynamic threshold; Among them, when mutation data appears in any parameter collected by the multi-parameter acquisition module, the central processing module sends a calibration instruction to the multi-parameter acquisition module to perform zero calibration of the temperature sensor, saturated salt solution calibration of the humidity sensor, standard vibration table calibration of the vibration intensity sensor, and weight loading calibration of the structural stress sensor. After the multi-parameter acquisition module completes the calibration, the central processing module automatically corrects the reference value of the corresponding parameter dynamic threshold according to the calibration deviation of any sensor.

[0027] In the above embodiment, the multi-parameter acquisition module includes a temperature sensor, a humidity sensor, a vibration intensity sensor, and a structural stress sensor. The measurement range of the temperature sensor is set to -20°C to 80°C, and the accuracy is ±0.5°C. The Honeywell HTU21D model temperature sensor can be used; the resolution of the humidity sensor is selected as 0.1%RH, and the Sensirion SHT45 humidity sensor can be used; the frequency response range of the vibration sensor covers 10Hz~1kHz, and the linear error <±2%. The PCB Piezotronics 352C33 vibration sensor can be selected; the range of the stress sensor is set to 0~50MPa, and the error ≤±0.5%FS. The HBM U9C stress sensor can be used. The sensor housing material can be selected as ABS engineering plastic with an IP67 protection level, the internal circuit board can use FR-4 substrate, and the connecting wire uses high-temperature resistant silicone wire.

[0028] In the above embodiment, taking the construction process of a certain bridge as an example, the temperature sensor can be installed in the embedded groove inside the bridge pile foundation concrete, the humidity sensor can be fixed on the outside of the concrete curing film, the vibration sensor can be arranged at the steel beam connection node, and the stress sensor can be welded to the key stress points of the main beam; each sensor is connected to the central processing unit through a wireless communication module (such as the Semtech SX1276 LoRa chip), collects data in real time and transmits it to the central processing unit through the LoRa protocol.

[0029] In the above embodiments, the central processing module uses an STM32H750 microcontroller. One implementation of the dynamic threshold adjustment algorithm is as follows: Through preprocessing of historical data, such as calculating the mean and variance of the data in the 600 s before the current time, a calculation formula for the dynamic threshold with respect to the mean and variance is designed to calculate the dynamic thresholds of each parameter, and the statistical parameters are updated every 30 - 90 s through exponentially weighted moving average. When a parameter mutation is detected (such as the stress change rate > 5% / s), zero - point calibration of the temperature sensor, saturated salt solution calibration of the humidity sensor, standard vibration table calibration of the vibration intensity sensor, and weight loading calibration of the structural stress sensor are triggered. Among them, when the central processing module triggers the calibration process, the system stabilizes the temperature of the calibration chamber at 0°C ± 0.1°C through a semiconductor refrigeration sheet to correct zero - point drift; the temperature sensor automatically switches to the built - in saturated salt solution chamber, and the temperature in the chamber is kept constant at 25°C ± 0.5°C by a bimetallic strip to achieve three - point calibration; the vibration sensor is built with a micro - electromagnetic vibration table, which generates a 100 Hz sine wave with an acceleration of 1g ± 0.5% after triggering the calibration instruction, and the vibration amplitude is fed back by a laser displacement sensor. The calibration frequency covers 10 Hz - 2 kHz and supports multi - axis vibration calibration; the node where the structural stress sensor is located automatically loads a 10% FS standard weight, and the strain value is measured through a Wheatstone bridge. The loading device integrates a stepper motor and a pressure sensor, supports stepped loading, and automatically compensates for temperature drift during the calibration process.

[0030] In the above embodiments, by using the precise environmental monitoring and early warning system provided by the present invention, in the actual measurement values during the construction of the bridge, the dynamic threshold increases the stress early warning accuracy rate to 98.2%, and reduces the false alarm rate by 73% compared with the fixed - threshold system. The automatic calibration period is shortened from 48 hours of traditional manual operation to 25 minutes, and the calibration error is controlled within ±0.3% FS.

[0031] In the above embodiments, a MEMs sensor array can be used to replace discrete sensors. The integrated chip uses the MS8607 temperature - humidity + vibration six - axis sensor of TE Connectivity. Its wireless transmission is based on LoRaWAN and supports star - network deployment. Its automatic calibration method: The factory calibration coefficients are stored in the on - chip EEPROM and read in real time through the I2C interface. The replacement solution effectively reduces the layout volume of the sensors and is suitable for environmental monitoring in narrow spaces.

[0032] According to the above technical features, the present invention has at least the following beneficial effects: 1. The multi-parameter acquisition module integrates temperature, humidity, vibration intensity, and structural stress sensors, enabling real-time and comprehensive acquisition of various environmental parameter data. This allows the system to monitor the environmental conditions in multiple dimensions and in detail, avoiding the limitations of single-parameter monitoring, and being able to promptly capture changes in various environmental factors, providing an accurate and rich data basis for subsequent analysis and early warning. 2. The central processing module runs a dynamic threshold adjustment algorithm to determine the dynamic thresholds of each parameter. Compared with traditional fixed-threshold systems, the dynamic thresholds can be adjusted in real time according to the actual changes in the environment, better adapting to different environmental conditions and complex and changeable working conditions, reducing false alarms or missed alarms caused by environmental changes, and greatly improving the accuracy and reliability of early warning. 3. The warning execution module is triggered immediately when any parameter data exceeds its dynamic threshold, capable of promptly sending an alarm to relevant personnel, enabling the staff to learn about abnormal changes in environmental parameters in a timely manner and quickly take corresponding measures to prevent the possible deterioration of dangerous situations and ensure environmental safety and the safety of related equipment and personnel. 4. When mutation data appears in the parameters collected by the multi-parameter acquisition module, the central processing module will automatically send calibration instructions to calibrate each sensor. By performing zero calibration of the temperature sensor, saturated salt solution calibration of the humidity sensor, standard vibration table calibration of the vibration intensity sensor, and weight loading calibration of the structural stress sensor, it can effectively eliminate errors caused by factors such as long-term use and environmental interference of the sensors and ensure the measurement accuracy of the sensors. Moreover, after calibration is completed, the central processing module will automatically correct the reference value of the corresponding parameter dynamic threshold according to the calibration deviation, further optimizing the monitoring and early warning performance of the system and keeping the system in a high-precision and reliable operating state.

[0033] In another embodiment, any of the following situations for any parameter is considered mutation data: A. Within a continuous time window T, the absolute value of the first derivative of any parameter data exceeds K times the average derivative of the same historical time period, and the change direction of this data is opposite to the change direction of its associated parameter; B. When the data of a single parameter exceeds its dynamic threshold, its associated parameter does not change co -operatively; C. Based on the ARIMA model to predict the predicted parameter values at the next N time points, if the real-time monitored parameter value exceeds the prediction interval (the prediction interval is the predicted parameter value ± 10%) more than M times; D. When the noise level measured by any sensor through the self - test circuit exceeds 150% of the nominal value; Among them, T is 30 to 90 s; K takes 2 to 4; N takes 5 to 8, M takes 3 to 5, and M < N; the temperature-humidity is a set of associated parameters, and the vibration intensity-structural stress is a set of associated parameters.

[0034] In the above embodiment, the determination of mutation data only needs to meet one of the following four conditions: Condition A: Calculate the first derivative of the parameter using a sliding time window (such as T = 60 s), and then query the average derivative in the same time period (such as 12:00 - 13:00 every day) through the historical database. When the current derivative is greater than K (such as K = 3) times the average derivative, and the change directions of the associated parameters (such as temperature and humidity) are opposite (such as the temperature increases but the humidity decreases), a mutation is triggered. In Condition A, the Savitzky-Golay filter can be used to smooth the data and reduce noise interference. Condition B: When a single parameter (such as vibration intensity) exceeds the dynamic threshold, check whether the associated parameter (structural stress) satisfies the co-variation of △S / S > 0.5%. If the change rate of the associated parameter (structural stress) < 0.5%, it is determined as a mutation. Condition C: Based on the ARIMA(2,1,1) model, predict the parameter values at the next N (such as N = 5) time points, and set the prediction interval as ±10% of the predicted value. If the real-time data exceeds the interval continuously for M (such as M = 3) times, a mutation is triggered. In Condition C, the model parameters are automatically optimized daily, and residual analysis ensures the prediction accuracy. Condition D: Each sensor is built with an ADC self-checking circuit to calculate the noise RMS value in real time. When RMS > 150% of the nominal noise value, it is triggered. In Condition D, the noise is calculated using a 50 ms sliding window to exclude the high-frequency components of the signal itself.

[0035] In the above embodiment, the present invention designs multi-dimensional mutation monitoring, integrating four dimensions: derivative analysis, threshold co-variation, prediction deviation, and noise level. In building construction or water conservancy project monitoring, the picture recognition accuracy rate reaches over 98%, and the false alarm rate is extremely reduced. Among them, in the threshold co-variation dimension, the co-judgment mechanism of temperature-humidity and vibration intensity-structural stress effectively distinguishes real anomalies from environmental interference factors, and avoids false alarms of humidity sensors caused by temperature fluctuations under working conditions with large solar radiation temperature differences. In the prediction deviation dimension, the ARIMA model predicts several steps in advance. In the monitoring of deep foundation pit support, it successfully warns of the stress mutation time, 2 - 3 minutes earlier than the traditional threshold trigger. In the dimension of noise level, the noise level detection realizes early warning of sensor failures. In the construction environment, problems with the aging of vibration sensor circuits can be detected at least 72 h in advance.

[0036] In the above embodiment, an LSTM neural network can also be used to construct a multivariate anomaly detection model. The input includes temperature, humidity, vibration intensity, structural stress, and their derivatives, and the output is the mutation probability. The model is incrementally trained with new data every 24 hours. This replacement scheme can adaptively learn data features, and the mutation recognition accuracy can be increased to more than 99% under complex working conditions, and the prediction lead can reach more than 10 steps. However, an edge computing device needs to be added, the computing power consumption increases, a heat dissipation device needs to be equipped, and the model training data needs to include at least 3 complete construction cycles, and its advantages can only be highlighted in extremely complex working conditions.

[0037] In another embodiment, the central processing module dynamically adjusts the values of T, K, N, and M according to the false alarm rate of historical mutation data. For every 10% increase in the false alarm rate of the mutation data, the value of T increases by 30 s, and the values of K, N, and M increase synchronously by 1.

[0038] In the above embodiment, the central processing module continuously records historical mutation data. After each determination of mutation data and triggering of calibration, it will subsequently determine whether this calibration is a false alarm according to the actual situation. If it is a false alarm, it will be recorded. When the central processing module calculates the false alarm rate, it dynamically adjusts the values of T, K, N, and M according to the change of the false alarm rate. Specifically, a time interval is set to check the change of the false alarm rate. If the false alarm rate increases by 10%, the parameters are adjusted according to the rule: the value of T increases by 30 s, and the values of K, N, and M increase synchronously by 1. For example, initially T = 30 s, K = 2, N = 5, M = 3. When the false alarm rate increases by 10%, T becomes 60 s, K becomes 3, N becomes 6, and M becomes 4.

[0039] In the above embodiment, in different application scenarios and environmental conditions, the characteristics and false alarm situations of mutation data will be different. By dynamically adjusting the values of T, K, N, and M, the system can automatically optimize the determination criteria of mutation data according to the actual false alarm situation. Especially in some scenarios with complex environments and many interference factors, the initial false alarm rate may be relatively high. At this time, increasing the values of T, K, N, and M can make the determination of mutation data more stringent and reduce the occurrence of false alarms. At the same time, dynamically adjusting the parameters can enable the system to better balance the false alarm rate and the missed alarm rate. When the false alarm rate is high, the determination criteria are improved by adjusting the parameters to avoid the trouble caused by excessive false alarms to the staff; while when the false alarm rate is low, a relatively loose determination criteria is maintained to ensure that real mutation data can be captured in time, improving the accuracy and reliability of early warning. The need for manual intervention is reduced, and the labor cost and the possibility of human errors are lowered.

[0040] In another embodiment, the implementation of the central processing module obtaining the dynamic thresholds of each parameter through a dynamic threshold adjustment algorithm includes the following steps: S1. Establish a coupling relationship model for temperature T, humidity H, vibration intensity V, and structural stress S based on historical data: ; S2. Calculate the upper and lower limits of the dynamic thresholds of each parameter according to the current parameter values and the coupling relationship model in step S1: ; S3. Recalculate the dynamic thresholds of each parameter every 30 - 90 s; where, ΔS, ΔV, ΔT, and ΔH are the change amounts of structural stress, vibration intensity, temperature, and humidity respectively; A, b, c, d, and e are the weight coefficients of multiple regression fitting; μ t is the mean value of the parameters in the current time period, σ t is the standard deviation, and ΔP is the co - variation amount of the associated parameters; f is the correlation strength coefficient, with a value range of 0.1 - 0.5, and g is the safety factor, with a value range of 1.5 - 3.

[0041] In the above - mentioned embodiment, the implementation of the dynamic threshold adjustment algorithm first constructs a multi - parameter coupling model. In the construction of a certain bridge, the coupling relationship model is obtained by fitting historical data: ; In further construction, the model prediction error ≤ ±2.5%; then further calculate the dynamic thresholds. Calculate the upper and lower limits of the dynamic thresholds of each parameter through the formula and update the mean value μ t and the standard deviation σ t every 30 - 90 s through Kalman filtering. Among them, for a group of associated parameters of temperature - humidity, ΔP = ΔT×ΔH, and for a group of associated parameters of vibration intensity - structural stress, ΔP = ΔS×ΔV. In the construction of this bridge, , and it is updated every 60 s, using the exponentially weighted moving average (α = 0.2). In the above process, the Savitzky - Golay filter can be used to eliminate high - frequency noise, and the multi - parameter coupling model is automatically refitted at 0:00 on Sundays to adapt to long - term environmental changes.

[0042] In the above - mentioned embodiment, through multi - parameter collaborative compensation, the threshold deviation is extremely reduced. The design of the co - variation term of the associated parameters enables the threshold range to dynamically expand with the working conditions. The threshold is updated every 30 - 90 s to improve the system response speed. The system calibration is automatically updated weekly, making the long - term monitoring data drift amount also extremely reduced.

[0043] Similarly, the above technical solution can use an LSTM neural network to construct a multivariable threshold prediction model. By inputting temperature, humidity, vibration intensity, structural stress, and their derivatives, the upper and lower limits of the dynamic thresholds of each parameter are output. This replacement solution requires adding edge computing devices, resulting in increased computing power consumption. A heat dissipation device needs to be equipped, and the model training data needs to include at least 3 complete construction cycles, and its advantages can only be highlighted under extremely complex working conditions.

[0044] In another embodiment, the value of the safety factor g is dynamically adjusted through an LSTM neural network. Specifically: After calculating the health index j of the sensor through the LSTM neural network, the safety factor g is adjusted according to the health index j of the sensor: When the parameter is a set of associated parameters of vibration intensity - structural stress, g = 1.5 + 1.5j; When the parameter is a set of associated parameters of temperature - humidity, g = 1.2 + 1.2j; where the value of j ranges from 0 to 1, and 1 indicates that the sensor is in the best state.

[0045] In the above embodiment, the safety factor is further dynamically adjusted. The LSTM neural network receives real-time data (temperature, humidity, vibration intensity, and structural stress) from multiple sensors, as well as multi-dimensional features such as their historical calibration deviations and noise levels. The model uses a 3-layer bidirectional LSTM network with 2 layers, 128 hidden units, and an input window size of 100 time steps. The output of the health index j, j ∈ [0, 1], is normalized through the sigmoid activation function, and 1 indicates the best sensor performance. The dynamic adjustment rule of the safety factor g is as follows: for the vibration intensity - structural stress group: g = 1.5 + 1.5j; for the temperature - humidity group: g = 1.2 + 1.2j. By dynamically adjusting the safety factor g through the LSTM neural network, an accurate match between the health state of the sensor and the threshold safety margin is achieved. When the health index j of the sensor decreases, the value of g automatically increases (for example, from 1.5 to 3.0 in the vibration - stress group), expanding the threshold range to compensate for the attenuation of sensor performance and avoiding missed alarms caused by decreased sensitivity. At the same time, the introduction of the health index j realizes the quantitative evaluation of the sensor state. By analyzing the change trend of the j value, the system can predict the sensor failure time in advance. Different g-value adjustment formulas are set for different parameter groups to reasonably allocate safety resources and improve the accuracy of risk control.

[0046] In another embodiment, the rule for the central processing module to automatically correct the reference value of the dynamic threshold of the corresponding parameter according to the calibration deviation of any sensor is as follows: ; where △Q is the correction amount of the reference value of the parameter dynamic threshold; △c is the calibration deviation value of the sensor corresponding to the parameter; hi is the weight of the associated parameter. The weight of temperature - humidity is 0.3, and the weight of vibration intensity - structural stress is 0.7; △ci is the mean value of the historical calibration deviation of the sensor corresponding to the associated parameter.

[0047] In the above - mentioned embodiment, after each sensor calibration, record the current calibration deviation △c (such as the zero - drift amount of the temperature sensor) in the historical calibration deviation database, store the △ci values of the most recent 10 calibrations, assign the weights of the associated parameters. The weight of the temperature - humidity group h = 0.3: Considering the indirectness of the influence of environmental parameters on the structure, and the weight of the vibration intensity - structural stress group h = 0.7: Directly reflecting the relevance of the structural response. The above weight values are determined by the Analytic Hierarchy Process (AHP), and the expert scoring weight ratio is 2.3:1. Finally, calculate the correction amount according to the formula, and the corrected threshold reference value is equal to the original reference value plus the correction amount.

[0048] In the above - mentioned embodiment, the weight distribution conforms to the engineering reality, is consistent with the finite - element analysis results, and the introduction of the historical calibration deviation reduces the drift amount of the reference value during long - term monitoring, extending the calibration period of the sensor.

[0049] In another embodiment, the correction amount of the dynamic threshold reference value of the parameter needs to consider the influence of the fuzzy - logic compensation term: ; where ρ is the fuzzy compensation coefficient, and its value ranges from 0.8 to 1.5.

[0050] In the above - mentioned embodiment, introducing the fuzzy compensation coefficient in the calculation of the correction amount of the threshold reference value effectively solves the non - linear problem of the calibration deviation. When the relative deviation of the sensor is in the range of 2% - 5%, the compensation coefficient ρ automatically increases to 1.2 - 1.3. After compensation, the adjustment accuracy of the reference value is significantly improved, and the quantified fuzzy compensation rule eliminates the subjective setting deviation. In a certain water - conservancy project, this rule makes the compensation consistency under different working conditions reach more than 95%, and the system robustness is significantly enhanced.

[0051] In another embodiment, the value - taking rule of the fuzzy compensation coefficient ρ is as follows: When the relative deviation of the sensor < ±2%, ρ takes a value in the range of 0.8 - 1.1; When ±2% ≤ the relative deviation of the sensor ≤ ±5%, ρ takes a value in the range of 1.2 - 1.3; When the relative deviation of the sensor > ±5%, ρ takes a value in the range of 1.4 - 1.5.

[0052] In the above embodiments, the present invention further realizes hierarchical value taking of the fuzzy compensation coefficient according to the newness and oldness of the sensor, so as to achieve refined processing of the calibration deviation. The general rule is as follows: For a new sensor (deviation < 2%): ρ = 0.8 - 1.1, to avoid overcorrection; for a sensor in mid - use (2% - 5% deviation): ρ = 1.2 - 1.3, for linear compensation; for an aging sensor (deviation > 5%): ρ = 1.4 - 1.5, for enhanced compensation. This mechanism effectively extends the service life of the sensor and reduces the average annual maintenance cost of the sensor.

[0053] In another embodiment, the warning execution module includes an audible and visual alarm, a GSM short - message sending unit, and a relay control unit. The central processing unit sends a pulse signal with a frequency of 2 - 3 kHz to the audible and visual alarm, simultaneously sends a text message containing the parameter exceeding - standard value to the GSM short - message sending unit, and sends a signal to cut off the power supply of the high - risk construction equipment to the relay control unit. A multi - modal emergency response system including on - site warning, remote notification, and equipment control is realized. Among them, the central processing unit triggers audible and visual alarm, short - message sending, and power - off control simultaneously, shortening the response time compared with the traditional sequential execution system.

[0054] In another embodiment, when the audible and visual alarm receives a pulse signal of 2 kHz to 3 kHz, it drives the buzzer and the LED flashing module with a duty cycle of 1:3. The sound pressure level of the buzzer is ≥90 dB@10 m, and the LED flashing frequency is 5 Hz ± 0.5 Hz; the text message generated by the GSM short - message sending module contains the name of the exceeding - standard parameter, the real - time value, the GPS geographical coordinates, and the associated construction equipment code. The positioning accuracy of the geographical coordinates is ±1 m, and the equipment code is bound to the equipment topology map pre - stored in the cloud server; the relay control unit is built - in with a buffer circuit. When cutting off the power supply of the high - risk equipment, the control current decline slope ≤ 30 A / s, and an RC absorption circuit is connected in parallel at both ends of the relay contact. The resistance value of the RC absorption circuit is 15 Ω ± 5%, and the capacitance value is 0.2 μF ± 10%. In the above embodiments, the on - site warning includes auditory warning and visual warning. The central processing unit triggers audible and visual alarm, short - message sending, and power - off control synchronously, shortening the response time compared with the traditional sequential execution system; the GPS geographical coordinates of ±1 m combined with the equipment code enable rescue personnel to quickly lock the accident location through the cloud equipment topology map, and the integrity of the data provides effective evidence for accident liability tracing.

[0055] The number of devices and the processing scale described here are used to simplify the description of the present invention. The application, modification, and variation of the precise environmental monitoring and warning system of the present invention are obvious to those skilled in the art.

[0056] Although the embodiments of the present invention have been disclosed as above, they are not limited to the applications listed in the specification and embodiments. It can be fully applied to various fields suitable for the present invention. For those skilled in the art, additional modifications can be easily made. Therefore, without departing from the general concept defined by the claims and the equivalent scope, the present invention is not limited to the specific details and the embodiments shown and described herein.

Claims

1. Precision environmental monitoring and early warning system, characterized by: Including: A multi-parameter acquisition module, which includes a temperature sensor, a humidity sensor, a vibration intensity sensor, and a structural stress sensor to monitor various parameter data in real time; A central processing module, which runs a dynamic threshold adjustment algorithm to obtain the dynamic thresholds of various parameters, and the multi-parameter acquisition module is connected to the central processing module; An early warning execution module, which is triggered when any parameter data monitored in real time by the multi-parameter acquisition module exceeds its dynamic threshold; Among them, when any parameter collected by the multi-parameter acquisition module shows mutation data, the central processing module sends a calibration instruction to the multi-parameter acquisition module to perform zero calibration of the temperature sensor, saturated salt solution calibration of the humidity sensor, standard vibration table calibration of the vibration intensity sensor, and weight loading calibration of the structural stress sensor. After the multi-parameter acquisition module completes the calibration, the central processing module automatically corrects the reference value of the corresponding parameter dynamic threshold according to the calibration deviation of any sensor.

2. The precise environmental monitoring and early warning system according to claim 1, characterized in that: Any of the following situations for any parameter is mutation data: A. Within a continuous time window T, the absolute value of the first derivative of any parameter data exceeds K times the average derivative of the same historical time period, and the data change direction is opposite to the change direction of its associated parameter; B. When the single parameter data exceeds its dynamic threshold, its associated parameter does not change co - ordinately; C. Based on the ARIMA model to predict the predicted parameter values at the next N time points, if the parameter value monitored in real time exceeds the prediction interval more than M times, the prediction interval is the predicted parameter value ± 10%; D. When the noise level measured by any sensor through the self - test circuit exceeds 150% of the nominal value; Among them, T is 30 - 90s; K takes 2 - 4; N takes 5 - 8, M takes 3 - 5, and M < N; the temperature - humidity is a set of associated parameters, and the vibration intensity - structural stress is a set of associated parameters.

3. The precise environmental monitoring and early warning system according to claim 2, characterized in that: The central processing module dynamically adjusts the values of T, K, N, and M according to the false alarm rate of historical mutation data. For every 10% increase in the false alarm rate of the mutation data, the value of T increases by 30s, and the values of K, N, and M increase by 1 synchronously.

4. The precise environmental monitoring and early warning system according to claim 3, characterized in that: The implementation of the central processing module obtaining the dynamic thresholds of various parameters through the dynamic threshold adjustment algorithm includes the following steps: S1. Establish a coupling relationship model of temperature T, humidity H, vibration intensity V, and structural stress S based on historical data: ； S2. Calculate the upper and lower limits of the dynamic thresholds of various parameters according to the current parameter values and the coupling relationship model in step S1: ； S3. Recalculate the dynamic thresholds of various parameters every 30 - 90s; Among them, △S, △V, △T, and △H are the change amounts of structural stress, vibration intensity, temperature, and humidity respectively; A, b, c, d, and e are the weight coefficients of multiple regression fitting; μ t is the mean value of the parameter in the current time period, σ t is the standard deviation, △P is the co-variation of the associated parameters; f is the correlation strength coefficient, taking values from 0.1 to 0.5, and g is the safety factor, taking values from 1.5 to 3.

5. The precise environmental monitoring and early warning system according to claim 4, characterized in that: The value of the safety factor g is dynamically adjusted through the LSTM neural network. Specifically: After calculating the health index j of the sensor through the LSTM neural network, adjust the safety factor g according to the health index j of the sensor: When the parameter is a set of related parameters of vibration intensity-structural stress, g=1.5+1.5j; When the parameters are not a set of temperature-humidity related parameters, g=1.2+1.2j; The value of j ranges from 0 to 1, and 1 means that the sensor is in the best state.

6. The precise environmental monitoring and early warning system according to claim 4, characterized in that: The rule for the central processing module to automatically correct the reference value of the dynamic threshold value of the corresponding parameter according to the calibration deviation of any sensor is as follows: ； Among them, △Q is the correction value of the parameter dynamic threshold reference value; △c is the calibration deviation value of the sensor corresponding to the parameter; h i is the weight of the associated parameters, the weight of temperature-humidity is 0.3, and the weight of vibration intensity-structural stress is 0.7; △ci is the mean of the historical calibration deviations of the sensor corresponding to the associated parameter.

7. The precise environmental monitoring and early warning system according to claim 6, characterized in that: The correction amount of the dynamic threshold reference value of the parameter needs to take into account the influence of the fuzzy logic compensation term: ； Among them, ρ is the fuzzy compensation coefficient, and its value is 0.8~1.

5.

8. The precise environmental monitoring and early warning system according to claim 7, characterized in that: The value of the fuzzy compensation coefficient ρ is as follows: When the relative deviation of the sensor is <±2%, ρ is 0.8~1.1; When ±2%≤relative deviation of sensor≤±5%, ρ is 1.2~1.3; When the relative deviation of the sensor is >±5%, ρ is taken as 1.4~1.

5.

9. The precise environmental monitoring and early warning system according to claim 5, characterized in that: The early warning execution module includes an audible and visual alarm, a GSM text message sending unit and a relay control unit. The central processing unit sends a pulse signal with a frequency of 2~3kHz to the audible and visual alarm, and at the same time sends a text message containing the parameter exceeding the standard value to the GSM text message sending unit, and sends a signal to the relay control unit to cut off the power supply of high-risk construction equipment.

10. The precise environmental monitoring and early warning system according to claim 9, characterized in that: When the sound and light alarm receives a pulse signal of 2kHz to 3kHz, it drives the buzzer and LED flashing module with a duty cycle of 1:

3. The sound pressure level of the buzzer is ≥90dB@10 meters, and the LED flashing frequency is 5Hz±0.5Hz; the text information generated by the GSM SMS sending module contains the name of the exceeded parameter, real-time value, GPS geographic coordinates and the associated construction equipment code. The geographic coordinate positioning accuracy is ±1 meter, and the equipment code is bound to the equipment topology map pre-stored in the cloud server; the relay control unit has a built-in buffer circuit. When the power supply of the high-risk equipment is cut off, the current drop slope is controlled to be ≤30A / s, and an RC absorption circuit is connected in parallel at both ends of the relay contacts. The resistance value of the RC absorption circuit is 15Ω±5%, and the capacitance value is 0.2μF±10%.

Images