Smoke alarm method and system based on quantum light detection

Abstract

The invention relates to a smoke alarm method and system based on quantum light detection, and the method comprises the steps: obtaining the current environment information, and matching quantum light source control parameters according to the current environment information; by adjusting quantum light source control parameters, dual-wavelength detection light with quantum entanglement characteristics is generated; in the process of generating the dual-wavelength detection light, inputting current environment information into an environment prediction model, and obtaining an environment prediction result of a future target time period; the environment prediction model is obtained by training a neural network model by using a first training set; taking an environment prediction result and keeping initial quantum entanglement characteristics as decision basis, and adaptively adjusting quantum light source control parameters; on the basis of dual-wavelength detection light with quantum entanglement characteristics, scattered light signals are collected, a corresponding fire smoke scattering spectrum is established, the mapping relation between the fire smoke scattering spectrum and smoke concentration is determined, fire smoke and non-fire aerosol are distinguished according to the mapping relation, and when fire smoke appears, smoke alarm is triggered.

Description

Technical Field

[0001] The present invention relates to the technical field of smoke alarm technology, and in particular to a smoke alarm method and system based on quantum light detection. Background Art

[0002] In smoke alarm systems, photoelectric sensors are the most commonly used detection devices. However, traditional photoelectric sensors are easily interfered by non-fire factors, such as dust and water mist, resulting in a high false alarm rate. In order to improve the accuracy of smoke alarms, it is necessary to develop a photoelectric sensing technology that can effectively distinguish between fire smoke and non-fire aerosols.

[0003] Quantum light detection technology provides a new approach to solving this problem. By utilizing quantum entanglement or the super-localization of quantum states, the detection sensitivity of photoelectric sensors can be improved. At the same time, by using dual-wavelength photoelectric sensing technology, the interference of non-fire factors can be reduced by measuring the difference in scattered light of different wavelengths. However, in practical applications, quantum light detection technology faces some challenges.

[0004] First, the generation and maintenance of quantum light sources require extremely precise control, which places high demands on the stability and reliability of the detection device. Secondly, the super-localization of quantum states is very sensitive to environmental conditions, and how to maintain the coherence of quantum states in a complex and ever-changing fire scene is a difficult problem. In addition, dual-wavelength photoelectric sensing technology requires precise control of the phase and intensity of the two wavelengths of light, which places higher demands on the design and manufacture of the detection device.

[0005] Therefore, how to ensure the high sensitivity and anti-interference ability of quantum light detection technology while improving the stability, reliability and environmental adaptability of the detection device is a key technical problem that needs to be solved urgently. This requires in-depth research in quantum light sources, quantum state control, dual-wavelength photoelectric sensing, etc., and combined with the actual application of smoke alarm systems, in order to ultimately achieve efficient and reliable smoke alarms. Summary of the invention

[0006] In order to solve the problems existing in the above-mentioned prior art, the purpose of the present invention is to provide a smoke alarm method and system based on quantum light detection, which can improve the stability, reliability and environmental adaptability of the detection device while ensuring the high sensitivity and anti-interference ability of quantum light detection technology.

[0007] To achieve the above object, the present invention provides the following solutions:

[0008] A smoke alarm method based on quantum light detection, comprising:

[0009] Acquire current environment information, and match quantum light source control parameters according to the current environment information; the quantum light source control parameters include: frequency, phase and intensity of the quantum light source;

[0010] By adjusting the control parameters of the quantum light source, a dual-wavelength detection light with quantum entanglement characteristics is generated; in the process of generating the dual-wavelength detection light, the current environmental information is input into the environmental prediction model to obtain the environmental prediction result of the future target period; the environmental prediction model is obtained by training the neural network model using the first training set; wherein the first training set includes: original environmental information;

[0011] Adaptively adjust the control parameters of the quantum light source by using the environmental prediction results and maintaining the initial quantum entanglement characteristics as a decision basis;

[0012] Based on the dual-wavelength detection light with quantum entanglement characteristics, scattered light signals are collected, a corresponding fire smoke scattering spectrum is established, a mapping relationship between the fire smoke scattering spectrum and smoke concentration is determined, fire smoke and non-fire aerosols are distinguished according to the mapping relationship, and a smoke alarm is triggered when fire smoke appears.

[0013] Optionally, the quantum entanglement characteristics include: entanglement degree and number of entangled photons.

[0014] Optionally, adaptively adjusting the quantum light source control parameters includes:

[0015] Step 1: Using the environmental prediction results and the initial quantum entanglement characteristics as the decision basis, randomly generate an initial particle group according to the quantum light source control parameters; each particle represents a scheme for adjusting the quantum light source control parameters, and a random speed is assigned to each particle to represent the change trend of the adjustment scheme;

[0016] Step 2, mapping the adjustment scheme into a position vector of the particle;

[0017] Step 3: construct a fitness function based on the decision basis to evaluate the pros and cons of the scheduling scheme represented by each particle;

[0018] Step 4: updating the velocity and position of the particle and performing a crossover mutation operation;

[0019] Step 5: preset a maximum number of iterations, repeat steps 3-4 according to the maximum number of iterations, and when the maximum number of iterations is reached, obtain the particle with the best fitness in the particle swarm as the optimal quantum light source control parameter, and adjust the quantum light source control parameter to the optimal quantum light source control parameter.

[0020] Optionally, updating the velocity of the particle includes:

[0021]

[0022] in, represents the new velocity of particle i in the jth dimension, w represents the inertia factor, V i j represents the current velocity of particle i in the jth dimension, a1 and a2 represent acceleration constants, r1 and r2 represent random numbers between [0,1], represents the individual optimal position of particle i in the jth dimension, represents the current position of particle i in the jth dimension, G represents the interference factor, represents the positions of other randomly selected particles, representing random exploration behavior.

[0023] Optionally, updating the position of the particle includes:

[0024]

[0025] in, represents the new position of particle i in the jth dimension.

[0026] Optionally, performing a crossover mutation operation on the particles includes:

[0027] The crossover operation on the particles includes:

[0028] Select two or more particles from the current particle group as intersection objects, randomly select an intersection point or multiple intersection points in the position vector of the particles, exchange the information of the intersection object position at the intersection point according to the selected intersection method, and obtain multiple new position vectors, which are new adjustment schemes;

[0029] The mutation operation on the particle includes:

[0030] Two or more particles are selected from the current particle group as mutation objects, a mutation method is preset, and position information of a pair of mutation objects is modified.

[0031] Optionally, establishing a corresponding fire smoke scattering spectrum includes:

[0032] Based on the dual-wavelength detection light with quantum entanglement characteristics, scattered light signals are collected, and an initial fire smoke scattering spectrum is obtained according to the scattered light signals. The initial fire smoke scattering spectrum is subjected to noise and outlier processing to obtain a standardized fire smoke scattering spectrum.

[0033] Optionally, distinguishing fire smoke and non-fire aerosol according to the mapping relationship includes:

[0034] Extracting key parameters reflecting the spectral characteristics of fire smoke and non-fire aerosol from the fire smoke scattering spectrum, constructing a scattering spectrum feature vector, inputting the scattering spectrum feature vector into a classification model of fire smoke and non-fire aerosol, and obtaining a classification result; the classification model is obtained by training with a second training set; the second training set includes: a scattering spectrum feature vector and a mapping relationship between the scattering spectrum feature vector and the smoke concentration;

[0035] In the process of training the model, the corresponding classification models of fire smoke and non-fire aerosol are trained according to the scattering spectrum data of different types of building environments.

[0036] To achieve the above object, the present invention also provides a smoke alarm system based on quantum light detection, comprising:

[0037] An environmental information acquisition module is used to acquire current environmental information and match quantum light source control parameters according to the current environmental information; the quantum light source control parameters include: frequency, phase and intensity of the quantum light source;

[0038] A quantum light generation module is used to generate a dual-wavelength detection light with quantum entanglement characteristics by adjusting the control parameters of the quantum light source; in the process of generating the dual-wavelength detection light, the current environmental information is input into the environmental prediction model to obtain the environmental prediction result of the future target period; the environmental prediction model is obtained by training the neural network model with the first training set; wherein the first training set includes: original environmental information;

[0039] A control parameter adjustment module, used to adaptively adjust the control parameters of the quantum light source by taking the environmental prediction result and maintaining the initial quantum entanglement characteristics as a decision basis;

[0040] The smoke alarm module is used to collect scattered light signals based on the dual-wavelength detection light with quantum entanglement characteristics, establish a corresponding fire smoke scattering spectrum, determine the mapping relationship between the fire smoke scattering spectrum and the smoke concentration, distinguish between fire smoke and non-fire aerosols according to the mapping relationship, and trigger a smoke alarm when fire smoke appears.

[0041] The beneficial effects of the present invention are:

[0042] The present invention discloses a smoke alarm method and system based on quantum light detection. The system uses a quantum entangled light source to generate dual-wavelength detection light, and adjusts the quantum state in real time through an adaptive algorithm to adapt to the complex and changeable fire scene environment.

[0043] The present invention collects parameters such as temperature, humidity, and air pressure, dynamically compensates for the deviations of quantum light sources and quantum state control, combines machine learning algorithms, extracts and classifies features of scattered light signals, constructs a classification model for fire smoke and non-fire aerosols, and realizes efficient and reliable smoke alarm. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0045] Figure 1 The figure is a flow chart of a smoke alarm method based on quantum light detection according to an embodiment of the present invention. DETAILED DESCRIPTION

[0046] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0047] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0048] like Figure 1 As shown, this embodiment discloses a smoke alarm method based on quantum light detection, including: obtaining current environmental information, matching quantum light source control parameters according to the current environmental information; the quantum light source control parameters include: frequency, phase and intensity of the quantum light source; by adjusting the quantum light source control parameters, generating dual-wavelength detection light with quantum entanglement characteristics; in the process of generating the dual-wavelength detection light, inputting the current environmental information into the environmental prediction model to obtain the environmental prediction result of the future target period; the environmental prediction model is obtained by training a neural network model using a first training set; wherein the first training set includes: original environmental information; using the environmental prediction result and maintaining the initial quantum entanglement characteristics as a decision basis, adaptively adjusting the quantum light source control parameters; based on the dual-wavelength detection light with quantum entanglement characteristics, collecting scattered light signals, establishing a corresponding fire smoke scattering spectrum, determining the mapping relationship between the fire smoke scattering spectrum and the smoke concentration, distinguishing fire smoke and non-fire aerosol according to the mapping relationship, and triggering a smoke alarm when fire smoke appears.

[0049] Generating dual-wavelength probe light with quantum entanglement properties includes:

[0050] Obtain current environmental condition information, including temperature, humidity, air pressure and other parameters, and use these parameters as input. According to the obtained environmental condition parameters, match them in the pre-established quantum entangled light source database to find the quantum light source control parameter record closest to the current environmental conditions. Extract the frequency, phase and intensity parameters of the quantum light source from the matched quantum light source control parameter record. Use the extracted quantum light source frequency, phase and intensity parameters as input, and accurately adjust the frequency, phase and intensity of the quantum light source through the quantum light source control unit. The quantum light source operates under the adjusted frequency, phase and intensity parameters to generate dual-wavelength detection light with quantum entanglement characteristics. Perform quantum state analysis on the generated dual-wavelength detection light to extract its quantum entanglement characteristic parameters, including entanglement degree, number of entangled photons, etc. The extracted quantum entanglement characteristic parameters are compared with the preset threshold. If the threshold is reached or exceeded, it is determined that the currently generated dual-wavelength detection light meets the quantum entanglement characteristic requirements and can be used as the initial quantum entanglement characteristics for subsequent quantum detection tasks; otherwise, return to step 1, re-acquire the current environmental conditions and match the quantum light source control parameters, and adjust and generate dual-wavelength detection light again until the quantum entanglement characteristic requirements are met.

[0051] Specifically, monitoring the environmental conditions of quantum entangled light sources is the key to achieving stable quantum detection. Environmental parameters include temperature, humidity, air pressure, etc., which directly affect the stability of quantum states. For example, in a laboratory environment, the temperature is controlled at around 20 degrees Celsius, the relative humidity is maintained at 40% to 60%, and the air pressure is maintained near a standard atmospheric pressure. Under these conditions, quantum entanglement works best. A large amount of historical operation data is stored in the quantum light source control parameter database. For example, at a specific temperature of 20 degrees and humidity of 50%, a stable quantum entangled state can be generated when the light source frequency is set to 500 terahertz, the phase difference is 90 degrees, and the light intensity is 100 milliwatts. By matching the current environment with the database records in real time, the optimal control parameters can be quickly found. The adjustment process of the quantum light source requires precise control. The frequency adjustment range is between 400 and 600 terahertz, the phase can be fine-tuned in the range of 0 to 360 degrees, and the light intensity varies in the range of 50 to 200 milliwatts. The combination of these parameters determines the quantum characteristics of the generated dual-wavelength detection light. The quantum entanglement characteristics of the dual-wavelength detection light are reflected in the degree of entanglement and the number of entangled photons. The degree of entanglement is represented by a value from zero to one, and it is usually required to be greater than 0.8 to meet the detection requirements. The number of entangled photons reflects the scale of the quantum state, and generally needs to reach more than one million pairs per second. These parameters can be measured by methods such as quantum state tomography. In practical applications, if the measured degree of entanglement is 0.75, which is lower than the preset threshold of 0.8, the system will automatically re-collect environmental parameters and adjust the light source. For example, when it is found that the temperature rises to 22 degrees, the light intensity can be reduced to 90 milliwatts and the frequency can be increased to 520 terahertz, so that the entanglement degree is increased to 0.85 to meet the detection requirements. This adaptive adjustment mechanism ensures the reliability of the quantum detection system. The stable generation of quantum entangled states is crucial to subsequent quantum detection. High-quality entangled photons significantly improve detection accuracy and sensitivity. Through environmental parameter monitoring, database matching, parameter optimization and other means, the stable and controllable operation of quantum entangled light sources can be achieved, laying the foundation for quantum detection applications.

[0052] Obtaining environmental forecast results for the future target period includes:

[0053] By utilizing dual-wavelength detection light, the detection sensitivity and anti-interference ability of the photoelectric sensor are improved, and a micro environmental sensor is integrated into the quantum light detection module to collect the temperature, humidity, and air pressure parameters of the environment in which the detection device is located in real time.

[0054] According to the characteristics of dual-wavelength detection light, the optical system of the photoelectric sensor is designed, the optical path structure and the photoelectric conversion circuit are optimized, and the response sensitivity and signal-to-noise ratio of the photoelectric sensor to the dual-wavelength light signal are improved. Through the signal processing and algorithm optimization of the dual-wavelength detection light, effective detection information is extracted, the influence of environmental interference and noise is reduced, and the anti-interference ability of the photoelectric sensor is enhanced. The micro-environmental sensor is integrated in the quantum light detection module to collect the temperature, humidity and air pressure parameters of the environment where the detection device is located in real time. The collected environmental parameters are fused with the quantum light detection signal, and the adaptive calibration and compensation of the detection signal are realized by establishing the association model between the environmental parameters and the detection signal. The quantum light detection signal and environmental parameters are trained and learned by machine learning algorithms, such as support vector machines or neural networks, and an environmentally adaptive detection model is established to improve the detection accuracy and reliability. According to the changing trend of environmental parameters, the working state of the detection device in the future is predicted, and the detection performance is dynamically optimized by adjusting the detection parameters and algorithms. The information of the dual-wavelength detection light and the environmental parameters are comprehensively utilized to construct an intelligent detection system with multi-source information fusion. Through decision-making algorithms and strategies, the optimal detection scheme is independently selected to realize adaptive control and optimization of the detection process.

[0055] Specifically, the optical system design of the photoelectric sensor needs to consider the characteristics of the dual-wavelength detection light. A beam splitter can be used to separate the dual-wavelength light, and the optical signal can be transmitted to different photodetectors through a fiber coupler. Taking dual-wavelength light with wavelengths of 800 nanometers and 1,000 nanometers as an example, a bandpass filter is used to filter different wavelengths, and the filter bandwidth is set to 10 nanometers, which can effectively improve the signal-to-noise ratio. The photoelectric conversion circuit uses a low-noise transimpedance amplifier with an adjustable gain range of 1,000 to 10,000 times and a dark current of less than one picoampere to achieve high-sensitivity photoelectric conversion. In terms of signal processing, wavelet transform is used to denoise and extract features from the dual-wavelength detection light signal. By setting a suitable threshold, high-frequency noise and environmental interference in the signal can be effectively suppressed. For example, for noise with a frequency above 1,000 Hz, a four-layer wavelet decomposition can be used to increase the signal-to-noise ratio by more than ten decibels. The integration of micro environmental sensors uses digital temperature and humidity sensors and air pressure sensors. The temperature measurement range is -40 degrees to 80 degrees, with an accuracy of 0.1 degree, the humidity measurement range is 0% to 100%, with an accuracy of 2%, and the air pressure measurement range is 300 to 1,100 Pa, with an accuracy of 100 Pa. The sensor communicates with the main controller through a serial interface, and the sampling period is one second. The fusion processing of environmental parameters and detection signals uses the Kalman filter algorithm to establish a state space model of environmental parameters and detection signals. For example, every degree of temperature change will cause the photodetector response to change by 0.5%. Through real-time compensation, the measurement error caused by temperature can be controlled within 0.1%. The machine learning model uses a three-layer neural network. The input layer contains environmental parameters such as temperature, humidity, and air pressure and dual-wavelength optical signal characteristics. The number of hidden layer neurons is 20, and the output layer is the corrected detection result. By collecting sample data under different environmental conditions for training, the detection accuracy after the model converges can be increased by 30%. The trend prediction of environmental parameter changes uses the time series analysis method to perform trend extrapolation on the environmental parameter data of the last hour and predict the parameter changes in the next ten minutes.

[0056] Furthermore, the characteristics of quantum entanglement include: the degree of entanglement and the number of entangled photons.

[0057] Furthermore, adaptively adjusting the quantum light source control parameters includes:

[0058] Step 1: Using the environmental prediction results and the initial quantum entanglement characteristics as the decision basis, the initial particle group is randomly generated according to the quantum light source control parameters; each particle represents a scheme for adjusting the quantum light source control parameters, and a random speed is assigned to each particle to represent the changing trend of the adjustment scheme;

[0059] Step 2: Map the adjustment scheme to the position vector of the particle;

[0060] Step 3: Construct a fitness function based on the decision-making basis to evaluate the pros and cons of the scheduling scheme represented by each particle;

[0061] Step 4: Update the speed and position of the particles and perform crossover and mutation operations;

[0062] Step 5: Preset the maximum number of iterations, repeat steps 3-4 according to the maximum number of iterations, and when the maximum number of iterations is reached, obtain the particle with the best fitness in the particle swarm as the optimal quantum light source control parameter, and adjust the quantum light source control parameter to the optimal quantum light source control parameter.

[0063] Constructing a fitness function based on decision-making criteria includes:

[0064] The maximum target thresholds of the frequency, phase and intensity of the quantum light source are clearly defined as the constraints for decision-making. The constraints are integrated into the decision-making basis, and a fitness function is defined to evaluate the objective function. Each adjustment scheme is scored. If the adjustment scheme violates the constraints, its fitness score is reduced by introducing a penalty term. The penalty term ensures that adjustment schemes that violate the constraints are avoided during the search process. The fitness score is normalized to guide the search behavior of particles.

[0065] Furthermore, updating the particle velocity includes:

[0066]

[0067] in, represents the new velocity of particle i in the jth dimension, w represents the inertia factor, V i j represents the current velocity of particle i in the jth dimension, a1 and a2 represent acceleration constants, r1 and r2 represent random numbers between [0,1], represents the individual optimal position of particle i in the jth dimension, represents the current position of particle i in the jth dimension, G represents the interference factor, represents the positions of other randomly selected particles, representing random exploration behavior.

[0068] Furthermore, updating the position of the particle includes:

[0069]

[0070] in, represents the new position of particle i in the jth dimension.

[0071] Furthermore, the crossover mutation operation on particles includes:

[0072] The crossover operations on particles include:

[0073] Select two or more particles from the current particle group as intersection objects, randomly select an intersection point or multiple intersection points in the particle position vector, exchange the information of the intersection object position at the intersection point according to the selected intersection method, and obtain multiple new position vectors, which are new adjustment schemes;

[0074] The mutation operations on particles include:

[0075] Select two or more particles from the current particle group as mutation objects, preset the mutation method, and modify the position information of a pair of mutation objects.

[0076] Furthermore, establishing the corresponding fire smoke scattering spectrum includes:

[0077] Based on dual-wavelength detection light with quantum entanglement characteristics, scattered light signals are collected. According to the scattered light signals, the initial fire smoke scattering spectrum is obtained. The initial fire smoke scattering spectrum is processed to remove noise and outliers to obtain a standardized fire smoke scattering spectrum.

[0078] Furthermore, the fire smoke and non-fire aerosol are distinguished according to the mapping relationship, including:

[0079] Extracting key parameters reflecting the spectral characteristics of fire smoke and non-fire aerosol from the fire smoke scattering spectrum, constructing a scattering spectrum feature vector, inputting the scattering spectrum feature vector into a classification model of fire smoke and non-fire aerosol, and obtaining a classification result; the classification model is trained using the second training set; the second training set includes: the scattering spectrum feature vector and its mapping relationship with smoke concentration;

[0080] In the process of training the model, the corresponding classification models of fire smoke and non-fire aerosol are trained according to the scattering spectrum data of different types of building environments.

[0081] The machine learning algorithm is used to extract and classify the scattered light signals collected by the dual-wavelength photoelectric sensor, and a scattered light difference model between fire smoke and non-fire aerosol is constructed. According to different types of building environments, a corresponding fire smoke scattering spectrum database is established.

[0082] The scattered light signals at the fire scene and in non-fire environments are collected by dual-wavelength photoelectric sensors to obtain initial scattered light spectrum data; the initial scattered light spectrum data is preprocessed to remove noise and outliers to obtain a standardized scattered light spectrum data set; key parameters reflecting the spectral characteristics of fire smoke and non-fire aerosols are extracted from the standardized scattered light spectrum data set to construct a scattered light spectrum feature vector; based on the scattered light spectrum feature vector, a support vector machine algorithm is used to train a classification model for fire smoke and non-fire aerosols; a decision tree algorithm is used to establish a mapping relationship between scattered light spectrum characteristics and smoke concentration to achieve quantitative analysis of fire smoke concentration; corresponding fire smoke recognition models are trained for scattered light spectrum data of different types of building environments to form a scattered light spectrum database for building fire detection; the trained machine learning model and scattered light spectrum database are deployed in the fire alarm controller to achieve real-time monitoring and early warning of fire smoke.

[0083] Specifically, the dual-wavelength photoelectric sensor distinguishes fire smoke from non-fire aerosols by collecting scattered light signals of different wavelengths. Taking the dual-wavelength sensor with wavelengths of 400 nanometers and 900 nanometers as an example, the ratio of the scattering intensity of fire smoke to these two wavelengths is usually between 0.6 and 0.8, while the scattering ratio of non-fire aerosols such as dust is between 0.3 and 0.5. This difference provides an important basis for fire identification. In the preprocessing of the initial scattering spectrum data, the median filtering method is used to remove sudden noise. For example, when the scattering intensity at a certain moment suddenly appears abnormally high, exceeding three times the previous and next data, the system will determine it as noise and remove it. For the long-term drift baseline noise, it is corrected by the polynomial fitting method to make the data baseline return to zero. The construction of the scattering spectrum feature vector includes multiple feature parameters. For example, the peak position of the scattering intensity usually appears around 500 nanometers in fire smoke, while the peak of dust is around 700 nanometers. The half-peak width of the scattering spectrum is also an important feature. The half-peak width of fire smoke is about 120 nanometers, which is significantly larger than the 80 nanometers of dust. The radial basis kernel function is used in the training of the support vector machine algorithm, and the classification accuracy is improved by adjusting the kernel function parameters. In the experiment, when the kernel function parameter value is 0.1, the model's recognition accuracy of fire smoke can reach 95%, and the false alarm rate is controlled below 3%.

[0084] The quantitative analysis of smoke concentration uses a decision tree algorithm to establish a corresponding relationship between the scattering spectrum characteristics and the smoke concentration. For example, when the scattering intensity increases by one unit, the corresponding smoke concentration increases by 0.05 mg / m3. Through this mapping relationship, real-time monitoring of smoke concentration can be achieved. The construction of scattering spectrum databases for different building environments needs to take into account the characteristics of the building. In the underground garage environment, due to the presence of a large amount of automobile exhaust, the model needs to pay special attention to the scattering characteristics of hydrocarbons. In the kitchen environment, it is necessary to focus on distinguishing the scattering characteristics of oil smoke and fire smoke. By establishing a targeted database, the accuracy of fire identification in different scenarios is improved. The real-time monitoring of the fire alarm controller adopts a sliding time window method to analyze the data of the last ten seconds every one second. The system will only trigger an alarm when three consecutive time windows show fire characteristics. This method effectively reduces the false alarm rate. The warning level is set according to the smoke concentration. When the concentration exceeds 0.1 mg / m3, a warning is issued, and when it exceeds 0.5 mg / m3, an emergency alarm is issued.

[0085] According to the scattering spectrum feature vector, the support vector machine algorithm is used to train the classification model of fire smoke and non-fire aerosol.

[0086] Obtain the scattering spectrum data of fire smoke and non-fire aerosol, pre-process the spectrum data, eliminate noise interference, and extract the spectrum feature vector. According to the scattering spectrum feature vector, use the support vector machine algorithm to train fire smoke and non-fire aerosol to obtain a classification model. For the newly collected unknown aerosol scattering spectrum data, extract its feature vector and input it into the trained classification model for prediction. If the output result of the classification model is fire smoke, it is judged that there is a fire risk in the current environment and a fire warning is triggered. If the output result of the classification model is non-fire aerosol, it is judged that the current environment is safe and no warning is triggered. By continuously collecting scattering spectrum data of fire smoke and non-fire aerosol, the classification model is continuously optimized and trained to improve the accuracy of fire smoke recognition. Deploy the trained classification model to embedded devices or cloud servers to achieve real-time monitoring and early warning of fire smoke.

[0087] Specifically, obtaining the scattered light spectrum data of fire smoke and non-fire aerosol is a basic step. For example, in a laboratory environment, the scattered light signals of smoke generated by wood burning and kitchen fume are collected by dual-wavelength photoelectric sensors. The sensor uses light sources with wavelengths of 450 nanometers and 700 nanometers, respectively, to record the scattered light intensity at different wavelengths. Assume that in the wood burning experiment, the scattered light intensity at a wavelength of 450 nanometers is 0.8, and the scattered light intensity at a wavelength of 700 nanometers is 0.5; while in the kitchen fume experiment, the scattered light intensity at a wavelength of 450 nanometers is 0.6, and the scattered light intensity at a wavelength of 700 nanometers is 0.4. Preprocessing the spectral data is a key link. Preprocessing includes denoising and standardization. For example, the wavelet transform method is used to remove high-frequency noise. Assuming that there is a high-frequency noise of 0.1 in the original data, the scattered light intensity becomes 0.7 and 0.4 (wood burning), 0.5 and 0.3 (kitchen fume) after processing. Next, standardization is performed to scale the data to the interval [0,1]. Assume that the wood burning data after standardization is [0.8,0.5] and the kitchen fume data is [0.6,0.4]. Extracting spectral feature vectors is a prerequisite for building a classification model. For example, characteristic parameters such as scattered light intensity ratio and spectral slope are extracted from standardized data. Assume that the feature vector of wood burning is [1.6,0.8] and the feature vector of kitchen fume is [1.5,0.75]. The support vector machine (SVM) algorithm is used to train fire smoke and non-fire aerosols to obtain a classification model. For example, the LibSVM tool is used to select the radial basis function (RBF) as the kernel function, and the optimal parameters are determined by cross-validation. Assume that during the training process, there are 100 wood burning samples and 100 kitchen fume samples, and the model training accuracy reaches 95%. For the newly collected unknown aerosol scattering spectrum data, its feature vector is extracted and input into the trained classification model for prediction. For example, if the feature vector of a newly collected set of data is [1.55, 0.78] after standardization, and the model output is "fire smoke" after input into the model, it is judged that there is a fire risk in the current environment and a fire warning is triggered. If the classification model outputs non-fire aerosol, the current environment is judged to be safe and no warning is triggered. For example, if the feature vector of another set of new data is [1.45, 0.72] and the model output is "non-fire aerosol", the system remains silent and does not issue an alarm. By continuously collecting scattering spectrum data of fire smoke and non-fire aerosol, the classification model is continuously optimized and trained to improve the accuracy of fire smoke identification. For example, 200 samples are added every quarter and the model is retrained. Assume that the accuracy of the new model is increased to 97%. Deploy the trained classification model to embedded devices or cloud servers to achieve real-time monitoring and early warning of fire smoke.For example, the model is embedded in the smart home system, connected to the cloud server via Wi-Fi, and sensor data is uploaded in real time. The server responds quickly and sends an alarm message to the user's mobile phone immediately once fire smoke is detected. The advantage of this method is that it can effectively distinguish fire smoke and non-fire aerosols through spectral feature extraction and machine learning algorithms, reducing the false alarm rate. The deployment of real-time monitoring and early warning systems improves the timeliness and accuracy of fire prevention and control, and ensures the safety of personnel and property. By continuously optimizing the model, the system can adapt to changes in different building environments, enhancing universality and reliability. In the specific implementation process, it is also necessary to consider the impact of environmental factors on sensors, such as temperature and humidity. For example, in a high temperature environment, the performance of the sensor may decline, and temperature compensation processing is required to ensure data accuracy. In addition, during the model training process, the number of samples of various types needs to be balanced to avoid overfitting problems and ensure the generalization ability of the model. Through the above examples and analysis from various aspects, it can be seen that the various links of the fire smoke detection system are interrelated, forming a complete monitoring and early warning system. The technical details and implementation methods of each link jointly ensure the stability and efficiency of the system, providing strong technical support for fire prevention and control.

[0088] This embodiment also discloses that whether a fire has occurred is determined by comparing the matching degree of the detection signal with the optimized fire discrimination model in real time, combining quantum light detection technology with traditional smoke sensing technology, and using a multi-sensor data fusion algorithm to comprehensively determine the possibility of fire occurrence.

[0089] The quantum light detection signal and smoke sensor data are obtained and input into the pre-established multi-sensor data fusion model for processing. According to the characteristics of the quantum light detection signal, the key parameters related to fire discrimination are extracted and the fire discrimination feature vector is constructed. The fire discrimination model is optimized by using the machine learning algorithm to improve the accuracy of the model in distinguishing fires. The optimized fire discrimination model is matched with the quantum light detection signal obtained in real time and the matching degree is calculated. At the same time, the smoke sensor data is analyzed to extract the parameters related to the fire, such as the smoke concentration. The matching degree of the quantum light detection signal and the smoke sensor parameters are input into the multi-sensor data fusion algorithm to comprehensively calculate the probability of fire occurrence. If the calculated probability of fire occurrence exceeds the preset threshold, it is determined that a fire may have occurred in the current environment and the fire alarm is triggered immediately.

[0090] Specifically, obtaining quantum light detection signals and smoke sensor data is the first step of the fire detection system. The quantum light detection signal captures the slight changes of photons through a highly sensitive photodetector, reflecting the spectral characteristics of a small flame in the early stage of a fire. For example, in a laboratory environment, the spectral characteristics of the flame are captured by a quantum light detector, and it is found that the flame has a significant light intensity change at a specific wavelength (such as 500 nanometers). This change can be used as one of the key parameters for fire identification. The acquired quantum light detection signal and smoke sensor data are input into a pre-established multi-sensor data fusion model for processing in order to integrate the information of multiple sensors and improve the accuracy of fire identification. The multi-sensor data fusion model can be a complex system based on a neural network, which can process and integrate data from different sensors to eliminate the possible false alarm or missed alarm problems of a single sensor. For example, in an office environment, the quantum light detector captures the spectral characteristics of a suspected flame, and the smoke sensor simultaneously detects the increase in smoke concentration in the air. After the data of the two are fused, the system can more reliably identify the fire. According to the characteristics of the quantum light detection signal, the key parameters related to fire identification are extracted to construct a fire identification feature vector. These key parameters may include the rate of change of light intensity, spectral distribution characteristics, etc. For example, in an experiment, the quantum light detector recorded that the intensity of the flame spectrum at 500 nanometers rose rapidly, with a change rate of 10% per second. This parameter was included in the fire discrimination feature vector. Using machine learning algorithms to optimize the fire discrimination model is an important means to improve the accuracy of model discrimination. A large amount of historical data can be trained and the parameters of the model can be optimized through algorithms such as support vector machines (SVM) or random forests. For example, through training, it was found that when the intensity change rate of the quantum light detection signal exceeded 5% per second and the smoke concentration exceeded 0.1 mg / m3, the model's accuracy in distinguishing fires increased from 80% to 95%. The optimized fire discrimination model is matched with the real-time quantum light detection signal to calculate the matching degree. The matching degree can be obtained by calculating the similarity between the actual signal and the model's expected signal. For example, the intensity change rate of the real-time signal at 500 nanometers is 8% per second, which is compared with the expected change rate of 10% per second of the model, and the matching degree is calculated to be 80%. At the same time, the smoke sensor data is analyzed to extract fire-related parameters such as smoke concentration. Smoke concentration is an important indicator for fire detection and is usually measured by electrochemical sensors or optical sensors. For example, in a fire simulation experiment, the smoke sensor detected that the smoke concentration rose rapidly from 0.05 mg / m3 to 0.2 mg / m3. This change was recorded and used for subsequent analysis. The matching degree of quantum light detection signals and smoke sensor parameters are input into the multi-sensor data fusion algorithm to comprehensively calculate the probability of fire. The multi-sensor data fusion algorithm can calculate the comprehensive probability of fire by combining the information of each sensor through weighted averaging, Bayesian network and other methods.For example, the matching degree of quantum light detection signal is 80%, and the smoke concentration parameter is 0.15 mg / m3. The probability of fire is 70% calculated by the fusion algorithm. If the calculated probability of fire exceeds the preset threshold, it is determined that a fire may have occurred in the current environment, and the fire alarm is triggered immediately. The preset threshold is usually set according to the actual environment and safety requirements. For example, the threshold is set to 60%. When the calculated probability of fire exceeds 60%, the system automatically triggers the alarm and notifies relevant personnel to take emergency measures. Through the above steps, the fire detection system can comprehensively utilize quantum light detection signals and smoke sensor data to improve the accuracy and reliability of fire identification, issue early warnings in time, and ensure the safety of personnel and property. This multi-sensor data fusion method not only improves the system's anti-interference ability, but also can detect and alarm in time at the early stage of a fire, which has important application value.

[0091] The fire identification model is optimized using machine learning algorithms to improve the model's accuracy in identifying fires.

[0092] According to the business needs of fire identification, historical data related to fire is obtained, the data is preprocessed, and features related to fire identification are extracted. According to the extracted fire features, a suitable machine learning algorithm, such as support vector machine, random forest or neural network, is selected to build a fire identification model. The preprocessed data is divided into a training set and a test set. The fire identification model is trained using the training set data to obtain the trained model parameters. The trained fire identification model is tested using the test set data to obtain the model's discrimination accuracy, and the model performance is evaluated based on the accuracy. If the model's discrimination accuracy does not reach the preset threshold, the model's feature importance is analyzed to screen out feature subsets that have a greater impact on the discrimination results. Using the data of the feature subset, the hyperparameters of the fire identification model are optimized, and the optimal model hyperparameter combination is obtained through methods such as grid search. The fire identification model is retrained using the optimized hyperparameters to obtain a model with better performance and improve the model's accuracy in fire identification.

[0093] Specifically, the business requirements for fire identification mainly include key indicators such as temperature changes, smoke concentration, and carbon monoxide concentration when a fire occurs. Historical data can be obtained from real fire case records, fire drill data, and laboratory simulation data. In the data preprocessing stage, outliers need to be processed. For example, when temperature data has negative values ​​or extreme values ​​exceeding one thousand degrees, it needs to be eliminated or corrected based on the actual situation. In the feature extraction process, dynamic features such as the temperature rise rate and the smoke concentration change rate can be calculated. For example, in a certain experiment, it was recorded that the temperature at the ignition point rose rapidly from 25 degrees to 120 degrees within three minutes. This drastic temperature change is an important characteristic indicator of fire. At the same time, the carbon monoxide concentration rose rapidly from zero to 50 milligrams per cubic meter. This change trend is also an important basis for identification. The selection of machine learning algorithms needs to consider data characteristics and actual application scenarios. Support vector machines are suitable for processing high-dimensional feature data and can effectively distinguish between normal environments and fire environments. Random forests have good feature importance evaluation capabilities and can identify key indicators that play a decisive role in fire identification. Neural networks are good at processing nonlinear relationships and can learn complex fire feature patterns. In the model training stage, the cross-validation method can be used to evaluate the model performance. For example, the data can be divided into training set and test set in a ratio of seven to three, and the risk of overfitting of the model can be reduced by multiple random divisions. The model performance evaluation should not only focus on the overall accuracy, but also focus on the false alarm rate and false alarm rate. In the field of fire safety, false alarms may have serious consequences, so the false alarm rate can be reduced by adjusting the model parameters. Feature importance analysis shows that the contribution of temperature change rate, smoke concentration and carbon monoxide concentration to fire discrimination is about 40%, 30% and 20% respectively. Based on this analysis result, the weight of important features can be appropriately increased, or the combination of these features can be added to improve the model performance. In the process of model hyperparameter optimization, different parameter combinations can be set for comparative experiments. For example, for support vector machines, different kernel functions and penalty parameters can be tried; for random forests, the number and depth of trees can be adjusted; for neural networks, the structure and activation function of the hidden layer can be changed. Through these optimization methods, the accuracy of the model on the test set was finally increased from the original 85% to 92%, and the false alarm rate was reduced to less than 0.5%.

[0094] Step S106, multiple quantum light detection modules are configured in the smoke alarm system, and redundant design and fault diagnosis technology are used to timely detect and isolate the faulty module by comparing the detection results of each module. When an abnormality is detected, it automatically switches to the backup module to ensure the reliability and continuous working capability of the system.

[0095] Acquire the real-time detection data of the quantum light detection module, pre-process the data, eliminate noise interference, and improve data quality. Input the pre-processed detection data into the anomaly detection model, use the isolation forest algorithm for anomaly detection, and determine whether there is a faulty module. If an anomaly is detected, the fault diagnosis process is triggered, and the location and type of the faulty module are determined through correlation analysis and causal reasoning. According to the results of the fault diagnosis, the faulty module is automatically isolated and removed from the system to avoid affecting the system performance. At the same time, the backup module is started to ensure the continuous working ability of the system through redundant design and maintain the reliability of the smoke alarm function. The incremental learning method is used to continuously optimize the anomaly detection model and the fault diagnosis model to improve the accuracy and adaptability of the model. The quantum light detection module is calibrated and maintained regularly to ensure the accuracy and stability of the detection results and reduce the probability of failure.

[0096] Specifically, the quantum light detection module will be disturbed by factors such as ambient light and temperature fluctuations during real-time detection, and the original data needs to be preprocessed. Taking a fire warning system as an example, the detection signal is denoised by wavelet transform, and the signal-to-noise ratio is increased from the initial 15 decibels to 25 decibels, which significantly improves the data quality. In terms of anomaly detection, the isolation forest algorithm identifies abnormal points by constructing a decision tree. For example, in a factory's fire warning system, the normal quantum light detection signal intensity should fluctuate between 0.5 and 1.5 milliwatts. When the signal intensity is detected to suddenly drop to 0.1 milliwatt, the system immediately determines it as abnormal. Through correlation analysis, it is found that the anomaly is related to the performance degradation of the detector's photoelectric conversion module. Fault diagnosis uses causal reasoning methods to establish a fault tree model. For example, in a shopping mall's early warning system, after the system detects an abnormal optical signal, it locates the failure of the gain circuit of the photoelectric detector by analyzing multiple parameters such as signal waveform characteristics and voltage changes. The system immediately starts the backup detector to ensure that the monitoring function is not interrupted. In order to improve the reliability of the system, a triple redundant design is adopted. Each monitoring point is equipped with three independent quantum light detection modules, and the final result is determined by a voting mechanism. When a module fails, the other two modules can still ensure the normal operation of the system. In the actual application of a high-rise building, this design makes the system reliability reach 99.9%. In terms of incremental learning, the system continuously optimizes the model by accumulating new fault cases. For example, in the application of a chemical plant, the system could only identify five common fault types in the early stage. After one year of operation, it could accurately identify twelve different types of faults through learning the accumulated data, and the detection accuracy was increased from 85% to 95%. Regular calibration and maintenance are the key to ensuring the stability of the system. The detector is calibrated with a standard light source to ensure measurement accuracy. In the application case of a data center, calibration is performed once a month to control the detection error within 0.5%. At the same time, the system automatically records the operating parameters of each module. When some parameters show abnormal trends, maintenance warnings are issued in advance, which effectively reduces the occurrence of faults.

[0097] To achieve the above object, the present invention also provides a smoke alarm system based on quantum light detection, comprising:

[0098] The environment information acquisition module is used to obtain the current environment information and match the quantum light source control parameters according to the current environment information; the quantum light source control parameters include: the frequency, phase and intensity of the quantum light source;

[0099] The quantum light generation module is used to generate dual-wavelength detection light with quantum entanglement characteristics by adjusting the control parameters of the quantum light source; in the process of generating the dual-wavelength detection light, the current environmental information is input into the environmental prediction model to obtain the environmental prediction result of the future target period; the environmental prediction model is obtained by training the neural network model with the first training set; wherein the first training set includes: original environmental information;

[0100] A control parameter adjustment module is used to adaptively adjust the control parameters of the quantum light source by taking the environmental prediction results and maintaining the initial quantum entanglement characteristics as the basis for decision-making;

[0101] The smoke alarm module is used to collect scattered light signals based on dual-wavelength detection light with quantum entanglement characteristics, establish the corresponding fire smoke scattering spectrum, determine the mapping relationship between the fire smoke scattering spectrum and the smoke concentration, distinguish between fire smoke and non-fire aerosols according to the mapping relationship, and trigger the smoke alarm when fire smoke appears.

[0102] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. A smoke alarm method based on quantum light detection, characterized in that: include: Acquire current environment information, and match quantum light source control parameters according to the current environment information; The quantum light source control parameters include: frequency, phase and intensity of the quantum light source; By adjusting the control parameters of the quantum light source, a dual-wavelength detection light with quantum entanglement characteristics is generated; in the process of generating the dual-wavelength detection light, the current environmental information is input into the environmental prediction model to obtain the environmental prediction result of the future target period; the environmental prediction model is obtained by training the neural network model using the first training set; wherein the first training set includes: original environmental information; Adaptively adjust the control parameters of the quantum light source by using the environmental prediction results and maintaining the initial quantum entanglement characteristics as a decision basis; Based on the dual-wavelength detection light with quantum entanglement characteristics, scattered light signals are collected, a corresponding fire smoke scattering spectrum is established, a mapping relationship between the fire smoke scattering spectrum and smoke concentration is determined, fire smoke and non-fire aerosols are distinguished according to the mapping relationship, and a smoke alarm is triggered when fire smoke appears.

2. The smoke alarm method based on quantum light detection according to claim 1 is characterized in that: The quantum entanglement characteristics include: entanglement degree and number of entangled photons.

3. The smoke alarm method based on quantum light detection according to claim 1 is characterized in that: Adaptively adjusting the quantum light source control parameters includes: Step 1: Using the environmental prediction results and the initial quantum entanglement characteristics as the decision basis, randomly generate an initial particle group according to the quantum light source control parameters; each particle represents a scheme for adjusting the quantum light source control parameters, and a random speed is assigned to each particle to represent the change trend of the adjustment scheme; Step 2, mapping the adjustment scheme into a position vector of the particle; Step 3: construct a fitness function based on the decision basis to evaluate the pros and cons of the scheduling scheme represented by each particle; Step 4: updating the velocity and position of the particle and performing a crossover mutation operation; Step 5: preset a maximum number of iterations, repeat steps 3-4 according to the maximum number of iterations, and when the maximum number of iterations is reached, obtain the particle with the best fitness in the particle swarm as the optimal quantum light source control parameter, and adjust the quantum light source control parameter to the optimal quantum light source control parameter.

4. The smoke alarm method based on quantum light detection according to claim 3 is characterized in that: Updating the velocity of the particle includes: Among them, V i j+1 represents the new velocity of particle i in the jth dimension, w represents the inertia factor, V i j represents the current velocity of particle i in the jth dimension, a1 and a2 represent acceleration constants, r1 and r2 represent random numbers between [0,1], represents the individual optimal position of particle i in the jth dimension, represents the current position of particle i in the jth dimension, G represents the interference factor, represents the positions of other randomly selected particles, representing random exploration behavior.

5. The smoke alarm method based on quantum light detection according to claim 3 is characterized in that: Updating the position of the particle includes: in, represents the new position of particle i in the jth dimension.

6. The smoke alarm method based on quantum light detection according to claim 3 is characterized in that: The crossover mutation operation on the particles includes: The crossover operation on the particles includes: Select two or more particles from the current particle group as intersection objects, randomly select an intersection point or multiple intersection points in the position vector of the particles, exchange the information of the intersection object position at the intersection point according to the selected intersection method, and obtain multiple new position vectors, which are new adjustment schemes; The mutation operation on the particle includes: Two or more particles are selected from the current particle group as mutation objects, a mutation method is preset, and position information of a pair of mutation objects is modified.

7. The smoke alarm method based on quantum light detection according to claim 1 is characterized in that: Establishing the corresponding fire smoke scattering spectrum includes: Based on the dual-wavelength detection light with quantum entanglement characteristics, scattered light signals are collected, and an initial fire smoke scattering spectrum is obtained according to the scattered light signals. The initial fire smoke scattering spectrum is subjected to noise and outlier processing to obtain a standardized fire smoke scattering spectrum.

8. The smoke alarm method based on quantum light detection according to claim 1 is characterized in that: Distinguishing fire smoke and non-fire aerosol according to the mapping relationship includes: Extracting key parameters reflecting the spectral characteristics of fire smoke and non-fire aerosol from the fire smoke scattering spectrum, constructing a scattering spectrum feature vector, inputting the scattering spectrum feature vector into a classification model of fire smoke and non-fire aerosol, and obtaining a classification result; the classification model is obtained by training with a second training set; the second training set includes: a scattering spectrum feature vector and a mapping relationship between the scattering spectrum feature vector and the smoke concentration; In the process of training the model, the corresponding classification models of fire smoke and non-fire aerosol are trained according to the scattering spectrum data of different types of building environments.

9. A smoke alarm system based on quantum light detection, characterized in that: include: An environmental information acquisition module is used to acquire current environmental information and match quantum light source control parameters according to the current environmental information; The quantum light source control parameters include: frequency, phase and intensity of the quantum light source; A quantum light generation module is used to generate a dual-wavelength detection light with quantum entanglement characteristics by adjusting the control parameters of the quantum light source; in the process of generating the dual-wavelength detection light, the current environmental information is input into the environmental prediction model to obtain the environmental prediction result of the future target period; the environmental prediction model is obtained by training the neural network model with the first training set; wherein the first training set includes: original environmental information; A control parameter adjustment module, used to adaptively adjust the control parameters of the quantum light source by taking the environmental prediction result and maintaining the initial quantum entanglement characteristics as a decision basis; The smoke alarm module is used to collect scattered light signals based on the dual-wavelength detection light with quantum entanglement characteristics, establish a corresponding fire smoke scattering spectrum, determine the mapping relationship between the fire smoke scattering spectrum and the smoke concentration, distinguish between fire smoke and non-fire aerosols according to the mapping relationship, and trigger a smoke alarm when fire smoke appears.

Images