Hrr-based acoustic fire type identification method, device, equipment and medium

Abstract

The application discloses an HRR-based acoustic fire type identification method and device, equipment and medium, and relates to the technical field of fire detection, which comprises the following steps: obtaining an original acoustic signal by emitting high-frequency modulated ultrasonic waves and receiving flame area reflected echoes, obtaining channel impulse response data through channel estimation, extracting flame features and calculating a heat release rate curve, and finally completing fire type identification and fire stage determination based on the curve. The method can resist smoke, shielding and light interference, realize accurate quantification of heat release rate, finely identify fire types and determine fire development stages, and provide stable and reliable technical support for fire monitoring and emergency decision-making.

Description

Technical Field

[0001] This invention relates to the field of fire detection technology, and in particular to an acoustic fire type identification method, apparatus, equipment and medium based on HRR. Background Technology

[0002] Existing fire analysis technologies mainly include three categories: traditional fire monitoring methods, visual imaging-based quantitative fire analysis systems, and ultrasonic time-of-flight detection schemes. Traditional fire monitoring methods, such as smoke detection and temperature monitoring, collect environmental parameters by deploying smoke and temperature sensors. When these parameters exceed preset thresholds, an alarm is triggered, enabling qualitative fire detection. Visual imaging-based quantitative fire analysis systems use cameras to capture images of the fire scene. Image processing algorithms extract features such as flame area, color, and texture, and combined with thermal radiation models to calculate the heat release rate, achieving a quantitative assessment of the fire's intensity. Ultrasonic time-of-flight threshold fire detection schemes measure the change in sound wave propagation time along a fixed monitoring path. Utilizing the direct proportionality between sound speed and temperature, a fire is detected and an alarm is triggered when the time difference exceeds a preset threshold.

[0003] However, traditional sensing methods cannot assess the scale, development trend, and type of fire; visual perception is easily affected by dense smoke, obstruction, and high temperatures, and heat release rate (HRR) data is prone to distortion, and there is a risk of privacy leakage; ultra-time-of-flight-based fire detection schemes can only provide qualitative alarms and cannot analyze spatial characteristics such as channel impulse response, nor can they establish a precise correlation between acoustic signals and HRR; existing HRR analysis dimensions are limited, and the differences in HRR morphology among different fires are not explored, which cannot support refined fire situation assessment and cannot meet the needs of intelligent monitoring and decision-making in complex fire scenes.

[0004] Therefore, how to achieve accurate quantitative analysis of fire heat release rate based on acoustic perception, and extract multi-dimensional fire characteristic information to identify fire type and determine fire stage has become an urgent problem to be solved. Summary of the Invention

[0005] The main objective of this application is to provide an acoustic fire type identification method, device, equipment, and medium based on HRR, aiming to solve the technical problem of how to accurately quantify HRR and identify fire types and determine fire stages based on acoustic perception.

[0006] To achieve the above objectives, this application proposes an acoustic fire type identification method based on HRR, comprising:

[0007] The system transmits high-frequency modulated ultrasonic signals to the monitoring area and receives the echo signals reflected by the flame area to obtain the original acoustic signals.

[0008] Channel estimation is performed based on the original acoustic signal and the ultrasonic signal to obtain channel impulse response data;

[0009] Flame feature extraction and heat release rate calculation are performed on the channel impulse response data to obtain the heat release rate curve;

[0010] Fire type identification and fire stage determination are performed based on the heat release rate curve to obtain the fire type identification result.

[0011] In one embodiment, the step of performing channel estimation based on the original acoustic signal and the ultrasonic signal to obtain channel impulse response data includes:

[0012] Retrieve the transmitted signal frame that matches the ultrasonic signal;

[0013] The original acoustic signal is subjected to carrier demodulation processing to obtain a baseband received signal of the same standard as the transmitted signal frame;

[0014] The baseband received signal and the transmitted signal frame are synchronized and aligned on the time axis to obtain the aligned transmitted signal frame and the aligned baseband received signal.

[0015] The aligned baseband received signal and the aligned transmitted signal frame are processed to obtain the initial channel impulse response data.

[0016] The initial channel impulse response data is subjected to amplitude normalization processing to remove clutter components with amplitudes lower than a preset effective threshold, thereby obtaining the processed response data.

[0017] The processed response data is divided into frames and encapsulated according to a preset frame length to obtain channel impulse response data.

[0018] In one embodiment, the step of performing calculations on the aligned baseband received signal and the aligned transmitted signal frame to obtain initial channel impulse response data includes:

[0019] Perform point-by-point cross-correlation operation between the aligned baseband received signal and the aligned transmitted signal frame to obtain a cross-correlation result sequence.

[0020] Peak detection is performed on the cross-correlation result sequence to extract multipath peak components that characterize the signal transmission path;

[0021] Based on the amplitude and delay information of the multipath peak components, an initial response sequence for the multipath transmission channel is constructed.

[0022] The initial response sequence is subjected to time-domain zero-padding and amplitude normalization to generate fixed-length channel response data;

[0023] The fixed-length channel response data is used as the initial channel impulse response data.

[0024] In one embodiment, the step of extracting flame features and calculating the heat release rate from the channel impulse response data to obtain the heat release rate curve includes:

[0025] Perform adjacent frame differential operation on the channel impulse response data to obtain differential channel data containing flame characteristics;

[0026] The time delay range of the flame effect is parsed from the differential channel data, and signal taps with intensity exceeding a preset effective threshold within the time delay range are selected and the corresponding total energy is calculated.

[0027] The total energy is corrected by combining the fuel specific heat capacity parameter to obtain the corrected energy data;

[0028] The corrected energy data is smoothed using a moving average algorithm to obtain smoothed energy data.

[0029] Based on the physical correlation model, the smoothed energy data is converted into corresponding time series heat release rate values.

[0030] The heat release rate values ​​are subjected to curve fitting and normalization along the time axis to generate a heat release rate curve characterizing the energy release pattern of the fire scene.

[0031] In one embodiment, the step of performing adjacent frame differential operations on the channel impulse response data to obtain differential channel data containing flame characteristics includes:

[0032] The first frame of channel impulse response data and the second frame of channel impulse response data are retrieved from the channel impulse response data in the order of time acquisition.

[0033] Perform tap-by-tap amplitude differential operation on the first frame of channel impulse response data and the second frame of channel impulse response data to obtain the original differential channel data;

[0034] Extract the spatial angle information from the original differential channel data;

[0035] Based on the angle information, beamforming technology is used to focus and enhance the signal in the direction of the fire source to obtain the processed data.

[0036] Data with amplitudes lower than the effective amplitude threshold of the signal in the processed data are removed to obtain the differential channel data after removal.

[0037] The removed differential channel data is then subjected to time-domain normalization to obtain differential channel data containing flame characteristics.

[0038] In one embodiment, the step of converting the smoothed energy data into corresponding time-series heat release rate values ​​based on a physical correlation model includes:

[0039] Establish a physical correlation model between ultrasonic propagation delay, fire temperature distribution, air density, and heat release rate;

[0040] The smoothed energy data is time-matched with the flame impact delay range to generate a time-aligned energy-delay sequence.

[0041] The time-aligned energy-delay sequence is input into the physical correlation model to calculate the initial value of the single-frame heat release rate.

[0042] The initial value of the single-frame heat release rate is calibrated by combining preset fuel type parameters to obtain the calibrated heat release rate value.

[0043] All the calibrated heat release rate values ​​are arranged in time frame order to form the corresponding time series of heat release rate values.

[0044] In one embodiment, the step of identifying the fire type and determining the fire stage based on the heat release rate curve to obtain the fire type identification result includes:

[0045] The heat release rate curve is subjected to moving average to remove fluctuations, and morphological features are extracted, wherein the morphological features include peak heat release rate, time to peak, average heat release rate, and decay rate.

[0046] The morphological features are input into a preset neural network classifier to obtain type discrimination feature values;

[0047] The type discrimination feature value is compared with a preset fire type feature library to determine the fire type matching result, wherein the fire type matching result includes fuel-controlled fire, ventilation-limited fire, pulse fire and ventilation-induced flashover fire;

[0048] The fire stage is determined by the temporal change trend of the heat release rate curve, wherein the fire stage includes the initial growth stage, the flashover stage, the complete combustion stage, or the decay stage.

[0049] Based on the fire type matching result and the fire stage, the fire type identification result is output.

[0050] Furthermore, to achieve the above objectives, this application also proposes an acoustic fire type identification device based on HRR, wherein the HRR-based acoustic fire type identification device includes:

[0051] The acquisition module is used to transmit high-frequency modulated ultrasonic signals to the monitoring area and receive the echo signals reflected by the flame area to obtain the original acoustic signals.

[0052] The estimation module is used to perform channel estimation based on the original acoustic signal and the ultrasonic signal to obtain channel impulse response data;

[0053] The calculation module is used to extract flame features and calculate heat release rate from the channel impulse response data to obtain a heat release rate curve.

[0054] The results module is used to identify the fire type and determine the fire stage based on the heat release rate curve, and obtain the fire type identification result.

[0055] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the HRR-based acoustic fire type identification method described above.

[0056] In addition, to achieve the above objectives, this application also provides a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the HRR-based acoustic fire type identification method described above.

[0057] This application acquires the original acoustic signal by transmitting high-frequency modulated ultrasonic waves and receiving the reflected echoes from the flame area. Channel impulse response data is obtained through channel estimation, flame features are extracted, and a heat release rate curve is calculated and generated. Finally, based on the curve, fire type identification and fire stage determination are completed. It can resist interference from dense smoke, obstruction, and light, and achieve precise quantification of heat release rate. It can accurately identify fire type and determine fire development stage, providing stable and reliable technical support for fire monitoring and emergency decision-making. Attached Figure Description

[0058] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0059] Figure 1 This is a flowchart illustrating the first embodiment of the acoustic fire type identification method based on HRR in this application;

[0060] Figure 2 HRR curve for a fuel-controlled fire according to the first embodiment of the HRR-based acoustic fire type identification method of this application;

[0061] Figure 3 HRR curve for a ventilation-restricted fire according to the first embodiment of the acoustic fire type identification method based on HRR in this application;

[0062] Figure 4 HRR curve of ventilation-induced flashover fire in the first embodiment of the acoustic fire type identification method based on HRR of this application;

[0063] Figure 5 The HRR curve of a pulsed fire is shown in the first embodiment of the acoustic fire type identification method based on HRR of this application.

[0064] Figure 6 This is a flowchart illustrating the second embodiment of the acoustic fire type identification method based on HRR in this application;

[0065] Figure 7 This is a schematic diagram of the module structure of the acoustic fire type identification device based on HRR in this application;

[0066] Figure 8 This is a schematic diagram of the device structure of the hardware operating environment involved in the HRR-based acoustic fire type identification method in the embodiments of this application.

[0067] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0068] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0069] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0070] Existing fire analysis technologies mainly include three categories: traditional fire monitoring methods, visual imaging-based quantitative fire analysis systems, and ultrasonic time-of-flight detection schemes. Traditional fire monitoring methods, such as smoke detection and temperature monitoring, collect environmental parameters by deploying smoke and temperature sensors. When these parameters exceed preset thresholds, an alarm is triggered, enabling qualitative fire detection. Visual imaging-based quantitative fire analysis systems use cameras to capture images of the fire scene. Image processing algorithms extract features such as flame area, color, and texture, and combined with thermal radiation models to calculate the heat release rate, achieving a quantitative assessment of the fire's intensity. Ultrasonic time-of-flight threshold fire detection schemes measure the change in sound wave propagation time along a fixed monitoring path. Utilizing the direct proportionality between sound speed and temperature, a fire is detected and an alarm is triggered when the time difference exceeds a preset threshold.

[0071] However, traditional sensing methods cannot assess the scale, development trend, and type of fire; visual perception is easily affected by dense smoke, obstruction, and high temperatures, and heat release rate (HRR) data is prone to distortion, and there is a risk of privacy leakage; ultra-time-of-flight-based fire detection schemes can only provide qualitative alarms and cannot analyze spatial characteristics such as channel impulse response, nor can they establish a precise correlation between acoustic signals and HRR; existing HRR analysis dimensions are limited, and the differences in HRR morphology among different fires are not explored, which cannot support refined fire situation assessment and cannot meet the needs of intelligent monitoring and decision-making in complex fire scenes.

[0072] Therefore, how to achieve accurate quantitative analysis of fire heat release rate based on acoustic perception, and extract multi-dimensional fire characteristic information to identify fire type and determine fire stage has become an urgent problem to be solved.

[0073] Based on the above, this application also provides an acoustic fire type identification method based on HRR, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the acoustic fire type identification method based on HRR of this application.

[0074] In this embodiment, the HRR-based acoustic fire type identification method includes steps S10 to S40:

[0075] Step S10: Transmit a high-frequency modulated ultrasonic signal to the monitoring area and receive the echo signal reflected by the flame area to obtain the original acoustic signal.

[0076] It should be noted that the monitoring area is a pre-defined spatial range for fire detection and data collection, covering the target location where flames appear. High-frequency modulated ultrasonic signals are acoustic signals formed by loading a basic signal frame onto a high-frequency carrier wave. This signal's frequency is higher than the range of human hearing, thus avoiding noise interference with the environment. The echo signal is the acoustic signal that propagates back after the emitted ultrasonic signal passes through the flame area and contacts a reflecting surface; it carries all the characteristic information about how the flame affects sound wave transmission.

[0077] Specifically, firstly, a pre-designed set of signal frame sequences with strong autocorrelation is retrieved. This set of signal frame sequences uses pseudo-random noise sequences or linear frequency modulation sequences as training sequence codes. The set of signal frame sequences is saved as an audio file and stored in a local buffer. The digital signal in the audio file format is converted into an analog electrical signal by a digital-to-analog converter. The analog electrical signal is then input into a power amplifier for signal amplification. The amplified analog electrical signal is transmitted to a loudspeaker, which modulates the amplified analog electrical signal onto a high-frequency carrier wave with a frequency higher than 18 kHz to form an ultrasonic signal, which is continuously transmitted to the monitoring area. This allows the ultrasonic signal to propagate within the monitoring area and be modulated by the temperature field when passing through the flame area. Simultaneously, the echo signal reflected by the flame area and the reflective surface is received by a microphone. The echo signal is a modulated acoustic signal carrying information about the temperature distribution of the fire area. The echo signal is input into a low-noise amplifier for pre-amplification. The amplified echo signal is then transmitted to an analog-to-digital converter for sampling to obtain a digital echo signal. The digital echo signal is digitally down-converted and demodulated at the same carrier frequency as the transmitter. After filtering out the high-frequency carrier component, the baseband signal is extracted and stored as the original acoustic signal in a data buffer. This is done because using training sequence codes with strong autocorrelation helps improve the accuracy of subsequent channel estimation. Modulating the signal to above 18 kHz ensures that the ultrasonic signal is outside the range of human hearing, avoiding noise interference to personnel. Acoustic signal transmission and reception are achieved through the cooperation of loudspeakers and microphones. The modulation effect of the flame temperature field on the speed of sound makes the echo signal carry fire field information, while demodulation is to recover the baseband signal that can be used for subsequent processing.

[0078] Step S20: Channel estimation is performed based on the original acoustic signal and ultrasonic signal to obtain channel impulse response data.

[0079] It should be noted that channel estimation is a process of calculating the characteristics of the signal transmission path by comparing the differences between the transmitted and received signals, and is used to fully reflect the changes in the signal in the propagation space. Channel impulse response data is a dataset describing all the characteristics of the acoustic transmission channel, used to record all physical changes in the signal, such as the time delay amplitude, during propagation.

[0080] Further, step S20 also includes: First, retrieving the transmitted signal frame that matches the ultrasonic signal. Specifically, retrieving the original transmitted signal frame corresponding to the currently transmitted ultrasonic signal from the local memory. This transmitted signal frame is a baseband digital signal sequence without carrier modulation, and its data format is consistent with the baseband signal after demodulation of the original acoustic signal. The transmitted signal frame is loaded into the memory buffer for subsequent processing, and the sampling rate and number of sampling points of the transmitted signal frame are recorded as reference parameters for synchronization alignment. This is done because channel estimation requires a known transmitted signal as a reference benchmark. The channel characteristics can only be accurately extracted by comparing the differences between the received signal and the transmitted signal. Maintaining the consistency of data format and parameters is a prerequisite for ensuring that subsequent cross-correlation operations and synchronization alignment can be performed correctly.

[0081] Next, the original acoustic signal undergoes carrier demodulation processing to obtain a baseband received signal of the same standard as the transmitted signal frame. Specifically, a sampled digital echo signal is obtained from an analog-to-digital converter. This digital echo signal is a modulated signal carrying a high-frequency carrier component. The digital echo signal is input into a digital mixer to generate a local oscillator signal, ensuring that the frequency of the local oscillator signal is consistent with the frequency of the modulated carrier at the transmitting end. The digital echo signal and the local oscillator signal are multiplied to achieve spectrum shifting. The mixed signal is input into a low-pass filter to filter out the high-frequency carrier component and the image frequency component, retaining the effective signal components within the baseband frequency range. The filtered baseband signal is downsampled to reduce the data volume and match the sampling rate of the transmitted signal frame. The downsampled baseband signal is then normalized to eliminate amplitude fluctuations caused by channel attenuation. The processed baseband signal is output as the baseband received signal to the next processing unit. This is done because digital downconversion demodulation can move the modulated signal carrying fire information from the high-frequency carrier to the baseband, filter out irrelevant high-frequency components, and extract the baseband received signal with the same format as the transmitted signal frame. Downsampling and normalization are to ensure that the baseband received signal and the transmitted signal frame are consistent in sampling rate and amplitude range, providing a data basis for format matching for subsequent synchronization alignment and cross-correlation operations.

[0082] Next, the baseband received signal and the transmitted signal frame are synchronized and aligned on the time axis to obtain aligned transmitted signal frames and aligned baseband received signals. Specifically, the baseband received signal and the transmitted signal frame are compared in the same time dimension, and the time positions of the two signals are adjusted according to the correspondence of signal feature points to obtain aligned transmitted signal frames and aligned baseband received signals with perfectly matched time positions. This ensures that the transmitted signal frame and the baseband received signal are strictly aligned on the time axis, providing a time-synchronized data foundation for subsequent channel impulse response calculations.

[0083] Subsequently, the aligned baseband received signal and the aligned transmitted signal frame are processed to obtain the initial channel impulse response data. Specifically, point-by-point cross-correlation is performed on the aligned baseband received signal and the aligned transmitted signal frame to obtain a cross-correlation result sequence; peak detection is performed on the cross-correlation result sequence to extract the multipath peak components that characterize the signal transmission path; based on the amplitude and delay information of the multipath peak components, the initial response sequence of the multipath transmission channel is constructed; the initial response sequence is then subjected to time-domain zero-padding and amplitude normalization to generate fixed-length channel response data; and this fixed-length channel response data is used as the initial channel impulse response data. Understandably, a point-by-point cross-correlation operation is performed between the aligned baseband received signal and the aligned transmitted signal frame. Specifically, the transmitted signal frame is used as a reference template, and the baseband received signal is shifted in steps of a single sampling point. After each shift, the values ​​of the corresponding sampling points of the two signals are multiplied and all products are accumulated to obtain the correlation value at that shift position. After traversing all shift positions, a cross-correlation result sequence is formed. The length of this sequence is the sum of the lengths of the two signals minus one, and each value in the sequence represents the signal similarity under different relative time delays. Next, peak detection is performed on the cross-correlation result sequence. A preset peak threshold is set, and a sliding window is used to traverse each data point in the sequence. When the value of a data point is greater than all adjacent points in the window and greater than the preset peak threshold, the point is determined to be a local peak point. The amplitude of all local peak points and their index positions in the sequence are recorded. Each local peak point is used as a representation of the multipath peak components of the signal reaching the receiver through different transmission paths. Then, an initial response sequence for the multipath transmission channel is constructed based on the amplitude and delay information of each multipath peak component. A discrete sequence is established with sampling points as the time unit. The amplitude of each multipath peak component is filled into the corresponding index position according to its index position, and the remaining index positions are padded with zeros, forming a sparse discrete sequence with delay as the horizontal axis and amplitude as the vertical axis. Subsequently, the initial response sequence is zero-padded in the time domain, and zero-value sampling points are added to the end of the sequence to make its total length reach the preset frame length. The zero-padded sequence is then normalized by dividing all amplitudes in the sequence by the maximum amplitude to make the maximum amplitude a unit amplitude. Finally, the fixed-length sequence after amplitude normalization is output as the initial channel impulse response data to the storage unit. This is done because pointwise cross-correlation can accurately estimate the channel's response intensity to signals with different delays. Peak detection is used to identify and extract multipath components generated by sound waves propagating through multiple paths. Constructing an initial response sequence transforms multipath information into a standard discrete form. Time-domain zero-padding and amplitude normalization are used to unify the data format and eliminate amplitude uncertainty, providing standardized channel impulse response data for subsequent flame feature extraction.

[0084] Then, the initial channel impulse response data is normalized to remove clutter components with amplitudes below a preset effective threshold, resulting in processed response data. Finally, the processed response data is framed and encapsulated according to a preset frame length to obtain the channel impulse response data. Specifically, the initial channel impulse response data is normalized by iterating through the amplitudes of all sampling points in the sequence, identifying the maximum amplitude, and dividing the amplitude of each sampling point by this maximum amplitude to map the amplitude range in the sequence to the interval between 0 and 1, resulting in normalized channel impulse response data. Next, a preset effective threshold is set, and the normalized channel impulse response data is iterated through. The amplitude of each sampling point is compared with the preset effective threshold. When the amplitude of a sampling point is lower than the preset effective threshold, the amplitude of that sampling point is set to zero. When the amplitude of a sampling point is greater than or equal to the preset effective threshold, the amplitude of that sampling point is kept unchanged. The processed sequence is used as the processed response data to remove clutter components generated by environmental noise and weak reflections. Then, the preset frame length parameter is obtained, and the ratio of the total length of the processed response data to the preset frame length is calculated. Based on the ratio, the number of frames to be divided is determined. Starting from the beginning of the sequence, data segments are sequentially extracted according to the preset frame length. When the last data segment is shorter than the preset frame length, zeros are padded to the end to the preset frame length. Each extracted data segment is encapsulated into an independent frame structure, with each frame containing a frame number, timestamp, and valid data payload. Finally, all the frame-encapsulated frame structures are arranged in chronological order to form a continuous channel impulse response data stream, which is then output to the next processing module. This is done because amplitude normalization can eliminate the influence of signal strength fluctuations at different measurement times, making each frame data comparable. Threshold-based clutter removal can filter out weak reflections and noise interference unrelated to the flame, retaining effective multipath signal components. Frame encapsulation is used to divide the continuous channel response data into standardized time units, facilitating subsequent frame-by-frame differential operations and flame feature extraction processing.

[0085] Step S30: Extract flame features and calculate heat release rate from the channel impulse response data to obtain the heat release rate curve.

[0086] It should be noted that heat release rate calculation is a process based on flame characteristic signals, combined with a physical correlation model, to convert signal energy data into a quantitative value characterizing the intensity of fire energy release. The core is to establish the correspondence between acoustic signal characteristics and fire heat release intensity. The heat release rate curve is a continuous curve formed by fitting the heat release rate values ​​calculated at different time points in chronological order along the time axis, which can intuitively reflect the changing patterns and development trends of fire energy release.

[0087] Specifically, firstly, adjacent frame differential operations are performed on the channel impulse response data. By comparing the amplitude differences between two consecutive frames, fixed signal components generated by static objects such as walls and equipment within the monitoring area are eliminated, while the dynamic signal changes caused by flame combustion are retained, resulting in differential channel data containing flame characteristics. Then, the time delay interval of the flame influence is extracted from the differential data. Signal taps with amplitudes higher than the effective threshold within this interval are selected, and their amplitudes are squared and summed to obtain the total energy representing the intensity of the flame disturbance. To improve calculation accuracy, the fuel specific heat capacity parameter is introduced to correct the total signal energy, offsetting the influence of different fuel types on sound wave propagation and energy conversion. Next, a moving average algorithm is used to smooth the corrected energy data, reducing numerical fluctuations caused by random noise. Finally, a pre-established physical correlation model is used to convert the smoothed signal energy into the corresponding heat release rate value. This model integrates ultrasonic propagation delay, fire temperature distribution, and air density variation, achieving a precise mapping from acoustic characteristics to heat release parameters. Finally, the heat release rate values ​​at each time point are fitted sequentially along the time axis and the amplitude is normalized to form a continuous, smooth heat release rate curve that can completely reflect the energy release pattern of the fire, providing a unified and reliable analytical basis for subsequent fire type identification and fire stage determination.

[0088] Step S40: Based on the heat release rate curve, identify the fire type and determine the fire stage to obtain the fire type identification result.

[0089] It should be noted that step S40 includes: performing a moving average to remove fluctuations from the heat release rate curve and extracting morphological features; inputting the morphological features into a preset neural network classifier to obtain type discrimination feature values; comparing the type discrimination feature values ​​with a preset fire type feature library to determine the fire type matching result, wherein the fire type matching result includes fuel-controlled fires, ventilation-limited fires, pulsating fires, and ventilation-induced flashover fires; determining the fire stage based on the temporal change trend of the heat release rate curve, wherein the fire stage includes the initial growth stage, flashover stage, complete combustion stage, or decay stage; and combining the fire type matching result with the fire stage to output the fire type identification result.

[0090] It's important to understand that the morphological features are the core set of parameters used to characterize the changing patterns of the heat release rate curve, including peak heat release rate, time to peak, average heat release rate, and decay rate. The preset neural network classifier can be a backpropagation (BP) neural network model or a shallow fully connected neural network model, etc.

[0091] Specifically, firstly, a database of standard heat release rate curves for four typical fire types is established, corresponding to fuel-controlled fires, ventilation-limited fires, pulsating fires, and ventilation-induced flashover fires, respectively. Based on combustion theory and experimental data, the evolution law of heat release rate for each type of fire is determined. Then, mathematical functions are used to model the standard curves for each type of fire, such as... Figure 2 The HRR curve of the fuel-controlled fire shown is as follows: Figure 3 The HRR curve shown is for ventilation-limited fires. Figure 4 The HRR curves for ventilation-induced flashover fires and Figure 5The HRR curves for pulsating fires shown are used to categorize fire types based on their HRR curves: fuel-controlled fires, ventilation-limited fires, ventilation-induced flashover fires, and pulsating fires. Fuel-controlled fires exhibit a monotonically increasing and then steadily decreasing function form; ventilation-limited fires exhibit a peak-limited function form modulated by the ventilation coefficient; pulsating fires exhibit a periodically oscillating function form; and ventilation-induced flashover fires exhibit a delayed, abruptly increasing function form. Next, the heat release rate curve is processed by moving average to remove fluctuations. A first preset window length and a second preset window length are set. The heat release rate curve is processed twice by moving average to obtain a first smooth curve and a second smooth curve. The fluctuation suppression effects of the first smooth curve and the second smooth curve are compared. The smooth curve with smaller fluctuation amplitude is selected as the target smooth curve. All data points in the target smooth curve are traversed, and the data point with the largest amplitude and its corresponding time index are identified. The largest amplitude is taken as the peak heat release rate, and the time index is taken as the peak time. The arithmetic mean of all data points in the target smooth curve is calculated as the average heat release rate. The data segment after the time index corresponding to the peak heat release rate is extracted, and the data segment is linearly fitted to obtain the descending slope. The absolute value of the descending slope is taken as the decay rate. The peak heat release rate, peak time, average heat release rate and decay rate are summarized as morphological features. Secondly, the morphological features are input into a pre-defined neural network classifier. The peak heat release rate, time to peak, average heat release rate, and decay rate are combined into a feature vector. This feature vector is normalized to ensure consistent numerical ranges across all dimensions. The normalized feature vector is then input into the trained neural network classifier. Taking a backpropagation (BP) neural network model as an example, the input layer receives four morphological features: peak heat release rate, time to peak, average heat release rate, and decay rate. The hidden layer performs weighted summation and nonlinear transformation on the input features, extracting the correlation patterns between features layer by layer. The output layer outputs the discrimination feature values ​​corresponding to four types of fires: fuel-controlled, ventilation-restricted, pulsating, and ventilation-induced flashover. Next, a pre-defined fire type feature library is retrieved. This library stores the standard feature vectors and classification thresholds for each type of fire: fuel-controlled, ventilation-restricted, pulsating, and ventilation-induced flashover. The type discrimination feature values ​​are compared with the standard feature vectors of each category in the feature library to calculate similarity. The category with the highest similarity is selected as the fire type matching result.Subsequently, the temporal trend of the heat release rate curve is determined. The first difference of each point on the curve is calculated to obtain the rate of change sequence. When multiple consecutive values ​​in the rate of change sequence are positive and gradually increase, it is determined to be the initial growth stage. When a sudden peak occurs in the rate of change sequence and the heat release rate reaches the global maximum value, it is determined to be the flashover stage. When the value in the rate of change sequence fluctuates slightly near zero and the heat release rate remains at a high level, it is determined to be the complete combustion stage. When multiple consecutive values ​​in the rate of change sequence are negative and the heat release rate continues to decrease, it is determined to be the decay stage. The determination results are used as the fire stage. Finally, the fire type matching results are fused with the fire stage to generate structured data containing fire type, fire stage, confidence level and timestamp. The structured data is output as the fire type identification result to the display terminal and storage unit. This is done because the moving average process can eliminate random fluctuations in the heat release rate curve and highlight the evolution trend. The four morphological features characterize the fire energy release characteristics from different dimensions. The neural network classifier uses machine learning algorithms to achieve automatic pattern recognition of fire types. The time-series trend determination identifies key turning points in fire development by analyzing the rate of change of the curve. The combination of the two forms a comprehensive understanding of the fire situation, providing accurate information on fire type and development stage for firefighting decisions.

[0092] This embodiment acquires the original acoustic signal by transmitting high-frequency modulated ultrasonic waves and receiving the reflected echoes from the flame area. Channel impulse response data is obtained through channel estimation, flame features are extracted, and a heat release rate curve is calculated and generated. Finally, based on the curve, fire type identification and fire stage determination are completed. It can resist interference from dense smoke, obstruction, and light, and achieve precise quantification of heat release rate. It can accurately identify fire type and determine fire development stage, providing stable and reliable technical support for fire monitoring and emergency decision-making.

[0093] Based on the first embodiment of this application, in the second embodiment of this application, the content that is the same as or similar to that in Embodiment 1 above can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 6 The HRR-based acoustic fire type identification method further includes steps S201-S206 in step S30:

[0094] Step S201: Perform adjacent frame differential operation on the channel impulse response data to obtain differential channel data containing flame characteristics.

[0095] It should be noted that step S201 includes: retrieving the first frame and the second frame of channel impulse response data from the channel impulse response data in the order of time acquisition; performing tap-by-tap amplitude differential operation on the first frame and the second frame of channel impulse response data to obtain the original differential channel data; extracting the spatial angle information from the original differential channel data; using beamforming technology based on the angle information to focus and enhance the signal in the direction of the fire source to obtain the processed data; removing data with amplitudes lower than the effective amplitude threshold of the signal from the processed data to obtain the removed differential channel data; and performing time-domain normalization processing on the removed differential channel data to obtain differential channel data containing flame characteristics.

[0096] It's important to understand that tap-by-tap amplitude difference calculation is an operation that calculates the amplitude difference of two temporally adjacent channel impulse response data points, sequentially using taps at the same positions. This operation preserves the flame characteristics that change over time while eliminating fixed signal components introduced by the static environment. A tap is a discrete sampling point in the time dimension of the channel impulse response data. Each tap corresponds to a fixed signal propagation delay position, used to record the signal amplitude information at that delay. In the channel impulse response data, all taps are arranged sequentially in time, forming a complete time-domain response sequence. Spatial angle information is the signal incident azimuth correlation parameter obtained from the differential channel data, reflecting the directional characteristics of the sound wave propagation within the monitoring area. Beamforming technology is a technique that weights and superimposes multiple signals based on target direction parameters, enhancing the signal energy in a specified direction. This technique can highlight the effective signal from the fire source direction and suppress interference signals from other directions.

[0097] Specifically, firstly, two adjacent frames of data are read from the channel impulse response data buffer in chronological order. The first frame is recorded as the first frame of channel impulse response data, and the second frame as the second frame, ensuring that both frames have the same number of sampling points and data format. Secondly, tap-by-tap amplitude difference calculation is performed on the first and second frames of channel impulse response data. The amplitudes at corresponding sampling index positions of the two frames are subtracted, and the absolute or signed difference is calculated. The calculation results of each sampling point are arranged in the original time sequence to form the original differential channel data, which suppresses static multipath components and highlights channel changes caused by dynamic flames. Then, spatial angle information is extracted from the original differential channel data. Based on the geometric coordinates of each element of the microphone array, the phase difference between the received signals of each element is calculated. The direction-of-arrival (DOA) estimation algorithm is used to convert the phase difference into azimuth and elevation angles to determine the spatial angle of the fire source relative to the array. Subsequently, beamforming weight vectors are calculated based on the spatial angles, according to the distance of each element to the reference point. The phase compensation value is calculated based on the difference and signal wavelength, generating complex-form weighting coefficients. The sequence corresponding to each element of the original differential channel data is multiplied by the corresponding weighting coefficients using complex multiplication. The weighted results of all elements are coherently accumulated to obtain a single-channel output beamforming signal. This enhances the focusing of the signal in the fire source direction and suppresses the signal in the interference direction. The beamforming signal is used as the processed data. Next, the amplitude of each sampling point in the processed data is compared with a preset effective signal amplitude threshold. When the amplitude of a sampling point is lower than the preset effective signal amplitude threshold, the amplitude of that sampling point is set to zero or marked as invalid. When the amplitude of a sampling point is greater than or equal to the preset effective signal amplitude threshold, the amplitude of that sampling point is retained unchanged. The filtered sequence is used as the discarded differential channel data to eliminate low-amplitude noise and residual clutter components. Finally, the maximum amplitude of the discarded differential channel data is calculated. The amplitude of each sampling point is divided by this maximum amplitude for normalization, mapping the sequence amplitude range to the 0-1 interval. The normalized sequence is output as differential channel data containing flame characteristics. This is done because differential operation between adjacent frames can effectively eliminate interference from static environmental reflections, retain channel change information caused by dynamic flames, and spatial angle extraction provides directional guidance for beamforming. Beamforming technology enhances the signal power in the target direction and improves the signal-to-noise ratio through array signal processing. Amplitude threshold removal further filters out irrelevant weak signals and noise substrates, and time-domain normalization unifies the data amplitude range, providing high-quality differential channel data for subsequent extraction of flame impact delay range and calculation of heat release rate.

[0098] Step S202: Extract the time delay range of the flame effect from the differential channel data, filter out the signal taps whose intensity exceeds the preset effective threshold within the time delay range, and calculate the corresponding total energy.

[0099] It should be noted that step S202 includes: traversing all sampling points in the differential channel data, calculating the square of the signal amplitude of each sampling point as the instantaneous energy value, arranging the instantaneous energy values ​​of each sampling point in order of sampling index to form an energy distribution sequence; setting a preset energy threshold, comparing the instantaneous energy value of each sampling point in the energy distribution sequence with the preset energy threshold, marking the sampling point positions where the instantaneous energy value exceeds the preset energy threshold, determining the start delay and end delay according to the minimum and maximum sampling indices of the marked sampling points, and taking the interval between the start delay and end delay as the delay range of the flame effect; within the delay range of the flame effect, filtering signal taps whose instantaneous energy values ​​exceed the preset effective threshold, summing the instantaneous energy values ​​corresponding to the filtered signal taps, and obtaining the total energy within the delay range.

[0100] Specifically, the process involves iterating through all sampling points in the differential channel data, calculating the square of the signal amplitude at each sampling point as the instantaneous energy value, and arranging the instantaneous energy values ​​of each sampling point in order of sampling index to form an energy distribution sequence. Next, a preset energy threshold is set, and the instantaneous energy value of each sampling point in the energy distribution sequence is compared with the preset energy threshold. The positions of all sampling points exceeding the preset energy threshold are marked, and the minimum and maximum sampling indices among the marked sampling points are identified. The delay value corresponding to the minimum sampling index is used as the starting delay, and the delay value corresponding to the maximum sampling index is used as the ending delay. The interval between the starting and ending delays is determined as the delay range affected by the flame. Then, within the delay range affected by the flame, the instantaneous energy values ​​of each sampling point are iterated again, and signal taps whose instantaneous energy values ​​exceed the preset effective threshold are selected. The instantaneous energy values ​​of the selected signal taps are summed to obtain the total energy within the delay range. Finally, the total energy is output as the flame energy feature value of the current frame. This is done because energy distribution analysis can identify the channel response enhancement region caused by the flame. Dual threshold processing first determines the time delay boundary of the flame's presence and then filters effective energy taps, which can accurately locate the spatial range of the flame's influence and quantify its energy intensity, providing reliable feature input for subsequent heat release rate calculation.

[0101] Step S203: Correct the total energy by combining the fuel specific heat capacity parameter to obtain the corrected energy data.

[0102] Specifically, firstly, a preset fuel type parameter library is retrieved, and the corresponding fuel specific heat capacity parameter is looked up based on the fuel type identifier of the current monitoring scene. This parameter represents the heat released by the complete combustion of a unit mass of fuel. Secondly, the time delay range of the flame's influence in the current frame is obtained. Based on the physical relationship between sound wave propagation speed and temperature, the time delay range is converted into an equivalent flame diameter, and the flame volume estimate is calculated in combination with the equivalent flame diameter. Then, the estimated mass of fuel participating in combustion is calculated based on the fuel density and the estimated flame volume. The estimated fuel mass is multiplied by the fuel specific heat capacity parameter to obtain a theoretical heat release reference value. Next, the ratio of the total energy to the theoretical heat release reference value is calculated to obtain an energy correction coefficient. The total energy is multiplied by the energy correction coefficient for scale calibration to eliminate energy calculation deviations caused by differences in the combustion characteristics of different fuel types. Finally, the calibrated value is output as the corrected energy data. This is done because the calorific value of different fuel types varies significantly, and directly accumulating acoustic energy cannot accurately reflect the actual heat release level. By introducing the fuel specific heat capacity parameter to establish a mapping relationship between acoustic measurements and actual thermophysical quantities, adaptive correction of the total energy for fuel type can be achieved, making the subsequently generated heat release rate curve comparable and physically consistent across fuel types.

[0103] Step S204: The corrected energy data is smoothed using a moving average algorithm to obtain smoothed energy data.

[0104] It should be noted that step S204 includes: setting a preset sliding window length, the preset sliding window length being determined based on the flame combustion fluctuation characteristics and real-time requirements; arranging the corrected energy data in time frame order to form an energy sequence; starting from the beginning position of the energy sequence, extracting the current position and multiple subsequent consecutive data points to form a sliding window; calculating the arithmetic mean of all data points within the sliding window; using the arithmetic mean as the smoothed value corresponding to the center position of the current window; moving the sliding window backward by one data point position; repeating the above operations of extracting the sliding window and calculating the arithmetic mean; sequentially traversing all data points in the energy sequence; for data points at both ends of the energy sequence that are insufficient to form a complete sliding window, processing is performed by padding with zeros or shortening the window to obtain a smoothed value; arranging the smoothed values ​​in time order to form smoothed energy data of the same length as the original energy sequence.

[0105] Specifically, firstly, a preset sliding window length is set, determined based on the characteristics of flame combustion fluctuations and real-time requirements. The corrected energy data is arranged in time frame order to form an energy sequence. Secondly, starting from the beginning of the energy sequence, the current position and several subsequent consecutive data points are extracted to form a sliding window. The arithmetic mean of all data points within the window is calculated, and this average is used as the smoothed value corresponding to the center position of the current window. Then, the sliding window is moved forward by one data point, and the operations of extracting window data and calculating the average are repeated, traversing all data points in the energy sequence in turn. For data points at the ends of the sequence that are insufficient to form a complete window, zero padding or window shortening are used. Finally, all the calculated smoothed values ​​are arranged in time order to form a smoothed energy data sequence of the same length as the original energy sequence. This is done because the flame combustion process has inherent instability, causing the energy data calculated frame by frame to exhibit rapid random fluctuations. By using a moving average algorithm to perform low-pass filtering on the time series, short-term fluctuations can be effectively suppressed while preserving long-term evolution trends, making the energy change curve smoother and more continuous, which facilitates the subsequent identification of the stage characteristics and overall patterns of fire development.

[0106] Step S205: Based on the physical correlation model, the smoothed energy data is converted into corresponding time series heat release rate values.

[0107] It should be noted that step S205 includes: establishing a physical correlation model of ultrasonic wave propagation delay, fire temperature distribution, air medium density, and heat release rate; performing time-series matching between the smoothed energy data and the flame influence delay range to generate a time-aligned energy-delay sequence; inputting the time-aligned energy-delay sequence into the physical correlation model to calculate the initial value of the heat release rate for a single frame; calibrating the initial value of the heat release rate for a single frame in combination with preset fuel type parameters to obtain the calibrated heat release rate value; and arranging all calibrated heat release rate values ​​in time frame order to form the corresponding time series heat release rate values.

[0108] Specifically, firstly, a physical correlation model is established between ultrasonic propagation delay, fire temperature distribution, air density, and heat release rate. This involves determining the square root proportionality between sound velocity and temperature based on gas dynamics theory, establishing a correlation between temperature and air density using the ideal gas law, and coupling the heat released by the flame with the temperature field distribution through the principle of energy conservation, ultimately forming a mathematical conversion relationship from the measured sound wave delay to the heat release rate. Secondly, the smoothed energy data is time-series matched with the flame influence delay range. The energy values ​​corresponding to each time frame and their associated start and end values ​​of the delay range are retrieved. The energy values ​​and delay range parameters are then aligned and combined according to timestamps to generate a time-aligned energy data containing both energy intensity and spatial scale information. The energy-time delay sequence is then input into the physical correlation model. The energy values ​​for each time frame are converted to a time delay range scale for numerical calculation, yielding an initial value for the single-frame heat release rate. Subsequently, preset fuel type parameters are retrieved, and the corresponding combustion efficiency coefficient and stoichiometry are queried based on the fuel type of the current monitoring scenario. The initial value of the single-frame heat release rate is multiplied by the combustion efficiency coefficient for correction, and then divided by the stoichiometry for standardization, eliminating systematic biases caused by differences in the combustion characteristics of different fuels, resulting in a calibrated heat release rate value. Finally, all data frames are traversed sequentially according to time frame order, and the calibrated heat release rate values ​​for each frame are stored sequentially in the sequence buffer, forming a time-series heat release rate value arranged by time. This is done because the physical correlation model establishes a theoretical bridge between acoustic measurement parameters and thermophysical quantities, enabling energy data to be converted into heat release rates with clear physical meaning. Time sequence matching ensures synchronous correspondence between energy intensity and spatial scale information, and fuel type parameter calibration achieves standardized output under different combustion scenarios, ultimately forming a standardized heat release rate time series that can be used for fire type identification and fire stage determination.

[0109] Step S206: Perform curve fitting and normalization on the heat release rate values ​​along the time axis to generate a heat release rate curve characterizing the energy release pattern of the fire.

[0110] Specifically, firstly, the heat release rate values ​​of the corresponding time series are sorted by timestamp, and the time intervals of each data point are checked for uniformity. For data points with uneven time intervals, an interpolation algorithm is used for resampling to ensure that the time intervals between adjacent data points are consistent, forming a discrete heat release rate sequence with equal time intervals. Secondly, the discrete heat release rate sequence is subjected to curve fitting processing. Polynomial fitting or spline interpolation methods are used to smooth the discrete data points, generating a continuous curve function expression. The derivative of the curve function at each time point is calculated to identify characteristic points of change trends. Then, the fitted heat release rate data is normalized, the maximum heat release rate value in the sequence is identified, and the heat release rate value at each time point is divided by this maximum value to map the value range of all data points to the interval between 0 and 1, or it is standardized according to a preset upper limit value of heat release rate. Finally, the normalized heat release rate data is graphically plotted with time as the horizontal axis and normalized heat release rate as the vertical axis to generate a heat release rate curve characterizing the energy release law of the fire scene. The curve data is saved in vector format or bitmap format and output to the display unit and storage unit. This is done because equal time interval processing ensures the uniformity of the time axis, which facilitates subsequent analysis. Curve fitting eliminates the step-like discontinuities caused by discrete sampling, making the curve smoother. Normalization eliminates the absolute numerical differences caused by different fire scales and highlights the relative change patterns. The resulting heat release rate curve can intuitively reflect the temporal evolution characteristics of fire energy release, providing a standardized and visualized data foundation for fire type identification and fire stage determination.

[0111] This embodiment extracts differential channel data containing flame characteristics by performing adjacent-frame differential operations on the channel impulse response data; determines the time delay range of the flame impact, filters valid signal taps and calculates the total energy, combines the energy data corrected by fuel specific heat capacity, smooths it with a moving average, and converts it into a heat release rate value through a physical correlation model. Finally, it fits and normalizes to generate a heat release rate curve. This achieves accurate extraction of flame characteristics, ensures accurate heat release rate calculation, provides reliable data support for fire type and stage determination, and improves the stability and accuracy of fire situation assessment.

[0112] Based on the first embodiment of this application, this application also provides an acoustic fire type identification device based on HRR, please refer to... Figure 7 The device includes:

[0113] The acquisition module 10 is used to transmit high-frequency modulated ultrasonic signals to the monitoring area and receive the echo signals reflected by the flame area to obtain the original acoustic signals.

[0114] The estimation module 20 is used to perform channel estimation based on the original acoustic signal and ultrasonic signal to obtain channel impulse response data.

[0115] The calculation module 30 is used to extract flame features and calculate heat release rate from the channel impulse response data to obtain the heat release rate curve.

[0116] Result module 40 is used to identify the fire type and determine the fire stage based on the heat release rate curve, and obtain the fire type identification result.

[0117] The HRR-based acoustic fire type identification device provided in this application employs the HRR-based acoustic fire type identification method described in the above embodiments, and can solve the technical problem of how to accurately quantify HRR and identify fire types and determine fire stages based on acoustic perception. Compared with the prior art, the beneficial effects of the HRR-based acoustic fire type identification device provided in this application are the same as those of the HRR-based acoustic fire type identification method provided in the above embodiments, and other technical features in the HRR-based acoustic fire type identification device are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0118] This application provides an acoustic fire type identification device based on HRR. The HRR-based acoustic fire type identification device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the HRR-based acoustic fire type identification method in the above embodiment 1.

[0119] The following is for reference. Figure 8 This document illustrates a structural schematic diagram of an HRR-based acoustic fire type identification device suitable for implementing embodiments of this application. The HRR-based acoustic fire type identification device in this application embodiment may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), and in-vehicle terminals (e.g., in-vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers. Figure 8 The HRR-based acoustic fire type identification device shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0120] like Figure 8As shown, the HRR-based acoustic fire type identification device may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage device 1003 into a random access memory (RAM) 1004. The RAM 1004 also stores various programs and data required for the operation of the HRR-based acoustic fire type identification device. The processing unit 1001, ROM 1002, and RAM 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following can be connected to I / O interface 1006: input devices 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. Communication device 1009 allows the HRR-based acoustic fire type identification device to communicate wirelessly or wiredly with other devices to exchange data. Although various HRR-based acoustic fire type identification devices are shown in the figures, it should be understood that implementation or possession of all of them is not required. More or fewer of them may be implemented alternatively.

[0121] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from ROM 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0122] The HRR-based acoustic fire type identification device provided in this application, employing the HRR-based acoustic fire type identification method described in the above embodiments, solves the technical problem of how to accurately quantify HRR and identify fire types and determine fire stages based on acoustic perception. Compared with the prior art, the beneficial effects of the HRR-based acoustic fire type identification device provided in this application are the same as those of the HRR-based acoustic fire type identification method provided in the above embodiments, and other technical features of this HRR-based acoustic fire type identification device are the same as those disclosed in the previous embodiment method, and will not be repeated here.

[0123] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0124] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0125] This application provides a computer-readable medium having computer-readable program instructions (i.e., a computer program) stored thereon, the computer-readable program instructions being used to execute the HRR-based acoustic fire type identification method in the above embodiments.

[0126] The computer-readable medium provided in this application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor devices, or any combination thereof. More specific examples of computer-readable media may include, but are not limited to: electrical connections with one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable medium may be any tangible medium containing or storing a program that can be executed by instructions, used by a device, or used in conjunction with it. The program code contained on the computer-readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0127] The aforementioned computer-readable medium may be included in an HRR-based acoustic fire type identification device; or it may exist independently and not assembled into an HRR-based acoustic fire type identification device.

[0128] The aforementioned computer-readable medium carries one or more programs that, when executed by the HRR-based acoustic fire type identification device, enable the HRR-based acoustic fire type identification device to write computer program code for performing the operations of this application in one or more programming languages ​​or a combination thereof. These programming languages ​​include object-oriented programming languages—such as Java, Smalltalk, and C++—and conventional procedural programming languages—such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0129] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of methods and computer program products according to various embodiments of this application. In this regard, all blocks in the flowcharts or block diagrams may represent a module, segment, or portion of code containing one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that all blocks in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using dedicated hardware-based implementations that perform the specified functions or operations, or using a combination of dedicated hardware and computer instructions.

[0130] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0131] The readable medium provided in this application is a computer-readable medium, which stores computer-readable program instructions (i.e., a computer program) for executing the above-described HRR-based acoustic fire type identification method. This solves the technical problem of how to accurately quantify HRR and identify fire types and determine fire stages based on acoustic sensing. Compared with the prior art, the beneficial effects of the computer-readable medium provided in this application are the same as those of the HRR-based acoustic fire type identification method provided in the above embodiments, and will not be repeated here.

[0132] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the HRR-based acoustic fire type identification method described above.

[0133] The computer program product provided in this application can solve the technical problem of how to accurately quantify HRR and identify fire types and determine fire stages based on acoustic sensing. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as those of the HRR-based acoustic fire type identification method provided in the above embodiments, and will not be repeated here.

[0134] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A method for acoustic fire type identification based on HRR, characterized in that, The method includes: The system transmits high-frequency modulated ultrasonic signals to the monitoring area and receives the echo signals reflected by the flame area to obtain the original acoustic signals. Channel estimation is performed based on the original acoustic signal and the ultrasonic signal to obtain channel impulse response data; Flame feature extraction and heat release rate calculation are performed on the channel impulse response data to obtain the heat release rate curve; Fire type identification and fire stage determination are performed based on the heat release rate curve to obtain fire type identification results; The step of extracting flame features and calculating the heat release rate from the channel impulse response data to obtain the heat release rate curve includes: Perform adjacent frame differential operation on the channel impulse response data to obtain differential channel data containing flame characteristics; The time delay range of the flame effect is parsed from the differential channel data, and signal taps with intensity exceeding a preset effective threshold within the time delay range are selected and the corresponding total energy is calculated. The total energy is corrected by combining the fuel specific heat capacity parameter to obtain the corrected energy data. Specifically, this includes: retrieving a preset fuel type parameter library and querying the corresponding fuel specific heat capacity parameter based on the fuel type identifier of the current monitoring scene; obtaining the time delay range of the flame's influence in the current frame, converting the time delay range into an equivalent flame diameter based on the physical relationship between sound wave propagation speed and temperature, and calculating the estimated flame volume based on the equivalent flame diameter; calculating the estimated mass of fuel participating in combustion based on the fuel density and the estimated flame volume to obtain a theoretical heat release reference value; calculating the ratio between the total energy and the theoretical heat release reference value to obtain the energy correction coefficient, and outputting the calibrated value as the corrected energy data. The corrected energy data is smoothed using a moving average algorithm to obtain smoothed energy data. Based on a physical correlation model, the smoothed energy data is converted into corresponding time-series heat release rate values. This model integrates ultrasonic wave propagation delay, fire temperature distribution, air density, and heat release rate. Specifically, it includes: time-series matching of the smoothed energy data with the flame influence delay range to generate a time-aligned energy-delay sequence; inputting the time-aligned energy-delay sequence into the physical correlation model to obtain an initial value for the heat release rate per frame; calibrating the initial value for the heat release rate per frame using preset fuel type parameters to obtain a calibrated heat release rate value; and arranging all the calibrated heat release rate values ​​in time-frame order to form the corresponding time-series heat release rate values. The heat release rate values ​​are subjected to curve fitting and normalization along the time axis to generate a heat release rate curve characterizing the energy release pattern of the fire scene. The steps of identifying the fire type and determining the fire stage based on the heat release rate curve to obtain the fire type identification result include: The heat release rate curve is subjected to moving average to remove fluctuations, and morphological features are extracted, wherein the morphological features include peak heat release rate, time to peak, average heat release rate, and decay rate. The morphological features are input into a preset neural network classifier to obtain type discrimination feature values; The type discrimination feature value is compared with a preset fire type feature library to determine the fire type matching result, wherein the fire type matching result includes fuel-controlled fire, ventilation-limited fire, pulse fire and ventilation-induced flashover fire; The fire stage is determined by the temporal change trend of the heat release rate curve, wherein the fire stage includes the initial growth stage, the flashover stage, the complete combustion stage, or the decay stage. Based on the fire type matching result and the fire stage, the fire type identification result is output.

2. The method as described in claim 1, characterized in that, The step of performing channel estimation based on the original acoustic signal and the ultrasonic signal to obtain channel impulse response data includes: Retrieve the transmitted signal frame that matches the ultrasonic signal; The original acoustic signal is subjected to carrier demodulation processing to obtain a baseband received signal of the same standard as the transmitted signal frame; The baseband received signal and the transmitted signal frame are synchronized and aligned on the time axis to obtain the aligned transmitted signal frame and the aligned baseband received signal. The aligned baseband received signal and the aligned transmitted signal frame are processed to obtain the initial channel impulse response data. The initial channel impulse response data is subjected to amplitude normalization processing to remove clutter components with amplitudes lower than a preset effective threshold, thereby obtaining the processed response data. The processed response data is divided into frames and encapsulated according to a preset frame length to obtain channel impulse response data.

3. The method as described in claim 2, characterized in that, The steps for processing the aligned baseband received signal and the aligned transmitted signal frame to obtain the initial channel impulse response data include: Perform point-by-point cross-correlation operation between the aligned baseband received signal and the aligned transmitted signal frame to obtain a cross-correlation result sequence. Peak detection is performed on the cross-correlation result sequence to extract multipath peak components that characterize the signal transmission path; Based on the amplitude and delay information of the multipath peak components, an initial response sequence for the multipath transmission channel is constructed. The initial response sequence is subjected to time-domain zero-padding and amplitude normalization to generate fixed-length channel response data; The fixed-length channel response data is used as the initial channel impulse response data.

4. The method as described in claim 1, characterized in that, The step of performing adjacent frame differential operations on the channel impulse response data to obtain differential channel data containing flame characteristics includes: The first frame of channel impulse response data and the second frame of channel impulse response data are retrieved from the channel impulse response data in the order of time acquisition. Perform tap-by-tap amplitude differential operation on the first frame of channel impulse response data and the second frame of channel impulse response data to obtain the original differential channel data; Extract the spatial angle information from the original differential channel data; Based on the angle information, beamforming technology is used to focus and enhance the signal in the direction of the fire source to obtain the processed data. Data with amplitudes lower than the effective amplitude threshold of the signal in the processed data are removed to obtain the differential channel data after removal. The removed differential channel data is then subjected to time-domain normalization to obtain differential channel data containing flame characteristics.

5. The method as described in claim 1, characterized in that, The step of converting the smoothed energy data into corresponding time-series heat release rate values ​​based on the physical correlation model includes: Establish a physical correlation model between ultrasonic propagation delay, fire temperature distribution, air density, and heat release rate; The smoothed energy data is time-matched with the flame impact delay range to generate a time-aligned energy-delay sequence. The time-aligned energy-delay sequence is input into the physical correlation model to calculate the initial value of the single-frame heat release rate. The initial value of the single-frame heat release rate is calibrated by combining preset fuel type parameters to obtain the calibrated heat release rate value. All the calibrated heat release rate values ​​are arranged in time frame order to form the corresponding time series of heat release rate values.

6. An acoustic fire type identification device based on HRR, characterized in that, The device is applied to the HRR-based acoustic fire type identification method as described in any one of claims 1-5, and the device comprises: The acquisition module is used to transmit high-frequency modulated ultrasonic signals to the monitoring area and receive the echo signals reflected by the flame area to obtain the original acoustic signals. The estimation module is used to perform channel estimation based on the original acoustic signal and the ultrasonic signal to obtain channel impulse response data; The calculation module is used to extract flame features and calculate heat release rate from the channel impulse response data to obtain a heat release rate curve. The results module is used to identify the fire type and determine the fire stage based on the heat release rate curve, and obtain the fire type identification result.

7. An acoustic fire type identification device based on HRR, characterized in that, The device includes: a memory, a processor, and an HRR-based acoustic fire type identification program stored on the memory and running on the processor, the HRR-based acoustic fire type identification program being configured to implement the steps of the HRR-based acoustic fire type identification method as described in any one of claims 1-5.

8. A storage medium, characterized in that, The storage medium stores an HRR-based acoustic fire type identification program, which, when executed by a processor, implements the steps of the HRR-based acoustic fire type identification method as described in any one of claims 1-5.

Images