Intelligent monitoring method and system for identifying boiling heat exchange state based on sound signal

Abstract

The invention provides an intelligent monitoring method and system for identifying a boiling heat exchange state based on a sound signal, and the method comprises the steps: obtaining a sound signal in a boiling process, and carrying out the preprocessing, and obtaining a preprocessed sound signal; combining Bayesian online change point detection and a state space model to construct a boiling state recognition model; based on the boiling state recognition model, the preprocessed sound signals are classified and recognized, a boiling state label is obtained, and intelligent monitoring of the boiling heat exchange state is completed. The method has the advantages of non-contact, high real-time performance and strong robustness, is suitable for boiling heat exchange state monitoring and energy efficiency optimization under complex or extreme working conditions, and has good engineering practical value and popularization prospect.

Description

Technical Field

[0001] This invention belongs to the field of heat transfer process monitoring and intelligent identification technology, specifically relating to an intelligent monitoring method and system for identifying boiling heat transfer state based on sound signals. Background Technology

[0002] Boiling heat transfer is widely used in nuclear cooling, chemical reactions, electronic heat dissipation, and thermal equipment. The boiling process typically involves stages of natural convection, nucleus boiling, transitional boiling, and film boiling. The bubble formation, growth, coalescence, and detachment behaviors differ significantly at each stage, directly affecting wall superheat, heat transfer coefficient, and critical heat flux. If these state changes are not identified and controlled in a timely manner, heat transfer degradation, localized drying, and operational instability can easily occur, leading to decreased energy efficiency and increased safety risks.

[0003] Existing monitoring methods mainly rely on thermocouple temperature measurement, infrared thermal imaging, optical visualization, and pressure fluctuation analysis. However, under high-temperature, opaque, or strongly disturbed conditions, these methods suffer from problems such as response lag, limited installation conditions, signal obstruction, and insufficient noise immunity. Faced with rapid switching between multiple states and complex background noise, common solutions struggle to meet requirements in terms of identification accuracy, state boundary clarity, and online adaptability, particularly in early warning of transition states and membrane states.

[0004] Acoustic signal methods offer advantages such as being non-contact, flexible in deployment, and sensitive to bubble behavior, enabling them to reflect sound pressure changes caused by microscopic events during boiling. However, approaches based on traditional acoustic features and shallow classifiers have limited feature representation capabilities, making it difficult to simultaneously address long-term scale dependencies and identify abrupt state changes. The decision-making process lacks traceable explanation and has insufficient generalization ability across operating conditions. Furthermore, many systems still operate on a fragmented process of data acquisition and offline discrimination, failing to establish real-time linkage with industrial control, making it difficult to promptly convert identification results into executable adjustment commands such as heating power and cooling flow rate.

[0005] In view of this, how to provide a method and system for intelligent identification and control of boiling heat exchange state for engineering applications, which can realize real-time extraction of key features, rapid capture of state changes and traceable interpretation of classification results based on sound signals, establish an integrated closed loop connecting acquisition, analysis, visualization and edge control, meet the technical requirements of stable, safe and energy-saving operation in high heat flux density and high reliability scenarios, and promote the transformation of heat exchange process monitoring from experience-based to intelligent and refined, is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] This invention addresses the problems of insufficient accuracy, weak interpretability, low real-time performance, and lack of linkage with process control in existing technologies for identifying boiling heat transfer status. It proposes an intelligent monitoring method and system based on sound signals to identify boiling heat transfer status.

[0007] To achieve the above objectives, the present invention provides the following solution: A smart monitoring method for identifying boiling heat transfer state based on sound signals, comprising: Acquire the sound signal of the boiling process and preprocess it to obtain the preprocessed sound signal; A boiling state recognition model is constructed by combining Bayesian online change point detection and state space model; Based on the boiling state recognition model, the preprocessed sound signal is classified and identified to obtain boiling state labels, thus completing the intelligent monitoring of the boiling heat transfer state.

[0008] Preferably, the structure of the boiling state identification model includes: The WavLM feature extraction module is used to extract the depth acoustic features of the preprocessed sound signal using a multi-layer Transformer coding structure, and to fuse the depth acoustic features with traditional acoustic features to obtain a time-series feature vector; wherein, the depth acoustic features are used to characterize the acoustic patterns of bubble formation, bursting and state transition. The PCA dimensionality reduction module is used to perform principal component analysis on the time series feature vectors based on the covariance matrix to obtain dimensionality-reduced feature vectors. The BOCPD change point detection module is used to perform change point detection on the dimensionality-reduced feature vector based on Bayesian recursive calculation, calculate the posterior probability and identify the boiling state change point based on the posterior probability, and divide the dimensionality-reduced feature vector to obtain segmented feature vectors. The SSM classification and recognition module is used to classify and recognize the segmented feature vectors using a state-space model, obtain boiling state labels, and complete intelligent monitoring of boiling heat transfer state.

[0009] Preferably, in the BOCPD change point detection module, the expression for calculating the posterior probability is as follows: , in Indicates the execution length of the current segment; This represents the complete sequence of observed acoustic features up to time step t, used to describe the temporal evolution of the boiling process, i.e., from... , ,..., A set; Indicates time step The observed acoustic characteristics; The transition probability represents the length of the operation; This represents the likelihood function under the current hypothetical change point; This represents the posterior distribution at the previous time step.

[0010] Preferably, in the SSM classification and recognition module, the state update and observation relationship of the state space model satisfies: , The hidden state vector representing the time step t. The observation output represents the time step t. , , , It is a trainable parameter matrix; It is a non-linear activation function. These are segmented feature vectors.

[0011] Preferably, the structure of the boiling state identification model further includes a SHAP module, which is used to calculate the Shapley additive interpretation value of each depth acoustic feature based on the feature contribution analysis model, determine the degree of influence of each depth acoustic feature on the boiling state classification result, and obtain the global feature importance ranking result.

[0012] Preferably, the structure of the boiling state recognition model further includes a GUI module for displaying real-time waveform plotting, Mel spectrum plotting, state segmentation highlighting and tabular results, and outputting boiling state labels, confidence scores and global feature importance ranking results in chronological order.

[0013] The present invention also provides an intelligent monitoring system for identifying boiling heat transfer state based on sound signals, for implementing the method, comprising: The signal acquisition module is used to acquire the sound signal of the boiling process and preprocess it to obtain the preprocessed sound signal. The identification model construction module is used to combine Bayesian online change point detection and state space model to construct a boiling state identification model; The classification and recognition module is used to classify and recognize the preprocessed sound signal based on the boiling state recognition model, obtain boiling state labels, and complete the intelligent monitoring of boiling heat transfer state.

[0014] Preferably, in the identification model construction module, the structure of the boiling state identification model includes: The WavLM feature extraction module is used to extract the depth acoustic features of the preprocessed sound signal using a multi-layer Transformer coding structure, and to fuse the depth acoustic features with traditional acoustic features to obtain a time-series feature vector; wherein, the depth acoustic features are used to characterize the acoustic patterns of bubble formation, bursting and state transition. The PCA dimensionality reduction module is used to perform principal component analysis on the time series feature vectors based on the covariance matrix to obtain dimensionality-reduced feature vectors. The BOCPD change point detection module is used to perform change point detection on the dimensionality-reduced feature vector based on Bayesian recursive calculation, calculate the posterior probability and identify the boiling state change point based on the posterior probability, and divide the dimensionality-reduced feature vector to obtain segmented feature vectors. The SSM classification and recognition module is used to classify and recognize the segmented feature vectors using a state-space model, obtain boiling state labels, and complete intelligent monitoring of boiling heat transfer state.

[0015] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention constructs an intelligent monitoring method and system for boiling heat transfer state based on sound signals. It integrates deep acoustic feature extraction, principal component dimensionality reduction, online change point detection, and state space recognition to achieve real-time identification and closed-loop control of key states in the boiling process, significantly reducing false positives and false negatives, and improving heat transfer efficiency and operational safety. The system introduces SHAP interpretable analysis, enhancing decision transparency and the accuracy and reliability of parameter calibration. Combining a graphical interface and edge computing, it achieves visualized presentation of identification results and automatic adjustment of heating power and cooling flow rate, improving on-site availability and response speed. The solution has multi-condition adaptability and noise resistance, supports multi-channel sensor expansion and sound source localization, and is suitable for applications such as high heat flux density cooling, electronic heat dissipation, and chemical heat exchange, possessing good economic value, energy-saving effects, and promising prospects for engineering promotion. Attached Figure Description

[0016] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart of a method according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the model and hardware device structure according to an embodiment of the present invention; Figure 3 This is a schematic diagram of a GUI (Graphical User Interface) according to an embodiment of the present invention.

[0018] Figure labels: 1-Sound acquisition device; 2-GUI visual interface. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0021] Example 1 like Figure 1 As shown, an intelligent monitoring method for identifying boiling heat transfer state based on sound signals includes: S1: Acquire the sound signal of the boiling process and preprocess it to obtain the preprocessed sound signal.

[0022] S2: Combining Bayesian online change point detection and state-space modeling, a boiling state recognition model is constructed. For example... Figure 2 As shown. Figure 2 In the model, the sound acquisition device 1 and the GUI visualization interface 2 are used for the acquisition of input signals and the display of output results, respectively.

[0023] The core innovation of this invention, which combines Bayesian online change point detection with a state-space model, lies in designing a collaborative "segmentation-recognition" architecture, rather than a simple combination. Specifically, the BOCPD change point detection module, acting as an adaptive time sequence segmenter, identifies the critical points of boiling state switching online and accurately by analyzing abrupt changes in the statistical distribution of acoustic features, dividing the continuous signal into time segments with consistent internal states. The state-space model, acting as a high-performance segment classifier, is specifically designed for fine-grained recognition of each homogeneous segment. The key improvement of this architecture is that the input to the state-space model is no longer the original sequence, but rather the aggregated features representing a stable state output by the change point detection module. This allows it to focus more on characterizing the dynamic evolution within the state without diverting resources to handle drastic changes, thus significantly improving the accuracy and robustness of state recognition overall. This deeply collaborative modular structure is a dedicated design for the characteristics of the boiling process and is not disclosed in existing technologies.

[0024] A further implementation method is that the structure of the boiling state recognition model includes: The WavLM feature extraction module is used to extract the deep acoustic features (multi-scale time-frequency and context features) of the preprocessed sound signal using a multi-layer Transformer coding structure, and to fuse the deep acoustic features with traditional acoustic features to obtain a time series feature vector; among them, the deep acoustic features are used to characterize the acoustic patterns of bubble generation, bursting and state transition.

[0025] The PCA dimensionality reduction module is used to perform principal component analysis on time series feature vectors based on the covariance matrix to obtain dimensionality-reduced feature vectors. Specifically, it receives time series feature vectors obtained by fusing WavLM deep features and traditional acoustic features; performs principal component analysis based on the covariance matrix, selects principal components according to the size of the eigenvalues, and ensures that the cumulative variance contribution rate is not less than 90%; generates dimensionality-reduced time series features and retains linear mapping parameters for online consistency processing, and outputs the dimensionality-reduced features to the BOCPD module.

[0026] We will utilize deep acoustic features extracted from a pre-trained audio model (WavLM) along with traditional acoustic features including 13-dimensional MFCC coefficients, spectral roll-off points, spectral centroids, short-time energy, and zero-crossing rate. We will also use in-window statistics to calculate the mean and standard deviation of these features as feature enhancements, forming a more stable high-dimensional feature combination. In the dimensionality reduction process, we will use PCA and set the principal components to retain more than 90% of the cumulative variance contribution rate to ensure the distinguishability of boiling states.

[0027] The BOCPD change point detection module is used to detect change points on the dimensionality-reduced feature vector based on Bayesian recursion calculation, calculate the posterior probability and identify boiling state abrupt change points based on the posterior probability, and divide the dimensionality-reduced feature vector to obtain segmented feature vectors. A further implementation method is that the expression for calculating the posterior probability in the BOCPD change point detection module is as follows: , in Indicates the execution length of the current segment; This represents the complete sequence of observed acoustic features up to time step t, used to describe the temporal evolution of the boiling process, i.e., from... , ,..., A set; Indicates time step The observed acoustic characteristics; The transition probability represents the length of the operation; This represents the likelihood function under the current hypothetical change point; This represents the posterior distribution at the previous time step. By continuously recursively applying this posterior probability, when the change-point confidence threshold δ satisfies 0.8 ≤ δ ≤ 0.9, the time step is determined. A sudden change in boiling state occurs at a certain point, and the acoustic signal is segmented accordingly for subsequent classification and identification. The formula yields the time step. The posterior probability distribution of the runtime.

[0028] The SSM classification and recognition module is used to classify and recognize segmented feature vectors using a state-space model, obtain boiling state labels, and complete intelligent monitoring of boiling heat transfer state.

[0029] A further implementation method is that, in the SSM classification and recognition module, the state update and observation relationship of the state space model satisfies: , The hidden state vector representing the time step t. The observation output represents the time step t. , , , It is a trainable parameter matrix; It is a non-linear activation function. This represents the piecewise feature vector. Based on the above formula, the time step can be obtained. Hidden state vector With observation output ; will be continuous time or The features are input into the classification layer to obtain the posterior probability of each boiling state and output the label.

[0030] In this embodiment, the SSM (Spatial State Model) model adopts a bidirectional state update mechanism, which captures the long-term temporal dependencies and state transition characteristics in the boiling sound signal through forward and backward recursive calculations; the output states include natural convection (BKG), nucleus boiling (NB), transitional boiling (TB), and film boiling (PreCHF).

[0031] A further implementation method involves the boiling state identification model further comprising a SHAP module. This module calculates the Shapley additive explanation value of each depth acoustic feature based on a feature contribution analysis model, determines the influence of each depth acoustic feature on the boiling state classification result, and obtains a global feature importance ranking result. Specifically, it receives the classification result and corresponding input features output by the SSM model, establishes a feature contribution analysis model, calculates the Shapley additive explanation value of each input acoustic feature to obtain the local feature contribution of each sample, and calculates the global feature importance through a time-weighted summation method. When the average absolute value of a feature exceeds a set threshold τ (0.05 ≤ τ ≤ 0.20), it is determined that the feature has a significant impact on the boiling state classification result, and the feature importance ranking result is output.

[0032] A further implementation of the boiling state identification model includes a GUI module for displaying real-time waveform plotting, Mel spectrum plotting, state segmentation highlighting, and tabular results. It also outputs boiling state labels, confidence scores, and global feature importance ranking results in chronological order. This enables dynamic monitoring and interactive operation of the boiling heat transfer state. Figure 3 As shown.

[0033] S3: Based on the boiling state recognition model, the pre-processed sound signal is classified and identified to obtain boiling state labels, thus completing the intelligent monitoring of boiling heat transfer state.

[0034] Example 2 The present invention also provides an intelligent monitoring system for identifying boiling heat transfer state based on sound signals, for implementing the method of Embodiment 1, comprising: The signal acquisition module is used to acquire the sound signal of the boiling process and preprocess it to obtain the preprocessed sound signal. Specifically, a non-contact sensor is used, which is adhered and fixed to the surface of the metal heat exchange structure by thermally conductive acoustic coupling adhesive. The thermally conductive adhesive is a high-temperature resistant, low acoustic impedance silicon-based thermally conductive adhesive with a thickness of no more than 0.5 mm.

[0035] The identification model construction module is used to combine Bayesian online change point detection and state space model to construct a boiling state identification model; The classification and recognition module is used to classify and recognize the pre-processed sound signal based on the boiling state recognition model, obtain boiling state labels, and complete the intelligent monitoring of boiling heat transfer state.

[0036] A further implementation method is that, in the model construction module, the structure of the boiling state recognition model includes: The WavLM feature extraction module is used to extract the depth acoustic features of the preprocessed sound signal using a multi-layer Transformer coding structure, and to fuse the depth acoustic features with traditional acoustic features to obtain a time-series feature vector; among them, the depth acoustic features are used to characterize the acoustic patterns of bubble formation, bursting and state transition. The PCA dimensionality reduction module is used to perform principal component analysis on time series feature vectors based on the covariance matrix to obtain dimensionality-reduced feature vectors. The BOCPD change point detection module is used to perform change point detection on the dimensionality-reduced feature vector based on Bayesian recursive calculation, calculate the posterior probability and identify the boiling state change point based on the posterior probability, divide the dimensionality-reduced feature vector, and obtain segmented feature vectors. The SSM classification and recognition module is used to classify and recognize segmented feature vectors using a state-space model, obtain boiling state labels, and complete intelligent monitoring of boiling heat transfer state.

[0037] The system also includes an edge computing module that interacts with external control systems via the Modbus communication protocol to drive heating power adjustment, cooling medium flow rate control, or system alarm logic, achieving real-time closed-loop control. Furthermore, it supports visualizing the identified status on a human-machine interface (HMI). It supports parallel input from multiple acoustic sensors to form a multi-channel sound source array, improving spatial resolution and anti-interference capabilities; it can also be expanded for boiling sound source localization and sound pressure distribution monitoring.

[0038] Based on the identified boiling state, the system generates control commands through the edge control module. These commands include: heating power adjustment, cooling medium flow rate change, system alarm triggering, or emergency shutdown signal. The system interacts with external devices through standard industrial communication protocols to achieve automated closed-loop control of the heat exchange process.

[0039] The system internally sets up a boiling state-control action mapping rule table, defining the execution strategy corresponding to different recognition states, including: "Natural convection" status: Maintain current heating or moderately increase temperature; State "Nuclear boiling": Maintains stable operation; "Transitional boiling state": Activate the cooling system to lower the temperature in advance; Status "Membrane Boiling": Triggers an alarm and executes a safety shutdown.

[0040] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A smart monitoring method for identifying boiling heat transfer state based on sound signals, characterized in that, include: Acquire the sound signal of the boiling process and preprocess it to obtain the preprocessed sound signal; A boiling state recognition model is constructed by combining Bayesian online change point detection and state space model; Based on the boiling state recognition model, the preprocessed sound signal is classified and identified to obtain boiling state labels, thus completing the intelligent monitoring of the boiling heat transfer state.

2. The method according to claim 1, characterized in that, The structure of the boiling state identification model includes: The WavLM feature extraction module is used to extract the depth acoustic features of the preprocessed sound signal using a multi-layer Transformer coding structure, and to fuse the depth acoustic features with traditional acoustic features to obtain a time-series feature vector; wherein, the depth acoustic features are used to characterize the acoustic patterns of bubble formation, bursting and state transition. The PCA dimensionality reduction module is used to perform principal component analysis on the time series feature vectors based on the covariance matrix to obtain dimensionality-reduced feature vectors. The BOCPD change point detection module is used to perform change point detection on the dimensionality-reduced feature vector based on Bayesian recursive calculation, calculate the posterior probability and identify the boiling state change point based on the posterior probability, and divide the dimensionality-reduced feature vector to obtain segmented feature vectors. The SSM classification and recognition module is used to classify and recognize the segmented feature vectors using a state-space model, obtain boiling state labels, and complete intelligent monitoring of boiling heat transfer state.

3. The method according to claim 2, characterized in that, In the BOCPD change point detection module, the expression for calculating the posterior probability is as follows: ， in Indicates the execution length of the current segment; This represents the complete sequence of observed acoustic features up to time step t, used to describe the temporal evolution of the boiling process, i.e., from... , ,..., A set; Indicates time step The observed acoustic characteristics; The transition probability represents the length of the operation; This represents the likelihood function under the current hypothetical change point; This represents the posterior distribution at the previous time step.

4. The method according to claim 2, characterized in that, In the SSM classification and recognition module, the state update and observation relationship of the state space model satisfy: ， The hidden state vector representing the time step t. The observation output represents the time step t. , , , It is a trainable parameter matrix; It is a non-linear activation function. These are segmented feature vectors.

5. The method according to claim 2, characterized in that, The structure of the boiling state identification model also includes a SHAP module, which is used to calculate the Shapley additive interpretation value of each depth acoustic feature based on the feature contribution analysis model, determine the degree of influence of each depth acoustic feature on the boiling state classification result, and obtain the global feature importance ranking result.

6. The method according to claim 5, characterized in that, The structure of the boiling state recognition model also includes a GUI module, which is used to display real-time waveform plotting, Mel spectrum plotting, state segmentation highlighting and tabular results, and output boiling state labels, confidence scores and global feature importance ranking results in chronological order.

7. An intelligent monitoring system for identifying boiling heat transfer state based on sound signals, used to implement the method described in any one of claims 1-6, characterized in that, include: The signal acquisition module is used to acquire the sound signal of the boiling process and preprocess it to obtain the preprocessed sound signal. The identification model construction module is used to combine Bayesian online change point detection and state space model to construct a boiling state identification model; The classification and recognition module is used to classify and recognize the preprocessed sound signal based on the boiling state recognition model, obtain boiling state labels, and complete the intelligent monitoring of boiling heat transfer state.

8. The system according to claim 7, characterized in that, In the identification model construction module, the structure of the boiling state identification model includes: The WavLM feature extraction module is used to extract the depth acoustic features of the preprocessed sound signal using a multi-layer Transformer coding structure, and to fuse the depth acoustic features with traditional acoustic features to obtain a time-series feature vector; wherein, the depth acoustic features are used to characterize the acoustic patterns of bubble formation, bursting and state transition. The PCA dimensionality reduction module is used to perform principal component analysis on the time series feature vectors based on the covariance matrix to obtain dimensionality-reduced feature vectors. The BOCPD change point detection module is used to perform change point detection on the dimensionality-reduced feature vector based on Bayesian recursive calculation, calculate the posterior probability and identify the boiling state change point based on the posterior probability, and divide the dimensionality-reduced feature vector to obtain segmented feature vectors. The SSM classification and recognition module is used to classify and recognize the segmented feature vectors using a state-space model, obtain boiling state labels, and complete intelligent monitoring of boiling heat transfer state.

Images