A boiler voiceprint monitoring and combustion regulation method

Abstract

The application discloses a boiler voiceprint monitoring and combustion regulation method, relates to the technical field of boiler combustion monitoring, and solves the problems of low monitoring precision and poor regulation working condition adaptability of the prior art. The method comprises the steps of self-adaptive acquisition of a voiceprint signal, hierarchical noise reduction and feature extraction, multi-modal fusion identification of a combustion state, self-adaptive combustion regulation, and closed-loop feedback optimization, and is matched with a fine adjustment auxiliary mechanism and a dynamic synchronous regulation and monitoring mechanism. A coupling regulation model is constructed based on a model predictive control algorithm, a coal quality soft measurement algorithm is introduced, differentiated regulation is performed for different abnormalities, a closed-loop control is formed in combination with a regularization coefficient optimization model, and the mechanism can dynamically adjust an acquisition position and a distance. The method is suitable for multiple types of boilers, can be connected to an existing DCS system, does not need to transform a boiler body, realizes self-adaptive regulation without human intervention, and improves combustion efficiency and operation stability.

Description

Technical Field

[0001] This invention relates to the field of boiler monitoring technology, specifically a method for boiler acoustic monitoring and combustion control. Background Technology

[0002] As a core energy conversion device in industrial production and power supply, the combustion efficiency and operational stability of boilers directly affect production energy consumption, equipment lifespan, and safe production. Therefore, real-time monitoring and precise control of boiler combustion status have become key to industry research and application.

[0003] Existing boiler combustion monitoring and control technologies still suffer from significant drawbacks, including limited monitoring dimensions and poor adaptability to various operating conditions. They struggle to adapt to the complex operating environment of boilers, characterized by high temperatures, high noise levels, and high dust concentrations, and cannot accurately capture the dynamic changes in the combustion process. Traditional technologies rely solely on parameters such as temperature and oxygen levels, failing to reflect the microscopic processes of fuel combustion within the furnace. They struggle to identify latent problems such as uneven fuel burning, localized coking in the furnace, and fan malfunctions in advance. Furthermore, parameter acquisition is easily distorted due to factors such as sensor placement and environmental interference, leading to inaccurate combustion status assessments. Simultaneously, some auxiliary monitoring technologies, such as image recognition, are easily affected by flue gas and dust within the furnace, significantly reducing their effectiveness. While acoustic signature monitoring technology has been attempted, it generally suffers from incomplete noise reduction of the acquired signal and inaccurate feature extraction, failing to provide effective evidence for combustion status identification.

[0004] Existing control technologies mostly employ fixed control logic and parameter ratios, lacking the ability to adapt to dynamic changes in boiler operating conditions and failing to achieve deep integration of monitoring data and control strategies. On the one hand, when operating conditions change due to fluctuations in coal quality or load variations, traditional control systems cannot adjust combustion parameters in real time, easily leading to problems such as incomplete combustion and abnormal furnace temperature. This not only reduces boiler thermal efficiency and increases energy consumption but also exacerbates equipment failures such as furnace coking and corrosion of heating surfaces. On the other hand, the control process lacks a closed-loop feedback optimization mechanism, making it impossible to verify and optimize the control effect in real time. Manual intervention to adjust parameters is often required, increasing maintenance workload and potentially exacerbating combustion faults due to delayed human judgment, even posing safety hazards. To address these issues, there is an urgent need to develop a multi-dimensional, high-precision, and self-adaptive boiler combustion monitoring and control method to achieve accurate identification of boiler combustion status and dynamic intelligent control of combustion parameters, thereby improving the efficiency, stability, and safety of boiler operation.

[0005] Therefore, a method for boiler acoustic monitoring and combustion control is proposed to solve the above problems. Summary of the Invention

[0006] 1. The technical problem to be solved by the present invention

[0007] The purpose of this invention is to propose a boiler acoustic monitoring and combustion control method to solve the following problems existing in the prior art: (1) Existing boiler combustion monitoring methods have the problem of being single in dimension and lacking in accuracy. Existing technologies mostly rely on monitoring traditional parameters such as temperature and oxygen content. Soundprint monitoring technology has problems such as incomplete signal noise reduction and inaccurate feature extraction. It also lacks the ability to fuse and identify multiple features, making it impossible to accurately capture hidden combustion anomalies such as fuel burning off-center and local coking in the furnace. At the same time, the monitoring signal is easily distorted due to the complex environment of high temperature, high noise and high dust in the boiler, resulting in a large deviation in the judgment of combustion status.

[0008] (2) Existing boiler combustion control logic is fixed, has poor adaptability to operating conditions, and lacks effective closed-loop optimization. Traditional control technology uses fixed parameter ratios and cannot dynamically adjust combustion parameters according to operating conditions such as coal quality fluctuations and load changes. Furthermore, it does not achieve deep integration of monitoring and identification results with control strategies. At the same time, it lacks a real-time closed-loop feedback optimization mechanism, and the control effect cannot be verified and the model optimized in a timely manner. It requires manual intervention and is prone to combustion failure due to control lag, which reduces boiler thermal efficiency and brings safety hazards.

[0009] 2. Technical Solution To achieve the above objectives, the present invention provides the following technical solution: a boiler acoustic signature monitoring and combustion control method, comprising the following steps: S1. Adaptive Acquisition of Soundprint Signal: An array of soundprint acquisition modules is deployed in different heights and quadrants of the combustion boiler (1). The soundprint acquisition module integrates a high-temperature and dust-resistant micro-sound sensor and a pre-signal conditioning unit to acquire the original soundprint signal generated during boiler combustion. At the same time, it simultaneously acquires the real-time operating parameters of the boiler and dynamically adjusts the soundprint acquisition frequency and gain according to the operating parameters to solve the problem of soundprint signal transmission attenuation and acquisition distortion under high temperature, high noise, and high dust conditions. S2. Layered noise reduction and feature extraction of voiceprint signal: The original voiceprint signal is preprocessed. First, the background noise and equipment vibration interference noise in the furnace environment are filtered out by the adaptive wavelet threshold noise reduction algorithm. Then, the nonlinear frequency mapping is performed by the Mel spectrum converter to enhance the low- and medium-frequency combustion features and compress the high-frequency redundant data. Subsequently, the time domain features, frequency domain features and Mel frequency cepstral coefficients of the voiceprint signal are extracted to form a multi-dimensional voiceprint feature set. S3. Multimodal Fusion Recognition of Combustion Status: Construct a deep learning recognition model based on attention mechanism, fuse the multi-dimensional voiceprint feature set extracted in step S2 with the real-time boiler operating parameters collected in S1, dynamically allocate the attention weight of each feature, and achieve accurate recognition of boiler combustion status such as normal combustion, fuel burning, incomplete combustion, furnace coking, and fan abnormality, with recognition time not exceeding 15ms. S4. Adaptive Combustion Control: Based on the combustion state identification results of S3 and combined with the boiler load demand, a combustion parameter coupled control model is constructed. The model is based on the model predictive control algorithm and dynamically outputs the control parameters of coal feed rate, primary air volume, secondary air volume, air supply angle and induced draft volume. At the same time, real-time coal quality information fed back by the coal quality soft measurement algorithm is introduced to realize the dynamic adaptive adjustment of combustion parameters and solve the problems of fixed control logic and poor adaptability to operating conditions in the existing technology. S5. Closed-loop feedback optimization: Real-time acquisition of boiler acoustic signals and operating parameters after regulation, comparison with preset optimal combustion benchmark, calculation of regulation error, and optimization of regulation model parameters using regularization coefficients negatively correlated with error and iteration number, forming a closed-loop control of "acquisition-identification-regulation-feedback-optimization", which can adapt to changes in operating conditions without manual intervention and improve the long-term reliability of the system.

[0010] Preferably, the array-type acoustic signature acquisition module in S1 has no fewer than 8 groups, which are respectively arranged in the four quadrants of the boiler furnace: front, rear, left, and right. At least 2 groups are arranged in each quadrant. The spacing between two adjacent acoustic signature acquisition modules is 1.5-2.5m. The acquisition frequency can be dynamically adjusted within the range of 100Hz-10kHz. The pre-amplifier signal conditioning unit integrates signal amplification, filtering, and anti-interference circuits. The acoustic sensor is made of a special cavity structure, acoustic vibrator, and specially made piezoelectric ceramic chip. A bracket (11) is installed at the bottom of the combustion boiler (1). A control module (12) is installed on one side of the bracket (11), and a boiler cover (13) is installed on the other side of the bracket (11). The array-type acoustic fingerprint acquisition module also includes a fine-tuning auxiliary mechanism (2) and a dynamic synchronous control monitoring mechanism (3). The fine-tuning auxiliary mechanism (2) is located below the control module (12). The fine-tuning auxiliary mechanism (2) is used to assist in the adjustment of the dynamic synchronous control monitoring mechanism (3). The dynamic synchronous control monitoring mechanism (3) is located on the combustion boiler (1). The dynamic synchronous control monitoring mechanism (3) is used for dynamic control during boiler monitoring.

[0011] Preferably, the fine-tuning auxiliary mechanism (2) includes an arc-shaped support (21), which is fixedly installed on the combustion boiler (1). A driving device (22) is fixedly installed on the outer surface of the middle part of the arc-shaped support (21). A guide plate (23) is installed on the drive shaft of the driving device (22). A sliding rod is fixedly installed in the middle of the guide plate (23), and the outer surface of the sliding rod is slidably installed in the arc-shaped support (21). A limiting arc-shaped top plate (24) is fixedly installed at the end of the sliding rod away from the driving device (22). The limiting arc-shaped top plate (24) is set on the inner surface of the arc-shaped support (21), and the limiting arc-shaped top plate (24) is the same as the arc surface of the combustion boiler (1).

[0012] Preferably, the dynamic synchronous control and monitoring mechanism (3) includes a stator track (31), which is fixedly installed on the outer surface of the combustion boiler (1) by a bracket. Guide rods (32) are evenly installed on the stator track (31), and a moving slide plate (33) is evenly slidably installed in the guide rod (32). An axial adjustment rod (34) is installed on the moving slide plate (33). The moving slide plates (33) are evenly arranged horizontally and vertically around the stator track (31). A positioning block (35) is slidably installed at the intersection of the horizontal and vertical directions of the axial adjustment rod (34). The height of the axial adjustment rod (34) fixed horizontally on the moving slide plate (33) is higher than the height of the moving slide plate (33) vertically. The center of the positioning block (35) is eccentric. A pressure rod (36) is slidably installed at the pressure rod (36). A return spring (37) is sleeved on the outer surface of the pressure rod (36). One end of the return spring (37) is fixedly installed on the positioning block (35), and the other end of the return spring (37) is fixedly installed on the spherical pressure block (38). A pressure sensor is embedded in the spherical pressure block (38) for real-time monitoring of the pressure information of the spherical pressure block (38). A monitoring sensor (39) is movably connected to the end of the pressure rod (36) away from the return spring (37). The monitoring sensor (39) is used for acoustic monitoring inside the combustion boiler (1). An electromagnetic coil is set on the outer ring of the contact surface between the bottom of the monitoring sensor (39) and the combustion boiler (1). The monitoring sensor (39) can be tightly attached to the outer surface of the combustion boiler (1) by starting and energizing it.

[0013] Preferably, the real-time operating parameters of the boiler in S1 include superheater wall temperature, flue gas oxygen content, coal feed rate, primary air volume, secondary air volume, induced draft volume, and boiler load. These parameters are collected synchronously through the boiler's existing DCS system and newly added temperature and pressure sensors, with the collection cycle consistent with the acoustic signature collection cycle.

[0014] As a preferred embodiment, the deep learning recognition model based on the attention mechanism described in S3 includes a feature fusion layer, an attention allocation layer, a convolutional deep belief network layer, and a classification output layer. The feature fusion layer integrates voiceprint features and operating parameter features using a splicing fusion method. The attention allocation layer dynamically allocates weights according to feature credibility. The convolutional deep belief network layer is used for feature depth extraction and fault classification, and the recognition accuracy is not less than 98%.

[0015] As a preferred embodiment, the soft coal quality measurement algorithm described in S4 is based on the near real-time boiler heat production per ton of coal. It uses boiler efficiency model, heat release model, and heat storage model to shield interference factors, and combines offline test data to calibrate the online model, outputting the coal calorific value curve in real time, providing data support for combustion parameter control.

[0016] Preferably, in step S4, the combustion parameter coupling control model adopts a differentiated control strategy for different combustion states: when fuel is identified as being burned off-center, the coal feeding ratio of each coal feeder and the opening of the secondary small damper are adjusted first; when combustion is identified as being incomplete, the ratio of coal feed to primary air volume is adjusted, and the air supply angle is optimized; when coking in the furnace is identified, the coal feed is appropriately reduced, the primary air volume is increased, and the induced draft volume is adjusted to reduce the local temperature of the furnace.

[0017] Preferably, an abnormality warning step is also included: when S3 detects an abnormal combustion state and the duration of the abnormality exceeds a preset threshold, an early warning signal is automatically triggered, and the abnormality information, soundprint characteristics and control suggestions are pushed to the DCS system and operation and maintenance terminal. The early warning signal includes audible and visual warnings and SMS warnings. The preset threshold can be adjusted according to the boiler model and operating conditions.

[0018] Preferably, the method is applicable to coal-fired boilers, gas-fired boilers and circulating fluidized bed boilers, and can be directly connected to the existing DCS system of the boiler without large-scale modification of the boiler body, adapting to the operating requirements of different load ranges.

[0019] Compared with the prior art, the boiler acoustic signature monitoring and combustion control method provided by the present invention has the following beneficial effects: (1) It is compatible with the complex structure of the boiler, and the sampling location is close to the outline of the furnace without any modification to the main body; The limiting arc-shaped top plate of the fine-tuning auxiliary mechanism is perfectly adapted to the arc surface of the combustion boiler, and the stator track of the dynamic synchronous control monitoring mechanism is arranged circumferentially along the outer surface of the boiler. Both mechanisms are directly assembled on the outer wall of the boiler, without the need for drilling, welding or other modifications to the boiler body. This not only meets the data acquisition requirements of the arc-shaped furnace structure, but also avoids the risks of equipment damage and sealing failure caused by modifications to the boiler body. Compared with the existing fixed-installation acquisition mechanism, it has higher installation adaptability and less impact on the original boiler structure.

[0020] (2) Realize multi-dimensional dynamic adjustment of the acquisition module to adapt to the acquisition needs of all working conditions; The fine-tuning auxiliary mechanism drives the sliding rod and the limiting arc-shaped top plate through the driving device, which can realize the radial and circumferential position fine adjustment of the acquisition module. The dynamic synchronous control and monitoring mechanism, with the sliding cooperation of the moving slide and the axial control rod, can realize the flexible adjustment of the acquisition module in the lateral, longitudinal and height directions. The two work together to complete the multi-dimensional position adjustment of the acquisition module. Compared with the existing fixed position acquisition mechanism, the acquisition position can be dynamically adjusted according to the boiler load and combustion status, always keeping the sensor in the optimal acquisition area and ensuring the effectiveness of acoustic signal acquisition under all operating conditions.

[0021] (3) Achieve dynamic adaptation of the acquisition array spacing, balancing signal coverage and acquisition accuracy; The dynamic synchronous control monitoring mechanism can change the spacing between adjacent monitoring sensors by sliding the moving slide. When the combustion is intense under high load, the spacing is reduced to improve the coverage of acoustic signals, while when the combustion is gentle under low load, the spacing is increased to avoid signal interference between adjacent modules. Compared with the existing fixed-spacing acquisition array, it solves the problem that "fixed spacing cannot meet the signal acquisition needs of different working conditions". It improves the acquisition accuracy while ensuring the comprehensiveness of signals, and can also reduce the energy consumption of the equipment under low load.

[0022] (4) The sensor is firmly attached and has pressure monitoring to ensure stable data acquisition in complex environments; The monitoring sensor of the dynamic synchronous control monitoring mechanism is equipped with an electromagnetic coil at its bottom. After being powered on, it can be firmly attached to the outer surface of the boiler, effectively resisting vibration interference during boiler operation and avoiding signal acquisition distortion caused by sensor loosening. At the same time, the spherical pressure block has an embedded pressure sensor, which can monitor the contact pressure between the sensor and the boiler wall in real time and provide timely feedback on the installation status. Compared with the existing acquisition mechanism without fastening and pressure monitoring, it significantly improves the installation stability of the sensor and the continuity of signal acquisition in the complex environment of high temperature, high vibration and high dust in the boiler.

[0023] (5) Dual-mechanism linkage control and deep integration with the system to automate the data acquisition process; The fine-tuning auxiliary mechanism and the dynamic synchronous control and monitoring mechanism work together, and both are deeply integrated with the boiler acoustic fingerprint acquisition system and the closed-loop feedback optimization system. Based on the real-time operating parameters of the boiler and the quality of acoustic fingerprint signal acquisition, the system can automatically control the mechanism to dynamically adjust the acquisition position and spacing without manual operation. Compared with the existing mechanisms that require manual adjustment of the acquisition position, this realizes the automation and intelligence of the acquisition module adjustment, which greatly reduces the workload of operation and maintenance, and the adjustment response is more timely, which can quickly adapt to the dynamic changes in boiler operating conditions. Attached Figure Description

[0024] Figure 1 This is a three-dimensional structural diagram of the present invention; Figure 2 This is an auxiliary schematic diagram of the three-dimensional structure of the present invention; Figure 3 This is a schematic diagram of the connection relationship of the fine-tuning auxiliary mechanism structure of the present invention; Figure 4 This is a schematic diagram illustrating the connection relationship of the fine-tuning auxiliary mechanism structure of the present invention; Figure 5 For the present invention Figure 4 Enlarged view of point A in the middle; Figure 6 This is a schematic diagram of the structural connection relationship of the dynamic synchronous control and monitoring mechanism of the present invention; Figure 7 For the present invention Figure 6 Enlarged view at point B in the middle; Figure 8 This is a schematic diagram of the boiler acoustic signature monitoring and combustion control method of the present invention.

[0025] In the picture: 1. Combustion boiler; 11. Support frame; 12. Control module; 13. Boiler cover; 2. Fine-tuning auxiliary mechanism; 21. Arc-shaped support; 22. Drive device; 23. Guide plate; 24. Limiting arc-shaped top plate; 3. Dynamic synchronous control and monitoring mechanism; 31. Stator track; 32. Guide rod; 33. Mover slide; 34. Axial control rod; 35. Positioning block; 36. Pressure rod; 37. Return spring; 38. Spherical pressure block; 39. Monitoring sensor. Detailed Implementation

[0026] 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 of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0027] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.

[0028] Example 1, please refer to Figures 1 to 8 As shown: To address the problems mentioned in the technical solutions, this application provides a boiler acoustic signature monitoring and combustion control method, comprising the following steps: S1. Adaptive Acquisition of Acoustic Text Signal: An array of acoustic text acquisition modules is deployed at different heights and quadrants in the furnace of the combustion boiler. The acoustic text acquisition module integrates a high-temperature and dust-resistant micro-sound sensor and a pre-signal conditioning unit to acquire the original acoustic text signal generated during boiler combustion. At the same time, it simultaneously acquires the real-time operating parameters of the boiler and dynamically adjusts the acoustic text acquisition frequency and gain according to the operating parameters to solve the problems of acoustic text signal transmission attenuation and acquisition distortion in high-temperature, high-noise, and high-dust environments. S2. Layered noise reduction and feature extraction of voiceprint signal: The original voiceprint signal is preprocessed. First, the background noise and equipment vibration interference noise in the furnace environment are filtered out by the adaptive wavelet threshold noise reduction algorithm. Then, the nonlinear frequency mapping is performed by the Mel spectrum converter to enhance the low- and medium-frequency combustion features and compress the high-frequency redundant data. Subsequently, the time domain features, frequency domain features and Mel frequency cepstral coefficients of the voiceprint signal are extracted to form a multi-dimensional voiceprint feature set. S3. Multimodal Fusion Recognition of Combustion Status: Construct a deep learning recognition model based on attention mechanism, fuse the multi-dimensional voiceprint feature set extracted in step S2 with the real-time boiler operating parameters collected in S1, dynamically allocate the attention weight of each feature, and achieve accurate recognition of boiler combustion status such as normal combustion, fuel burning, incomplete combustion, furnace coking, and fan abnormality, with recognition time not exceeding 15ms. S4. Adaptive Combustion Control: Based on the combustion state identification results of S3 and combined with the boiler load demand, a combustion parameter coupled control model is constructed. The model is based on the model predictive control algorithm and dynamically outputs the control parameters of coal feed rate, primary air volume, secondary air volume, air supply angle and induced draft volume. At the same time, real-time coal quality information fed back by the coal quality soft measurement algorithm is introduced to realize the dynamic adaptive adjustment of combustion parameters and solve the problems of fixed control logic and poor adaptability to operating conditions in the existing technology. S5. Closed-loop feedback optimization: Real-time acquisition of boiler acoustic signals and operating parameters after regulation, comparison with preset optimal combustion benchmark, calculation of regulation error, and optimization of regulation model parameters using regularization coefficients negatively correlated with error and iteration number, forming a closed-loop control of "acquisition-identification-regulation-feedback-optimization", which can adapt to changes in operating conditions without manual intervention and improve the long-term reliability of the system.

[0029] The S1 array-type acoustic signature acquisition module has no fewer than 8 groups, evenly distributed in the four quadrants of the boiler furnace (front, rear, left, and right), with at least 2 groups in each quadrant. The distance between adjacent groups is controlled at 1.5-2.5m. The acquisition frequency can be dynamically adjusted within the range of 100Hz-10kHz. The pre-amplifier signal conditioning unit integrates signal amplification, filtering, and anti-interference circuits to ensure signal transmission stability. The acoustic sensor is made with a special cavity structure, acoustic vibrator, and specially made piezoelectric ceramic chip, which can adapt to the complex operating conditions of the boiler. A bracket 11 is installed at the bottom of the combustion boiler 1, and a control module 12 and a boiler cover 13 are installed on both sides of the bracket 11, respectively. The array-type acoustic signature acquisition module is equipped with a fine-tuning auxiliary mechanism 2 and a dynamic synchronous control monitoring mechanism 3. The fine-tuning auxiliary mechanism 2 is located below the control module 12 and is used to provide auxiliary adjustment for the dynamic synchronous control monitoring mechanism 3. The dynamic synchronous control monitoring mechanism 3 is mounted on the combustion boiler 1 and is responsible for the dynamic control operation during the boiler monitoring process.

[0030] Specifically, such as Figure 3 and Figure 4As shown, the arc-shaped support 21 is fixedly installed on the surface of the combustion boiler 1. A drive device 22 is fixedly mounted in the middle of its outer surface. A guide plate 23 is connected to the drive shaft of the drive device 22. A slide rod is fixedly provided in the middle of the guide plate 23, and the outer surface of the slide rod slides in cooperation with the arc-shaped support 21. A limiting arc-shaped top plate 24 is fixedly installed at the end of the slide rod away from the drive device 22. The limiting arc-shaped top plate 24 fits against the inner surface of the arc-shaped support 21, and its arc surface is completely adapted to the arc surface of the furnace of the combustion boiler 1. By setting the fine-tuning auxiliary mechanism 2, the acoustic signature acquisition module can be driven to achieve radial and circumferential distance adjustment along the furnace wall without damaging the boiler body structure. It can realize dynamic and rapid adjustment of the acquisition position and adapt to the furnace contour characteristics.

[0031] This solution mechanically connects the pitch-changing mechanism, including the fine-tuning auxiliary mechanism 2 and the dynamic synchronous control and monitoring mechanism 3, with the array-type acoustic fingerprint acquisition module containing high-temperature and dust-resistant micro-sound sensors. These are then synchronously deployed in the four quadrants of the boiler furnace: front, rear, left, and right. The pitch-changing mechanism can drive the acoustic fingerprint acquisition module to achieve radial and circumferential distance adjustment along the furnace wall, and precisely adapt to the boiler furnace contour without damaging the boiler body structure, thus reducing the difficulty of equipment modification.

[0032] In terms of linkage control, the dynamic synchronous control monitoring mechanism 3 is linked with the "adaptive control of acoustic text acquisition" of S1 and the "closed-loop feedback optimization" of S5. The control system can automatically control the variable pitch mechanism to adjust the spacing of the acoustic text acquisition module and the distance from the center of the furnace according to the real-time operating parameters of the boiler, such as the flue gas temperature at the furnace outlet, the boiler load, and the quality of acoustic text signal acquisition, such as the signal-to-noise ratio, so as to achieve dynamic adaptation of the acquisition position and ensure the acquisition effect under different operating conditions.

[0033] For different combustion conditions such as low load, high load, and fluctuating coal quality, the dynamic synchronous control and monitoring mechanism 3 in this solution can dynamically adjust the distribution density of the acquisition array: under high load and intense combustion conditions, the spacing between acquisition modules is reduced to improve the coverage of acoustic signals and ensure comprehensive signal acquisition; under low load and gentle combustion conditions, the spacing between modules is increased, which can reduce equipment energy consumption and effectively avoid signal interference between adjacent modules, ensuring the accuracy of the acquired signals.

[0034] The specific structure is as follows: Figures 5 to 7As shown, the stator track 31 is fixedly installed on the outer surface of the combustion boiler 1 by a bracket. Guide rods 32 are evenly assembled on the track, and a mover slide plate 33 is slidably installed on the guide rods 32. An axial adjustment rod 34 is provided on the mover slide plate 33. The mover slide plate 33 is evenly distributed around the stator track 31 and is arranged in a staggered pattern of transverse and longitudinal directions. The axial adjustment rods 34 arranged laterally are installed at a higher height than the axial adjustment rods 34 arranged longitudinally. A positioning block 35 is slidably installed at the intersection of the two. A pressure rod 36 is slidably assembled at the eccentric position in the middle of the positioning block 35. A return spring 37 is sleeved on the outer surface of the pressure rod 36. The two ends of the return spring 37 are fixedly connected to the positioning block 35 and the spherical pressure block 38, respectively. A pressure sensor is embedded in the spherical pressure block 38, which can monitor its pressure information in real time. The end of the pressure rod 36 away from the return spring 37 is movably connected to a monitoring sensor 39, which is used to collect the acoustic signal inside the combustion boiler 1. An electromagnetic coil is provided on the outer ring of the bottom of the monitoring sensor 39 and the contact surface with the boiler. When powered on, the monitoring sensor 39 can be firmly attached to the outer surface of the combustion boiler 1 to ensure the stability of data acquisition.

[0035] The coordinated setup of the fine-tuning auxiliary mechanism 2 and the dynamic synchronous control and monitoring mechanism 3 effectively solves the drawback of the fixed position of the acoustic fingerprint acquisition module in the existing technology. Through variable distance adjustment, the acquisition module can always be in the optimal acquisition area, ensuring the integrity and accuracy of acoustic fingerprint signal acquisition in the complex environment of boilers with high temperature, high noise, and high dust. At the same time, it can follow the dynamic changes of boiler load, coal quality, and combustion state, and adjust the spacing and position of the acquisition array in real time, making up for the defects of "fixed acquisition parameters and poor adaptability to operating conditions" in the existing technology, and providing more accurate acoustic fingerprint data support for feature extraction in step S2 and state recognition in step S3.

[0036] Furthermore, existing technologies require manual adjustment of the acquisition module position or the addition of multiple fixed acquisition modules to adapt to different boiler operating conditions, which is not only difficult to modify but also has high equipment costs. The variable pitch mechanism in this application can adapt to different loads and coal quality conditions of coal-fired, gas-fired, and circulating fluidized bed boilers through dynamic pitch variation of the same acquisition array, without the need to add additional acquisition modules, which greatly reduces the difficulty and cost of equipment modification.

[0037] The real-time operating parameters of the boiler in S1 include superheater wall temperature, flue gas oxygen content, coal feed rate, primary air volume, secondary air volume, induced draft volume, and boiler load. These parameters are collected synchronously through the boiler's existing DCS system and newly added temperature and pressure sensors, with the collection cycle consistent with the acoustic fingerprint collection cycle.

[0038] The deep learning recognition model based on the attention mechanism in S3 includes a feature fusion layer, an attention allocation layer, a convolutional deep confidence network layer, and a classification output layer. The feature fusion layer integrates voiceprint features and running parameter features through a splicing fusion method. The attention allocation layer dynamically allocates weights according to feature confidence. The convolutional deep confidence network layer is used for feature depth extraction and fault classification, and the recognition accuracy is no less than 98%.

[0039] The S4 coal quality soft measurement algorithm is based on the near real-time boiler heat production per ton of coal. It uses boiler efficiency model, heat release model, and heat storage model to shield interference factors, and combines offline test data to calibrate the online model. It outputs the coal calorific value curve in real time, providing data support for combustion parameter control.

[0040] In S4, the combustion parameter coupling control model adopts differentiated control strategies for different combustion states: when fuel is identified as being burned off-center, the coal feeding ratio of each coal feeder and the opening of the secondary small damper are adjusted first; when combustion is identified as being incomplete, the ratio of coal feed to primary air volume is adjusted, and the air supply angle is optimized; when coking in the furnace is identified, the coal feed is appropriately reduced, the primary air volume is increased, and the induced draft volume is adjusted to reduce the local temperature of the furnace.

[0041] It also includes an abnormal warning step: when S3 detects an abnormal combustion state and the duration of the abnormality exceeds a preset threshold, it automatically triggers a warning signal and pushes the abnormal information, soundprint characteristics and control suggestions to the DCS system and operation and maintenance terminal. The warning signals include audible and visual warnings and SMS warnings. The preset threshold can be adjusted according to the boiler model and operating conditions.

[0042] The method is applicable to coal-fired boilers, gas-fired boilers and circulating fluidized bed boilers. It can be directly connected to the existing DCS system of the boiler without large-scale modification of the boiler body, and is adapted to the operating requirements of different load ranges.

[0043] During boiler combustion, fuel combustion, airflow disturbance, and equipment operation generate sound waves with specific frequencies and amplitudes within the furnace. Different combustion states correspond to different sound wave characteristics: during normal combustion, the sound wave frequency is stable and the amplitude is uniform; under conditions such as uneven fuel burning, incomplete combustion, furnace coking, or fan malfunction, the time-domain waveform, frequency-domain distribution, and energy intensity of the sound waves will show significant shifts. This method collects these sound wave signals using an array of high-temperature resistant and dust-resistant sensors, extracts core acoustic fingerprint features such as Mel-frequency cepstral coefficients and time / frequency domain characteristics through layered noise reduction, and then combines these features with an attention-based deep learning model to fuse the acoustic fingerprint features with boiler operating parameters. This achieves accurate identification of the combustion state based on acoustic features, essentially utilizing the acoustic fingerprint of the combustion process to achieve all-time, high-sensitivity boiler status monitoring.

[0044] Example 2: Based on Embodiment 1, but with some differences, the present invention will now be described in conjunction with specific examples and accompanying drawings. The specific details are as follows.

[0045] This embodiment uses a 35t / h coal-fired circulating fluidized bed boiler as the application example. This boiler is equipped with an existing DCS control system and is mainly used for industrial steam supply. Its operating load range is 20%–100%, and it suffers from common problems such as incomplete combustion at low loads, localized coking in the furnace at high loads, and uneven fuel burning when coal quality fluctuates. The method of this invention directly interfaces with the boiler's existing DCS system without requiring any modification to the boiler itself, achieving fully automated monitoring and control. The detailed implementation process is as follows.

[0046] The specific installation process is as follows: Array-type acoustic signature acquisition module deployment: Two sets of acoustic signature acquisition modules are deployed in each of the four quadrants (front, rear, left, and right) of the boiler furnace, for a total of eight sets. The spacing between adjacent modules is set at 2m, and the modules are deployed at two levels at furnace heights of 6m and 10m to cover the dense phase and dilute phase regions. The acquisition module integrates a high-temperature resistant (≤300℃) and dust-resistant micro-sound sensor and a pre-amplifier signal conditioning unit. The micro-sound sensor adopts a special cavity structure + acoustic vibrator + specially made piezoelectric ceramic chip, and the pre-amplifier unit integrates signal amplification, filtering, and anti-interference circuits.

[0047] Installation of supporting mechanisms: A fine-tuning auxiliary mechanism 2 and a dynamic synchronous control monitoring mechanism 3 are installed at corresponding positions on the outer wall of the boiler. The arc support 21 of the fine-tuning auxiliary mechanism is precisely adapted to the arc surface of the boiler furnace. The stator track 31 of the dynamic synchronous control monitoring mechanism is fixed along the circumference of the furnace. The electromagnetic coil at the bottom of the monitoring sensor 39 is attached to the outer wall of the boiler and is tightened by energizing.

[0048] Parameter acquisition system: The existing DCS system of the boiler is used to collect parameters such as superheater wall temperature, flue gas oxygen content, coal feed rate, primary air volume, secondary air volume, induced draft volume, and boiler load. New temperature and pressure sensors and acoustic fingerprint acquisition modules are deployed synchronously to ensure that the operating parameters and acoustic fingerprint signal acquisition cycle are consistent and set to 50ms.

[0049] System integration: The acoustic fingerprint acquisition module, supporting mechanisms, and sensors are integrated with the boiler DCS system and the newly added control host. The control host incorporates the deep learning recognition model, combustion parameter coupling control model, and coal quality soft measurement algorithm of this invention.

[0050] Step 1: Adaptive acquisition of voiceprint signals; Initial acquisition parameter settings: Based on the boiler's rated load of 35t / h, the initial acoustic signature acquisition frequency is set to 5kHz and the gain is set to 20dB. The acoustic sensor is activated to acquire the raw acoustic signature signal during the combustion process in the furnace. At the same time, the DCS system synchronously acquires the boiler's real-time operating parameters.

[0051] Dynamic adjustment of acquisition parameters: The control host adaptively adjusts the acquisition frequency and gain according to real-time operating parameters: When the boiler load drops to the low load range of 20%~50% and the combustion is slow, the acquisition frequency is adjusted to 1kHz and the gain is reduced to 10dB to avoid signal redundancy; when the load rises to the high load range of 80%~100% and the combustion is intense, the acquisition frequency is adjusted to 10kHz and the gain is increased to 30dB to ensure acoustic signal coverage; when a sudden increase in local noise in the furnace is detected, the filtering level is automatically increased to solve the problems of signal transmission attenuation and acquisition distortion in high noise and high dust environments.

[0052] Dynamic Adaptation of Acquisition Location: The dynamic synchronous control monitoring mechanism 3 is linked with the acquisition system to adjust the position of the acquisition module according to the signal-to-noise ratio of the acoustic signal and the flue gas temperature at the furnace outlet: the module spacing is reduced to 1.5m under high load to improve signal coverage; the spacing is increased to 2.5m under low load to avoid signal interference; the fine-tuning auxiliary mechanism 2 drives the slide bar and the limiting arc surface top plate 24 to move through the drive device 22 to assist in adjusting the radial and circumferential positions of the monitoring sensor 39, so that the acquisition module is always in the optimal acquisition area and does not damage the boiler body structure throughout the process.

[0053] Step 2; Layered noise reduction and feature extraction of voiceprint signals; Adaptive wavelet threshold denoising: The control host preprocesses the collected raw acoustic signal and uses an adaptive wavelet threshold denoising algorithm to filter out background noise (such as boiler body vibration) and equipment interference noise (such as fan and coal feeder operating noise) in the furnace environment, retaining the effective combustion acoustic signal and improving the signal-to-noise ratio.

[0054] Mel spectrum nonlinear frequency mapping: The noise-reduced acoustic signal is input into the Mel spectrum converter for nonlinear frequency mapping to enhance the low-to-mid frequency combustion features of 50Hz to 5kHz (this frequency band is the core feature area of ​​boiler combustion acoustics), while compressing high-frequency redundant data above 5kHz to reduce the computational load of subsequent feature extraction.

[0055] Multi-dimensional feature extraction: Extract time-domain features (root mean square, peak factor, impulse factor), frequency-domain features (center frequency, band energy, spectral centroid), and Mel frequency cepstral coefficients (MFCC) (12th-order fundamental coefficients + 1st-order differential coefficients) from the processed signal. Integrate the three types of features to form a multi-dimensional voiceprint feature set, providing a data foundation for subsequent combustion state identification.

[0056] Step 3: Multimodal Fusion Recognition of Combustion State Feature fusion: The multi-dimensional voiceprint feature set extracted in step S2 and the boiler real-time operation parameter features collected in S1 are input into the feature fusion layer of the deep learning recognition model based on the attention mechanism. The two types of features are integrated by splicing and fusion to form a fused feature matrix.

[0057] Attention weight allocation: The attention allocation layer dynamically allocates weights based on feature credibility: high weights (0.7~0.9) are allocated to frequency bands with obvious combustion characteristics in the acoustic signal (such as 2~3kHz) and indicators with strong correlation in the operating parameters (such as flue gas oxygen content and superheater wall temperature), while low weights (0.1~0.3) are allocated to features with strong interference and low credibility, so as to achieve key mining of effective features.

[0058] Deep extraction and classification: The feature matrix is ​​input into the convolutional deep confidence network layer. Feature depth is extracted through multi-layer convolution and pooling operations. Then, the Softmax classifier is used for fault classification to identify five states: normal combustion, fuel burning, incomplete combustion, furnace coking, and fan malfunction.

[0059] Recognition result output: The recognition time of this model is controlled within 10ms (≤15ms), and the recognition accuracy reaches 98.5% (≥98%). The recognition results are transmitted to the control host and boiler DCS system in real time to provide a basis for subsequent control.

[0060] Step 4: Adaptive combustion control; Real-time analysis of coal quality soft measurement: The control host uses the coal quality soft measurement algorithm as the basis for the boiler's near real-time heat production per ton of coal. It combines the boiler efficiency model, heat release model, and heat storage model to shield against interference factors such as coal quality fluctuations and load changes. At the same time, it uses daily offline coal quality test data to calibrate the online model and outputs the coal calorific value curve in real time (e.g., when the calorific value drops from 20MJ / kg to 16MJ / kg, the algorithm identifies and provides feedback in real time).

[0061] Initialization of control model parameters: Combustion parameter coupled control model based on Model predictive control (MPC) The algorithm initializes the basic parameters of coal feed, primary air volume, secondary air volume, air supply angle, and induced draft volume based on the current load demand of the boiler (e.g., industrial steam demand of 30t / h, i.e., 85% load).

[0062] Execution of differentiated combustion control strategy: Based on the combustion state identification results in step S3 and combined with real-time coal quality information from soft coal quality measurement, the model outputs dynamic control parameters and executes differentiated control. Fuel imbalance: When it is detected that the combustion is intense on the left side of the furnace and weak on the right side, the coal feeding ratio of the four coal feeders is adjusted first (the coal feeding on the left side is reduced by 10% and the coal feeding on the right side is increased by 10%). At the same time, the opening of the secondary small air damper is adjusted (the small air damper on the left side is closed by 20% and the one on the right side is opened by 20%) to achieve uniform coal and air distribution.

[0063] Incomplete combustion: When the exhaust oxygen content is below 3% and the combustion pulse factor indicated by the audible signal is low, reduce the coal feed by 8%, increase the primary air volume by 12%, and adjust the air supply angle from horizontal to upward by 15° to enhance the mixing effect of pulverized coal and air and improve combustion efficiency.

[0064] Furnace coking: When the superheater wall temperature is locally overheated (≥550℃) and the low-frequency characteristics of the acoustic signal are abnormal, reduce the total coal feed by 15%, increase the primary air volume by 10%, increase the induced draft volume by 8%, reduce the local temperature of the furnace, and inhibit the further development of coking.

[0065] Fan malfunction: When the frequency deviation of the fan's operating sound pattern or the fluctuation of air volume are detected, an early warning is triggered first, and then the fan frequency conversion parameters are finely adjusted to stabilize the primary / secondary air volume and prevent the combustion conditions from deteriorating.

[0066] Normal combustion: Maintain the current combustion parameters and make minor adjustments based on changes in coal quality (e.g., when the calorific value decreases, slightly increase the coal feed rate and adjust the air volume simultaneously) to maintain the optimal combustion state.

[0067] Step 5: Closed-loop feedback optimization; Post-control parameter acquisition: After the control parameters are executed, the system collects the boiler's acoustic signature signal and operating parameters in real time as feedback data.

[0068] Control error calculation: The feedback data is compared with the preset optimal combustion benchmark (such as flue gas oxygen content of 3%~5%, boiler thermal efficiency ≥88%, and sound signature characteristics within the normal range) to calculate the control error of each parameter.

[0069] Model parameter optimization: A regularization coefficient negatively correlated with error and iteration count is used to optimize the combustion parameter coupling control model. The larger the error, the larger the regularization coefficient, and the greater the adjustment range of model parameters. The more iterations (the more stable the system operation), the smaller the regularization coefficient, and the more minor parameter adjustments are made. For example, if the initial control fails to alleviate furnace coking, the coefficient is set to 0.8, and the coal and air blending parameters are significantly adjusted. After multiple iterations, combustion stabilizes, and the coefficient is set to 0.2, with only minor adjustments based on coal quality.

[0070] Closed-loop control is formed: Through the cycle of "collection-identification-regulation-feedback-optimization", the closed-loop automatic control of boiler combustion is realized. Without manual intervention, it can adapt to scenarios such as load changes, coal quality fluctuations, and operating condition adjustments, thereby improving the long-term reliability of the system.

[0071] Step 5: Anomaly Warning Procedure; Abnormal duration determination: When step S3 identifies an abnormal combustion state, such as coking in the furnace or abnormal fan operation, the system starts timing to determine whether the duration of the abnormality exceeds a preset threshold. In this embodiment, the threshold is set to 10 seconds based on the characteristics of the boiler, and can be adjusted as needed.

[0072] Warning signal triggering: If the abnormal duration is ≥10s, an audible and visual warning (the audible and visual alarm in the boiler control room is activated) and an SMS warning (the abnormal information is pushed to the mobile phones of operation and maintenance personnel) will be automatically triggered. At the same time, the abnormal type, soundprint characteristics, real-time operating parameters and recommended control suggestions will be pushed to the boiler DCS system and operation and maintenance terminal.

[0073] Manual intervention: If the anomaly persists after automatic system adjustment, maintenance personnel can intervene manually based on the warning information to prevent the fault from escalating.

[0074] The method in this embodiment is also applicable to gas-fired boilers and other tonnage coal-fired boilers. It is only necessary to adjust the number / spacing of the acoustic fingerprint acquisition modules, the preset threshold, the optimal combustion benchmark, and other parameters according to the boiler model and furnace structure to achieve the adaptation application.

[0075] This design employs a closed-loop control system encompassing "collection-identification-regulation-feedback-optimization," unlike existing technologies that lack effective feedback optimization and require frequent manual intervention. This solution collects the acoustic signature signal and operating parameters after regulation in real time, compares them with the optimal combustion benchmark to calculate the regulation error, and dynamically optimizes the regulation model parameters using a regularization coefficient negatively correlated with the error and iteration count. Simultaneously, a dynamic synchronous regulation and monitoring mechanism is included, which adaptively adjusts the position and parameters of the acquisition module according to changes in operating conditions. This achieves end-to-end self-optimization of monitoring, identification, and regulation, adapting to various operating conditions such as full load ranges and coal quality fluctuations without manual intervention, significantly improving the long-term stability and reliability of the boiler.

[0076] Please refer to the above work process. Figures 1 to 8 .

[0077] It should be noted that the term "comprising" or any other variation thereof is intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0078] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for boiler acoustic monitoring and combustion control, characterized in that, Includes the following steps: S1. Adaptive acquisition of acoustic signature: An array of acoustic signature acquisition modules is deployed at different heights and quadrants of the furnace in the combustion boiler (1); S2. Layered noise reduction and feature extraction of voiceprint signal: The original voiceprint signal is preprocessed. First, the background noise and equipment vibration interference noise in the furnace environment are filtered out by the adaptive wavelet threshold noise reduction algorithm. Then, the nonlinear frequency mapping is performed by the Mel spectrum converter to enhance the low- and medium-frequency combustion features and compress the high-frequency redundant data. Subsequently, the time domain features, frequency domain features and Mel frequency cepstral coefficients of the voiceprint signal are extracted to form a multi-dimensional voiceprint feature set. S3. Multimodal Fusion Recognition of Combustion Status: Construct a deep learning recognition model based on attention mechanism, fuse the multi-dimensional voiceprint feature set extracted in step S2 with the real-time boiler operating parameters collected in S1, dynamically allocate the attention weight of each feature, and achieve accurate recognition of boiler combustion status such as normal combustion, fuel burning, incomplete combustion, furnace coking, and fan abnormality, with recognition time not exceeding 15ms. S4. Adaptive Combustion Control: Based on the combustion state identification results of S3 and combined with the boiler load demand, a combustion parameter coupled control model is constructed. The model is based on the model predictive control algorithm and dynamically outputs the control parameters of coal feed rate, primary air volume, secondary air volume, air supply angle and induced draft volume. At the same time, real-time coal quality information fed back by the coal quality soft measurement algorithm is introduced to realize the dynamic adaptive adjustment of combustion parameters and solve the problems of fixed control logic and poor adaptability to operating conditions in the existing technology. S5. Closed-loop feedback optimization: Real-time acquisition of boiler acoustic signals and operating parameters after regulation, comparison with preset optimal combustion benchmark, calculation of regulation error, and optimization of regulation model parameters using regularization coefficients negatively correlated with error and iteration number, forming a closed-loop control of "acquisition-identification-regulation-feedback-optimization", which can adapt to changes in operating conditions without manual intervention and improve the long-term reliability of the system.

2. The boiler acoustic signature monitoring and combustion control method according to claim 1, characterized in that, The array-type acoustic signature acquisition module described in S1 shall be deployed in no fewer than 8 groups, respectively deployed in the four quadrants of the boiler furnace: front, rear, left, and right. At least 2 groups shall be deployed in each quadrant. The spacing between two adjacent acoustic signature acquisition modules shall be 1.5-2.5m. The acquisition frequency can be dynamically adjusted within the range of 100Hz-10kHz. The pre-amplifier signal conditioning unit integrates signal amplification, filtering, and anti-interference circuits. The micro-sound sensor is made of a special cavity structure, acoustic vibrator, and specially made piezoelectric ceramic chip. A bracket (11) is installed at the bottom of the combustion boiler (1). 11) A control module (12) is installed on one side, and a boiler cover (13) is installed on the other side of the bracket (11). The array-type acoustic fingerprint acquisition module also includes a fine-tuning auxiliary mechanism (2) and a dynamic synchronous control monitoring mechanism (3). The fine-tuning auxiliary mechanism (2) is located below the control module (12). The fine-tuning auxiliary mechanism (2) is used to assist in the adjustment of the dynamic synchronous control monitoring mechanism (3). The dynamic synchronous control monitoring mechanism (3) is located on the combustion boiler (1). The dynamic synchronous control monitoring mechanism (3) is used for dynamic control during boiler monitoring.

3. The boiler acoustic signature monitoring and combustion control method according to claim 2, characterized in that, The fine-tuning auxiliary mechanism (2) includes an arc-shaped support (21), which is fixedly installed on the combustion boiler (1). A driving device (22) is fixedly installed on the outer surface of the middle part of the arc-shaped support (21). A guide plate (23) is installed on the drive shaft of the driving device (22). A sliding rod is fixedly installed in the middle of the guide plate (23), and the outer surface of the sliding rod is slidably installed in the arc-shaped support (21). A limiting arc-shaped top plate (24) is fixedly installed at the end of the sliding rod away from the driving device (22). The limiting arc-shaped top plate (24) is set on the inner surface of the arc-shaped support (21), and the limiting arc-shaped top plate (24) is the same as the arc surface of the combustion boiler (1).

4. The boiler acoustic signature monitoring and combustion control method according to claim 2, characterized in that, The dynamic synchronous control and monitoring mechanism (3) includes a stator track (31), which is fixedly installed on the outer surface of the combustion boiler (1) by a bracket. Guide rods (32) are evenly installed on the stator track (31), and a moving slide plate (33) is evenly slidably installed in the guide rod (32). An axial adjustment rod (34) is installed on the moving slide plate (33). The moving slide plates (33) are evenly arranged horizontally and vertically around the stator track (31). A positioning block (35) is slidably installed at the intersection of the horizontal and vertical directions of the axial adjustment rod (34). The height of the axial adjustment rod (34) fixed horizontally on the moving slide plate (33) is higher than the height of the vertical installation on the moving slide plate (33). The positioning block (35) slides at the eccentric part in the middle. A pressure rod (36) is installed, and a return spring (37) is sleeved on the outer surface of the pressure rod (36). One end of the return spring (37) is fixedly installed on the positioning block (35), and the other end of the return spring (37) is fixedly installed on a spherical pressure block (38). A pressure sensor is embedded in the spherical pressure block (38) for real-time monitoring of the pressure information of the spherical pressure block (38). A monitoring sensor (39) is movably connected to the end of the pressure rod (36) away from the return spring (37). The monitoring sensor (39) is used for acoustic monitoring inside the combustion boiler (1). An electromagnetic coil is set on the outer ring of the contact surface between the bottom of the monitoring sensor (39) and the combustion boiler (1). The monitoring sensor (39) can be tightly attached to the outer surface of the combustion boiler (1) by starting and energizing it.

5. The boiler acoustic signature monitoring and combustion control method according to claim 1, characterized in that, The real-time operating parameters of the boiler in S1 include superheater wall temperature, flue gas oxygen content, coal feed rate, primary air volume, secondary air volume, induced draft volume, and boiler load. These parameters are collected synchronously through the boiler's existing DCS system and newly added temperature and pressure sensors, with the collection cycle consistent with the acoustic fingerprint collection cycle.

6. The boiler acoustic signature monitoring and combustion control method according to claim 1, characterized in that, The deep learning recognition model based on the attention mechanism described in S3 includes a feature fusion layer, an attention allocation layer, a convolutional deep belief network layer, and a classification output layer. The feature fusion layer integrates voiceprint features and running parameter features through a splicing fusion method. The attention allocation layer dynamically allocates weights according to feature credibility. The convolutional deep belief network layer is used for feature depth extraction and fault classification, and the recognition accuracy is no less than 98%.

7. The boiler acoustic signature monitoring and combustion control method according to claim 1, characterized in that, The soft coal quality measurement algorithm described in S4 is based on the near real-time boiler heat production per ton of coal. It uses boiler efficiency model, heat release model, and heat storage model to shield interference factors, and combines offline test data to calibrate the online model. It outputs the calorific value curve of coal in real time, providing data support for combustion parameter control.

8. The boiler acoustic signature monitoring and combustion control method according to claim 1, characterized in that, In S4, the combustion parameter coupling control model adopts differentiated control strategies for different combustion states: when fuel is identified as being burned off-center, the coal feeding ratio of each coal feeder and the opening of the secondary small damper are adjusted first; when combustion is identified as being incomplete, the ratio of coal feed to primary air volume is adjusted, and the air supply angle is optimized; when coking in the furnace is identified, the coal feed is appropriately reduced, the primary air volume is increased, and the induced draft volume is adjusted to reduce the local temperature of the furnace.

9. The boiler acoustic signature monitoring and combustion control method according to claim 1, characterized in that, It also includes an abnormal warning step: when S3 detects an abnormal combustion state and the duration of the abnormality exceeds a preset threshold, it automatically triggers a warning signal and pushes the abnormal information, soundprint characteristics and control suggestions to the DCS system and operation and maintenance terminal. The warning signals include audible and visual warnings and SMS warnings. The preset threshold can be adjusted according to the boiler model and operating conditions.

Images