Intelligent monitoring device for ball mill based on multi-modal voiceprint analysis

Abstract

The invention discloses a ball mill intelligent monitoring device based on multi-mode voiceprint analysis, and belongs to the technical field of ball mill equipment monitoring. Comprising a mechanical fixing assembly. And the energy supply unit is integrated on one side of the mechanical fixing assembly through a waterproof sealing structure and is used for providing electric energy for the multi-mode sensing module and the data preprocessing module. Through the innovative multi-mode sensing fusion technology and the acoustic guided wave structure design, the leap-type improvement of the monitoring precision and the early warning capability is realized. The device integrates a voiceprint acquisition unit, a vibration sensing unit and a temperature sensing unit, and constructs a'voiceprint-vibration-temperature 'three-in-one multi-modal monitoring system.

Description

Technical Field

[0001] This application relates to the field of ball mill equipment monitoring technology, and more specifically, to an intelligent monitoring device for ball mills based on multimodal acoustic signature analysis. Background Technology

[0002] Accurate monitoring of the working status of a ball mill is crucial for ensuring production efficiency, improving product quality, and reducing equipment wear and tear during operation. Traditional monitoring methods have several shortcomings. For example, traditional vibration monitoring is insensitive to early, subtle changes in the internal structure of the equipment, resulting in a high false alarm rate; infrared thermography, due to time delays, struggles to capture instantaneous anomalies during ball mill operation; and existing acoustic detection methods rely solely on single-frequency threshold alarms, failing to comprehensively and accurately reflect the true working status of the ball mill. Furthermore, in terms of mechanical structure design, existing monitoring devices have deficiencies in sensor layout, signal transmission structure, and coordination with the ball mill's main structure, leading to unsatisfactory monitoring results.

[0003] In view of this, an intelligent monitoring device for ball mills based on multimodal acoustic signature analysis is proposed. Summary of the Invention

[0004] 1. Technical problems to be solved

[0005] The purpose of this application is to provide an intelligent monitoring device for ball mills based on multimodal acoustic signature analysis, which solves the technical problems mentioned in the background art.

[0006] 2. Technical Solution

[0007] This application provides an intelligent monitoring device for ball mills based on multimodal acoustic signature analysis, including:

[0008] A mechanical fixing assembly includes an arc-shaped mounting base, an elastic damping layer, and a fastening unit. The arc-shaped mounting base is configured to match the curvature of the outer wall of the ball mill cylinder and is attached to the outer wall of the cylinder through the elastic damping layer. The fastening unit is configured to fix the arc-shaped mounting base to the ball mill cylinder.

[0009] A multimodal sensing module, integrated on the arc-shaped mounting base, includes an acoustic signature acquisition unit, a vibration sensing unit, and a temperature sensing unit. The acoustic signature acquisition unit is configured to acquire acoustic wave signals transmitted through the ball mill cylinder.

[0010] A data preprocessing module is disposed inside the mechanical fixing component and electrically connected to the multimodal sensing module, and is configured to collect acoustic wave signals transmitted through the ball mill cylinder;

[0011] An energy supply unit, which is integrated into one side of the mechanical fixing component through a waterproof sealing structure, is used to provide power to the multimodal sensing module and the data preprocessing module.

[0012] As an optional solution to the technical solution of this application, the voiceprint acquisition unit includes:

[0013] An acoustic resonant cavity is formed inside the arc-shaped mounting base;

[0014] An elastic diaphragm is embedded at the bottom of the acoustic resonant cavity and configured to contact the surface of the ball mill cylinder to pick up vibrational sound waves.

[0015] A metal acoustic conduit, one end of which is connected to the acoustic resonant cavity;

[0016] A microphone array connected to the other end of the metal acoustic conduit, and the microphone array includes at least two microphones optimized for different frequency bands;

[0017] The elastic diaphragm, acoustic resonant cavity, and metal sound transmission duct together constitute an acoustic waveguide structure.

[0018] As an optional solution to the technical solution of this application, the mechanical fixing component further includes a dustproof and rainproof cover, which is sealed to the arc-shaped mounting base to form a sealed cavity. The multimodal sensing module and the data preprocessing module are housed in the sealed cavity. The dustproof and rainproof cover is provided with an acoustic transmission window, which faces the acoustic path of the acoustic signature acquisition unit.

[0019] As an optional solution to the technical solution of this application, the fastening unit is a mounting ear extending from both sides of the arc-shaped mounting base, and the mounting ear is provided with a through hole for connection to the ball mill cylinder by bolts.

[0020] As an optional solution to the technical solution of this application, a ball mill intelligent monitoring method based on multimodal acoustic signature analysis is provided. The method is implemented using the apparatus described in any one of claims 1-4, comprising:

[0021] S1. The multimodal signals during the operation of the ball mill are collected through the multimodal sensing module. The multimodal signals include acoustic wave signals, vibration signals and temperature signals transmitted through the acoustic waveguide structure.

[0022] S2. Perform synchronization and noise reduction processing on the acquired multimodal signals;

[0023] S3. Extract feature parameters from the processed signal, including the Mel frequency cepstral coefficients of the acoustic signature signal, the peak frequency of the vibration signal, and the rate of change of the temperature signal.

[0024] S4. Input the feature parameters into the trained multimodal fusion model to obtain the operating status evaluation result of the ball mill, and generate early warning information when the evaluation result is abnormal.

[0025] As an optional solution to the technical solution in this application, the multimodal fusion model adopts an attention mechanism to weightedly fuse voiceprint features and vibration features, and the attention weights are dynamically adjusted according to the contribution of each modal feature in historical fault data.

[0026] As an optional solution to the technical solution in this application, it also includes a model self-updating step: when an actual fault is detected in the ball mill, the multimodal data within a set time period before the fault occurs is automatically marked as samples for optimizing the parameters of the multimodal fusion model.

[0027] As an optional solution to the technical solution in this application, a ball mill intelligent monitoring system based on multimodal acoustic signature analysis includes:

[0028] The intelligent monitoring device as described in any one of claims 1-4 is configured to be installed in the ball mill cylinder and collect multimodal sensing data;

[0029] An edge computing unit, connected to the intelligent monitoring device via wireless communication, is configured to perform the method of any one of claims 5-7 to process and analyze the multimodal sensing data;

[0030] The cloud management platform communicates with the edge computing unit and is configured to store analysis results, perform trend analysis, and generate remote early warning commands.

[0031] The human-computer interaction terminal is connected to the cloud management platform and is configured to display the ball mill's operating status and early warning information to the user.

[0032] As an optional solution to the technical solution in this application, the intelligent monitoring device is connected to the edge computing unit via an industrial Ethernet or LoRa wireless network, and the wireless communication adopts frequency hopping spread spectrum technology.

[0033] The cloud management platform includes a fault diagnosis knowledge base, which stores multimodal feature templates corresponding to different ball mill types and fault states, used to assist in the verification of real-time monitoring results.

[0034] As an optional solution to the technical solution of this application, it also includes an adaptive adjustment module, which adjusts the feeding rate or grinding media filling rate of the ball mill through an electromagnetic actuator based on the analysis results of the cloud management platform.

[0035] 3. Beneficial effects

[0036] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0037] 1. This application achieves a significant leap forward in monitoring accuracy and early warning capabilities through innovative multimodal sensor fusion technology and acoustic waveguide structure design. The device integrates an acoustic signature acquisition unit, a vibration sensing unit, and a temperature sensing unit, constructing a three-in-one multimodal monitoring system encompassing acoustic signature, vibration, and temperature. The acoustic signature acquisition unit employs an acoustic waveguide structure composed of an elastic diaphragm, an acoustic resonant cavity, and a metal sound transmission conduit, coupled with at least two microphone arrays optimized for different frequency bands, breaking through the frequency band limitations of traditional sound wave detection. The elastic diaphragm directly contacts the cylinder surface to pick up vibrational sound waves. After the acoustic resonant cavity amplifies the signal in a specific frequency band, it is efficiently transmitted to the microphone array through the metal sound transmission conduit, achieving accurate capture of wide-frequency acoustic signature features. This not only solves the problem of incomplete monitoring information in a single frequency band but also amplifies weak acoustic signature signals of early faults, enabling the identification of subtle anomalies such as gear wear and bearing deterioration within the equipment at their nascent stage.

[0038] 2. This application constructs a highly stable monitoring hardware platform adapted to the operating conditions of a ball mill through an integrated mechanical fixing component design. The coordinated design of the arc-shaped mounting base and the elastic damping layer: the arc-shaped mounting base precisely matches the curvature of the outer wall of the ball mill cylinder, achieving a tight fit through mounting ears on both sides and bolt fastening units; the elastic damping layer uses high-damping material to effectively filter environmental interference noise outside the cylinder vibration, ensuring an improved signal-to-noise ratio of the sensing signal. Simultaneously, the mechanical fixing component integrates a dustproof and rainproof cover and a sealed cavity structure, encapsulating the multimodal sensing module and data preprocessing module within. Combined with the directional design of the acoustic wave transmission window, it not only blocks external erosion such as dust and rainwater but also ensures lossless transmission of the acoustic signal.

[0039] 3. This application achieves intelligent, adaptive, and closed-loop control of fault diagnosis by constructing a multimodal fusion algorithm and an "edge-cloud" collaborative architecture. At the algorithm level, the multimodal fusion model uses an attention mechanism to dynamically weight acoustic and vibration features, with the weight values ​​adjusted in real time according to the contribution of each mode in historical fault data. At the system architecture level, the edge computing unit realizes localized real-time data processing, quickly extracting feature parameters through algorithms such as wavelet threshold denoising and Kalman filtering. The cloud management platform relies on the fault diagnosis knowledge base and big data analysis to perform trend analysis and remote early warning, and drives the electromagnetic actuator through an adaptive adjustment module to dynamically optimize the feeding rate and grinding media filling rate. Attached Figure Description

[0040] Figure 1This is a schematic diagram of the overall structure and installation of a ball mill intelligent monitoring device based on multimodal acoustic signature analysis, as disclosed in a preferred embodiment of this application.

[0041] Figure 2 This is a schematic diagram of the mechanical fixing component structure of the intelligent monitoring device for ball mills based on multimodal acoustic signature analysis disclosed in a preferred embodiment of this application;

[0042] Figure 3 This is a schematic diagram of the multimodal sensing module structure of the intelligent monitoring device for ball mills based on multimodal acoustic signature analysis, as disclosed in a preferred embodiment of this application.

[0043] Figure 4 This is a schematic diagram of the acoustic signature acquisition unit structure of the intelligent monitoring device for ball mills based on multimodal acoustic signature analysis disclosed in a preferred embodiment of this application;

[0044] Figure 5 This is a schematic flowchart of a ball mill intelligent monitoring method based on multimodal acoustic signature analysis disclosed in a preferred embodiment of this application.

[0045] Figure 6 This is a schematic diagram of a ball mill intelligent monitoring system based on multimodal acoustic signature analysis, as disclosed in a preferred embodiment of this application.

[0046] The following are the labeling instructions in the diagram: 10. Mechanical fixing component; 11. Arc-shaped mounting base; 12. Elastic damping layer; 13. Fastening unit; 14. Acoustic resonant cavity; 15. Elastic diaphragm; 16. Dustproof and rainproof cover; 17. Acoustic wave-transmitting window; 20. Multimodal sensing module; 21. Acoustic fingerprint acquisition unit; 22. Vibration sensing unit; 23. Temperature sensing unit; 24. Microphone array; 25. Metal acoustic conduit; 30. Data preprocessing module; 40. Energy supply unit. Detailed Implementation

[0047] The present application will be further described in detail below with reference to the accompanying drawings.

[0048] Example 1

[0049] Reference Figure 1-4 A ball mill intelligent monitoring device based on multimodal acoustic signature analysis includes:

[0050] Mechanical fixing assembly 10 includes an arc-shaped mounting base 11, an elastic damping layer 12, and a fastening unit 13. The arc-shaped mounting base 11 is configured to match the curvature of the outer wall of the ball mill cylinder and is attached to the outer wall of the cylinder through the elastic damping layer 12. The fastening unit 13 is configured to fix the arc-shaped mounting base 11 to the ball mill cylinder.

[0051] The multimodal sensing module 20 is integrated on the arc-shaped mounting base 11 and includes an acoustic signature acquisition unit 21, a vibration sensing unit 22 and a temperature sensing unit 23. The acoustic signature acquisition unit 21 is configured to acquire acoustic wave signals transmitted through the ball mill cylinder.

[0052] The data preprocessing module 30 is located inside the mechanical fixing component 10 and is electrically connected to the multimodal sensing module 20. It is configured to collect the acoustic wave signal transmitted through the ball mill cylinder.

[0053] The energy supply unit 40 is integrated into one side of the mechanical fixing component 10 through a waterproof sealing structure, and is used to provide power to the multimodal sensing module 20 and the data preprocessing module 30.

[0054] This intelligent monitoring device for ball mills based on multimodal acoustic signature analysis achieves a tight fit with the ball mill cylinder through the cooperation of the arc-shaped mounting base 11 of the mechanical fixing component 10 and the elastic damping layer 12, ensuring the stability of sensor signal acquisition.

[0055] Reference Figure 4 and Figure 3 The acoustic signature acquisition unit 21 in the intelligent monitoring device for ball mills based on multimodal acoustic signature analysis described in this application embodiment includes:

[0056] An acoustic resonant cavity 14 is formed inside the arc-shaped mounting base 11;

[0057] An elastic diaphragm 15 is embedded in the bottom of the acoustic resonant cavity 14 and is configured to contact the surface of the ball mill cylinder to pick up vibrational sound waves.

[0058] A metal acoustic conduit 25, one end of which is connected to an acoustic resonant cavity 14;

[0059] Microphone array 24 is connected to the other end of metal sound transmission duct 25, and microphone array 24 includes at least two microphones optimized for different frequency bands;

[0060] The elastic diaphragm 15, the acoustic resonant cavity 14, and the metal sound transmission duct 25 together constitute an acoustic waveguide structure.

[0061] This intelligent monitoring device for ball mills based on multimodal acoustic signature analysis effectively enhances acoustic signals in specific frequency bands through the design of an acoustic waveguide structure, and the multi-band configuration of the microphone array 24 enables comprehensive capture of wideband acoustic signature features.

[0062] Reference Figure 2 and Figure 3The mechanical fixing component 10 of the ball mill intelligent monitoring device based on multimodal acoustic signature analysis described in this application embodiment also includes a dustproof and rainproof cover 16. The dustproof and rainproof cover 16 is sealed to the arc-shaped mounting base 11 to form a sealed cavity. The multimodal sensing module 20 and the data preprocessing module 30 are housed in the sealed cavity. The dustproof and rainproof cover 16 is provided with an acoustic transmission window 17, which faces the acoustic path of the acoustic signature acquisition unit 21.

[0063] This intelligent monitoring device for ball mills based on multimodal acoustic signature analysis, through the combined design of a sealed cavity and an acoustically transparent window 17, provides effective protection while ensuring the transmission efficiency of acoustic signature signals.

[0064] Reference Figure 2 and Figure 1 In the ball mill intelligent monitoring device based on multimodal acoustic analysis described in this application embodiment, the fastening unit 13 consists of mounting ears extending from both sides of the arc-shaped mounting base 11. The mounting ears are provided with through holes to be connected to the ball mill cylinder by bolts.

[0065] This intelligent monitoring device for ball mills based on multimodal acoustic signature analysis achieves rapid disassembly and secure fixation through the connection method of mounting ears and bolts.

[0066] Example 2

[0067] Reference Figure 5 A method for intelligent monitoring of ball mills based on multimodal acoustic signature analysis, the method being implemented using the apparatus of any one of claims 1-4, comprising:

[0068] S1. The multimodal signals during the operation of the ball mill are collected by the multimodal sensing module 20. The multimodal signals include acoustic wave signals, vibration signals and temperature signals transmitted through the acoustic waveguide structure.

[0069] S2. Perform synchronization and noise reduction processing on the acquired multimodal signals;

[0070] S3. Extract feature parameters from the processed signal. Feature parameters include the Mel frequency cepstral coefficients of the acoustic signal, the peak frequency of the vibration signal, and the rate of change of the temperature signal.

[0071] S4. Input the feature parameters into the trained multimodal fusion model to obtain the operating status evaluation results of the ball mill, and generate early warning information when the evaluation results are abnormal.

[0072] Reference Figure 5 In the ball mill intelligent monitoring method based on multimodal acoustic signature analysis described in the embodiments of this application, the multimodal fusion model adopts an attention mechanism to weightedly fuse acoustic signature features and vibration features, and the attention weights are dynamically adjusted according to the contribution of each modal feature in historical fault data.

[0073] It also includes a model self-updating step: when an actual fault is detected in the ball mill, the multimodal data within a set time period before the fault occurs is automatically marked as samples to optimize the parameters of the multimodal fusion model.

[0074] This intelligent monitoring method for ball mills based on multimodal acoustic signature analysis employs an attention mechanism to achieve weighted fusion of acoustic signature and vibration features. Specifically, this mechanism dynamically weights the features of the two modes by constructing an attention weight matrix. The attention weights are not fixed but dynamically adjusted based on the contribution of each modal feature to fault diagnosis in historical fault data. During the training phase, the weight parameters are continuously optimized using a backpropagation algorithm, enabling the model to adaptively learn the importance of each modal feature under different operating conditions, thereby improving the accuracy of feature fusion.

[0075] To further enhance the model's adaptability and diagnostic accuracy, this method also introduces a model self-updating step. When the system detects an actual fault in the ball mill, it automatically triggers a data acquisition and labeling process: using the fault occurrence time as a baseline, it retrospectively collects and labels multimodal data such as acoustic signatures and vibration data from a time period T prior to the fault occurrence as fault samples. These newly generated samples are incorporated into the training dataset to optimize the network parameters of the multimodal fusion model. Through continuous iterative updates, the model can learn new fault modes in a timely manner, effectively reducing the false alarm and missed detection rates.

[0076] Through innovative multimodal feature fusion strategies and intelligent model self-updating mechanisms, not only is high-precision real-time assessment of the ball mill's operating status achieved, but potential fault hazards can also be identified in advance, providing reliable fault warning information for equipment maintenance personnel and significantly improving the ball mill's operational reliability and maintenance efficiency.

[0077] Example 3

[0078] Reference Figure 6 The ball mill intelligent monitoring system based on multimodal acoustic signature analysis described in this application includes:

[0079] The intelligent monitoring device as described in any one of claims 1-4 is configured to be installed in the ball mill cylinder and to collect multimodal sensing data;

[0080] An edge computing unit, connected wirelessly to an intelligent monitoring device, is configured to perform the method of any one of claims 5-7 to process and analyze multimodal sensing data;

[0081] The cloud management platform communicates and connects with the edge computing unit, and is configured to store analysis results, perform trend analysis, and generate remote early warning commands.

[0082] The human-machine interface terminal communicates with the cloud management platform and is configured to display the ball mill's operating status and early warning information to the user.

[0083] This intelligent monitoring system for ball mills based on multimodal acoustic signature analysis uses any one of the intelligent monitoring devices described in claims 1-4 as the core of the system's data sensing. It employs a modular design and can be easily installed on key parts of the ball mill cylinder. The device integrates high-precision vibration sensors, acoustic wave acquisition arrays, temperature and pressure sensors, and other multimodal sensing components, enabling real-time acquisition of multi-dimensional data such as mechanical vibration signals, internal grinding acoustic signatures, equipment surface temperature, and operating pressure during ball mill operation, ensuring the comprehensiveness and accuracy of the monitoring data.

[0084] The edge computing unit and the intelligent monitoring device establish a connection via low-power, high-bandwidth 5G or Wi-Fi 6 wireless communication protocols, constructing a localized data processing hub. This unit integrates a high-performance processor and a deep learning acceleration chip, configured to execute the intelligent analysis method of any one of claims 5-7, capable of real-time noise reduction, feature extraction, pattern recognition, and other preprocessing and deep analysis of multimodal sensor data. Through edge computing technology, rapid data processing and preliminary anomaly diagnosis are achieved, effectively reducing data transmission latency and cloud load.

[0085] The cloud-based management platform relies on distributed storage and cloud computing architecture, interacting with edge computing units via secure and encrypted communication links. The platform not only undertakes the long-term storage of analysis results but also utilizes big data analytics and machine learning algorithms to perform trend mining and correlation analysis on historical data, predicting potential failure risks of the ball mill. Simultaneously, based on preset thresholds and risk levels, the platform can automatically generate remote early warning commands, ensuring that maintenance personnel can promptly grasp the equipment's operating status.

[0086] The human-machine interface terminal serves as the interface between the user and the system, establishing a secure communication connection with the cloud management platform via a web or mobile application. The terminal features a large, visually appealing screen and an interactive interface, intuitively displaying ball mill operating parameters, health status assessment reports, fault warning information, and historical data curves. Furthermore, the terminal supports user-defined monitoring indicators and warning rules, enabling personalized operation and maintenance management and providing users with an efficient and convenient equipment management experience.

[0087] Through an innovative architecture design that combines edge computing and cloud collaboration, the system enables real-time processing, in-depth analysis, and global management of monitoring data. It not only responds quickly to equipment anomalies but also optimizes ball mill operating parameters through big data analysis, improving equipment efficiency and reliability, reducing maintenance costs, and providing strong support for the intelligent upgrading of industrial production.

[0088] Reference Figure 6In the ball mill intelligent monitoring system based on multimodal acoustic signature analysis described in this application embodiment, the intelligent monitoring device and the edge computing unit are connected through an industrial Ethernet or LoRa wireless network, and the wireless communication adopts frequency hopping spread spectrum technology.

[0089] The cloud management platform includes a fault diagnosis knowledge base, which stores multimodal feature templates corresponding to different ball mill types and fault states, used to assist in the verification of real-time monitoring results.

[0090] It also includes an adaptive adjustment module, which adjusts the feeding rate or grinding media filling rate of the ball mill through electromagnetic actuators based on the analysis results of the cloud management platform.

[0091] This intelligent monitoring system for ball mills, based on multimodal acoustic signature analysis, establishes a dual-redundant communication link between the intelligent monitoring device and the edge computing unit: Industrial Ethernet provides a high-speed, stable wired data transmission channel, supporting low-latency interaction of real-time monitoring data; the LoRa wireless network serves as a backup channel, employing frequency-hopping spread spectrum (FHSS) technology to achieve interference-resistant wireless communication, effectively avoiding signal conflicts in complex industrial environments. The communication protocol follows Modbus TCP / IP and LoRaWAN standards, ensuring data transmission compatibility and reliability.

[0092] The cloud-based management platform integrates a fault diagnosis knowledge base that uses a hierarchical data architecture to store ball mill operation and maintenance data. The system employs the Dynamic Time Warping (DTW) algorithm to perform similarity matching between real-time monitoring data and feature templates, achieving a dual verification mechanism for fault diagnosis.

[0093] As a key node in the closed-loop control of the system, the adaptive adjustment module has multi-dimensional adjustment capabilities: the system supports manual / automatic dual-mode switching, and operators can set adjustment parameters through the HMI interface and view the equipment operating status feedback in real time;

[0094] By constructing an anti-interference redundant communication network, building an intelligent fault diagnosis model, and implementing an adaptive closed-loop control strategy, a complete technical link from data acquisition and intelligent analysis to optimization and adjustment is formed, effectively improving the intelligent operation and maintenance level and production efficiency of ball mills.

[0095] Working Principle: During installation, the device is attached to the surface of the ball mill cylinder via the arc-shaped mounting base 11 and the elastic damping layer 12, and secured with bolts using the mounting ears of the fastening unit 13. During operation, the acoustic signature acquisition unit 21 picks up the cylinder vibration through the elastic diaphragm 15, transmitting it to the microphone array 24 via the acoustic resonant cavity 14 and the metal sound transmission duct 25; the vibration sensing unit 22 and temperature sensing unit 23 synchronously acquire corresponding signals. The data preprocessing module 30 performs noise reduction and synchronization processing on the multimodal signals, extracting characteristic parameters such as MFCC coefficients and peak frequencies. The multimodal fusion model uses an attention mechanism to weightedly fuse features, combining them with a fault diagnosis knowledge base to output state assessment results. In case of anomalies, the cloud platform generates an early warning and dynamically optimizes operating parameters through an adaptive adjustment module. The model continuously improves performance through self-updating of fault data, while the energy supply unit 40 and the sealed protection structure ensure stable system operation.

[0096] In the description of this invention, it should be understood that the terms "center", "front", "rear", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limiting the scope of protection of this invention.

[0097] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A ball mill intelligent monitoring device based on multimodal acoustic signature analysis, characterized in that, include: A mechanical fixing assembly (10) includes an arc-shaped mounting base (11), an elastic damping layer (12), and a fastening unit (13). The arc-shaped mounting base (11) is configured to match the curvature of the outer wall of the ball mill cylinder and is attached to the outer wall of the cylinder through the elastic damping layer (12). The fastening unit (13) is configured to fix the arc-shaped mounting base (11) to the ball mill cylinder. A multimodal sensing module (20) is integrated on the arc-shaped mounting base (11) and includes an acoustic signature acquisition unit (21), a vibration sensing unit (22) and a temperature sensing unit (23). The acoustic signature acquisition unit (21) is configured to acquire acoustic wave signals transmitted through the ball mill cylinder. The data preprocessing module (30) is located inside the mechanical fixing assembly (10) and electrically connected to the multimodal sensing module (20), and is configured to collect the acoustic wave signal transmitted through the ball mill cylinder; An energy supply unit (40) is integrated into one side of the mechanical fixing component (10) through a waterproof sealing structure, and is used to provide power to the multimodal sensing module (20) and the data preprocessing module (30).

2. The apparatus according to claim 1, characterized in that: The voiceprint acquisition unit (21) includes: An acoustic resonant cavity (14) is formed inside the arc-shaped mounting base (11); An elastic diaphragm (15) is embedded at the bottom of the acoustic resonant cavity (14) and is configured to contact the surface of the ball mill cylinder to pick up vibrational sound waves. A metal acoustic conduit (25), one end of which is connected to the acoustic resonant cavity (14); A microphone array (24) is connected to the other end of the metal acoustic conduit (25), and the microphone array (24) includes at least two microphones optimized for different frequency bands; The elastic diaphragm (15), acoustic resonant cavity (14), and metal sound transmission duct (25) together constitute an acoustic waveguide structure.

3. The apparatus according to claim 1, characterized in that: The mechanical fixing assembly (10) also includes a dustproof and rainproof cover (16), which is sealed to the arc-shaped mounting base (11) to form a sealed cavity. The multimodal sensing module (20) and the data preprocessing module (30) are housed in the sealed cavity. The dustproof and rainproof cover (16) is provided with an acoustic wave transmission window (17), which faces the acoustic path of the acoustic pattern acquisition unit (21).

4. The apparatus according to claim 1, characterized in that: The fastening unit (13) consists of mounting ears extending from both sides of the arc-shaped mounting base (11), and the mounting ears are provided with through holes for connection to the ball mill cylinder by bolts.

5. A method for intelligent monitoring of ball mills based on multimodal acoustic signature analysis, characterized in that, The method is implemented using the apparatus described in any one of claims 1-4, comprising: S1. The multimodal signals of the ball mill during operation are collected by the multimodal sensing module (20). The multimodal signals include acoustic wave signals, vibration signals and temperature signals transmitted through the acoustic waveguide structure. S2. Perform synchronization and noise reduction processing on the acquired multimodal signals; S3. Extract feature parameters from the processed signal, including the Mel frequency cepstral coefficients of the acoustic signature signal, the peak frequency of the vibration signal, and the rate of change of the temperature signal. S4. Input the feature parameters into the trained multimodal fusion model to obtain the operating status evaluation result of the ball mill, and generate early warning information when the evaluation result is abnormal.

6. The method according to claim 5, characterized in that: The multimodal fusion model employs an attention mechanism to weightedly fuse acoustic and vibration features, with the attention weights dynamically adjusted based on the contribution of each modal feature in historical fault data.

7. The method according to claim 5, characterized in that, It also includes a model self-updating step: when an actual fault is detected in the ball mill, the multimodal data within a set time period before the fault occurs is automatically marked as samples to optimize the parameters of the multimodal fusion model.

8. A ball mill intelligent monitoring system based on multimodal acoustic signature analysis, characterized in that, include: The intelligent monitoring device as described in any one of claims 1-4 is configured to be installed in the ball mill cylinder and collect multimodal sensing data; An edge computing unit, connected to the intelligent monitoring device via wireless communication, is configured to perform the method of any one of claims 5-7 to process and analyze the multimodal sensing data; The cloud management platform communicates with the edge computing unit and is configured to store analysis results, perform trend analysis, and generate remote early warning commands. The human-computer interaction terminal is connected to the cloud management platform and is configured to display the ball mill's operating status and early warning information to the user.

9. The system according to claim 8, characterized in that: The intelligent monitoring device is connected to the edge computing unit via an industrial Ethernet or LoRa wireless network, and the wireless communication adopts frequency hopping spread spectrum technology. The cloud management platform includes a fault diagnosis knowledge base, which stores multimodal feature templates corresponding to different ball mill types and fault states, used to assist in the verification of real-time monitoring results.

10. The system according to claim 8, characterized in that: It also includes an adaptive adjustment module, which adjusts the feeding rate or grinding media filling rate of the ball mill through an electromagnetic actuator based on the analysis results of the cloud management platform.

Images