A cloud-edge collaborative-based device fault voiceprint recognition monitoring system and method
By using a cloud-edge collaborative device fault voiceprint recognition and monitoring system, bandwidth is dynamically allocated and combined with fault recognition priority and data segment feature tags, the problems of latency and energy consumption in device fault data transmission are solved, and efficient, real-time fault monitoring and analysis are achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- 华颐昌能(北京)科技有限公司
- Filing Date
- 2025-07-23
- Publication Date
- 2026-07-21
AI Technical Summary
In existing technologies, the transmission mode of equipment fault data is fixed, which results in invalid data being transmitted along with valid data. This squeezes the transmission space of critical data, increases transmission delay and network congestion, affects the real-time performance of fault monitoring, and increases data processing costs and energy consumption of edge devices.
A cloud-edge collaborative device fault voiceprint recognition monitoring system is adopted. By dynamically allocating bandwidth through the cloud server and combining fault identification priority and data segment feature tags, the data transmission process is optimized to achieve fine-grained scheduling.
It reduces transmission latency, improves data transmission efficiency, ensures timely uploading and analysis of critical data, and reduces equipment fault identification delays and energy consumption of edge devices.
Smart Images
Figure CN120602518B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of equipment fault monitoring technology, and in particular to a cloud-edge collaborative equipment fault voiceprint recognition monitoring system and method. Background Technology
[0002] In industrial equipment fault monitoring scenarios, existing technologies often employ relatively fixed transmission patterns for equipment fault data. Typically, edge devices collect data from industrial equipment during operation, and regardless of whether the equipment is actually generating valid fault diagnosis data, the collected data is uniformly uploaded to the cloud. For example, regarding equipment audio signals, even if the equipment is not operating and has no effective fault identification value, the edge device will still transmit this redundant audio data to the cloud. Subsequently, the cloud performs filtering and analysis of a large amount of invalid data, hoping to select data that can be used for fault diagnosis during the equipment's operating period.
[0003] However, existing technologies have significant drawbacks. On the one hand, invalid data is transmitted along with valid data during device operation, squeezing the transmission space for critical data and leading to increased overall transmission latency. This is especially problematic in scenarios where multiple devices are monitoring simultaneously, easily causing network congestion and affecting the real-time performance of fault monitoring. On the other hand, the cloud requires significant computing power to process redundant data, increasing data processing costs and fault identification latency. Furthermore, the continuous data collection and uploading by edge devices puts high energy consumption and storage pressure on them, reducing the stability and battery life of device operation (especially for portable edge devices). This makes it difficult to efficiently adapt to the actual needs of dynamic and accurate transmission of device fault data. Summary of the Invention
[0004] This application provides a device fault voiceprint recognition monitoring system and method based on cloud-edge collaboration, which can reduce transmission latency and further improve data transmission efficiency.
[0005] To achieve the above objectives, this application adopts the following technical solution: In a first aspect, this application provides a cloud-edge collaborative device fault voiceprint recognition and monitoring system. The system includes a cloud server and multiple edge devices. Data transmission between the multiple edge devices and the cloud server shares the same bandwidth. The cloud server is used to receive data transmission requests sent by target edge devices, the data transmission requests including the amount of voiceprint data, and the target edge device is one of the multiple edge devices. The cloud server determines a first bandwidth required to transmit the voiceprint data based on latency and data volume. The cloud server determines whether the first bandwidth is less than the remaining available bandwidth and obtains a first determination result. If the first determination result indicates that the first bandwidth is less than or equal to the remaining available bandwidth, the cloud server allocates the first bandwidth to the target edge device. The cloud server receives the voiceprint data through the first bandwidth and then performs fault voiceprint recognition and monitoring on industrial equipment based on the voiceprint data.
[0006] In some possible implementations, the cloud server determines whether the first bandwidth is less than the remaining available bandwidth and obtains a first determination result; if the first determination result indicates that the first bandwidth is greater than the remaining available bandwidth, then a second bandwidth is allocated to the target edge device, wherein the second bandwidth is less than or equal to the remaining available bandwidth.
[0007] In some possible implementations, the cloud server is also used to dynamically schedule based on the fault identification priority of voiceprint data when allocating the second bandwidth to the target edge device; The dynamic scheduling of fault identification priorities based on voiceprint data includes: The cloud server pre-sets priority levels for industrial equipment fault voiceprint recognition scenarios; When allocating the second bandwidth, the fault identification priority corresponding to the voiceprint data transmitted by the target edge device is extracted. If it is the first priority, the second bandwidth is allocated according to the first allocation method; if it is the second priority, the second bandwidth is allocated according to the second allocation method; if it is the third priority, the second bandwidth is allocated according to the third allocation method.
[0008] In some possible implementations, the target edge device divides the voiceprint data obtained from the industrial equipment into multiple data segments according to time series or frequency characteristics, and obtains a feature label for each data segment. Before transmission, the target edge device reports the feature labels of multiple data segments to the cloud server. The cloud server dynamically allocates sub-bandwidth to data segments with different feature labels based on bandwidth usage. During transmission, the utilization rate of each data segment's sub-bandwidth is evaluated in real time. If the utilization rate of the first sub-bandwidth is continuously lower than the first utilization rate threshold, the first sub-bandwidth is reclaimed and bandwidth is reallocated.
[0009] In some possible implementations, the cloud server performs fault voiceprint recognition monitoring on industrial equipment based on the voiceprint data to obtain the voiceprint data recognition result; based on the voiceprint data recognition result, the cloud server determines whether the recognition accuracy loss due to bandwidth allocation exceeds the loss threshold; if the recognition accuracy loss exceeds the loss threshold, a reverse bandwidth adjustment process is triggered to reserve bandwidth by a preset multiple for subsequent voiceprint data transmission of the same type; simultaneously, based on the recognition results of historical bandwidth allocation, a dynamic bandwidth prediction model is trained, and bandwidth is pre-allocated for the voiceprint data to be transmitted according to the training results.
[0010] In some possible implementations, the cloud server is also used to monitor the transmission progress of multiple edge devices in real time. When a first-priority data transmission request occurs, the third-priority data transmission can be interrupted and bandwidth can be reallocated.
[0011] In some possible implementations, after acquiring the initial voiceprint data of the industrial equipment, the target edge device performs noise filtering on the voiceprint data based on wavelet transform to obtain voiceprint data; wherein, the filtering intensity of the noise filtering is dynamically adjusted according to the operating status of the industrial equipment; the target edge device divides the voiceprint data into multiple data segments according to time series or frequency characteristics, and sets feature overlap areas for adjacent data segments; the target edge device transmits multiple data segments to the cloud server.
[0012] In some possible implementations, when the cloud server manages multiple edge devices, it summarizes the remaining available bandwidth and device failure risk level of the multiple edge devices in real time; if the first priority voiceprint data transmission of the target edge device requires additional bandwidth, and multiple edge devices other than the target edge device have idle bandwidth, then cross-regional bandwidth scheduling is triggered to temporarily allocate idle bandwidth to the target edge device; after the transmission is completed, the cross-regional allocated bandwidth is automatically recovered and the original regional bandwidth allocation benchmark is restored.
[0013] Secondly, this application provides a device fault voiceprint recognition and monitoring method based on cloud-edge collaboration, the method being applied to a cloud server; the method includes: A data transmission request is received from a target edge device, the data transmission request including the amount of voiceprint data, and the target edge device is one of the plurality of edge devices; The first bandwidth required to transmit the voiceprint data is determined based on latency and data volume; Determine whether the first bandwidth is less than the remaining available bandwidth to obtain the first determination result; If the first determination result indicates that the first bandwidth is less than or equal to the remaining available bandwidth, then the first bandwidth is allocated to the target edge device; The voiceprint data is received through the first bandwidth, and then the industrial equipment is monitored for fault voiceprint identification based on the voiceprint data.
[0014] Thirdly, this application provides a computing device, including a memory and a processor; The memory stores one or more computer programs, the one or more computer programs including instructions; when the instructions are executed by the processor, the computing device performs the method as described in any one of the first aspects.
[0015] Fourthly, this application provides a computer-readable storage medium for storing a computer program for performing the method as described in any one of the first aspects.
[0016] Fifthly, this application provides a computer program product comprising one or more computer instructions, wherein when the computer instructions are executed by a computer, the computer performs the method as described in any one of the first aspects.
[0017] As can be seen from the above technical solution, this application has at least the following beneficial effects: In this application, a cloud server receives a data transmission request from a target edge device, the data transmission request including the amount of voiceprint data, and the target edge device is one of multiple edge devices. Based on latency and data volume, a first bandwidth required for transmitting the voiceprint data is determined. It is then determined whether the first bandwidth is less than the remaining available bandwidth, resulting in a first judgment result. If the first judgment result indicates that the first bandwidth is less than or equal to the remaining available bandwidth, the first bandwidth is allocated to the target edge device. Voiceprint data is received through the first bandwidth, and then fault voiceprint identification monitoring of industrial equipment is performed based on the voiceprint data. In the field of industrial equipment fault monitoring, existing technologies often employ relatively fixed transmission modes, where edge devices upload the collected data to the cloud regardless of whether the industrial equipment is in a state of generating valid fault diagnosis data. Therefore, the cloud-edge collaborative equipment fault voiceprint identification monitoring system provided in this application, by having the cloud server determine the required bandwidth based on latency and voiceprint data volume, dynamically allocate bandwidth, and perform fine-grained scheduling combined with fault identification priority and data segment feature tags, achieves the goal of reducing transmission latency and improving data transmission efficiency.
[0018] It should be understood that the descriptions of technical features, technical solutions, beneficial effects, or similar language in this application do not imply that all features and advantages can be achieved in any single embodiment. Rather, it is understood that the description of a feature or beneficial effect means that a specific technical feature, technical solution, or beneficial effect is included in at least one embodiment. Therefore, the descriptions of technical features, technical solutions, or beneficial effects in this specification do not necessarily refer to the same embodiment. Furthermore, the technical features, technical solutions, and beneficial effects described in this embodiment can be combined in any suitable manner. Those skilled in the art will understand that embodiments can be implemented without one or more specific technical features, technical solutions, or beneficial effects of a particular embodiment. In other embodiments, additional technical features and beneficial effects may be identified in specific embodiments that do not embody all embodiments. Attached Figure Description
[0019] Figure 1 A schematic diagram illustrating an application scenario provided in an embodiment of this application; Figure 2 A flowchart illustrating a device fault acoustic signature recognition and monitoring method based on cloud-edge collaboration provided in this application embodiment; Figure 3 This is a schematic diagram of a computing device provided in an embodiment of this application. Detailed Implementation
[0020] The terms "first," "second," and "third," etc., used in this application specification and accompanying drawings are used to distinguish different objects, not to limit a specific order.
[0021] In the embodiments of this application, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design that is described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design. Specifically, the use of the terms "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.
[0022] Currently, existing technologies using statically allocated bandwidth transmit invalid data along with valid data during device operation. This not only squeezes the transmission space for critical data, but also, due to the fixed bandwidth allocation lacking dynamic adjustment, fails to prioritize critical data, leading to increased overall transmission latency. This can easily cause network congestion when multiple devices are monitoring simultaneously, affecting the real-time performance of fault monitoring. Furthermore, it forces the cloud to expend significant computing power to process redundant data, increasing data processing costs and fault identification latency. Simultaneously, the continuous data collection and uploading by edge devices places high demands on their own energy consumption and storage, reducing operational stability and battery life, and failing to efficiently meet the actual needs for dynamic and accurate transmission of device fault data.
[0023] In view of this, embodiments of this application provide a cloud-edge collaborative device fault acoustic signature recognition and monitoring system. In this system, a cloud server receives a data transmission request from a target edge device, the data transmission request including the amount of acoustic signature data, and the target edge device is one of multiple edge devices. A first bandwidth required for transmitting the acoustic signature data is determined based on latency and data volume. A first judgment result is obtained by determining whether the first bandwidth is less than the remaining available bandwidth. If the first judgment result indicates that the first bandwidth is less than or equal to the remaining available bandwidth, the first bandwidth is allocated to the target edge device. The acoustic signature data is received through the first bandwidth, and then fault acoustic signature recognition and monitoring of industrial equipment is performed based on the acoustic signature data. It can be seen that the cloud-edge collaborative device fault acoustic signature recognition and monitoring system provided in this application, by determining the required bandwidth based on latency and acoustic signature data volume through a cloud server, dynamically allocating bandwidth, and combining fault identification priority, data segment feature tags, etc., for fine-grained scheduling, achieves the purpose of reducing transmission latency and improving data transmission efficiency.
[0024] To make the technical solution of this application clearer and easier to understand, the application scenarios of the technical solution of this application are described below with reference to the accompanying drawings. Figure 1 As shown in the figure, this figure is a schematic diagram of an application scenario provided by an embodiment of this application.
[0025] In this application scenario, there are three edge devices, three industrial devices, and one cloud server 4. The edge devices are nodes located close to the industrial devices and are responsible for the initial data collection and processing. The cloud server is responsible for data aggregation, storage, and further analysis. The three edge devices include edge device 1, edge device 2, and target edge device 3, and the three industrial devices include industrial device 11, industrial device 12, and industrial device 13.
[0026] Edge device 1 is used to receive voiceprint data from industrial device 11, edge device 2 is used to receive voiceprint data from industrial device 12, and target edge device 3 is used to receive voiceprint data from industrial device 13. The three edge devices transmit the received voiceprint data to cloud server 4. Voiceprint data is like the sound fingerprint of industrial equipment, which can reflect the operating status of the equipment. For example, the voiceprint characteristics are different in different states such as normal operation and early stage of failure.
[0027] After the three edge devices collect the voiceprint data of their respective industrial equipment, they will transmit this voiceprint data to the cloud server 4. In this way, the cloud server can centrally manage and analyze the voiceprint data of multiple industrial equipment, such as judging whether the equipment is faulty by comparing voiceprint anomalies and predicting maintenance needs, thus building an industrial Internet of Things data link from industrial equipment data collection to edge processing and then to cloud aggregation and analysis.
[0028] This application provides a cloud-edge collaborative device fault voiceprint recognition and monitoring system. The system determines the required bandwidth based on latency and voiceprint data volume through a cloud server, dynamically allocates bandwidth, and performs fine-grained scheduling by combining fault recognition priority and data segment feature tags, thereby achieving the goal of reducing transmission latency and improving data transmission efficiency.
[0029] To make the technical solution of this application clearer and easier to understand, the following describes a cloud-edge collaborative device fault voiceprint recognition and monitoring system provided in the embodiments of this application, in conjunction with the above application scenarios. In the embodiments of this application, the system includes: The equipment fault voiceprint recognition monitoring system includes a cloud server and multiple edge devices; the data transmission between the multiple edge devices and the cloud server shares the same bandwidth; the cloud server has powerful data processing, storage and computing capabilities, and is used to manage and analyze the data uploaded by the edge devices; the edge devices are deployed close to the industrial equipment and are responsible for collecting and initially processing voiceprint data and interacting with the cloud server.
[0030] The cloud server receives data transmission requests from target edge devices. These requests include the amount of voiceprint data. The target edge device is one of multiple edge devices. Specifically, the cloud server receives the data transmission request from the target edge device, which is the one among multiple edge devices that is currently receiving voiceprint data from the industrial equipment and initiating the data transmission request. The data transmission request includes the amount of voiceprint data. For example, if the target edge device detects that the industrial equipment has started operating and needs to transmit its operating sound data to the cloud server, it will send a data transmission request, which includes transmitting 100MB of voiceprint data.
[0031] The cloud server determines the initial bandwidth required for transmitting voiceprint data based on latency and data volume. Latency refers to the time delay for voiceprint data to travel from the target edge device to the cloud server, while the initial bandwidth refers to the ideal bandwidth for the target edge device to transmit the voiceprint data smoothly without congestion. The cloud server calculates the initial bandwidth required for transmitting the voiceprint data by combining the amount of data to be transmitted and the allowable latency. For example, to transmit 100MB of voiceprint data, if the transmission is to be completed within 1 second, considering network characteristics, a bandwidth of 100Mbps might be required, which is the initial bandwidth. If a transmission time of 3 seconds is allowed, the initial bandwidth can be less than 100Mbps.
[0032] The cloud server determines whether the first bandwidth is less than the remaining available bandwidth and obtains the first determination result. If the first determination result indicates that the first bandwidth is less than or equal to the remaining available bandwidth, then the first bandwidth is allocated to the target edge device; if the first determination result indicates that the first bandwidth is greater than the remaining available bandwidth, then the second bandwidth is allocated to the target edge device, and the second bandwidth is less than or equal to the remaining available bandwidth. The specific steps are as follows: Remaining available bandwidth refers to the total bandwidth minus the bandwidth already allocated to other edge devices and the bandwidth required by the cloud server to transmit requests to other edge devices (i.e., the fixed reserved bandwidth), leaving the portion that can be allocated. Second bandwidth refers to the bandwidth that can actually be allocated, but not exceeding the remaining available bandwidth, when the first bandwidth exceeds the remaining available bandwidth. Comparing the first bandwidth and the remaining available bandwidth, if the first bandwidth is less than or equal to the remaining available bandwidth, the first bandwidth is directly allocated to the target edge device; if the first bandwidth is greater than the remaining available bandwidth, the second bandwidth, less than or equal to the remaining available bandwidth, is allocated to ensure that bandwidth allocation is feasible within existing network resources. In some embodiments, the total bandwidth is 100Mbps, other edge devices and the fixed reserved bandwidth have used 60Mbps, leaving 40Mbps. If the first bandwidth is calculated to be 30Mbps, 30Mbps is directly allocated; if the first bandwidth is 50Mbps, only a second bandwidth less than or equal to 40Mbps can be allocated, for example, 40Mbps.
[0033] When the cloud server is specifically used to allocate secondary bandwidth to target edge devices, it dynamically schedules bandwidth allocation based on the fault identification priority of voiceprint data. This includes: the cloud server pre-setting priority levels for industrial equipment fault voiceprint identification scenarios; wherein, the priority levels are the urgency and importance levels that the cloud server pre-sets for industrial equipment fault voiceprint identification scenarios, determining which edge device the bandwidth allocation is tilted towards. The cloud server pre-sets priority levels for scenarios that use voiceprint to identify equipment faults. For example, abnormal noises during equipment startup (which may be a precursor to a serious fault) are set as the first priority, while minor noises during normal equipment operation are set as the third priority.
[0034] When allocating the second bandwidth, the fault identification priority corresponding to the voiceprint data transmitted by the target edge device is extracted. If it is the first priority, the second bandwidth is allocated according to the first allocation method; if it is the second priority, the second bandwidth is allocated according to the second allocation method; and if it is the third priority, the second bandwidth is allocated according to the third allocation method. The first allocation method is a strategy that allocates as much bandwidth as possible to ensure fast and stable transmission, such as prioritizing the allocation of most of the remaining available bandwidth. The second allocation method allocates less bandwidth than the first priority to ensure basic transmission. The third allocation method allocates less bandwidth, just enough to meet basic transmission needs. For example, if the remaining available bandwidth is 40Mbps, the first priority data might be allocated 20Mbps; the second priority 15Mbps; and the third priority 5Mbps, allowing critical fault data to be transmitted efficiently and with priority.
[0035] Specifically, if industrial equipment 11 is a stamping press in an automobile factory, the abnormal metallic friction sound when it is first started may be due to mold misalignment or loose parts, which could seriously cause the production line to stop. This is set as the first priority, requiring the cloud server to allocate high bandwidth to edge device 1 (the edge device that connects to the stamping press) when transmitting data to ensure that the data is returned for analysis within seconds. The slight airflow sound during the equipment's idling and preheating phase (normal startup process) is set as the third priority, and bandwidth allocation can be temporarily delayed.
[0036] Specifically, if industrial equipment 11 is an assembly line robot, edge device 1 collects friction sound patterns indicating joint jamming in the robot (this sound pattern data is set as the first priority because this fault may cause material jamming on the production line, affecting production), while edge device 12 (connected to the stamping machine) collects slight mechanical noises from normal operation of the equipment (this sound pattern data is the third priority). At this time, the cloud server will prioritize allocating the second bandwidth to edge device 1 to ensure that the fault sound pattern data of the assembly line robot is uploaded and analyzed first, while the normal data of the stamping machine is temporarily suspended from transmission.
[0037] Specifically, if industrial equipment 13 is a chemical reactor, when the target edge device 3 collects acoustic data, if an abnormal noise is detected during the startup phase (this acoustic data has the highest priority and is a precursor to a serious fault), the cloud server immediately allocates a second bandwidth to the edge device 3. This second bandwidth is the maximum available bandwidth that the cloud server can allocate, allowing fault warning data to be uploaded quickly. The bandwidth allocation during the stable operation phase is reduced to the minimum, for example, if there is only a slight cooling fan noise, this acoustic data has the fourth priority and only basic data transmission is maintained. When the reactor begins to unload or the equipment load changes suddenly, if abnormal pipe vibration acoustics occur (this acoustic data has the highest priority and may indicate material blockage), the cloud server dynamically increases the bandwidth to ensure that abnormal data is analyzed first.
[0038] The purpose of setting priority levels is to classify the urgency of different fault hazards and equipment operation phases for acoustic fingerprint analysis needs. These hierarchical levels directly determine which edge devices prioritize bandwidth allocation to when multiple devices transmit voiceprint data simultaneously, enabling faster data transmission and analysis of critical data. For factory operations and maintenance, prioritizing the transmission and analysis of high-risk fault voiceprints allows for early detection of equipment malfunctions. For example, for first-priority events like abnormal noises during press startup, data transmission and fault diagnosis can be completed within 10 seconds. Compared to traditional methods of waiting for faults to occur before addressing them, this reduces unplanned downtime losses by more than 80%. Regarding resource utilization, dynamic bandwidth scheduling avoids the waste of fixed bandwidth allocation regardless of data criticality, allowing limited network resources to be precisely targeted at the most critical fault identification scenarios. This is particularly suitable for smart factories with multiple production lines and concurrent devices. For continuously operating equipment experiencing load changes, such as the blast furnace blower in a steel plant, a sudden high-frequency vibration voiceprint during full-load operation may indicate bearing overload or impeller deformation. Failure to intervene promptly could lead to shutdown or explosion risks. Setting this voiceprint data as the first priority ensures data transmission for the target edge device 3 (connected to the blower). Compare the voiceprints of the device during a smooth transition from low load to full load, set this voiceprint data as the third priority, and allow bandwidth allocation to be postponed.
[0039] In some embodiments, the target edge device divides the voiceprint data acquired from the industrial equipment into multiple data segments according to time series or frequency characteristics, and obtains a feature label for each data segment. Here, time series refers to dividing the data according to the chronological order of voiceprint data acquisition, such as dividing voiceprint data within one minute into 10-second segments; frequency characteristics refer to dividing the data according to different frequency components, such as separating high-frequency and low-frequency voiceprint data into segments; feature labels are identifiers affixed to each data segment, marking the key characteristics of that segment for easy cloud identification. The target edge device divides the acquired industrial equipment voiceprint data into multiple segments either by time or by frequency components, and then assigns a feature label to each segment, such as [Time: 0-10 seconds, Frequency: High frequency, Equipment: Machine Tool A].
[0040] Before transmitting voiceprint data, the target edge device reports feature tags for multiple data segments to the cloud server. The cloud server dynamically allocates sub-bandwidth to data segments with different feature tags based on bandwidth usage. Sub-bandwidth is the total bandwidth allocated to the edge device further subdivided into smaller channels for each data segment, allowing different types of voiceprint data segments to be transmitted in parallel and in an orderly manner. During transmission, the utilization rate of each data segment's sub-bandwidth is evaluated in real time. If the utilization rate of the first sub-bandwidth consistently falls below a first utilization threshold, the first sub-bandwidth is reclaimed and bandwidth is reallocated. The first utilization threshold is used to determine whether sub-bandwidth is idle or used inefficiently; for example, a sub-bandwidth utilization rate below 30% is considered inefficient.
[0041] On the one hand, segmenting by time series allows for the capture of equipment operating status at different times. For example, acoustic fingerprint data from stages such as 0-10 seconds for industrial equipment startup, 10-20 seconds for stable operation, and 20-30 seconds for load changes can be used for precise analysis of health status at each stage. Time series segmentation also allows for retrospective analysis of equipment operating status, enabling maintenance personnel to review when anomalies occurred and the changes in status before and after the anomalies, thus assisting in optimizing maintenance strategies, such as adjusting equipment startup procedures and maintenance cycles. On the other hand, segmenting by frequency characteristics is also beneficial. High-frequency acoustic fingerprint data is often associated with wear and tear on industrial equipment components (such as bearing noise), while low-frequency acoustic fingerprint data corresponds to structural vibration. Separate analysis allows for rapid location of fault sources, improving diagnostic accuracy. Dividing continuous, massive amounts of acoustic fingerprint data into smaller segments also reduces the amount of data processed in the cloud, decreases computing resource consumption, increases data processing speed, and makes fault identification and status monitoring more timely.
[0042] Feature tags annotate key characteristics, enabling cloud servers to quickly identify abnormal data segments. For example, if a feature tag contains "high frequency, device: machine tool A, time: startup phase," combined with a historical fault database, it can provide an early warning that high-frequency abnormal noise during machine tool A startup may indicate a bearing failure, allowing for proactive maintenance and preventing downtime losses. Standardized feature tags also allow cloud servers to quickly understand the meaning of data uploaded by edge devices without complex decoding, improving cloud-edge collaboration efficiency and facilitating smoother unified management of data from multiple edge devices and industrial equipment in the cloud. Standardized data segmentation and tagging provide high-quality data for building equipment acoustic signature models and training AI fault identification algorithms, driving industrial operations towards predictive and intelligent development. For instance, based on a large amount of tagged data, models can be trained to automatically classify and predict faults.
[0043] The cloud server performs fault voiceprint recognition monitoring on industrial equipment based on voiceprint data, and obtains the recognition results of the voiceprint data. Based on the recognition results of the voiceprint data, the cloud server determines whether the recognition accuracy loss caused by bandwidth allocation exceeds the loss threshold. The recognition accuracy loss refers to the degree to which the accuracy and precision of fault recognition decrease due to insufficient bandwidth, incomplete data transmission, or loss. The loss threshold is the maximum allowable decrease in recognition accuracy; if it is exceeded, the bandwidth must be adjusted.
[0044] If the recognition accuracy loss exceeds the loss threshold, a reverse bandwidth adjustment process is triggered, prioritizing the reservation of bandwidth by a preset multiple for subsequent voiceprint data of the same type. Simultaneously, based on the recognition results of historical bandwidth allocation, a dynamic bandwidth prediction model is trained, and bandwidth is pre-allocated for the voiceprint data to be transmitted according to the training results. The preset multiple is a proportion of bandwidth reserved in advance to ensure the transmission of subsequent critical data; for example, reserving twice the current required bandwidth for subsequent data of the same type. The dynamic bandwidth prediction model learns patterns by analyzing historical bandwidth allocation and recognition results, and predicts the amount of bandwidth required for future transmissions.
[0045] In some embodiments, the cloud server is also used to monitor the transmission progress of multiple edge devices in real time. When a first-priority data transmission request occurs, the third-priority data transmission can be interrupted and bandwidth reallocated. By monitoring the transmission progress of edge devices in real time, interrupting the third-priority voiceprint data transmission and reallocating bandwidth when a first-priority voiceprint data transmission request occurs ensures that critical fault warning data is uploaded to the cloud in a prioritized and timely manner. This allows the cloud server to obtain high-value data as quickly as possible, buying time for rapid fault diagnosis and emergency handling, reducing the risk of equipment failure escalation caused by delays in critical data transmission, and minimizing production line downtime losses. Furthermore, the cloud server no longer allocates bandwidth in a fixed manner but dynamically schedules bandwidth based on the priority of voiceprint data, avoiding long-term occupation of bandwidth resources by low-value data, improving the overall network resource utilization efficiency in industrial equipment status monitoring scenarios, and allowing limited bandwidth to accurately serve high-priority, high-value data transmission needs.
[0046] In some embodiments, after collecting initial acoustic signature data from industrial equipment, the target edge device performs noise filtering on the acoustic signature data based on wavelet transform to obtain acoustic signature data. The noise filtering intensity is dynamically adjusted according to the operating status of the industrial equipment. The target edge device divides the acoustic signature data into multiple data segments according to time series or frequency characteristics, and sets feature overlap areas between adjacent data segments. The target edge device transmits multiple data segments to the cloud server. Setting feature overlap areas aims to avoid the loss of edge features of data segments, achieving both structured data management and ensuring feature integrity. This facilitates in-depth analysis of equipment status from time series and frequency dimensions in the cloud, providing more reliable data support for fault diagnosis and predictive maintenance.
[0047] The specific steps are as follows: After the edge device collects the initial voiceprint data of the industrial equipment, it uses wavelet transform to filter out noise, obtaining clean voiceprint data. Furthermore, the filtering intensity is not fixed; it depends on the current state of the industrial equipment. For example, when the equipment is running, the noise is high, so the filtering intensity is increased; when it is running stably, the noise is low, so the filtering intensity is decreased. Wavelet transform is a signal processing method that can separate useful signals from noise in the voiceprint, retaining the useful signals and filtering out the noise. The filtering intensity refers to the strength of noise filtering, such as how much low-frequency / high-frequency noise is filtered out, which is adjusted according to the equipment's operating status. The feature overlap region refers to the overlapping part of two adjacent data segments, avoiding overly rigid segmentation of the voiceprint data, which could lead to the loss of key features during segmentation.
[0048] After receiving voiceprint data through the allocated first bandwidth (or adjusted bandwidth), the cloud server uses this data to perform fault voiceprint recognition monitoring on industrial equipment, determine whether the equipment has a fault and the type of fault, and complete the complete closed loop from data transmission to fault diagnosis.
[0049] The cloud server receives voiceprint data through the first bandwidth, and then uses the voiceprint data to perform fault voiceprint identification and monitoring of industrial equipment.
[0050] Based on the above, the cloud server receives a data transmission request from a target edge device, the data transmission request including the amount of voiceprint data, and the target edge device is one of multiple edge devices; it determines the first bandwidth required to transmit the voiceprint data based on latency and data volume; it determines whether the first bandwidth is less than the remaining available bandwidth, obtaining a first judgment result; if the first judgment result indicates that the first bandwidth is less than or equal to the remaining available bandwidth, the first bandwidth is allocated to the target edge device; the voiceprint data is received through the first bandwidth, and then fault voiceprint identification monitoring of industrial equipment is performed based on the voiceprint data. The cloud-edge collaborative equipment fault voiceprint identification monitoring system provided in this application determines the required bandwidth based on latency and voiceprint data volume through the cloud server, dynamically allocates bandwidth, and performs fine-grained scheduling by combining fault identification priority, data segment feature tags, etc., thereby achieving the purpose of reducing transmission latency and improving data transmission efficiency.
[0051] This application also provides a method for monitoring device fault voiceprint recognition based on cloud-edge collaboration, such as... Figure 2 As shown, this figure is a flowchart of a device fault voiceprint recognition and monitoring method based on cloud-edge collaboration provided in an embodiment of this application. The method is applied to a cloud server and includes: S201. A data transmission request was received from the target edge device.
[0052] The data transmission request includes the amount of voiceprint data, and the target edge device is one of multiple edge devices; S202. Determine the first bandwidth required for transmitting voiceprint data based on latency and data volume.
[0053] S203. Determine whether the first bandwidth is less than the remaining available bandwidth, and obtain the first judgment result.
[0054] If the first judgment result indicates that the first bandwidth is less than or equal to the remaining available bandwidth, then execute S204; If the first judgment result indicates that the first bandwidth is greater than the remaining available bandwidth, then execute S206.
[0055] S204. Allocate the first bandwidth to the target edge device.
[0056] S205. Receive voiceprint data through the first bandwidth, and then perform fault voiceprint identification and monitoring on industrial equipment based on the voiceprint data.
[0057] S206. Allocate a second bandwidth to the target edge device, where the second bandwidth is less than or equal to the remaining available bandwidth.
[0058] After executing S206, the second bandwidth is determined as the first bandwidth, and then S204 is executed.
[0059] In some possible implementations, the method further includes: when the cloud server allocates the second bandwidth to the target edge device, it performs dynamic scheduling based on the fault identification priority of the voiceprint data; The dynamic scheduling of fault identification priorities based on voiceprint data includes: The cloud server pre-sets priority levels for industrial equipment fault voiceprint recognition scenarios; When allocating the second bandwidth, the fault identification priority corresponding to the voiceprint data transmitted by the target edge device is extracted. If it is the first priority, the second bandwidth is allocated according to the first allocation method; if it is the second priority, the second bandwidth is allocated according to the second allocation method; if it is the third priority, the second bandwidth is allocated according to the third allocation method.
[0060] In some possible implementations, the method is also applied to a target edge device, the method comprising: the target edge device dividing voiceprint data obtained from industrial equipment into multiple data segments according to time series or frequency characteristics, and obtaining a feature label for each data segment; the target edge device reporting the feature labels of multiple data segments to the cloud server before transmission; the cloud server dynamically allocating sub-bandwidth to data segments with different feature labels based on bandwidth usage; and evaluating the utilization rate of each data segment's sub-bandwidth in real time during transmission, if the utilization rate of the first sub-bandwidth is continuously lower than a first utilization rate threshold, then reclaiming the first sub-bandwidth and reallocating bandwidth.
[0061] In some possible implementations, the method further includes: performing fault voiceprint recognition monitoring on industrial equipment based on the voiceprint data to obtain the recognition result of the voiceprint data; the cloud server, based on the recognition result of the voiceprint data, determining whether the recognition accuracy loss due to bandwidth allocation exceeds a loss threshold; if the recognition accuracy loss exceeds the loss threshold, then triggering a reverse bandwidth adjustment process, reserving bandwidth by a preset multiple for subsequent voiceprint data transmission of the same type; simultaneously, based on the recognition results of historical bandwidth allocation, training a dynamic bandwidth prediction model, and pre-allocating bandwidth for the voiceprint data to be transmitted according to the training results.
[0062] In some possible implementations, the method further includes: real-time monitoring of the transmission progress of multiple edge devices, and when a first-priority data transmission request occurs, interrupting the third-priority data transmission and reallocating bandwidth.
[0063] In some possible implementations, the method is also applied to a target edge device, the method comprising: after the target edge device collects initial voiceprint data from an industrial device, performing noise filtering on the voiceprint data based on wavelet transform to obtain voiceprint data; wherein the filtering intensity of the noise filtering is dynamically adjusted according to the operating status of the industrial device; the target edge device divides the voiceprint data into multiple data segments according to time series or frequency characteristics, and sets feature overlap areas for adjacent data segments; the target edge device transmits the multiple data segments to the cloud server.
[0064] In some possible implementations, the method further includes: when the cloud server manages multiple edge devices, it summarizes the remaining available bandwidth and device failure risk level of the multiple edge devices in real time; if the first priority voiceprint data transmission of the target edge device requires additional bandwidth, and multiple edge devices other than the target edge device have idle bandwidth, it triggers cross-regional bandwidth scheduling to temporarily allocate the idle bandwidth to the target edge device; after the transmission is completed, it automatically reclaims the cross-regional allocated bandwidth and restores the original regional bandwidth allocation benchmark.
[0065] This application also provides a computing device. For example... Figure 3 As shown in the figure, this is a schematic diagram of a computing device provided in an embodiment of this application. The computing device 400 includes a bus 401, a processor 402, a communication interface 403, and a memory 404. The processor 402, the memory 404, and the communication interface 403 communicate with each other via the bus 401.
[0066] Bus 401 can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of representation, Figure 3 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.
[0067] Processor 402 can be any one or more of the following processors: central processing unit (CPU), graphics processing unit (GPU), microprocessor (MP), or digital signal processor (DSP).
[0068] The communication interface 403 is used for communication with external devices. For example, if the computing device is a first switch, the communication interface 403 can be used for communication between the first switch and a first user terminal, or for communication between the first switch and a second switch.
[0069] Memory 404 may include volatile memory, such as random access memory (RAM). Memory 404 may also include non-volatile memory, such as read-only memory (ROM), flash memory, hard disk drive (HDD), or solid state drive (SSD).
[0070] The memory 404 stores executable code, and the processor 402 executes the executable code to perform the aforementioned cloud-edge collaborative device fault voiceprint recognition and monitoring method.
[0071] This application also provides a computer-readable storage medium. The computer-readable storage medium can be any available medium that a computing device can store, or a data storage device such as a data center containing one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state drive). The computer-readable storage medium includes instructions that instruct the computing device to execute the aforementioned cloud-edge collaborative device fault voiceprint recognition and monitoring method.
[0072] This application also provides a computer program product comprising one or more computer instructions. When the computer instructions are loaded and executed on a computing device, all or part of the processes or functions described in this application are generated.
[0073] The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, or data center to another website, computer, or data center via wired (e.g., coaxial cable, fiber optic) or wireless (e.g., infrared, wireless, microwave, etc.) means.
[0074] When the computer program product is executed by a computer, the computer executes any of the aforementioned methods of the cloud-edge collaborative device fault voiceprint recognition and monitoring method. The computer program product can be a software installation package; when any of the aforementioned cloud-edge collaborative device fault voiceprint recognition and monitoring methods needs to be used, the computer program product can be downloaded and executed on the computer.
[0075] The descriptions of the processes or structures corresponding to the above figures each have their own emphasis. For parts of a process or structure that are not described in detail, please refer to the relevant descriptions of other processes or structures.
[0076] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions within the technical scope disclosed in this application should be covered within the scope of protection of this application.
Claims
1. A device fault voiceprint recognition and monitoring system based on cloud-edge collaboration, characterized in that, The system includes a cloud server and multiple edge devices; data transmission between the multiple edge devices and the cloud server shares the same bandwidth. The target edge device is used to collect initial acoustic print data of industrial equipment, perform noise filtering on the initial acoustic print data to obtain acoustic print data; increase the filtering intensity when the industrial equipment is turned on; and decrease the filtering intensity when the industrial equipment is running stably. The cloud server is configured to receive a data transmission request sent by a target edge device, the data transmission request including the amount of voiceprint data, and the target edge device being one of the plurality of edge devices; The cloud server is used to determine the first bandwidth required to transmit the voiceprint data based on latency and data volume. The cloud server is used to determine whether the first bandwidth is less than the remaining available bandwidth, and to obtain a first determination result; If the first determination result indicates that the first bandwidth is less than or equal to the remaining available bandwidth, the first bandwidth is allocated to the target edge device. The cloud server is used to receive the voiceprint data through the first bandwidth, and then perform fault voiceprint identification and monitoring on industrial equipment based on the voiceprint data to obtain the identification result of the voiceprint data. Based on the recognition results of the voiceprint data, determine whether the recognition accuracy loss due to bandwidth allocation exceeds the loss threshold; If the recognition loss accuracy exceeds the loss threshold, a preset multiple of bandwidth will be reserved for the transmission of the same type of voiceprint data; The cloud server is used to train a dynamic bandwidth prediction model based on the identification results of historical bandwidth allocation, and to pre-allocate bandwidth for the voiceprint data that needs to be transmitted according to the training results.
2. The system according to claim 1, characterized in that, The cloud server is configured to allocate a second bandwidth to the target edge device if the first determination result indicates that the first bandwidth is greater than the remaining available bandwidth, wherein the second bandwidth is less than or equal to the remaining available bandwidth.
3. The system according to claim 2, characterized in that, The cloud server is also used to dynamically schedule based on the fault identification priority of voiceprint data when allocating the second bandwidth to the target edge device. The dynamic scheduling of fault identification priorities based on voiceprint data includes: The cloud server pre-sets priority levels for industrial equipment fault voiceprint recognition scenarios; When allocating the second bandwidth, the fault identification priority corresponding to the voiceprint data transmitted by the target edge device is extracted. If it is the first priority, the second bandwidth is allocated according to the first allocation method; if it is the second priority, the second bandwidth is allocated according to the second allocation method; if it is the third priority, the second bandwidth is allocated according to the third allocation method.
4. The system according to any one of claims 1-3, characterized in that, The target edge device will divide the voiceprint data obtained from the industrial equipment into multiple data segments according to time series or frequency characteristics, and obtain the feature label of each data segment. Before transmission, the target edge device reports feature tags of multiple data segments to the cloud server. The cloud server dynamically allocates sub-bandwidth to data segments with different feature tags based on bandwidth usage. During transmission, the bandwidth utilization of each data segment is evaluated in real time. If the utilization of the first sub-bandwidth continues to be lower than the first utilization threshold, the first sub-bandwidth is reclaimed and the bandwidth is reallocated.
5. The system according to claim 3, characterized in that, The cloud server is also used to monitor the transmission progress of multiple edge devices in real time. When a first-priority data transmission request occurs, it can interrupt the third-priority data transmission and reallocate bandwidth.
6. The system according to claim 4, characterized in that, After acquiring the initial voiceprint data of the industrial equipment, the target edge device performs noise filtering on the voiceprint data based on wavelet transform to obtain voiceprint data; wherein, the filtering intensity of the noise filtering is dynamically adjusted according to the operating status of the industrial equipment. The target edge device divides the voiceprint data into multiple data segments according to time series or frequency characteristics, and sets feature overlap areas for adjacent data segments; The target edge device transmits multiple data segments to the cloud server.
7. A method for monitoring device faults using voiceprint recognition based on cloud-edge collaboration, characterized in that, The method is applied to cloud servers; Data transmission between multiple edge devices and the cloud server shares the same bandwidth; the method includes: The target edge device collects initial voiceprint data from the industrial equipment, performs noise filtering on the initial voiceprint data to obtain voiceprint data; increases the filtering intensity when the industrial equipment is turned on; and decreases the filtering intensity when the industrial equipment is running stably. A data transmission request is received from a target edge device, the data transmission request including the amount of voiceprint data, and the target edge device is one of multiple edge devices; The first bandwidth required to transmit the voiceprint data is determined based on latency and data volume; Determine whether the first bandwidth is less than the remaining available bandwidth to obtain the first determination result; If the first determination result indicates that the first bandwidth is less than or equal to the remaining available bandwidth, then the first bandwidth is allocated to the target edge device; The voiceprint data is received through the first bandwidth, and then the industrial equipment is monitored for fault voiceprint recognition based on the voiceprint data to obtain the recognition result of the voiceprint data. Based on the recognition results of the voiceprint data, determine whether the recognition accuracy loss due to bandwidth allocation exceeds the loss threshold; if the recognition accuracy loss exceeds the loss threshold, then reserve a preset multiple of bandwidth for the same type of voiceprint data transmission. Based on the identification results of historical bandwidth allocation, a dynamic bandwidth prediction model is trained, and bandwidth is pre-allocated for the voiceprint data that needs to be transmitted according to the training results.