Sealing-tape machine state detection method and device, computer equipment and storage medium

Abstract

The invention relates to an adhesive tape machine state detection method and device, computer equipment and a storage medium. The method comprises the following steps: acquiring basic operation data generated by the adhesive tape machine under different operation dimensions; carrying out characterization processing on the basic operation data to obtain operation characteristic data of the sealing-tape machine; inputting the operation characteristic data into a preset multi-algorithm fusion decision model, and performing reasoning decision on a fault state of an operation part of the sealing-tape machine to obtain a fault state decision result of the sealing-tape machine; and detecting the state of the sealing-tape machine according to the fault state decision result. By adopting the method, the effect of detecting the state of the sealing-tape machine is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of adhesive tape machine state detection, in particular to an adhesive tape machine state detection method and device, computer equipment and a computer readable storage medium. BACKGROUND

[0002] With the development of science and technology, the application scenarios of adhesive tape machines are becoming more and more extensive, covering many fields such as heavy industry, logistics and manufacturing, and food and beverage. The adhesive tape machine, with the core advantages of high continuous conveying efficiency, strong carrying capacity, and adaptability to complex working conditions, can stably undertake long-distance and uninterrupted transportation tasks of bulk materials such as ores, coal and building materials, or various types of assembled articles, and has become a key equipment in the production and circulation links of various industries. This also makes real-time state detection and fault warning of the adhesive tape machine an indispensable part.

[0003] At present, the state of the adhesive tape machine is usually detected by visual monitoring. Specifically, the work staff visually inspects whether the adhesive tape of the adhesive tape machine has faults such as deviation, tearing or slipping. However, due to the subjectivity and hysteresis of the work staff's inspection, it is difficult to accurately and timely respond to rapid changes in operating conditions and accurately detect various faults, which may cause material transportation interruption and production line stagnation, etc. Therefore, the current state detection effect of the adhesive tape machine is poor. SUMMARY

[0004] In view of the above technical problems, a kind of adhesive tape machine state detection method, device, computer equipment and computer readable storage medium for improving the effect of state detection of adhesive tape machine are provided.

[0005] In the first aspect, the present application provides an adhesive tape machine state detection method, comprising:

[0006] Obtaining basic operation data generated by the adhesive tape machine under different operation dimensions;

[0007] Performing feature processing on the basic operation data to obtain operation feature data of the adhesive tape machine;

[0008] Inputting the operation feature data into a preset multi-algorithm fusion decision model to infer and decide the fault state of the operation components of the adhesive tape machine, and obtaining a fault state decision result of the adhesive tape machine;

[0009] According to the fault state decision result, the state of the adhesive tape machine is detected.

[0010] In one of the embodiments, the basic operation data includes vibration time series data of the vibration components of the adhesive tape machine; the feature processing on the basic operation data to obtain the operation feature data of the adhesive tape machine comprises:

[0011] performing Fourier transform on the vibration time sequence data to obtain vibration frequency domain data of the vibration component;

[0012] extracting a vibration fault feature matching the vibration frequency domain data from the preset fault feature frequency library;

[0013] generating operation feature data of the tape machine according to the vibration fault feature.

[0014] In one of the embodiments, the basic operation data includes an operation image of the tape machine; the basic operation data is subjected to feature processing to obtain operation feature data of the tape machine, including:

[0015] performing fault detection on the operation image through a preset fault detection model to obtain a fault detection result of the tape machine;

[0016] cutting the operation image into an operation feature image retaining a target fault area according to the fault detection result;

[0017] generating operation feature data of the tape machine according to the operation fault feature extracted from the operation image.

[0018] In one of the embodiments, the operation feature data includes vibration spectrum data and image fault feature data, the preset multi-algorithm fusion decision model includes a multi-layer LSTM time sequence prediction model, an image detection model and a federated learning model; the operation feature data is input into the preset multi-algorithm fusion decision model to make inference decision on a fault state of an operation component of the tape machine to obtain a fault state decision result of the tape machine, including:

[0019] inputting the vibration spectrum data into the multi-layer LSTM time sequence prediction model to predict a bearing fault condition of the tape machine to obtain a fault prediction result;

[0020] inputting the image fault feature data into the image detection model to predict a tape fault condition of the tape machine to obtain a fault detection result;

[0021] inputting the fault prediction result and the fault detection result into the federated learning model to make fusion decision on a multi-modal fault condition of the tape machine to obtain the fault state decision result of the tape machine.

[0022] In one of the embodiments, the vibration spectrum data is input into the multi-layer LSTM time sequence prediction model to predict a bearing fault condition of the tape machine to obtain a fault prediction result, including:

[0023] obtaining fused time sequence data by fusing the vibration spectrum data and temperature historical data;

[0024] Using fused time series data as input, after normalization processing, the remaining bearing life of the conveyor belt machine is predicted by a multi-layer LSTM time series prediction model.

[0025] Based on the predicted remaining life of the bearing and the preset confidence interval, the fault status decision result of the conveyor belt is generated.

[0026] In one embodiment, image fault feature data is input into an image detection model to predict the tape fault conditions of the tape machine, and fault detection results are obtained, including:

[0027] By performing multi-scale feature extraction on the image fault feature data, multi-scale image fault features of the conveyor belt machine are obtained. The multi-scale image fault features include image contour fault features and image semantic fault features.

[0028] Image contour fault features are input into the first model structure of the image detection model to classify the fault types of the conveyor belt machine and obtain the fault classification result of the conveyor belt machine. Image semantic fault features are input into the second model structure of the image detection model to locate the fault location of the conveyor belt machine and obtain the fault location result of the conveyor belt machine.

[0029] The fault classification results and fault location results are integrated into the fault detection results.

[0030] In one embodiment, the fault prediction results and fault detection results are jointly input into a federated learning model to perform a fusion decision on the multimodal fault conditions of the conveyor belt machine, resulting in a fault state decision result for the conveyor belt machine, including:

[0031] By inputting both the fault prediction results and the fault detection results into the federated learning model, the bearing fault features in the fault prediction results and the conveyor belt fault features in the fault detection results are spliced ​​and fused, and a weighted voting mechanism is used to determine the overall failure probability of the conveyor belt machine.

[0032] Based on the remaining bearing life data and historical operating records in the fault prediction results, the failure and deterioration trend of the conveyor belt is determined.

[0033] Based on the overall failure probability, failure degradation trend and failure type, the corresponding maintenance suggestions are matched from the preset maintenance knowledge base;

[0034] By integrating the overall failure probability, failure degradation trend, and maintenance suggestions, the failure status decision results of the conveyor belt machine are obtained.

[0035] Secondly, this application also provides a tape machine condition detection device, comprising:

[0036] The acquisition module is used to acquire basic operational data generated by the conveyor belt machine under different operational dimensions;

[0037] The processing module is used to perform feature processing on the basic operating data to obtain the operating feature data of the conveyor belt machine;

[0038] The reasoning and decision-making module is used to input the operating feature data into the preset multi-algorithm fusion decision model, to reason and make decisions on the fault status of the operating components of the conveyor belt, and to obtain the fault status decision result of the conveyor belt.

[0039] The detection module is used to perform status detection on the conveyor belt based on the fault status decision results.

[0040] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0041] Acquire basic operational data generated by the conveyor belt machine under different operational dimensions; perform feature processing on the basic operational data to obtain operational feature data of the conveyor belt machine; input the operational feature data into a preset multi-algorithm fusion decision model to infer and decide on the fault status of the operating components of the conveyor belt machine, and obtain the fault status decision result of the conveyor belt machine; perform status detection on the conveyor belt machine based on the fault status decision result.

[0042] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:

[0043] Acquire basic operational data generated by the conveyor belt machine under different operational dimensions; perform feature processing on the basic operational data to obtain operational feature data of the conveyor belt machine; input the operational feature data into a preset multi-algorithm fusion decision model to infer and decide on the fault status of the operating components of the conveyor belt machine, and obtain the fault status decision result of the conveyor belt machine; perform status detection on the conveyor belt machine based on the fault status decision result.

[0044] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, performs the following steps:

[0045] Acquire basic operational data generated by the conveyor belt machine under different operational dimensions; perform feature processing on the basic operational data to obtain operational feature data of the conveyor belt machine; input the operational feature data into a preset multi-algorithm fusion decision model to infer and decide on the fault status of the operating components of the conveyor belt machine, and obtain the fault status decision result of the conveyor belt machine; perform status detection on the conveyor belt machine based on the fault status decision result.

[0046] The aforementioned conveyor belt machine condition detection method, device, computer equipment, and computer-readable storage medium first acquire basic operating data generated by the conveyor belt machine under different operating dimensions; then, the basic operating data is characterized to obtain the operating characteristic data of the conveyor belt machine; then, the operating characteristic data is input into a preset multi-algorithm fusion decision model to infer and decide on the fault state of the conveyor belt machine's operating components, obtaining the fault state decision result of the conveyor belt machine; finally, based on the fault state decision result, the condition of the conveyor belt machine is detected. Since the operating characteristic data can reflect the actual operation of the conveyor belt machine from multiple operating dimensions, by inputting the operating characteristic data into the preset multi-algorithm fusion decision model, the fault state of the conveyor belt machine can be automatically and timely inferred and decided, thereby achieving the purpose of real-time detection of the conveyor belt machine's condition, rather than relying solely on the delayed inspection by personnel. This overcomes the technical defects of personnel inspection, which is subjective and delayed, leading to an inability to respond promptly and accurately to rapid changes in operating conditions and difficulty in accurately detecting various faults, thus easily resulting in material conveying interruptions and production line stoppages. Therefore, the effect of conveyor belt machine condition detection is improved. Attached Figure Description

[0047] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1 This is a flowchart illustrating a tape machine condition detection method in one embodiment;

[0049] Figure 2 This is a flowchart illustrating a tape machine condition detection method in one embodiment;

[0050] Figure 3 This is a schematic diagram of the process of performing condition detection on a conveyor belt machine according to one embodiment of the condition detection method.

[0051] Figure 4 This is a structural block diagram of a tape machine condition detection device in one embodiment;

[0052] Figure 5 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0053] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0054] First, it should be understood that belt conveyors, with their core advantages of high continuous conveying efficiency, strong load-bearing capacity, and adaptability to complex working conditions, can reliably undertake long-distance, uninterrupted transportation of bulk materials such as ores, coal, and building materials, or various packaged goods, becoming key equipment in the production and flow processes of various industries. For example, belt conveyors used in coal mines employ flame-retardant belts and explosion-proof motors to ensure safe operation in high-risk environments; fully automatic belt conveyors achieve unmanned operation through sensors and control systems, significantly improving the efficiency of packaging in logistics and manufacturing, and reducing labor costs; in logistics and manufacturing, belt conveyors are used for product sealing and labeling, ensuring the consistency and aesthetics of packaging; with the increasing applications of belt conveyors... The widespread use of conveyor belt machines in various scenarios has made real-time status monitoring and fault early warning an indispensable part of the process. However, the current problem of insufficient fault detection and early warning capabilities for conveyor belt machines is becoming increasingly prominent. Traditional monitoring methods are unable to monitor and warn of faults such as belt misalignment, tearing, and slippage in real time, and lack detection and early warning mechanisms for risks such as material blockage or the ingress of sharp objects. For example, in one feasible scenario, if a steel group fails to detect belt tearing in time, it will cause production to be interrupted for several hours, seriously affecting the company's production costs and operational efficiency. Therefore, there is an urgent need to provide a conveyor belt machine status monitoring method that can improve the effectiveness of status monitoring of conveyor belt machines.

[0055] Therefore, this application provides a method for detecting the condition of a tape machine. This embodiment uses the application of this method to a computer device as an example, including but not limited to personal computers, laptops, smartphones, and tablets.

[0056] The aforementioned computer equipment includes an acquisition module, a processing module, an inference and decision-making module, and a detection module. The acquisition module acquires basic operational data generated by the conveyor belt machine under different operational dimensions. The processing module performs feature processing on the basic operational data to obtain the operational feature data of the conveyor belt machine. The inference and decision-making module inputs the operational feature data into a preset multi-algorithm fusion decision model to infer and decide on the fault states of the conveyor belt machine's operating components, obtaining the fault state decision result. The detection module performs state detection on the conveyor belt machine based on the fault state decision result, enabling automatic and timely inference and decision-making on the fault states of the conveyor belt machine. This achieves real-time state detection of the conveyor belt machine, rather than relying solely on delayed inspection by personnel. Therefore, the effectiveness of state detection of the conveyor belt machine is improved.

[0057] It is understood that this method can also be applied to servers, and can also be applied to systems including computer devices and servers, and can be implemented through the interaction between computer devices and servers. This application embodiment does not specifically limit this.

[0058] like Figure 1 As shown, the tape machine status detection method provided in this embodiment includes the following steps:

[0059] Step 202: Obtain the basic operating data generated by the conveyor belt machine under different operating dimensions.

[0060] Specifically, basic operational data generated by the conveyor belt machine under different operating dimensions can be collected using various types of sensors.

[0061] The aforementioned acquisition and processing modules can be integrated into the multimodal sensing layer of a computer device. The core function of this multimodal sensing layer is to complete the holographic acquisition of conveyor belt operation data through various types of sensors such as vibration, temperature, pressure, sound, lidar, and industrial cameras.

[0062] Specifically, the acquisition module collects different basic operational data through corresponding types of sensors and transmits it to an edge AI box integrated into the edge perception layer of the computer device. The edge AI box then performs real-time preprocessing of the raw signals. After receiving the preprocessed edge data, the processing module integrated into the multimodal perception layer performs FFT spectrum analysis on the vibration time series data to extract fault feature frequency bands. Simultaneously, it performs YOLOv8 target detection and precompression operations on the images captured by the industrial camera to achieve feature structuring of multi-source data.

[0063] The computer device's transport layer is configured with a wireless low-power sensor network (LoRaWAN + self-powered design) to encrypt and transmit pre-processed structured data. Simultaneously, based on a dynamic priority queuing mechanism, urgent data such as bearing failures is prioritized for uploading via a dedicated 5G network. Finally, the computer device encapsulates all data into structured messages containing semantic features and sends them to its intelligent transport layer, providing high-quality data support for subsequent status analysis by upper-layer applications.

[0064] In addition, the computer equipment is equipped with an intelligent transmission layer. The intelligent transmission layer can efficiently compress structured messages using time-series compression algorithms (such as PAA), and combine AES-256 encryption to ensure transmission security. Then, it intelligently allocates channels according to data type, transmits key fault characteristics back through a 5G private network, and periodically synchronizes regular status data via Wi-Fi 6 / LoRa. At the same time, it deploys dynamic priority queues to schedule emergency alarms such as bearing failures and tape tears. Finally, the structured messages are sent to the AI ​​analysis layer.

[0065] Step 204: Perform feature processing on the basic operating data to obtain the operating feature data of the conveyor belt machine.

[0066] Operational characteristic data characterizes the operating status and abnormal tendencies of key components of the conveyor belt machine, serving as the core data foundation for subsequent fault detection and condition assessment. Understandably, different characteristic processing methods are used for different basic operational data.

[0067] Step 206: Input the running feature data into the preset multi-algorithm fusion decision model to reason about the fault status of the running parts of the conveyor belt and obtain the fault status decision result of the conveyor belt.

[0068] The computer equipment inputs the operating characteristic data into the preset multi-algorithm fusion decision model, and uses the preset multi-algorithm fusion model to reason about the fault status of the operating components of the conveyor belt and obtain the fault status decision result of the conveyor belt.

[0069] The aforementioned reasoning and decision-making module can be deployed on the AI ​​analysis layer of a computer device. This module can perform deep analysis of structured messages through a multi-algorithm fusion engine. Specifically, the reasoning and decision-making module drives a pre-defined multi-algorithm fusion decision-making model to perform deep analysis of the input structured operational feature messages (operational feature data). The operational feature data can specifically include vibration spectrum data, historical temperature data, conveyor belt image data, and sound signal data collected during the operation of the conveyor belt machine.

[0070] Specifically, the inference and decision-making module first uses an LSTM time-series model to model vibration spectrum data and historical temperature data to predict the remaining life of the bearing. Simultaneously, it deploys an improved EfficientDet-D7 image model to detect tape tearing based on tape image data and employs a Transformer architecture to classify MFCC-characteristic sound signals. A federated learning framework is constructed for multi-device model co-evolution, and transfer learning techniques are used to inject fault data from other scenarios into the local model. Finally, multi-dimensional feature cross-validation generates a decision message containing fault probability, degradation trend, and maintenance suggestions, which is then sent to the application service layer of the computer equipment.

[0071] Step 208: Based on the fault status decision results, perform status detection on the conveyor belt.

[0072] The aforementioned detection module can be deployed at the application service layer of a computer device. Based on the decision messages output by the reasoning and decision-making module, this module can construct a closed-loop decision-making system for the conveyor belt machine's operating status.

[0073] Specifically, the execution process for condition monitoring of the conveyor belt machine is as follows: A digital twin engine renders a 3D virtual model of the conveyor belt machine in real time on the web, overlaying a fault heatmap to intuitively present the fault location and severity. It also supports the time-series playback and multi-dimensional visualization of historical operating data, enabling visualized monitoring of fault status. A multi-level alarm engine is run, classifying fault levels based on the fault probability and severity in the decision message. Correspondingly, it automatically triggers shutdown protection, generates structured work orders, or pushes warning information to mobile devices. Simultaneously, NLP technology is introduced to transform standardized diagnostic codes into natural language maintenance guidelines, lowering the barrier to maintenance operations. Based on the predictive maintenance module, a dynamic maintenance plan is output. Combining the equipment degradation trend and spare parts inventory in the decision message, the optimal maintenance window is intelligently recommended, achieving closed-loop management of the entire process from fault warning to maintenance execution.

[0074] As an example, step 208 includes: detecting the status of the conveyor belt based on the fault status decision result, and obtaining the status detection result of the conveyor belt.

[0075] The aforementioned conveyor belt machine condition detection method first acquires basic operational data generated by the conveyor belt machine under different operational dimensions; then, it performs feature processing on the basic operational data to obtain operational feature data of the conveyor belt machine; next, it inputs the operational feature data into a preset multi-algorithm fusion decision model to infer and decide on the fault status of the conveyor belt machine's operating components, obtaining the fault status decision result of the conveyor belt machine; finally, based on the fault status decision result, it performs condition detection on the conveyor belt machine. Since the operational feature data can reflect the actual operation of the conveyor belt machine from multiple operational dimensions, by inputting the operational feature data into the preset multi-algorithm fusion decision model, it can automatically and promptly infer and decide on the fault status of the conveyor belt machine, thereby achieving the goal of real-time detection of the conveyor belt machine's status, rather than relying solely on the delayed inspection by personnel. This overcomes the technical defects of personnel inspection, which is subjective and delayed, leading to an inability to respond promptly and accurately to rapid changes in operating conditions and difficulty in accurately detecting various faults, thus easily resulting in material conveying interruptions and production line stoppages. Therefore, it improves the effectiveness of conveyor belt machine condition detection.

[0076] In one embodiment, such as Figure 2 As shown, the basic operating data includes the vibration time series data of the vibrating components of the conveyor belt; the basic operating data is then subjected to feature processing to obtain the operating feature data of the conveyor belt, including:

[0077] Step 302: Perform Fourier transform on the vibration time series data to obtain the vibration frequency domain data of the vibrating component.

[0078] In the process of characterizing the vibration time series data of the vibrating components of the conveyor belt machine, the FFT spectrum analysis of the vibration time series data is first performed, and then the relevant vibration fault features are extracted by matching the vibration frequency domain data obtained by the preset fault feature frequency library and the spectrum analysis.

[0079] As an example, step 302 includes: obtaining the vibration frequency domain data of the vibrating component by performing a fast Fourier transform on the vibration time series data.

[0080] Step 304: Extract vibration fault features that match the vibration frequency domain data from the preset fault feature frequency library.

[0081] It should be noted that the preset fault characteristic frequency library can store different fault-related frequency bands, which can be used as the basis for screening vibration fault characteristics.

[0082] As an example, step 304 includes: selecting vibration fault features from a preset fault feature frequency library of vibration frequency domain data.

[0083] Step 306: Generate the operating characteristic data of the conveyor belt based on the vibration fault characteristics.

[0084] As an example, step 306 includes converting vibration fault characteristics into operating characteristic data of the conveyor belt.

[0085] In one feasible approach, the vibration time-series signal of the vibrating components of the conveyor belt (such as motors, reducers, or roller bearings) is acquired in real time using vibration sensors and denoted as x(t). A Fast Fourier Transform (FFT) is performed on x(t) to transform it from the time domain to the frequency domain, yielding the spectrum X(f).

[0086] X(f) = ∫-∞∞x(t)e-j2πftdt

[0087] Where X(f) is the vibration frequency domain data and ft is the vibration time series data; then, based on the equipment fault characteristic frequency library, frequency bands related to the fault (such as gear meshing frequency fm and its harmonics 2fm, 3fm, bearing outer ring fault frequency fBPFO, etc.) are selected, and the amplitude information of these frequency bands is retained as fault characteristics.

[0088] In this embodiment, the vibration frequency domain data of the vibrating component is obtained by fast Fourier transform, and then the vibration fault features matching the vibration frequency domain data are extracted from the preset fault feature frequency library. Finally, the vibration fault features are converted into the operating feature data of the conveyor belt, so as to facilitate the subsequent accurate condition detection of the conveyor belt. Therefore, it lays the foundation for improving the condition detection effect of the conveyor belt.

[0089] In one embodiment, the basic operating data includes images of the conveyor belt in operation; the basic operating data is then subjected to feature processing to obtain the operating feature data of the conveyor belt, including:

[0090] The fault detection results of the conveyor belt are obtained by performing fault detection on the running image through a preset fault detection model; based on the fault detection results, the running image is cropped into a running feature image that retains the target fault area; and the running feature data of the conveyor belt is generated based on the running fault features extracted from the running image.

[0091] Images of the tape conveyor operating are captured at a fixed frame rate using an industrial camera, denoted as I(x,y). I(x,y) is input into a YOLOv8 object detection model, which outputs fault detection results. These results include the bounding box coordinates (xmin, ymin, xmax, ymax), category labels (e.g., "tear", "foreign_object"), and confidence scores s for the target (e.g., a torn tape, a foreign object). Based on the fault detection results, the image is pre-compressed: the region within the bounding box is retained, irrelevant background is cropped to reduce the data volume, and the resolution of the cropped image is reduced from the original value (e.g., 4K) to the resolution required by the model input (e.g., 640×640). The original pixel information of the target region is retained or its features (e.g., texture, color histogram) are extracted for subsequent analysis.

[0092] As an example, a running image is input into a YOLOv8 object detection model. The YOLOv8 model performs fault detection on the running image, obtaining fault detection results for the conveyor belt. Based on the fault detection results, the running image is pre-compressed to obtain a running feature image that retains the target fault region. Running fault features are extracted from the running image and converted into running feature data for the conveyor belt. This provides image-based running feature data for the conveyor belt, facilitating subsequent accurate condition detection and laying the foundation for improving the effectiveness of conveyor belt condition detection.

[0093] Understandably, when transmitting basic operational data at the intelligent transport layer, specifically, the structured messages generated by the multimodal sensing layer (including vibration characteristic frequency bands, image target detection results, etc.) are encapsulated into the frame format specified by the LoRaWAN protocol. The LoRaWAN frame structure includes a preamble, a physical layer header (PHY header), and a physical layer payload (PHY payload), with the PHY payload carrying structured data. The PHY payload of the LoRaWAN frame is encrypted using the AES-256 algorithm.

[0094] C = EK(M), where M is the plaintext data (structured message), K is the 256-bit encryption key, EK(·) is the AES-256 encryption function, and C is the ciphertext data. The encrypted data is sent to the LoRaWAN gateway using LoRa modulation technology. The spreading factor (SF) and bandwidth (BW) of LoRa modulation affect the data transmission rate and distance, and their relationship can be approximately expressed as: Rb = SF·BW²SF·CRR, where Rb is the bit rate and CR is the coding rate. Sensor nodes collect ambient energy through solar panels or supercapacitors to power the LoRaWAN module, reducing reliance on traditional batteries. In addition, dynamic priority queues can be set up, marking priorities according to data type at the LoRaWAN gateway or edge computing node; for example, bearing fault characteristics are marked as high priority (QoS 2), and normal state data is marked as low priority (QoS 2). 0); Weighted Fair Queuing (WFQ) or Strict Priority Queuing (SPQ) algorithms are used to schedule data of different priorities; high-priority data (such as bearing fault characteristics) occupies the transmission channel first, ensuring that its transmission delay is below the threshold (such as 50ms); high-priority data is allocated dedicated bandwidth and low-latency channels through 5G private network slicing technology, and is quickly uploaded to the cloud or data center. The user plane latency (UPdelay) of the 5G network can be as low as 1ms, meeting the needs of emergency data transmission.

[0095] Furthermore, structured messages are efficiently compressed using time-series compression algorithms (such as PAA), and AES-256 encryption is used to ensure transmission security. Subsequently, channels are intelligently allocated based on data type, and backhaul is performed via a 5G private network based on key fault characteristics. This includes the following steps:

[0096] Time Series Compression: Time series data (such as vibration spectrum amplitude, temperature history record) in structured messages is segmented into equal-length windows. Assuming the original time series is X=[x1,x2,...,xn], it is divided into w segments, each with a length of l=nw. The average value of each segment is calculated to obtain the compressed sequence Xˉ=[xˉ1,xˉ2,...,xˉw], where xˉi=1lj=(i-1)l+1ilxj (i=1,2,...,w). The compressed sequence xˉ retains the trend characteristics of the original data while reducing the data volume to 1w of the original. AES-256 Encryption: A 256-bit encryption key K is generated to ensure the randomness and security of the key. The compressed data xˉ is encrypted using the AES-256 algorithm to generate ciphertext C. Intelligent Channel Assignment: Priority is marked according to data type. For example, critical fault data such as bearing fault characteristics and tape tear detection results are marked as high priority (QoS). 2) Routine status data (such as temperature and pressure) are marked as low priority (QoS 0); high priority data is allocated dedicated bandwidth and low-latency channels through 5G private network slicing technology to ensure real-time backhaul of key fault characteristics; low priority data is periodically synchronized through Wi-Fi 6 or LoRa networks to reduce communication costs; 5G private network backhaul: network slicing is configured in the 5G private network to allocate independent logical channels for high priority data to avoid competing for resources with other services; by utilizing the short frame structure of 5G (such as slot duration as low as 125μs) and flexible scheduling mechanism, key fault characteristics can be quickly uploaded to the cloud or data center with transmission latency below the threshold (such as 50ms).

[0097] In one embodiment, the operational feature data includes vibration spectrum data and image fault feature data. The preset multi-algorithm fusion decision model includes a multi-layer LSTM time-series prediction model, an image detection model, and a federated learning model. The operational feature data is input into the preset multi-algorithm fusion decision model to infer and decide on the fault state of the conveyor belt's operating components, resulting in a fault state decision for the conveyor belt, including:

[0098] Vibration spectrum data is input into a multilayer LSTM time series prediction model to predict bearing failures in a conveyor belt machine, resulting in a failure prediction result. Image fault feature data is input into an image detection model to predict conveyor belt failures, resulting in a failure detection result. The failure prediction result and the failure detection result are then input into a federated learning model to perform fusion decision-making on the multimodal failures of the conveyor belt machine, resulting in a failure state decision result for the conveyor belt machine.

[0099] The vibration spectrum data is input into a multilayer LSTM time series prediction model to predict bearing failures in a conveyor belt machine, yielding failure prediction results, including:

[0100] By fusing vibration spectrum data and historical temperature data, fused time series data is obtained. Using the fused time series data as input, after normalization processing, the remaining bearing life of the conveyor belt is predicted by a multi-layer LSTM time series prediction model. Based on the predicted remaining bearing life and a preset confidence interval, the fault status decision result of the conveyor belt is generated.

[0101] As an example, vibration spectrum data (such as gear meshing frequency) (and its harmonic amplitude) and historical temperature data (such as bearing outer ring temperature) Construct a joint time series by aligning with timestamps. S=[ X vib , X temp ] That is, to obtain fused time series data, where, This represents the characteristics of the vibration spectrum over n time steps and f frequency bands. S represents the scalar values ​​of temperature at n time steps; S is normalized by min and max to scale the features to the [0,1] interval, as shown in the following formula:

[0102]

[0103] Then, LSTM time series modeling is performed. Specifically, a multi-layer LSTM network is constructed, with the normalized time series Snorm as input and the bearing remaining life prediction value y as output. The LSTM cell state update formula is:

[0104] Forget Gate: Decides which information to discard;

[0105] ft = σ (Wf ⋅ [ht - 1 ,xt ]+bf )

[0106] Input gate: Determines which parts of the new information need to be updated;

[0107] it = σ (Wi ⋅ [ht - 1 ,xt ]+bi )

[0108] C ̃ t =tanh(WC ⋅ [ht - 1 ,xt ]+bC )

[0109] Cell status update:

[0110]

[0111] Output gate: Generates a hidden state;

[0112] ot = σ (Wo ⋅ [ht - 1 ,xt ]+bo )

[0113]

[0114] Where σ(*) is the Sigmoid function, ⊙ represents element-wise multiplication, and Wf,Wi,WC,Wo and bf,bi,bC,bo are trainable parameters. Enter the current time. This is the hidden state from the previous moment. This represents the cell state at the previous moment. Output for the forget gate. For input gate output, Candidate cell state, The output is then used as the output gate; subsequently, a multi-layer LSTM time series prediction model is trained and optimized. Specifically, the mean squared error (MSE) is used as the loss function to measure the predicted value. Deviation from the true value y:

[0115]

[0116] Update network parameters using the Adam optimizer, with the learning rate set to... Weight decay terms are used to prevent overfitting; a sliding window method is employed to generate training samples, with a window length L covering 80% of the bearing's entire lifespan data, and a step size of L / 2 to increase data diversity; real-time collected vibration spectrum and temperature data are input into the trained LSTM model, which outputs the predicted bearing lifespan at the current moment. That is, the predicted remaining life of the bearings of the output conveyor belt machine; the uncertainty of the prediction result is estimated by Monte Carlo dropout (MC Dropout), that is, some neurons are randomly dropped during the forward propagation of the model, and the prediction distribution N(μ, ), where μ is the expected lifespan and σ is the confidence interval width, that is, the final decision result of the failure state of the conveyor belt is generated by combining the predicted remaining lifespan of the bearing and the preset confidence interval.

[0117] The process involves inputting image fault feature data into an image detection model to predict conveyor belt faults in the conveyor belt machine, thereby obtaining fault detection results, including:

[0118] Multi-scale feature extraction is performed on the image fault feature data to obtain multi-scale image fault features of the conveyor belt machine. These multi-scale image fault features include image contour fault features and image semantic fault features. The image contour fault features are input into the first model structure of the image detection model to classify the conveyor belt fault types and obtain the fault classification results. The image semantic fault features are input into the second model structure of the image detection model to locate the conveyor belt fault locations and obtain the fault location results. The fault classification results and fault location results are integrated into the fault detection results.

[0119] As an example, the image detection model can be an improved EfficientDet-D7 image model, which then uses an industrial camera to acquire images of a conveyor belt in operation at a high frame rate (e.g., 60 FPS) with a resolution of at least 4K (3840×2160). Pixel-level annotations are performed on the torn areas of the conveyor belt to generate semantic segmentation labels. Data augmentation techniques (random rotation, contrast adjustment, noise injection) are used to expand the sample size to 10 times that of the original data. Based on the EfficientNet-B7 backbone, a Squeeze-and-Excitation (SE) attention module is introduced to enhance the representation of channel-dimensional features. The SE module then applies the feature map... Perform adaptive weight allocation:

[0120]

[0121]

[0122] Where GAP(*) is global average pooling, δ(*) is ReLU activation, and σ(*) is the Sigmoid function; W1 and W2 are dimensionality reduction parameters. The feature map after recalibration. Input feature map;

[0123] Bidirectional Feature Pyramid Network (BiFPN) is employed to enhance multi-scale detection capabilities through weighted feature fusion; nodes The output is:

[0124]

[0125] in, , For learnable weights, For numerically stable terms, By adjusting the feature map size, the image contour fault features and image semantic fault features of the tape machine can be obtained. Then, a DecoupledHead structure is used to handle the classification and regression tasks respectively. The classification branch outputs the confidence score p∈[0,1] of the tape tear, and the regression branch outputs the bounding box coordinates (x,y,w,h). Focal Loss is used to alleviate the sample imbalance problem, and then the fault classification results and fault location results of the tape machine are output respectively.

[0126] Where α=0.25 is the balance factor and γ=2 is the modulation coefficient; GIoU Loss is used to optimize the bounding box positioning accuracy.

[0127] ,in, Regression loss, GIoU is the generalized intersection-union ratio, Bpred is the predicted bounding box, and Bgt is the ground truth bounding box;

[0128] Transformer architecture:

[0129] Pre-emphasize the audio signal s(t) (boost the high-frequency components):

[0130]

[0131] in, The original speech signal is sampled at time t. ′( The signal is after pre-emphasis, and α is the pre-emphasis coefficient. For discrete-time indexing;

[0132] Divide the signal into frames (frame length 25ms, frame shift 10ms), and multiply by the Hamming window w(n):

[0133]

[0134] Where n=0,1,...,N-1, is the intra-frame sample index, and N is the window length;

[0135] Perform FFT on each frame of signal, calculate the energy spectrum using a Mel filter bank (40 triangular filters), take the logarithm, and then perform DCT transform to obtain the 13-dimensional MFCC feature M∈ (T is the number of frames); linearly project the MFCC feature M onto the dimension. =512, and add position encoding PE∈ :

[0136] =sin(

[0137] =cos(

[0138] A stacked encoder with L=6 layers is used, each layer containing a multi-head attention (MHA) network and a feedforward network (FFN); the self-attention mechanism is calculated as follows:

[0139] Where Q,K,V∈ For querying key-value matrices, =64 represents the single-head dimension; the [CLS] label feature output from the last layer of the encoder is taken, and the class probability p∈ is output through a fully connected layer. (C represents the number of categories, such as normal / friction / impact), which is the final integrated result of the fault detection of the conveyor belt machine.

[0140] The fault prediction results and fault detection results are jointly input into the federated learning model to perform fusion decision-making on the multimodal fault conditions of the conveyor belt machine, resulting in the fault state decision results of the conveyor belt machine, including:

[0141] By inputting both fault prediction and fault detection results into a federated learning model, the bearing fault features from the fault prediction results and the conveyor belt fault features from the fault detection results are spliced ​​and fused. A weighted voting mechanism is used to determine the overall failure probability of the conveyor belt. Based on the remaining bearing life data and historical operating records in the fault prediction results, the failure degradation trend of the conveyor belt is determined. Based on the overall failure probability, failure degradation trend, and fault type, corresponding maintenance suggestions are matched from a pre-set maintenance knowledge base. By integrating the overall failure probability, failure degradation trend, and maintenance suggestions, the fault state decision result of the conveyor belt is obtained.

[0142] It should be noted that by constructing a federated learning framework for multi-device model co-evolution, and combining transfer learning technology to inject fault data from other scenarios into the local model, a decision message containing fault probability, degradation trend and maintenance suggestions is generated through multi-dimensional feature cross-validation, and the decision message is used as the fault status decision result of the conveyor belt machine.

[0143] As an example, the federated learning model is first built by having a computer device (such as an edge AI box) register with the federated learning server and download the initial global model parameters. Initial model The model is pre-trained on a public dataset (such as the CWRU bearing dataset) and includes an LSTM time series module, an EfficientDet-D7 image module, and a Transformer sound module. Each device trains its model using local data (vibration, temperature, images, and sound) and calculates the parameter gradients ∇. Differential privacy (DP) technology is used to add noise to the gradients to protect data privacy.

[0144] Where σ is the noise scale, inversely proportional to the privacy budget ϵ; the server uses the FedAvg algorithm to aggregate gradients uploaded from each device and update the global model. :

[0145] = -η· Where η is the learning rate and N is the number of participating devices;

[0146] Repeat the local training and global aggregation steps until the global model converges (e.g., validation set loss). The rate of decline was <1% for five consecutive rounds. Transfer learning techniques were employed, specifically collecting fault data from other scenarios (such as coal mines and ports), including vibration spectrum (Xsrc), image (Isrc), and sound (Ssrc). The source data underwent the same preprocessing as the target scenario (e.g., FFT, MFCC, normalization) to generate structured features. Maximum mean difference (MMD) loss was used to align the feature distributions of the source and target scenarios.

[0147] Where ϕ(*) is the kernel function (such as the RBF kernel), and H is the regenerating kernel Hilbert space;

[0148] Freeze the low-level parameters of the global model (such as the early convolutional layers of EfficientNet-B7), and fine-tune only the high-level parameters (such as the classification head); train jointly using source scene data and target scene data, with a total loss of:

[0149] Where λ=0.5 and γ=0.1 are hyperparameters;

[0150] Multi-dimensional feature cross-validation:

[0151] The vibration-temperature joint characteristics output by LSTM Image features output by EfficientDet-D7 Sound characteristics output by Transformer splicing into multimodal features =[ , , ];

[0152] A weighted voting mechanism is used to synthesize the prediction results from the three modalities:

[0153] Where wvib=0.5, image=0.3, sound=0.2 is the preset weight, σ(*) is the Sigmoid function;

[0154] The degradation rate is calculated based on the remaining lifetime prediction value output by LSTM and combined with historical data:

[0155]

[0156] That is, after obtaining the failure and deterioration trend of the conveyor belt machine, further analysis is conducted based on the type of failure (such as bearing wear, conveyor belt tearing) and severity (such as...). (fault > 0.9), matches maintenance strategies (such as replacing bearings or repairing tape) from the knowledge base, generates a decision message, specifically, the fault probability is... The fault, degradation trend (vdeg), and maintenance suggestion (Srepair) are encapsulated into a structured message (Mdecision), which is then sent to the application service layer via the intelligent transport layer to trigger digital twin engine updates, alarm pushes, and other operations. Furthermore, the digital twin engine renders a 3D model of the conveyor belt machine in real-time on the web and overlays a fault heatmap, synchronously supporting historical data time-series playback, including the following steps:

[0157] Digital twin engine 3D model rendering:

[0158] Based on the CAD drawings of the conveyor belt machine, a high-precision 3D model was built using Blender or SolidWorks, including key components such as motors, reducers, rollers, and idlers. The model underwent lightweight processing (e.g., reducing polygon count and merging materials), and was exported in glTF 2.0 format, compressing the file size to 30% of the original model. The glTF model was loaded into a browser using the Three.js or Babylon.js engine, and hardware-accelerated rendering was performed via WebGL 2.0. A PBR (Physically Based Rendering) material system was adopted, dynamically adjusting the model's surface properties (such as metallicity and roughness) based on real-time data to enhance realism. Fault heatmap overlay was performed: the fault probability pfault and degradation trend vdeg output from the AI ​​analysis layer were received and mapped to color gradients; for example, the Jet color gamut was used. The Jet(*) function maps scalar values ​​[-1,1] to RGB colors; dynamic heatmap textures are generated based on the fault probability distribution. UV unwrapping is performed on the surface of the conveyor belt, mapping the heatmap textures to corresponding areas (such as bearings and reducers); in the WebGL shader, fragment colors are calculated by combining model vertex coordinates and heatmap textures; historical data time-series playback: structured messages (including vibration, temperature, and image features) are stored in a time-series database (such as InfluxDB) by timestamps, and a time index is designed to support fast queries; a timeline control is designed on the Web side, supporting slider adjustment of playback speed (0.1×~10×) and time period selection; data queries are triggered when the timeline is updated; based on the retrieved historical data, the 3D model status (such as component displacement and temperature color scale) and fault heatmap are updated; in addition, to enhance interactivity and usability... Visualization: When pfault > 0.8, an exploded view of the model is automatically triggered, shifting the faulty component (such as a bearing) by 10% along the normal direction and flashing a warning; fault code labels (such as "BPFO fault") are overlaid on the heatmap, and clicking the label displays detailed information (probability value, occurrence time, maintenance suggestions); split-screen display is supported (such as a 3D model on the left and a time-series curve on the right), and view parameters (such as camera position and zoom level) are synchronized via WebSockets; furthermore, a multi-level alarm engine is run to automatically trigger shutdown protection, generate structural work orders, or push mobile alerts based on the severity of the fault, and NLP technology is introduced to convert diagnostic codes into natural language maintenance guidelines, including the following steps:

[0159] Assessing fault severity: Receive the fault probability pfault and degradation trend vdeg from the AI ​​analysis layer, and calculate the comprehensive risk value R. ,in, p=0.7 v=0.3 is the weighting coefficient, max(0,- deg) increases the risk when degradation accelerates; set a three-level risk threshold: =0.8 (Red Alert) =0.5 (Orange Alert) =0.2 (Yellow Alarm); Multi-level Alarm Response: Level 1 Alarm (Red, R≥) ): Sends an emergency stop command to the PLC via the Modbus TCP protocol to cut off the motor power; Level 2 alarm (orange, ≤R< ): Create a maintenance work order, including the faulty equipment, type, recommended actions, and priority; send the work order to the EAM (Enterprise Asset Management) system via REST API and assign it to the maintenance team; three-level alarms (yellow, yellow, ... ≤R< ): Push notifications to maintenance personnel's mobile APP via MQTT protocol, including fault briefings and 3D model snapshots; Generate NLP maintenance guides: Receive fault codes (such as "ERR_BEARING_BPFO_0x1F") output by the AI ​​analysis layer, break them down into faulty parts (bearings), type (BPFO fault), and severity level (0x1F); Query the standardized maintenance process for the corresponding fault in the maintenance knowledge base; Use T5 or GPT models to convert the structured maintenance process into natural language guides, adding safety tips and spare parts codes.

[0160] In one feasible approach, refer to Figure 3 , Figure 3 This is a flowchart illustrating the process of condition detection for a conveyor belt machine. Initially, the multimodal perception layer of the computer device outputs structured feature messages (operational feature data). Then, the intelligent transmission layer of the computer device encrypts and classifies these structured feature messages before transmission. After the structured feature messages are transmitted to the AI ​​analysis layer of the computer device, the AI ​​analysis layer inputs them into a preset multi-algorithm fusion decision model to infer and decide on the fault states of the conveyor belt machine's operating components. This results in a fault state decision message, which is then output to the application service layer of the computer device. Based on the decision message, the application service layer performs condition detection on the conveyor belt machine, obtaining the condition detection result. Ultimately, the condition detection result can trigger shutdown protection, maintenance plans, or generate work orders.

[0161] In the above implementation, a fusion acquisition method using LiDAR, industrial cameras, and vibration sensors is adopted. Combined with FFT spectrum analysis and YOLOv8 target detection preprocessing by the edge AI box, structured feature messages are generated. Through a multi-algorithm fusion engine using LSTM time series model, EfficientDet-D7 image model, and Transformer sound classification model, accurate prediction of fault probability and degradation trend is achieved. Finally, the diagnostic code is transformed into a natural language maintenance guide by a digital twin engine and a multi-level alarm engine combined with NLP technology. This solves the problem of insufficient fault detection and early warning capabilities of traditional computer equipment, greatly reduces the failure risk of conveyor belt machines, and effectively improves the operational efficiency of enterprises, that is, it improves the effect of condition detection of conveyor belt machines.

[0162] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0163] Based on the same inventive concept, this application also provides a tape machine condition detection device for implementing the tape machine condition detection method described above. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations in one or more tape machine condition detection device embodiments provided below can be found in the limitations of the tape machine condition detection method described above, and will not be repeated here.

[0164] In one exemplary embodiment, such as Figure 4 As shown, a tape machine status detection device is provided, including: an acquisition module 401, a processing module 402, a reasoning and decision-making module 403, and a detection module 404, wherein:

[0165] The acquisition module 401 is used to acquire basic operating data generated by the conveyor belt machine under different operating dimensions;

[0166] The processing module 402 is used to perform feature processing on the basic operating data to obtain the operating feature data of the conveyor belt machine;

[0167] The reasoning and decision-making module 403 is used to input the operating feature data into the preset multi-algorithm fusion decision model, to reason and make decisions on the fault status of the operating components of the conveyor belt, and to obtain the fault status decision result of the conveyor belt.

[0168] The detection module 404 is used to perform status detection on the conveyor belt based on the fault status decision results.

[0169] In one embodiment, the basic operational data includes images of the conveyor belt in operation; the processing module 402 is further configured to:

[0170] The fault detection results of the conveyor belt are obtained by performing fault detection on the running image through a preset fault detection model; based on the fault detection results, the running image is cropped into a running feature image that retains the target fault area; and the running feature data of the conveyor belt is generated based on the running fault features extracted from the running image.

[0171] In one embodiment, the basic operational data includes images of the conveyor belt in operation; the processing module 402 is further configured to:

[0172] The fault detection results of the conveyor belt are obtained by performing fault detection on the running image through a preset fault detection model; based on the fault detection results, the running image is cropped into a running feature image that retains the target fault area; and the running feature data of the conveyor belt is generated based on the running fault features extracted from the running image.

[0173] In one embodiment, the running feature data includes vibration spectrum data and image fault feature data, and the preset multi-algorithm fusion decision model includes a multi-layer LSTM time series prediction model, an image detection model, and a federated learning model; the inference decision module 403 is also used for:

[0174] Vibration spectrum data is input into a multilayer LSTM time series prediction model to predict bearing failures in a conveyor belt machine, resulting in a failure prediction result. Image fault feature data is input into an image detection model to predict conveyor belt failures, resulting in a failure detection result. The failure prediction result and the failure detection result are then input into a federated learning model to perform fusion decision-making on the multimodal failures of the conveyor belt machine, resulting in a failure state decision result for the conveyor belt machine.

[0175] In one embodiment, the reasoning decision module 403 is further configured to:

[0176] By fusing vibration spectrum data and historical temperature data, fused time series data is obtained. Using the fused time series data as input, after normalization processing, the remaining bearing life of the conveyor belt is predicted by a multi-layer LSTM time series prediction model. Based on the predicted remaining bearing life and a preset confidence interval, the fault status decision result of the conveyor belt is generated.

[0177] In one embodiment, the reasoning decision module 403 is further configured to:

[0178] Multi-scale feature extraction is performed on the image fault feature data to obtain multi-scale image fault features of the conveyor belt machine. These multi-scale image fault features include image contour fault features and image semantic fault features. The image contour fault features are input into the first model structure of the image detection model to classify the conveyor belt fault types and obtain the fault classification results. The image semantic fault features are input into the second model structure of the image detection model to locate the conveyor belt fault locations and obtain the fault location results. The fault classification results and fault location results are integrated into the fault detection results.

[0179] In one embodiment, the reasoning decision module 403 is further configured to:

[0180] By inputting both fault prediction and fault detection results into a federated learning model, the bearing fault features in the fault prediction results and the conveyor belt fault features in the fault detection results are spliced ​​and fused, and a weighted voting mechanism is used to determine the overall failure probability of the conveyor belt. Based on the bearing remaining life data and historical operating records in the fault prediction results, the failure degradation trend of the conveyor belt is determined. Based on the overall failure probability, failure degradation trend, and fault type, corresponding maintenance suggestions are matched from a preset maintenance knowledge base. By integrating the overall failure probability, failure degradation trend, and maintenance suggestions, the failure status decision result of the conveyor belt is obtained.

[0181] Each module in the aforementioned conveyor belt condition detection device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0182] In one exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 5 As shown. The computer device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements a method for detecting the state of a conveyor belt machine. Those skilled in the art will understand that... Figure 5 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0183] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0184] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.

[0185] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0186] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0187] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0188] The above embodiments are merely illustrative of several implementation methods of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for detecting the condition of a tape conveyor machine, characterized in that, The method includes: Acquire basic operational data generated by the conveyor belt machine under different operational dimensions; The basic operating data is subjected to feature processing to obtain the operating feature data of the conveyor belt machine; The operational feature data is input into a preset multi-algorithm fusion decision model to infer and decide on the fault status of the operating components of the conveyor belt, thereby obtaining the fault status decision result of the conveyor belt. Based on the fault status decision results, the condition of the conveyor belt is monitored.

2. The method according to claim 1, characterized in that, The basic operating data includes the vibration time-series data of the vibrating components of the conveyor belt; the characteristic processing of the basic operating data to obtain the operating characteristic data of the conveyor belt includes: The vibration time series data is subjected to Fourier transform to obtain the vibration frequency domain data of the vibration component; Extract vibration fault features that match the vibration frequency domain data from a preset fault feature frequency library; Based on the vibration fault characteristics, the operating characteristic data of the conveyor belt is generated.

3. The method according to claim 1 or 2, characterized in that, The basic operating data includes the operating images of the conveyor belt machine; the feature processing of the basic operating data to obtain the operating feature data of the conveyor belt machine includes: The fault detection results of the conveyor belt machine are obtained by performing fault detection on the running image using a preset fault detection model. Based on the fault detection results, the running image is cropped into a running feature image that retains the target fault area; Based on the operational fault features extracted from the operational images, operational feature data of the conveyor belt machine is generated.

4. The method according to claim 3, characterized in that, The operational feature data includes vibration spectrum data and image fault feature data. The preset multi-algorithm fusion decision model includes a multi-layer LSTM time-series prediction model, an image detection model, and a federated learning model. The step of inputting the operational feature data into the preset multi-algorithm fusion decision model to infer and decide the fault state of the conveyor belt's operating components, and obtaining the fault state decision result of the conveyor belt, includes: The vibration spectrum data is input into the multilayer LSTM time series prediction model to predict the bearing failure of the conveyor belt machine and obtain the failure prediction result. The image fault feature data is input into the image detection model to predict the tape fault conditions of the tape machine and obtain the fault detection result. The fault prediction results and fault detection results are input together into the federated learning model to perform fusion decision on the multimodal fault conditions of the conveyor belt machine, and obtain the fault state decision result of the conveyor belt machine.

5. The method according to claim 4, characterized in that, The step of inputting the vibration spectrum data into the multilayer LSTM time series prediction model to predict the bearing failure of the conveyor belt machine and obtain the failure prediction result includes: By fusing the vibration spectrum data and temperature history data, fused time-series data is obtained; Using the fused time series data as input, after normalization processing, the remaining bearing life of the conveyor belt machine is predicted by the multilayer LSTM time series prediction model. Based on the predicted remaining life of the bearing and the preset confidence interval, the fault status decision result of the conveyor belt is generated.

6. The method according to claim 4, characterized in that, The step of inputting the image fault feature data into the image detection model to predict the tape fault condition of the tape machine and obtain the fault detection result includes: By performing multi-scale feature extraction on the image fault feature data, multi-scale image fault features of the tape machine are obtained, wherein the multi-scale image fault features include image contour fault features and image semantic fault features. The image contour fault features are input into the first model structure of the image detection model to classify the tape fault types of the tape machine and obtain the fault classification result of the tape machine. The image semantic fault features are input into the second model structure of the image detection model to locate the tape fault location of the tape machine and obtain the fault location result of the tape machine. The fault classification results and the fault location results are integrated into the fault detection results.

7. The method according to claim 4, characterized in that, The step of inputting the fault prediction results and fault detection results into the federated learning model to perform fusion decision-making on the multimodal fault conditions of the conveyor belt machine, and obtaining the fault state decision result of the conveyor belt machine, includes: By inputting the fault prediction results and fault detection results into the federated learning model, the bearing fault features in the fault prediction results and the conveyor belt fault features in the fault detection results are spliced ​​and fused, and a weighted voting mechanism is used to determine the overall failure probability of the conveyor belt machine. Based on the remaining bearing life data and historical operating records in the fault prediction results, the fault deterioration trend of the conveyor belt is determined; Based on the overall failure probability, the failure degradation trend, and the failure type, corresponding maintenance suggestions are matched from a preset maintenance knowledge base; By integrating the overall failure probability, failure degradation trend, and maintenance suggestions, the failure status decision result of the conveyor belt machine is obtained.

8. A condition detection device for a tape conveyor, characterized in that, The device includes: The acquisition module is used to acquire basic operational data generated by the conveyor belt machine under different operational dimensions; The processing module is used to perform feature processing on the basic operating data to obtain the operating feature data of the conveyor belt machine; The reasoning and decision-making module is used to input the operating feature data into a preset multi-algorithm fusion decision model, to reason and make decisions on the fault status of the operating components of the conveyor belt, and to obtain the fault status decision result of the conveyor belt. The detection module is used to perform status detection on the conveyor belt based on the fault status decision results.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images