Belt conveyor carrier roller fault diagnosis method based on vibration and temperature data fusion

Abstract

The invention belongs to the technical field of equipment state monitoring and fault diagnosis, and relates to a belt conveyor carrier roller fault diagnosis method based on vibration and temperature data fusion, which comprises the following steps of: driving a plurality of wireless sensor nodes deployed along a line through a unified synchronous pulse to carry out synchronous data acquisition, and generating a multi-element sensing data flow; reconstructing a space-time vibration data graph and a space-time temperature distribution graph according to the node physical position information, and identifying and filtering common-mode vibration interference to obtain a differential vibration signal set; the differential signals are subjected to mode recognition, spatial propagation parameters of abnormal fluctuation are extracted, a fault physical source is reversely traced and locked accordingly, and positioning of the carrier roller is achieved; according to the method, the defects that traditional single-point monitoring is prone to being interfered by global vibration and propagation paths are difficult to identify are overcome, the fault identification accuracy and positioning precision are effectively improved, and misinformation and maintenance resource waste are reduced.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of equipment state monitoring and fault diagnosis, and relates to a belt conveyor roller fault diagnosis method based on vibration and temperature data fusion. BACKGROUND

[0002] As a key equipment for material transportation, the running state of a large number of rollers deployed along the line of a belt conveyor is directly related to the stability and efficiency of the entire system. The vibration signals generated in the early stage of roller failure are usually extremely weak and easily mixed with strong background noise generated by the driving system fluctuation and the body resonance, so that it is a severe challenge to effectively extract fault features from complex mixed signals and achieve early and accurate positioning.

[0003] At present, the industry mainly relies on two ways for monitoring the state of the roller: one is regular manual inspection, which relies on maintenance personnel to make experience judgments through listening and touching the temperature; and the other is to install vibration or temperature sensors at key rollers, to independently analyze the data based on single-point data, and to determine whether a specific roller is abnormal by detecting characteristic frequencies or threshold events. Some systems also try to use acoustic or optical fiber sensing technology to achieve regional coverage.

[0004] However, manual inspection cannot achieve continuous monitoring, has poor timeliness and strong subjectivity. The monitoring method based on single-point sensors regards each monitoring point as an isolated unit, and lacks consideration of the spatio-temporal correlation characteristics of vibration signals among roller clusters, which makes it difficult to effectively distinguish and filter out global common mode vibration interference caused by driving system start-stop and belt fluctuation, and prone to false positives. When the fault vibration propagates along the belt, single-point analysis cannot identify the position and propagation path of the vibration, often misjudging normal rollers with strong vibration response as fault sources, resulting in inaccurate positioning and waste of maintenance resources. SUMMARY

[0005] To solve the above problems, the application provides a belt conveyor roller fault diagnosis method based on vibration and temperature data fusion.

[0006] A belt conveyor roller fault diagnosis method based on vibration and temperature data fusion, comprising the following steps:

[0007] S1, based on a unified synchronous pulse, driving a plurality of wireless sensor nodes deployed along the line of the belt conveyor to perform synchronous data acquisition and labeling, and generating a multi-element sensor data stream;

[0008] S2, according to the physical location information of the sensor nodes, combining and reconstructing the data of a plurality of nodes in the multi-element sensor data stream that are spatially continuous, and constructing a spatio-temporal vibration data graph and a spatio-temporal temperature distribution graph;

[0009] S3, analyze the space-time vibration data graph, identify and filter out the common mode vibration signals of the sensor node synchronous response, and generate a differential vibration signal set;

[0010] S4, performing pattern recognition on the differential vibration signal set, extracting spatial mode parameters representing the propagation characteristics of abnormal fluctuations among sensor nodes;

[0011] S5, according to the spatial mode parameters, reverse tracing and locking the physical source of abnormal fluctuations, and outputting the fault source roller positioning result;

[0012] S6, combining the fault source roller positioning result, its corresponding temperature data and the fault nature attribution obtained based on the spatial mode parameter analysis, generating a cooperative maintenance task instruction.

[0013] Further schemes of the present application, generating a multi-element sensor data stream, comprising the following steps:

[0014] Receiving the unified synchronization pulse periodically broadcast by the clock synchronization unit;

[0015] Each wireless sensor node calibrates its internal data sampling clock according to the received unified synchronization pulse;

[0016] Each wireless sensor node collects vibration and temperature data, and adds a synchronization timestamp based on the calibrated clock to each data point to generate a multi-element sensor data stream.

[0017] Further schemes of the present application, constructing a space-time vibration data graph and a space-time temperature distribution graph, comprising the following steps:

[0018] Obtaining the multi-element sensor data stream and the pre-calibrated sensor node physical position information;

[0019] Based on the physical position information, logically dividing a plurality of wireless sensor nodes continuously in space into a dynamic diagnostic array;

[0020] Sorting the data of all nodes in the dynamic diagnostic array according to the physical position, and aligning them based on the synchronization timestamp, to construct a space-time vibration data graph and a space-time temperature distribution graph, respectively.

[0021] Further schemes of the present application, generating a differential vibration signal set, comprising the following steps:

[0022] Analyzing the space-time vibration data graph, identifying the vibration components with the same phase and amplitude change trend of the sensor nodes as the common mode vibration signals;

[0023] Subtracting the common mode vibration signals from the original vibration signals of each sensor node point by point to generate a differential vibration signal set.

[0024] The further scheme of the present application generates the spatial mode parameter, comprising the following steps:

[0025] detecting whether there is a center-symmetrical attenuation mode in the differential vibration signal set, the center-symmetrical attenuation mode being represented as that the energy of abnormal fluctuation reaches a peak at a certain center sensor node and is symmetrically attenuated to the adjacent nodes on both sides;

[0026] detecting whether there is an asymmetric propagation mode in the differential vibration signal set, the asymmetric propagation mode being represented as that the energy of abnormal fluctuation presents a gradient change along the running direction or the opposite direction of the belt, and the time sequence of the signal reaching each node has a sequence;

[0027] quantifying the energy distribution gradient, the propagation direction and / or the time sequence delay between nodes detected to generate the spatial mode parameter.

[0028] The further scheme of the present application outputs the fault source idler positioning result, comprising the following steps:

[0029] if the spatial mode parameter represents the center-symmetrical attenuation mode, the idler corresponding to the center sensor node where the energy peak value is located is determined as the fault source idler positioning result;

[0030] if the spatial mode parameter represents the asymmetric propagation mode, the starting excitation point of the signal is inversely calculated according to the propagation time sequence and the propagation speed of the signal between the sensor nodes, and the idler corresponding to the starting excitation point is determined as the fault source idler positioning result.

[0031] The further scheme of the present application generates the cooperative maintenance task instruction, comprising the following steps:

[0032] associating the fault source idler positioning result with the preset equipment asset information to determine the position of the fault source;

[0033] extracting the temperature data corresponding to the fault source idler positioning result from the space-time temperature distribution map, and determining whether there is a temperature anomaly;

[0034] according to the spatial mode parameter, attributing the center-symmetrical attenuation mode to the intrinsic fault and attributing the asymmetric propagation mode to the conductive fault;

[0035] based on the fault nature attribution obtained by vibration analysis and the determination result of the temperature anomaly, combining the accurate position, the fault nature attribution and the temperature state to generate the cooperative maintenance task instruction.

[0036] The further scheme of the present application determines the common mode vibration signal by calculating the average value of the vibration signals of all wireless sensor nodes in the dynamic diagnosis array at each synchronization timestamp.

[0037] In a further aspect of the present application, the time-space vibration data graph is a two-dimensional data matrix, representing a sequence of sensor nodes arranged according to physical positions and a sequence of synchronous time stamps, respectively, and a matrix element value is a vibration energy value of a corresponding sensor node at a corresponding time.

[0038] In a further aspect of the present application, the determination condition of the center-symmetry attenuation mode includes that a signal arrives at a center sensor node in the dynamic diagnosis array earlier than other nodes in the array, and the determination condition of the asymmetric propagation mode includes that a time sequence order of the signal arriving at each node is the same as a direction of an energy gradient.

[0039] In summary, the present application has the following beneficial technical effects:

[0040] 1. By constructing a dynamic diagnosis array and performing differential analysis on synchronous data streams of multiple sensors in the array, common mode vibration signals of all sensors in the array caused by factors such as drive system fluctuations and fuselage resonance can be effectively identified and filtered out, reducing background interference on local real fault signals and helping to reduce false positives caused by system-wide vibration.

[0041] 2. By analyzing the energy distribution gradient and propagation time sequence characteristics of the differential signal in the dynamic diagnosis array, it can be determined whether the fault is a center-symmetry mode originating from the monitoring point itself or a directional propagation mode transmitted from the outside. Based on this time-space causal relationship analysis, the initial excitation point of abnormal vibration can be traced back and locked, overcoming the problem of misjudging the transmission point as the fault source when only the vibration amplitude is used for judgment, and improving the accuracy of fault location.

[0042] 3. By mapping the identified spatial propagation mode to a corresponding physical fault type, such as an intrinsic fault or a conductive fault, and combining with device asset information, a structured collaborative maintenance task instruction can be automatically generated, indicating the location of the fault roller and preliminarily attributing the fault nature, providing more targeted decision basis for maintenance personnel and helping to optimize the maintenance process. BRIEF DESCRIPTION OF DRAWINGS

[0043] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. The drawings are used to provide a further understanding of the present application, and for those skilled in the art, other drawings can be obtained without creative labor based on these drawings.

[0044] Figure 1 The flowchart of the embodiment of the present application is disclosed.

[0045] Figure 2 The structural schematic diagram of the embodiment of the present application is disclosed. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some but not all of the embodiments of the present application. All other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0047] The above and other embodiments of the present application will be described in detail below with reference to the accompanying drawings. Figure 1 Figure 2 The preferred embodiments of the present application are described in detail.

[0048] The above and other embodiments of the present application will be described in detail below with reference to the accompanying drawings. Figure 1 The present application proposes a belt conveyor roller fault diagnosis method based on vibration and temperature data fusion, comprising the following steps:

[0049] S1, based on a unified synchronization pulse, driving a plurality of wireless sensor nodes arranged along the belt conveyor to perform synchronous data acquisition and marking, and generating a multi-element sensor data stream;

[0050] S2, according to the physical position information of the sensor nodes, combining and reconstructing the data of a plurality of nodes in the multi-element sensor data stream, and constructing a space-time vibration data graph and a space-time temperature distribution graph;

[0051] S3, analyzing the space-time vibration data graph, identifying and filtering out the common mode vibration signals of the sensor nodes in synchronization, and generating a differential vibration signal set;

[0052] S4, performing pattern recognition on the differential vibration signal set, and extracting spatial mode parameters representing the propagation characteristics of abnormal fluctuations among sensor nodes;

[0053] S5, according to the spatial mode parameters, reversely tracing and locking the physical source of abnormal fluctuations, and outputting a fault source roller positioning result;

[0054] S6, combining the fault source roller positioning result, its corresponding temperature data, and the fault nature attribution obtained based on the spatial mode parameter analysis, and generating a cooperative maintenance task instruction.

[0055] In one embodiment of the present application, step S1 comprises the following steps:

[0056] Receiving a unified synchronization pulse periodically broadcast by a clock synchronization unit; each wireless sensor node calibrates its internal data sampling clock according to the received unified synchronization pulse; each wireless sensor node collects vibration and temperature data, and adds a synchronization timestamp based on the calibrated clock to each data point to generate a multi-element sensor data stream.

[0057] ​Specifically, first activate the clock synchronization unit starts, it should be noted that the clock synchronization unit is usually integrated in the belt machine monitoring master station or set as a hardware device independent master node, its built-in high stability crystal oscillator, as the unified time reference of the whole monitoring system. Activation is completed by sending a start instruction to the unit, after starting, the clock synchronization unit will be through its wireless communication module, with a fixed broadcast cycle, for example, 10 times per second, to the belt machine along the line deployed all wireless sensor nodes broadcast uniform synchronization pulse. The synchronization pulse is a special wireless data packet, its data content itself is not important, but its physical time is precisely controlled.

[0058] Deployed on each wireless sensor node of the line roller continuously listens to this particular broadcast channel. When the node's wireless module receives the signal front of the synchronization pulse, its internal microcontroller will immediately capture this event. The microcontroller will compare the time of receiving the pulse with its own internal running free oscillation clock, calculate the deviation of the local clock from the master station time reference. According to this deviation, the microcontroller will calibrate and adjust its internal timer to ensure that its timing is aligned with the reference time of the clock synchronization unit. This calibration process is repeated periodically, so as to dynamically compensate for the clock drift of each node caused by temperature changes or individual differences in crystal oscillator, so that the local timestamp of all wireless sensor nodes achieves sub-millisecond level synchronization accuracy.

[0059] After completing the clock synchronization, each wireless sensor node starts to perform data acquisition task, the vibration sensor on each node will collect vibration data at a preset sampling rate, the vibration sensor is usually a three-axis acceleration sensor; at the same time, the temperature sensor integrated in the same node collects the temperature data of the roller bearing seat, such as thermistor or semiconductor temperature sensor, the key is that when the microcontroller completes the sampling of vibration and temperature sensor each time, obtains the data point, also reads the internal synchronized and calibrated clock value, and packs this time stamp, which is usually accurate to sub-millisecond, i.e. less than one thousandth of a second, with the vibration and temperature data just collected. This data packet constitutes the basic data unit containing spatial position, timestamp and multi-element sensing quantity. All wireless sensor nodes continuously generate such data units and send them to the data aggregation center through the wireless network, so as to collectively collect into data stream, which is the multi-element sensing data stream.

[0060] Among them, the multi-element sensing data stream is a data set, and its data structure is a series of time-ordered data records, each record contains the identifier of the wireless sensor node, the synchronization timestamp, and the vibration measurement value and temperature measurement value corresponding to the timestamp.

[0061] It should be noted that the wireless sensor node is a small electronic device deployed on the belt conveyor roller, and its function is to integrate a vibration sensor, a temperature sensor, a microcontroller, and a wireless communication module, which can autonomously collect data, mark a timestamp, and perform wireless transmission; the synchronization pulse is a wireless signal as a time marker, and its role is not to transmit complex data, but to provide a highly time-determined reference event that all nodes can receive. The synchronization timestamp is a high-resolution time marker, and the accuracy setting is based on the need for a time resolution much smaller than the propagation time of the vibration wave in order to effectively analyze the propagation time of the vibration wave between adjacent rollers. The speed of the vibration wave propagating on the belt is usually between 10 and 50 meters per second, and the distance between adjacent rollers is about 1.5 meters, so the propagation time is between 30 and 150 milliseconds. Therefore, using a sub-millisecond, i.e. less than 1 millisecond, timestamp accuracy can ensure that the timing difference is captured with a high enough resolution.

[0062] For example, assume that two hundred wireless sensor nodes numbered from 001 to 200 are deployed along the belt conveyor. The clock synchronization unit is activated to start broadcasting the synchronization pulse. At time point T0, the microcontroller of node 007 receives the synchronization pulse and calibrates its internal clock. At T0+0.01025 seconds, the three-axis vibration values collected by the vibration sensor of node 007 are (X: 0.5g, Y: 0.2g, Z: 1.1g), and the temperature value collected by the temperature sensor is 45.2 degrees Celsius. The microcontroller of node 007 immediately packages the timestamp 1678886400.01025, which is assumed to be a Unix timestamp plus a sub-millisecond decimal part, with node 007, and the vibration and temperature values into a record. At the same physical time T0+0.01025 seconds, node 008 also completes data collection and marks the same timestamp 1678886400.01025 for its data. All 200 nodes perform this operation, and all the data records generated by them are collected together to form a multi-sensor data stream. The data collected by any two nodes in the data stream at the same instant has a timestamp difference of less than 1 millisecond, directly verifying the setting of the sub-millisecond synchronization timestamp.

[0063] In one embodiment of the present application, step S2 includes the following steps:

[0064] Obtaining the multi-sensor data stream and the pre-calibrated physical location information of the sensor nodes;

[0065] Based on the physical location information, logically dividing a plurality of wireless sensor nodes that are spatially continuous into a dynamic diagnostic array; sorting the data of all nodes in the dynamic diagnostic array according to the physical location and aligning them based on the synchronization timestamp to construct a spatio-temporal vibration data graph and a spatio-temporal temperature distribution graph, respectively.

[0066] Specifically, the generated continuous multi-sensor data stream is acquired on the data processing server in the background. The server receives and collects all data packets sent by the wireless sensor nodes through the wireless network. Then, the physical location information of each sensor node is called up. This information has been recorded in the system database during the sensor deployment phase, which maps the identifier of each wireless sensor node to its accurate physical coordinates on the belt conveyor line. The spatially continuous multiple sensor nodes are combined into a dynamic diagnostic array. The software module in the background reads the physical location information of the sensor nodes and sorts all the nodes according to the order of the physical coordinates.

[0067] Then, the combination is created in a sliding window manner. For example, the first sensor node with the most leading coordinates is selected as the starting point, and then it is combined with the nine subsequent physically adjacent nodes to form a dynamic diagnostic array containing ten nodes. Then, the window slides one position forward, starting from the second sensor node, and then selects the next nine consecutive nodes to form the second dynamic diagnostic array. This process continues until the last sensor node is also included in an array, thereby achieving seamless coverage of the entire belt conveyor line. This combination is dynamic, and the size of the array and the sliding step can be adjusted according to the diagnostic requirements.

[0068] Finally, based on the constructed dynamic diagnosis array, a one-dimensional space-time vibration data graph is constructed to show the propagation process of vibration energy and temperature in time and space dimensions. For a certain dynamic diagnosis array, for example, the array composed of nodes 101 to 110, all the data collected by these ten nodes in the same time period is extracted from the multi-element sensor data stream. Then, the data is reconstructed. The one-dimensional space-time data graph is a term understood by those skilled in the art, which one-dimensional refers to the linear spatial dimension along the belt conveyor, and the data graph itself is a two-dimensional data structure containing a spatial axis and a time axis. Specifically, it can be constructed as two independent two-dimensional data matrices. In the one-dimensional space-time vibration waveform graph, one dimension represents space, i.e., the ten sensor nodes arranged in physical order from 101 to 110; the other dimension represents time, i.e., continuous and synchronized sub-millisecond synchronization timestamps. The value of each cell in the matrix represents the vibration energy of the sensor at that time, which can be obtained by calculating the amplitude or effective value of the original vibration signal. This two-dimensional data matrix is presented in the form of an image, where the X-axis is the sensor position, the Y-axis is the time, and the color or brightness represents the size of the vibration energy, and an intuitive space-time vibration data graph is obtained. This graph shows how the vibration event propagates from one roller to another in space and the complete process of the energy evolution over time; the structure of the space-time temperature distribution graph is similar to the vibration waveform graph, and the rows and columns of the matrix represent time and space position, respectively, but the value of each cell is the temperature measurement value of the corresponding node. The two graphs together form a complete space-time data graph, providing a basis for subsequent vibration analysis and temperature data fusion.

[0069] The sensor node physical position information is a data correspondence table, and its data structure is a set of key-value pairs that bind the digital or string identifier of each wireless sensor node to a numerical value representing its position on the line. This information is measured manually and entered into the system when deploying the sensor, and its setting is based on the engineering drawings of the site construction and the actual measurement results.

[0070] It should be noted that the dynamic diagnosis array is a logical grouping of sensors, which is an ordered list containing multiple continuously arranged sensor node identifiers. Its function is to capture the propagation characteristics of vibration in the local area as a whole analysis unit. The space-time vibration data graph is a visual representation of data, which function is to visually display the dynamic process of vibration energy along the space path of the belt conveyor changing over time. Its data structure can be understood as a two-dimensional array, where the row index corresponds to the time step, the column index corresponds to the sensor node at the spatial position, and the array element value is the vibration energy.

[0071] For example, assume that the background server obtains a multi-element sensor data stream containing data of two hundred nodes, and loads the pre-calibrated sensor node physical position information, which indicates that the nodes numbered 001 to 200 are sequentially arranged at intervals of 1.5 meters along the running direction of the belt conveyor. Assuming that the array size is set to 10, the first dynamic diagnosis array is created as: [001, 002, …, 009, 010]. Then, all data generated by the 10 nodes in the time period T0 to T0+1 second are extracted. At T0+0.501 second, the node numbered 005 records an impact with a vibration energy of 5 units, while at T0+0.531 second, the adjacent nodes numbered 004 and 006 record vibrations with energies of 3 units, respectively, and the energies of other nodes are close to 0. The data points corresponding to this series of space-time are plotted to form an image with the sensor number as the X-axis, time as the Y-axis, and energy as the color depth, i.e., a space-time vibration data graph is obtained. This graph shows the process of the vibration event generated from the node numbered 005 and propagating to both sides, directly verifying the function of the space-time vibration data graph in showing the propagation process of vibration energy in the time and space dimensions.

[0072] In one embodiment of the present application, step S3 comprises the following steps:

[0073] The space-time vibration data graph is analyzed to identify vibration components with the same phase and amplitude change trend of the sensor nodes as common mode vibration signals. The common mode vibration signals are subtracted from the original vibration signals of each sensor node point by point to generate a set of differential vibration signals.

[0074] Specifically, the space-time vibration data graph is first taken as input. This graph is essentially a two-dimensional data matrix, which records the vibration energy values of each sensor in the dynamic diagnosis array at consecutive time points.

[0075] The process of identifying the common mode vibration signal is to analyze the response of all sensors in the array point by point along the time dimension. In this embodiment, the correlation coefficient analysis method is adopted to extract the vibration signal sequence of all sensor nodes in the dynamic diagnostic array within the current time window, and calculate the Pearson correlation coefficient between the mean values of each node signal. If the correlation coefficient between more than 80% of the nodes in the array is greater than a predetermined consistency threshold, for example 0.85, or the instantaneous variance of all node signals is less than a predetermined background noise limit, it is determined that there is a consistency trend at the current time. At any synchronous timestamp, check the vibration signal values of all sensor nodes in the array and analyze whether the change trend is highly consistent. The specific implementation method is to calculate the average value of all sensor vibration signals in the array at each time point. If the signal values of most sensors are close to this average value or within a predetermined error range at this time point, and this collective behavior remains consistent for a continuous period of time, then the average signal changing with time is identified as the common mode vibration signal. The characteristic of this signal is that it is not caused by local anomalies of a single roller, but is applied synchronously to the entire area covered by the diagnostic array as a systematic disturbance, and its physical source is usually the start-stop fluctuation, speed change of the belt conveyor driving system, or the structural resonance of the entire steel structure body.

[0076] After identifying the common mode vibration signal, a filtering operation is performed to generate a differential vibration signal set. The most direct filtering method is signal subtraction. Specifically, the time series representing the common mode vibration signal calculated in the previous step, i.e. the average value of the array signal at each time point, is subtracted from the original vibration signal time series of each sensor in the diagnostic array point by point. After this operation, each sensor obtains a new signal time series. This new sequence reflects the deviation of the sensor from the background vibration of the entire array, i.e. its non-consistent abnormal fluctuation. The new signal sequences generated by all sensors in the array are collected together to form a differential vibration signal set. This set eliminates systematic and ubiquitous interference, making the weak signal with unique propagation characteristics caused by local roller failure stand out, providing a pure data basis for subsequent accurate fault source localization.

[0077] Differential vibration signal calculation formula:

[0078]

[0079] wherein, represents a data point in the differential vibration signal set, indicating the differential vibration signal value of the th sensor at time point . represents the original signal value read from the spatio-temporal vibration data graph, indicating the vibration signal value of the a sensor at a time point The value is set directly from the data collected by the sensor and processed by the previous step. representing the common mode vibration signal value at a time point The value is set by averaging the raw signal values of all sensors in the diagnostic array at the time point, i.e. where is the total number of sensors in the diagnostic array.

[0080] It is noted that the common mode vibration signal is a vibration component representing the synchronous response of all sensors in the diagnostic array, and its data structure is a time series array, and its function is to serve as a quantitative representation of systematic disturbance, which is filtered out from the raw signal. The differential vibration signal set is the remaining signal set after the raw signal is filtered out of the common mode vibration signal, and its data structure is a two-dimensional array, which has the same structure as the original time-space vibration data graph, but its values only contain non-uniform local abnormal fluctuations, and its function is to provide pure input without systematic common mode interference for subsequent fault pattern recognition.

[0081] For example, analyze the time-space vibration data graph generated by the dynamic diagnostic array composed of nodes 001 to 010. At time point T0+0.800 seconds, the vibration energy of all 10 nodes is found to rise synchronously by 0.2 units, which is identified as a common mode vibration signal caused by motor start-up, so the at this moment is calculated and set to 0.2. Looking back at the T0+0.501 second moment, the raw signal value of the 005 node is 5.0 units, while the signal values of the remaining 9 nodes are all around 0.1, so the common mode vibration signal at this moment is calculated to be about 0.59 units. The differential calculation is performed on the signal of the 005 node at this time, and the is 4.41. After performing this operation on all node data at all time points, the new data set obtained is the differential vibration signal set. In this new data set, the motor start-up disturbance at T0+0.800 seconds is completely eliminated, i.e. the differential signal values of all nodes are close to 0, while the abnormal impact of the 005 node at T0+0.501 seconds is retained and highlighted, directly verifying that this step can generate a differential vibration signal set containing only non-uniform abnormal fluctuations.

[0082] In one embodiment of the present application, step S4 includes the following steps:

[0083] detecting whether a central symmetric attenuation mode exists in the differential vibration signal set, the central symmetric attenuation mode being represented as the energy of the abnormal fluctuation reaching a peak at a certain central sensor node and symmetrically attenuating to the adjacent nodes on both sides; detecting whether an asymmetric propagation mode exists in the differential vibration signal set, the asymmetric propagation mode being represented as the energy of the abnormal fluctuation presenting a gradient change along the running direction or the opposite direction of the belt, and the time sequence of the signals reaching each node having a sequence; quantifying the energy distribution gradient, the propagation direction and / or the time sequence delay between the nodes to generate a spatial mode parameter.

[0084] Specifically, the differential vibration signal set is subjected to mode recognition, the energy distribution and the time difference of the signals between the multiple sensors constituting the dynamic diagnosis array are analyzed, and then the spatial mode parameter capable of representing the propagation characteristics of the abnormal fluctuation is extracted.

[0085] The mode recognition process includes two parallel detection logics:

[0086] The first logic is to detect the central symmetric attenuation mode. The differential vibration signal set is traversed to find whether there is a time point at which the abnormal signal energy of a certain central sensor node in the diagnosis array appears a significant peak. At the same time, it is checked whether the two sensors adjacent to the central node in the physical position also detect the signals at almost the same time or slightly delayed time points, but the energy values are significantly lower than that of the central node, and the energy presents a symmetric attenuation trend with the increase of the distance from the central node. In addition, the sub-millisecond synchronization time stamp is used to accurately compare the time of the signals reaching each node, and it is confirmed that the time of the signals reaching the central node is the earliest among all related nodes. If these conditions are met at the same time, the central symmetric propagation mode is identified.

[0087] The second logic is to detect the asymmetric propagation mode. The differential vibration signal set is also scanned, but another rule is sought. It checks whether there is an abnormal signal whose energy distribution in the diagnosis array presents a significant asymmetry. Specifically, the signal energy presents a gradient of sequentially increasing or sequentially decreasing along the running direction of the belt or the opposite direction of the running direction. The gradient is calculated by using the least square method for linear fitting: the physical position coordinates of the continuous nodes whose energy exceeds the trigger threshold are selected as the independent variable x, and the energy peak of the differential signal is selected as the dependent variable y, and a straight line is fitted where the slope k is the energy distribution gradient. If the absolute value of k is greater than the set gradient threshold, it is confirmed that there is a gradient change.

[0088] At the same time, the time stamps of the signal arriving at each node in the array are strictly analyzed. If it is found that the time sequence of the signal arriving at each node has a clear order, and this order coincides with the direction of the energy gradient, for example, the signal first arrives at the node with weaker energy and then arrives at the node with stronger energy, the mode of directional propagation is identified.

[0089] After identifying any of the above modes, quantitative extraction is performed to generate spatial mode parameters. For the central symmetric mode, the size of the energy peak, the central sensor node number where the peak is located, and the speed of energy attenuation to both sides are quantified as parameters. For the directional propagation mode, the propagation direction of the abnormal fluctuation, which is the same as or opposite to the running direction of the belt, is quantified; the propagation speed of the abnormal fluctuation, which is calculated by dividing the distance between adjacent nodes by the time difference of signal arrival, is quantified; and the energy change gradient of the abnormal fluctuation along the propagation path is quantified. These quantified values, including the energy distribution gradient, the propagation direction, and the time sequence delay between nodes, together constitute a set of structured spatial mode parameters. This set of parameters highly summarizes the physical propagation behavior of the abnormal event and provides the core basis for accurately tracing the source of the fault.

[0090] The spatial mode parameter is a data set used to describe the propagation characteristics of the abnormal fluctuation in space, and its data structure can be an object containing multiple key-value pairs, such as {“mode type”:“central symmetric”,“peak energy”:5.0,“central node”:“005”,“attenuation gradient”:0.4} or {“mode type”:“directional propagation”,“direction”:“in the direction of the belt”,“speed”:15.8 m / s,“energy gradient”:0.2}. Its function is to convert the identified complex spatiotemporal dynamic behavior into simple, quantified indicators that can be used for logical judgment.

[0091] For example, the analysis only contains the differential vibration signal set of the abnormal impact of node 005. It is detected that the differential signal energy of node 005 reaches a peak value of 4.41 units at T0+0.501 seconds. At the same time, at T0+0.510 seconds, the adjacent nodes 004 and 006 detect signals with an energy of 2.1 units, and at T0+0.519 seconds, the more distant nodes 003 and 007 detect signals with an energy of 0.8 units. The signal reaches node 005 earliest, and the energy symmetrically decays to both sides. Therefore, a central symmetric mode is identified, and a set of spatial mode parameters is extracted, the content of which is {“mode type”:“central symmetric”,“peak energy”:4.41,“central node”:“005”,“decay gradient”: (calculated)}, which directly verifies that the spatial mode parameters characterizing the propagation characteristics of abnormal fluctuations can be extracted. Alternatively, in another scenario, it is found that nodes 008, 009, and 010 successively appear at T0+0.600 seconds, T0+0.694 seconds, and T0+0.788 seconds with energies of 1.5, 2.5, and 3.5 units, respectively. The energy increases along the running direction of the belt, and the timing is clear. A“directional propagation”mode is identified, and the spatial mode parameters {“mode type”:“directional propagation”,“direction”:“in the belt direction”,“speed”:16 m / s,“energy gradient”:+1.0 unit / node} are extracted.

[0092] In one embodiment of the present application, step S5 includes the following steps:

[0093] If the spatial mode parameters represent a central symmetric decay mode, the idler roll corresponding to the central sensor node where the energy peak value is located is determined as the fault source idler roll positioning result; if the spatial mode parameters represent an asymmetric propagation mode, the starting excitation point of the signal is inversely calculated according to the propagation timing and propagation speed of the signal between the sensor nodes, and the idler roll corresponding to the starting excitation point is determined as the fault source idler roll positioning result.

[0094] Specifically, the physical source of the abnormal vibration is calculated by backtracking, and finally the source is locked to the roller, and the fault source roller positioning result is output. The execution logic of this process is completely determined by the identified mode type. If the mode represented by the received spatial mode parameter is central symmetric decay, that is, the mode type is central symmetric, the positioning logic of the fault source is very direct. In this mode, the energy of the abnormal vibration reaches a peak at a certain central sensor node, and then symmetrically decays to both sides, and the signal reaches the central node at the earliest. This physical phenomenon points to the fact that the excitation source of the abnormal vibration is at the central sensor node where the energy peak is located. Therefore, the central node number recorded in the spatial mode parameter is directly determined as the physical source of this abnormal event. Since each sensor node is one-to-one corresponding to a roller in physics, determining the sensor node as the source is equivalent to locking the roller monitored by the sensor node as the fault source.

[0095] If the mode represented by the received spatial mode parameter is directional propagation, that is, the mode type is directional propagation, the positioning logic is more complex and needs to be back calculated. In this mode, the abnormal signal propagates in a clear direction in the diagnostic array, and its energy may increase or decrease in this direction, but the time sequence of the signal reaching each sensor node has a clear order. This shows that the signal source is not necessarily at the node with the maximum energy, but at the starting point of the propagation path. Using the propagation direction recorded in the spatial mode parameter and the propagation time delay and propagation speed information between the sensors.

[0096] By analyzing the time at which the signal reaches the first triggered sensor in the array and the propagation speed of the signal, the initial excitation point of the signal can be back calculated. For example, if the signal propagates from upstream to downstream and is detected at node 008 first, the system will preliminarily determine that the fault source is located at node 008 or upstream of it. More accurately, if the calculated theoretical excitation point falls within the monitoring range of node 008 by combining the propagation speed and the triggering time of the first node in the array, node 008 will be locked as the physical source of the fault. The key of this method is that it does not rely on the size of the signal energy as a basis for judgment, but strictly follows the cause-and-effect relationship, that is, the origin of the signal must be the earliest point in time, even if the vibration energy of the excitation point is not the largest due to propagation effects, resonance amplification, etc. It will still be locked as the most likely physical source of the fault.

[0097] No matter which logic is used, a clear positioning conclusion pointing to a single roller will eventually be reached. This conclusion exists in the form of an identifier of the roller in the entire system or its precise physical coordinates, constitutes the fault source roller positioning result, and is output to the next processing link.

[0098] The fault source roller positioning result is information identifying a specific fault roller, and its data structure can be a string or a numerical value containing the roller device number or its precise mileage coordinate on the line. Its function is to clearly indicate the physical device that needs to be repaired or paid attention to, and is the core output of the entire diagnostic process.

[0099] For example, upon receiving the spatial mode parameters with the content {“mode type”: “central symmetry”, “peak energy”: 4.41, “central node”: “005”}. According to the logical rule, because the mode type is “central symmetry”, the value of “central node”, i.e. “005”, is immediately determined as the fault physical source. Since the 005 sensor node is installed on the 005 roller, the output fault source roller positioning result is “005 roller”. Alternatively, in another scenario, the spatial mode parameters with the content {“mode type”: “directional propagation”, “direction”: “forward belt direction”, “speed”: 16 m / s} are received, and it is known that the signal was first triggered by node 008 at T0+0.600 seconds. According to the “directional propagation” logic, the starting point of the propagation path, i.e. the roller corresponding to node 008 where the signal was first detected, is locked as the fault physical source. Even if the energy of node 010 is greater, “008 roller” is still output as the fault source roller positioning result. This directly verifies the ability to trace back the physical source of abnormal vibration and lock the source to the roller.

[0100] In one embodiment of the present application, step S6 includes the following steps:

[0101] The fault source roller positioning result is associated with the preset device asset information to determine the location of the fault source; temperature data corresponding to the fault source roller positioning result is extracted from the time-space temperature distribution map, and it is determined whether there is a temperature anomaly; according to the spatial mode parameters, the central symmetric decay mode is attributed to the intrinsic fault, and the asymmetric propagation mode is attributed to the conductive fault; based on the fault nature attribution obtained by vibration analysis and the determination result of temperature anomaly, the accurate position, fault nature attribution and temperature state are combined to generate a collaborative maintenance task instruction.

[0102] Specifically, first, the fault source roller positioning result output by the previous link is received, which clearly indicates which roller has a problem. In implementation, this abstract positioning result is associated with specific geographic or device information. The internal device asset database is queried, and the roller number in the fault source roller positioning result is used as an index to find more detailed description information of the roller, such as the name of the belt conveyor line it belongs to, the specific geographic location description, and the asset number in the device management system. This association operation converts a logical positioning point into a physical location that maintenance personnel can actually find.

[0103] Next, data fusion diagnosis is performed. This process contains two parts:

[0104] The first part: attribution judgment of the nature of the fault. If the identified spatial pattern is central symmetry, it indicates that the vibration energy spreads out from the center point, which is a typical point source excitation feature. This situation usually corresponds to structural or mechanical problems of the roller itself, such as pitting, spalling or fracture inside the bearing, or severe wear of the roller cylinder skin leading to imbalance. Therefore, the nature of the fault is attributed to the intrinsic fault. On the contrary, if the identified spatial pattern is directional propagation, it indicates that the vibration is transmitted along the belt on the line, and the energy accumulates or decays in one direction. This situation usually means that the problem is not on the roller with the strongest vibration, but is caused by some factor upstream and transmits the vibration. For example, a roller upstream is stuck, causing the belt to rub and jump on it, and this vibration wave will propagate downstream. Therefore, the nature of the fault is attributed to the conductive fault.

[0105] The second part: using the fault source roller positioning result, in the time-space temperature distribution map, the temperature time series corresponding to the roller is extracted. Then, the latest value or recent average value of the sequence is compared with the preset alarm threshold, such as 60 degrees Celsius, and the change rate compared with the historical normal operation state temperature baseline is analyzed. If the temperature is significantly higher and exceeds the threshold, it is determined that the temperature is abnormal.

[0106] Finally, the information obtained in the previous two steps is combined to form a structured generated collaborative maintenance task instruction, and the rules include:

[0107] If the diagnosis is: intrinsic fault with temperature anomaly, the diagnosis conclusion will point to the mechanical fault caused by friction aggravation. The diagnosis book will clearly state: "Located in the line of No. 1 transfer station, product number is roller T-005, diagnosed as intrinsic fault, vibration mode is central symmetric decay, and accompanied by significant temperature rise, highly suspected bearing severe wear, lubrication failure or seizure, suggest immediate shutdown for inspection and preparation for replacement."

[0108] If the diagnosis is: intrinsic fault but normal temperature, it may be a non-friction fault. The diagnosis book describes: "Product number is roller T-005, diagnosed as intrinsic fault, vibration mode is central symmetric decay, current temperature is normal, may be roller cylinder skin imbalance, cracking or subjected to instantaneous impact, suggest to upgrade the priority of inspection."

[0109] If the diagnosis is a conduction fault, regardless of its associated temperature state, the core is to indicate the vibration source. The diagnosis report will state: "Roller T-008 with production number is the source of conduction fault, vibration propagates along the belt running direction, please check whether T-008 roller is stuck, not rotating, etc. Abnormalities, and check its upstream area." At this time, the temperature data can be used as auxiliary information, such as "the current temperature of the source roller is normal".

[0110] The diagnosis report is generated in a format that is easy for humans to read and machines to parse, such as a JSON object or a formatted text report. Its content indicates which roller has what kind of problem. For example, the diagnosis report will clearly state: "Roller T-005 located at the first transfer station line, asset number T-005, is diagnosed as a body fault, suspected bearing damage", or "Roller T-008 has a conduction fault in its surrounding area, and the vibration source is traced back to the roller, suspected to be caused by roller jam, causing upstream vibration transmission". This diagnosis report can then be automatically pushed to the maintenance personnel's mobile terminal, work order management system or central control room monitor through the system interface, providing direct, specific and decision-making reference value guidance information for subsequent accurate maintenance activities.

[0111] It should be noted that the generated collaborative maintenance task instruction is a comprehensive fault report and maintenance instruction, which is a text or data object containing multiple fields, including but not limited to detailed description of fault location, device asset number, determined fault nature, and recommended maintenance level or operation. It serves as a bridge connecting the intelligent diagnosis system and manual maintenance activities, converting abstract analysis results into specific and executable work instructions.

[0112] For example, the received fault source roller positioning result is "Roller 005". Query the database to associate "Roller 005" with its detailed information "A-line belt conveyor, K1+7.5 meters, asset number T-005". At the same time, it is known that the spatial pattern relied on for this positioning is center symmetry. According to the rules, the fault nature is attributed to a body fault. Finally, these information is integrated to generate a collaborative maintenance task instruction, the content is: "Emergency maintenance task - fault location: A-line belt conveyor K1+7.5 meters. Fault equipment: Roller T-005. Diagnosis conclusion: Body fault, vibration mode is center-symmetric decay, highly suspected bearing damage, please check and prepare for replacement immediately".

[0113] The diagnosis is pushed to the repair worker's mobile APP, or in another scenario, the positioning result is "008 idler", and the spatial mode is "directional propagation", and the diagnosis is "attention task - fault area: A line belt conveyor K1+12 meters. Suspected source: idler T-008. Diagnosis conclusion: conduction fault, vibration propagates along the belt running direction, please check whether the T-008 idler is stuck, not rotating and other abnormalities, and check the upstream area"; which directly verifies that a generated collaborative maintenance task instruction containing the accurate location of the fault source and the fault nature attribution can be generated and pushed.

[0114] Referring to the accompanying drawings Figure 2 The application also proposes a belt conveyor idler fault diagnosis system based on vibration and temperature data fusion, including the following modules:

[0115] The data stream generation module is based on a unified synchronization pulse to drive multiple wireless sensor nodes deployed along the belt conveyor to perform synchronous data acquisition and labeling, generating multi-element sensor data streams;

[0116] The waveform diagram construction module combines and reconstructs the data of multiple nodes in the multi-element sensor data stream according to the physical location information of the sensor nodes, and constructs a spatio-temporal vibration data graph and a spatio-temporal temperature distribution graph;

[0117] The signal set generation module is used to analyze the spatio-temporal vibration data graph, identify and filter out common mode vibration signals that are synchronously responded by the sensor nodes, and generate a differential vibration signal set;

[0118] The parameter extraction module is used to perform pattern recognition on the differential vibration signal set to extract spatial mode parameters representing the propagation characteristics of abnormal fluctuations among sensor nodes;

[0119] The fault source positioning module reversely traces and locks the physical source of abnormal fluctuations according to the spatial mode parameters, and outputs the fault source idler positioning result;

[0120] The diagnosis generation module combines the fault source idler positioning result, its corresponding temperature data, and the fault nature attribution obtained based on the spatial mode parameter analysis to generate a collaborative maintenance task instruction.

[0121] The various modules can be implemented in whole or in part by software, hardware and their combination, supporting hardware forms embedded in or independent of the processor in the computer device, and also supporting software forms stored in the memory in the computer device, facilitating the processor to call and execute the operations corresponding to the above various modules.

[0122] The above examples are only used to illustrate the technical solutions of the present application, and are not intended to limit the present application; although the present application is described in detail with reference to the foregoing examples, those skilled in the art should understand that the technical solutions recorded in the foregoing examples can be modified, or some technical features can be replaced by equivalent features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A method for fault diagnosis of belt conveyor idler rollers based on vibration and temperature data fusion, characterized in that, Includes the following steps: S1. Based on a unified synchronization pulse, drive multiple wireless sensor nodes deployed along the belt conveyor to perform synchronous data acquisition and labeling, generating a multi-sensor data stream; S2. Based on the physical location information of the sensor nodes, combine and reconstruct the data of multiple spatially continuous nodes in the multi-sensor data stream to construct a spatiotemporal vibration data map and a spatiotemporal temperature distribution map. S3. Analyze the spatiotemporal vibration data graph, identify and filter out the common-mode vibration signals of the synchronous response of sensor nodes, and generate a differential vibration signal set; S4. Perform pattern recognition on the differential vibration signal set and extract spatial pattern parameters that characterize the propagation characteristics of abnormal fluctuations between sensor nodes. S5. Based on the spatial pattern parameters, trace back and locate the physical source of the abnormal fluctuations, and output the fault source idler roller positioning result. S6. Based on the fault source idler location result, its corresponding temperature data, and the fault nature attribute obtained from spatial pattern parameter analysis, a collaborative maintenance task instruction is generated.

2. The method for fault diagnosis of belt conveyor idler rollers based on vibration and temperature data fusion according to claim 1, characterized in that, Generating a multi-sensor data stream includes the following steps: Receive unified synchronization pulses periodically broadcast by the clock synchronization unit; Each wireless sensor node calibrates its internal data sampling clock based on the received unified synchronization pulse; Each wireless sensor node collects vibration and temperature data, and adds a synchronization timestamp based on a calibrated clock to each data point to generate a multi-sensor data stream.

3. The method for fault diagnosis of belt conveyor idler rollers based on vibration and temperature data fusion according to claim 1, characterized in that, Constructing spatiotemporal vibration data maps and spatiotemporal temperature distribution maps includes the following steps: Acquire multi-sensor data streams and pre-calibrated physical location information of sensor nodes; Based on physical location information, multiple spatially continuous wireless sensor nodes are logically divided into a dynamic diagnostic array. The data of all nodes in the dynamic diagnostic array are sorted by physical location and aligned with the synchronization timestamp to construct spatiotemporal vibration data map and spatiotemporal temperature distribution map respectively.

4. The method for fault diagnosis of belt conveyor idler rollers based on vibration and temperature data fusion according to claim 1, characterized in that, Generating a differential vibration signal set includes the following steps: Analyze the spatiotemporal vibration data graph to identify vibration components of sensor nodes that have the same trend in phase and amplitude variation, and use them as common-mode vibration signals; The common-mode vibration signal is subtracted point by point from the original vibration signal of each sensor node to generate a differential vibration signal set.

5. The method for fault diagnosis of belt conveyor idler rollers based on vibration and temperature data fusion according to claim 1, characterized in that, Generating spatial model parameters includes the following steps: The detection function checks whether a centrally symmetric decay mode exists in the differential vibration signal set. The centrally symmetric decay mode is characterized by the energy of abnormal fluctuations reaching a peak at a certain central sensor node and decaying symmetrically towards the adjacent nodes on both sides. The detection of whether there is an asymmetric propagation mode in the differential vibration signal set is characterized by the energy of abnormal fluctuations changing in a gradient along the direction of belt running or in the opposite direction, and the timing of the signals arriving at each node having a certain order. The detected energy distribution gradient, propagation direction, and / or temporal delay between nodes are quantized to generate spatial model parameters.

6. The method for fault diagnosis of belt conveyor idler rollers based on vibration and temperature data fusion according to claim 5, characterized in that, Outputting the fault source idler roller positioning results includes the following steps: If the spatial mode parameters are characterized as a center-symmetric decay mode, then the idler corresponding to the central sensor node where the energy peak is located is determined as the fault source idler location result. If the spatial mode parameters represent an asymmetric propagation mode, then based on the propagation timing and speed of the signal between each sensor node, the starting excitation point of the signal is calculated in reverse, and the idler corresponding to the starting excitation point is determined as the fault source idler positioning result.

7. The method for fault diagnosis of belt conveyor idler rollers based on vibration and temperature data fusion according to claim 1, characterized in that, Generating collaborative maintenance task instructions includes the following steps: The location of the fault source idler is determined by associating the fault source idler location result with the preset equipment asset information. Extract temperature data corresponding to the positioning results of the fault source idler roller from the spatiotemporal temperature distribution map, and determine whether there is a temperature anomaly. Based on the spatial model parameters, the centrally symmetric decay mode is classified as a physical fault, and the asymmetric propagation mode is classified as a conductive fault. Based on the fault nature attribution and temperature anomaly determination results obtained from vibration analysis, the precise location, fault nature attribution, and temperature status are combined to generate collaborative maintenance task instructions.

8. The method for fault diagnosis of belt conveyor idler rollers based on vibration and temperature data fusion according to claim 4, characterized in that, The common-mode vibration signal is determined by calculating the average vibration signal of all wireless sensor nodes in the dynamic diagnostic array at each synchronization timestamp.

9. The method for fault diagnosis of belt conveyor idler rollers based on vibration and temperature data fusion according to claim 1, characterized in that, The spatiotemporal vibration data map is a two-dimensional data matrix, which represents the sequence of sensor nodes ordered by physical location and the sequence of synchronization timestamps, respectively. The matrix element values ​​are the vibration energy values ​​of the corresponding sensor nodes at the corresponding time.

10. A method for fault diagnosis of belt conveyor idler rollers based on vibration and temperature data fusion according to claim 1, characterized in that, The criteria for determining the centrally symmetric attenuation mode include that the signal arrives at the central sensor node in the dynamic diagnostic array earlier than other nodes in the array; the criteria for determining the asymmetric propagation mode include that the timing sequence of the signal arriving at each node is the same as the direction of the energy gradient.

Images