A roller temperature and load monitoring method and roller conveying system

Abstract

The present invention discloses a roller temperature and load monitoring method and a roller conveying system, wherein the roller conveying system comprises a roller, a data processor, a temperature and load multimodal data fusion prediction model, a temperature sensor and a load sensor arranged on the roller; the data processor receives the temperature information and load information collected by the sensor; the temperature and load multimodal data fusion prediction model is connected to the data processor in communication, and the prediction model estimates the evaluation scores of multiple modal variables related to the temperature information and the load information, wherein the modal variables include a real-time temperature value, a temperature change rate, a maximum continuous operating temperature value, a peak load, a load change rate, and a maximum continuous operating load; according to the evaluation scores of each modal variable and in combination with the weights of each modal variable, the current state comprehensive score of the roller is calculated. The present invention realizes AI monitoring of the working state of the roller by establishing a multimodal data fusion prediction model based on temperature and load.

Description

Technical Field

[0001] The present invention relates to the field of AI monitoring, and in particular to a roller temperature and load monitoring method and a roller conveying system. Background Art

[0002] In the field of industrial automation, roller conveyor systems are widely used due to their simple structure and convenient operation. In order to improve the service life and maintenance efficiency of rollers, it is necessary to monitor their working temperature and load. Therefore, smart rollers with pressure sensors and temperature sensors set in rollers are provided, such as the Chinese utility model patent with authorization announcement number CN210162688U.

[0003] However, existing monitoring systems still have deficiencies in temperature monitoring and load monitoring. For example, they monitor temperature and load independently and rely on simple threshold settings to trigger alarms: if the temperature exceeds the temperature threshold, or the load exceeds the load threshold, the monitoring alarm is triggered.

[0004] This monitoring and analysis method lacks intelligent analysis capabilities and ignores the multi-dimensional and multi-modal impact of temperature and load on the performance of the roller conveyor system, thus limiting the accuracy and preventive nature of maintenance work.

[0005] The disclosure of the above background technology content is only used to assist in understanding the concept and technical solution of the present application. It does not necessarily belong to the prior art of the present application, nor does it necessarily provide technical guidance. In the absence of clear evidence that the above content has been disclosed before the filing date of the present application, the above background technology should not be used to evaluate the novelty and creativity of the present application. Summary of the invention

[0006] The purpose of the present invention is to establish a multimodal data fusion prediction model based on temperature and load to intelligently monitor the working state of the roller.

[0007] To achieve the above object, the technical solution adopted by the present invention is as follows:

[0008] A roller temperature and load monitoring method comprises the following steps:

[0009] The temperature information and load information of the roller are collected by sensors respectively;

[0010] The collected temperature information and load information are sent to a data processor, which is in communication connection with a pre-trained temperature and load multimodal data fusion prediction model;

[0011] The prediction model estimates evaluation scores of a plurality of preset modal variables related to temperature information and load information;

[0012] According to the evaluation scores of each modal variable and the weights of each modal variable, the comprehensive score of the current state of the roller is calculated;

[0013] The temperature and load multimodal data fusion prediction model is trained in the following way:

[0014] Collect sample data: simulate different working conditions for the roller, and use sensors to collect the temperature sample information and load sample information of the roller respectively, and record the correlation information between temperature, load and corresponding working conditions;

[0015] Extracting one or more pieces of temperature characteristic information from the collected temperature sample information, and extracting one or more pieces of load characteristic information from the collected load sample information;

[0016] The extracted temperature characteristic information and load characteristic information are spliced ​​and fused to form a comprehensive characteristic vector of multi-modal variables, and the weight of each modal variable is determined;

[0017] Marking the comprehensive feature vector according to the corresponding associated working condition information;

[0018] The support vector machine algorithm is used to train the labeled comprehensive feature vector, with the training purpose of minimizing the error between the predicted state comprehensive score of the sample data and the labeled label, to generate the temperature and load multimodal data fusion prediction model.

[0019] Further, based on any one of the above-mentioned technical solutions or a combination of multiple technical solutions, the current comprehensive score of the roller state is calculated by the following formula:

[0020] S = w 1× f 1+ w 2× f 2+…+ w n × f n ,in, S is the comprehensive score of status, f 1 is the evaluation score of the first modal variable estimated by the prediction model, w 1 is the weight of the first modal variable, f 2 is the evaluation score of the second modal variable estimated by the prediction model, w 2 is the weight of the second modal variable, f n Estimate the evaluation score of the nth modal variable for the prediction model, w n is the weight of the nth modal variable.

[0021] Furthermore, according to any one of the technical solutions or a combination of multiple technical solutions mentioned above, if the calculated comprehensive status score is lower than a preset alarm threshold, an early warning prompt is given that the current working condition of the roller is not good.

[0022] Further, based on any one of the above-mentioned technical solutions or a combination of multiple technical solutions, the training objective of the support vector machine algorithm is to minimize the following function:

[0023] ,

[0024] The constraints are y i (w× x i + b )≥1- ξ i ,in ξ i ≥0;

[0025] Among them, w is the weight vector, b is the deviation term, C A penalty factor to balance the relationship between model complexity and classification error. ξ i For the i The slack variable for samples, M is the number of samples to be predicted, x i For the i The comprehensive feature vector of samples, y i is the label of the th sample.

[0026] Further, based on any one of the above-mentioned technical solutions or a combination of multiple technical solutions, the temperature characteristic information extracted from the collected temperature sample information includes one or more of the normalized real-time temperature value, the temperature change rate, and the maximum continuous operating temperature value;

[0027] And / or, the load characteristic information extracted from the collected load sample information includes one or more of peak load, load change rate, maximum continuous working load, load fluctuation amplitude, and load fluctuation frequency.

[0028] Further, based on any one of the above-mentioned technical solutions or a combination of multiple technical solutions, before extracting the temperature characteristic information and / or the load characteristic information, it also includes:

[0029] The temperature sample information and / or load sample information collected by the sensor is preprocessed, including one or more operations of data cleaning, normalization, and packaging into a preset standard data format.

[0030] Further, based on any one of the above-mentioned technical solutions or a combination of multiple technical solutions, the weight of each modal variable is determined by the following method:

[0031] The weights of each modal variable are determined by the recursive feature elimination method of support vector machine;

[0032] Alternatively, the weight of each modal variable is set through empirical values ​​or experiments.

[0033] Further, based on any one of the above-mentioned technical solutions or a combination of multiple technical solutions, the comprehensive feature vector is marked according to the working condition information in the following manner:

[0034] The factors for presetting the working conditions include one or more of the ambient temperature, the load pressure, the ambient humidity, the continuous working time, and the roller speed;

[0035] Different test conditions are set, and the value of at least one factor in different test conditions is different;

[0036] After running various test conditions on the roller, obtaining roller performance results corresponding to the various test conditions;

[0037] According to the roller performance results, the state scores are evaluated for various test conditions. The better the roller performance results, the higher the evaluation value of the state score. The evaluation value of the state score is associated with the test condition.

[0038] In the stage of collecting sample data, different working conditions are simulated for the roller and executed as various test conditions set for operation, and then the evaluation value of its associated state score is used as the label value for marking.

[0039] Further, based on any one of the above-mentioned technical solutions or a combination of multiple technical solutions, the roller temperature and load monitoring method provided by the present invention further includes adjusting the equipment where the roller is located according to the calculated comprehensive score of the current state of the roller:

[0040] Multiple levels are pre-divided, and each level corresponds to a different range of comprehensive status scores;

[0041] If the calculated comprehensive score of the current state of the roller belongs to the first level, the current motion parameters of the device where the roller is located are maintained;

[0042] and / or, if the calculated comprehensive score of the current state of the roller belongs to the second level, reducing the operating speed of the equipment where the roller is located;

[0043] and / or, if it is calculated that the current comprehensive score of the roller state belongs to the third level, the operating speed of the equipment where the roller is located is reduced and the refrigeration device is started;

[0044] And / or, if it is calculated that the current comprehensive score of the roller status belongs to the fourth level, the device where the roller is located is forcibly shut down.

[0045] According to another aspect of the present invention, there is provided a roller conveying system, comprising a roller, a data processor, a temperature and load multimodal data fusion prediction model, a temperature sensor and a load sensor disposed on the roller, the load sensor being configured to detect the weight of a material passing through the roller;

[0046] The temperature sensor sends the collected temperature information to the data processor, and the load sensor sends the collected load information to the data processor;

[0047] The temperature and load multimodal data fusion prediction model is communicatively connected to the data processor, and the prediction model estimates the evaluation scores of a plurality of preset modal variables related to the temperature information and the load information, wherein the modal variables include a plurality of the real-time temperature value, the temperature change rate, the maximum continuous operating temperature value, the peak load, the load change rate, the maximum continuous operating load, the load fluctuation amplitude, and the load fluctuation frequency;

[0048] According to the evaluation scores of each modal variable and the weights of each modal variable, the comprehensive score of the current state of the roller is calculated.

[0049] Further, based on any one of the technical solutions or a combination of multiple technical solutions mentioned above, the temperature sensor is arranged inside the roller shaft of the roller; and / or the load sensor is arranged on one side of the roller shaft.

[0050] Further, based on any one of the above technical solutions or a combination of multiple technical solutions, the roller conveying system has multiple rollers equipped with temperature sensors and load sensors;

[0051] An average value is calculated for the comprehensive scores of the states of the multiple rollers, and the average value of the comprehensive scores is used as the monitoring result data.

[0052] Further, based on any one of the above-mentioned technical solutions or a combination of multiple technical solutions, the roller conveyor system provided by the present invention further includes a communication module and a data receiving and processing module, and the data receiving and processing module is a host computer or a cloud server;

[0053] The data receiving and processing module is configured to receive the temperature information and load information sent by the data processor through the communication module, pre-process the received information, extract the temperature characteristic information and the load characteristic information, and then input the extracted characteristic information into the pre-trained temperature and load multimodal data fusion prediction model;

[0054] The temperature and load multimodal data fusion prediction model is integrated in the data processor, or is set in a cloud server.

[0055] Furthermore, based on any one of the technical solutions or a combination of multiple technical solutions mentioned above, the data processor determines whether the comprehensive status score is lower than a preset threshold. If so, the data processor triggers the early warning device of the roller conveyor system or shuts down the roller conveyor system.

[0056] Further, based on any one of the above-mentioned technical solutions or a combination of multiple technical solutions, the temperature and load multimodal data fusion prediction model is trained in the following manner:

[0057] Collect sample data: simulate different working conditions for the roller, and use sensors to collect the temperature sample information and load sample information of the roller respectively, and record the correlation information between temperature, load and corresponding working conditions;

[0058] Extracting one or more pieces of temperature characteristic information from the collected temperature sample information, and extracting one or more pieces of load characteristic information from the collected load sample information;

[0059] The extracted temperature characteristic information and load characteristic information are spliced ​​and fused to form a comprehensive characteristic vector of multi-modal variables, and the weight of each modal variable is determined;

[0060] Marking the comprehensive feature vector according to the corresponding associated working condition information;

[0061] The support vector machine algorithm is used to train the labeled comprehensive feature vector, with the training purpose of minimizing the error between the predicted state comprehensive score of the sample data and the labeled label, to generate the temperature and load multimodal data fusion prediction model.

[0062] The beneficial effects brought by the technical solution provided by the present invention are as follows:

[0063] a. A multimodal data fusion prediction model is established based on temperature and load to intelligently monitor the working status of the roller. By introducing machine learning algorithms for data analysis, the intelligence of data processing is improved, and intelligent analysis and prediction of the equipment operating status are realized, which helps to improve the safety, reliability and efficiency of equipment operation and effectively extend the service life of the equipment;

[0064] b. Once the AI ​​model identifies a potential problem, it automatically triggers an early warning mechanism and formulates a dynamic maintenance strategy based on the prediction results to ensure that the equipment operates within a reasonable working range. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0066] Figure 1 A schematic flow chart of a roller temperature and load monitoring method provided for an exemplary embodiment of the present invention;

[0067] Figure 2 A flow chart of a method for training a temperature and load multimodal data fusion prediction model provided by an exemplary embodiment of the present invention;

[0068] Figure 3 A schematic diagram of a process for labeling sample data provided for an exemplary embodiment of the present invention;

[0069] Figure 4 A structural block diagram of a roller conveyor system provided for an exemplary embodiment of the present invention. DETAILED DESCRIPTION

[0070] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0071] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, device, product or equipment that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or equipment.

[0072] In one embodiment of the present invention, a roller temperature and load monitoring method is provided. Figure 1 , the monitoring method comprises the following steps:

[0073] The first step is to use sensors to collect the temperature information and load information of the roller respectively;

[0074] Specifically, the temperature sensor is arranged inside the roller shaft of the roller, and the load sensor is arranged on one side of the roller shaft. By embedding the temperature sensor inside the roller shaft, the working temperature of the roller shaft can be directly monitored, thereby improving the accuracy and timeliness of temperature monitoring. At the same time, by installing the load sensor at the end of the roller shaft, the load status of the equipment can be monitored in real time to ensure that the equipment operates within a reasonable load range and avoid damage caused by overload.

[0075] The temperature sensor can specifically adopt PT100 high-precision resistance temperature detector. The temperature sensor converts the temperature signal into an electrical signal. These electrical signals are usually analog signals, such as voltage or current, and then converted into digital signals through an analog-to-digital converter (ADC); the load sensor can adopt HBM C9C piezoelectric load sensor, which is close to the material conveying path to monitor the weight of the material passing through the roller. The load sensor converts the pressure change it senses into an electrical signal, which also needs to be converted into a digital signal through ADC. The digital signal can be directly read by the data processor and further processed.

[0076] The second step is to send the collected temperature information and load information to the data processor, which is connected to the pre-trained temperature and load multimodal data fusion prediction model; the data processor can use an STM32F407VG microcontroller, which can be set near the sensor, and communicate with the temperature sensor and load sensor by wired or wireless means, and is responsible for receiving and processing data from the temperature sensor and load sensor; the data processor is responsible for receiving signals from the temperature sensor and the load sensor, and preprocessing them. Preprocessing includes signal filtering (noise removal), signal amplification, and signal bias adjustment. The preprocessed signal is further converted into a digital signal and stored in the memory of the data processor; and the data processor packages the preprocessed digital signal to form a standard data packet format, which is convenient for subsequent analysis and transmission, and the packaged data is sent to the wireless communication module through the serial communication interface.

[0077] In one embodiment, the temperature and load multimodal data fusion prediction model is integrated in the data processor; in another embodiment, the temperature and load multimodal data fusion prediction model is set in the cloud server.

[0078] Step 3: The prediction model estimates the evaluation scores of multiple modal variables related to temperature information and load information.

[0079] Step 4: Calculate the comprehensive score of the roller's current state based on the evaluation scores of each modal variable and the weights of each modal variable;

[0080] The specific formula for calculating the comprehensive score of the roller's current status is as follows:

[0081] S = w 1× f 1+ w 2× f 2+…+ w n × f n ,in, S is the comprehensive score of status, f 1 is the evaluation score of the first modal variable estimated by the prediction model, w 1 is the weight of the first modal variable, f 2 is the evaluation score of the second modal variable estimated by the prediction model, w 2 is the weight of the second modal variable, f n Estimate the evaluation score of the nth modal variable for the prediction model, w n is the weight of the nth modal variable.

[0082] Step 5. Control the operation of the equipment where the roller is located based on the calculated comprehensive status score. For example, if the calculated comprehensive status score is lower than the preset alarm threshold, an early warning will be issued that the roller is currently in poor working condition, indicating that the equipment is in poor condition or is about to be scrapped. The early warning information will be promptly notified to maintenance personnel via email, text message or application push, ensuring that the equipment operates within a reasonable working range and promptly warning of potential problems.

[0083] In addition, the system will formulate dynamic maintenance strategies based on the prediction results to ensure that the equipment operates within a reasonable working range and promptly warn of potential problems. In a specific embodiment, multiple levels are pre-divided, and each level corresponds to a different range of comprehensive status scores, such as the first level, the second level, the third level, and the fourth level with scores from high to low;

[0084] If the calculated comprehensive score of the roller's current state belongs to the first level, the current motion parameters of the device where the roller is located are maintained; if the calculated comprehensive score of the roller's current state belongs to the second level, the operating speed of the device where the roller is located is reduced; if the calculated comprehensive score of the roller's current state belongs to the third level, the operating speed of the device where the roller is located is reduced and the refrigeration device is started; if the calculated comprehensive score of the roller's current state belongs to the fourth level, the device where the roller is located is forcibly shut down.

[0085] The core of this embodiment is to train the temperature and load multi-modal data fusion prediction model used to evaluate each modal variable in the third step. The specific training steps are as follows: Figure 2 As shown:

[0086] Step ①, collect sample data (including samples and labels under normal operation and various fault conditions): simulate different working conditions for the roller, and use sensors to collect temperature sample information and load sample information of the roller respectively, and record the correlation information between temperature, load and corresponding working conditions. The working conditions are used to determine the labels, see the details below;

[0087] Step ②, pre-processing the temperature sample information and load sample information collected by the sensor, such as data cleaning, normalization, and packaging into a preset standard data format;

[0088] Step ③, extracting one or more temperature characteristic information from the collected temperature sample information, such as the normalized real-time temperature value, temperature change rate, and maximum continuous working temperature value; extracting one or more load characteristic information from the collected load sample information, such as peak load, load change rate, maximum continuous working load, load fluctuation amplitude, and load fluctuation frequency;

[0089] Step ④: splice the extracted temperature characteristic information and load characteristic information to form a comprehensive characteristic vector of the multimodal variable; in a specific embodiment, the comprehensive characteristic vector of the multimodal variable is formed by fusion as follows:

[0090] f combined =[ T norm , Δ T , T max_work , L peak , Δ L , L max_work ],in, f combined is the comprehensive feature vector of multimodal variables, T norm is the normalized temperature value, Δ T is the temperature change rate, T max_work is the maximum continuous operating temperature, L peak is the peak load, Δ L is the load change rate, L max_work is the maximum continuous workload.

[0091] Among them, the temperature characteristic information is as follows: the normalized temperature value T norm Calculated by the following formula:

[0092] ,in, T is the real-time temperature value, T min is the minimum temperature value within the most recent preset time period. T max is the maximum temperature value within the most recent preset time. Normalization ensures that the data from different sensors are in the same dimension and value range. The temperature change rate Δ T The calculation formula is: ,in, T ( t )for t The temperature value at the moment, T ( t- Δ t ) is the temperature value at the previous moment, Δ t is the time difference between the previous and the next moments; the maximum continuous operating temperature T max_work It is the maximum temperature value in the historical temperature data from power-on to the present.

[0093] Load characteristic information is as follows: Peak load L peak is the maximum load value within the most recent preset time period, and the load change rate Δ L The calculation formula is as follows: ,in, L ( t )for t The load value at the moment, L ( t- Δ t ) is the load value at the previous moment, Δ t is the time difference between the previous and the next moments; the maximum continuous workload L max_work is the maximum load value in the historical load data from power-on to the present.

[0094] Step ⑤: Determine the weight of each modal variable by the recursive feature elimination method of the support vector machine (SVM) to ensure the importance of each feature in evaluating the device status. It should be noted that the present invention is not limited to setting the weight of each modal variable according to empirical values ​​or experiments.

[0095] Step ⑥: Mark the comprehensive feature vector according to the corresponding associated working condition information; Figure 3As shown: pre-set working condition factors, such as one or more of ambient temperature, load pressure, ambient humidity, continuous working time, and roller speed; set different test conditions, and the value of at least one factor in different test conditions is different; after running various test conditions on the roller, obtain roller performance results corresponding to various test conditions; evaluate the state scores of various test conditions according to the roller performance results, the better the roller performance result, the higher the evaluation value of the state score, and associate the evaluation value of the state score with the test condition; in the stage of collecting sample data, simulate different working conditions for the roller to be executed as the various test conditions set for running, and then use the evaluation value of the associated state score as the label value for marking.

[0096] Step 7: Use the support vector machine algorithm to train the labeled comprehensive feature vector, with the training purpose of minimizing the error between the predicted state comprehensive score of the sample data and the labeled label, to generate the temperature and load multimodal data fusion prediction model.

[0097] Specifically, the training objective of the support vector machine algorithm is to minimize the following function:

[0098] ,

[0099] The constraints are y i (w× x i + b )≥1- ξ i ,in ξ i ≥0;

[0100] Among them, w is the weight vector, b is the deviation term, C A penalty factor to balance the relationship between model complexity and classification error. ξ i For the i The slack variable for samples, M is the number of samples to be predicted, x i For the i The comprehensive feature vector of samples, y i is the label of the th sample.

[0101] In one embodiment of the present invention, a roller conveying system is provided, such as Figure 4As shown, the roller conveying system includes a roller, a data processor, a temperature and load multimodal data fusion prediction model, a temperature sensor and a load sensor arranged on the roller, and the load sensor is configured to detect the weight of the material passing through the roller; wherein the temperature sensor is arranged inside the roller shaft of the roller, and the load sensor is arranged on one side of the roller shaft;

[0102] The temperature sensor sends the collected temperature information to the data processor, and the load sensor sends the collected load information to the data processor;

[0103] The temperature and load multimodal data fusion prediction model is communicatively connected to the data processor, and the prediction model estimates the evaluation scores of a plurality of preset modal variables related to the temperature information and the load information, wherein the modal variables include a plurality of the real-time temperature value, the temperature change rate, the maximum continuous operating temperature value, the peak load, the load change rate, the maximum continuous operating load, the load fluctuation amplitude, and the load fluctuation frequency;

[0104] According to the evaluation scores of each modal variable and the weights of each modal variable, the comprehensive score of the current state of the roller is calculated.

[0105] The roller conveying system provided by the present invention also includes a communication module and a data receiving and processing module, wherein the data receiving and processing module is a host computer or a cloud server; the communication module can adopt an ESP32-WROOM-32 Wi-Fi module, which is integrated with a data processor to form a compact unit, and uses the WiFi function to send the processed data to the host computer or the cloud server. The entire system ensures seamless connection from data acquisition, data processing to data transmission, so that the remote monitoring platform can obtain and analyze key data in real time, thereby effectively managing and maintaining the equipment where the roller is located. Before sending the data, the data is encapsulated into a data packet that complies with a specific wireless protocol (such as TCP / IP) to ensure the integrity and security of the data during network transmission. After receiving the data, the remote monitoring platform will first pre-process the data, including steps such as cleaning and normalization, in preparation for subsequent analysis. Then, the data will be input into a pre-trained machine learning model for in-depth analysis. The machine learning model is trained based on a historical data set to identify abnormal patterns that may affect the performance of the equipment. In the model training stage, a large amount of historical operating data and known fault cases are used as training sets to ensure the accuracy and robustness of the model. Once a potential problem is discovered, the system will automatically trigger an early warning mechanism so that maintenance personnel can take preventive measures to reduce unplanned downtime, ensure long-term stable operation of the equipment, and improve production efficiency.

[0106] The data receiving and processing module is configured to receive the temperature information and load information sent by the data processor through the communication module, pre-process the received information, extract the temperature characteristic information and the load characteristic information, and then input the extracted characteristic information into the pre-trained temperature and load multimodal data fusion prediction model;

[0107] The data processor determines whether the comprehensive status score is lower than a preset threshold value. If so, the data processor triggers the early warning device of the roller conveying system or shuts down the roller conveying system.

[0108] In one embodiment, the roller conveying system has a plurality of rollers equipped with temperature sensors and load sensors;

[0109] An average value is calculated for the comprehensive status scores of multiple rollers, and the average value of the comprehensive status scores is used as the monitoring result data, that is, if the average value of the comprehensive status scores is lower than a preset threshold value, then the data processor triggers the early warning device of the roller conveying system or shuts down the roller conveying system.

[0110] Alternatively, if the average value of the comprehensive status score belongs to the first level, the current motion parameters of the equipment where the roller is located are maintained; if the average value of the comprehensive status score belongs to the second level, the operating speed of the equipment where the roller is located is reduced; if the average value of the comprehensive status score belongs to the third level, the operating speed of the equipment where the roller is located is reduced and the refrigeration device is started; if the average value of the comprehensive status score belongs to the fourth level, the equipment where the roller is located is forcibly shut down.

[0111] The temperature and load multimodal data fusion prediction model is trained in the following way:

[0112] Collect sample data: simulate different working conditions for the roller, and use sensors to collect the temperature sample information and load sample information of the roller respectively, and record the correlation information between temperature, load and corresponding working conditions;

[0113] Extracting one or more pieces of temperature characteristic information from the collected temperature sample information, and extracting one or more pieces of load characteristic information from the collected load sample information;

[0114] The extracted temperature characteristic information and load characteristic information are spliced ​​and fused to form a comprehensive characteristic vector of multi-modal variables, and the weight of each modal variable is determined;

[0115] Marking the comprehensive feature vector according to the corresponding associated working condition information;

[0116] The support vector machine algorithm is used to train the labeled comprehensive feature vector, with the training purpose of minimizing the error between the predicted state comprehensive score of the sample data and the labeled label, to generate the temperature and load multimodal data fusion prediction model.

[0117] The roller conveying system provided in this embodiment and the roller temperature and load monitoring method provided in the above embodiment belong to the same inventive concept. The entire contents of the roller temperature and load monitoring method embodiment are incorporated into the roller conveying system embodiment by full reference here to avoid redundancy.

[0118] Traditional monitoring systems often rely on wired connections, which are susceptible to interference and difficult to expand. The present invention solves this problem by adopting wireless data transmission technology, and pays special attention to the layout of sensors, data processing algorithms, wireless data transmission protocols, and the specific application of machine learning models. After receiving the data, the remote monitoring platform will first pre-process the data, including cleaning, normalization and other steps, in preparation for subsequent analysis. Subsequently, the data will be input into a pre-trained machine learning model for in-depth analysis. The machine learning model is trained based on historical data sets and aims to identify abnormal patterns that may affect equipment performance.

[0119] The integrated monitoring system of the present invention can effectively overcome the limitations of traditional monitoring systems. Traditional monitoring systems often lack comprehensive monitoring of the operating status of equipment, resulting in delayed maintenance response, thereby increasing unplanned downtime. The present invention can not only monitor the key operating parameters of the equipment in real time, but also significantly reduce maintenance costs and downtime losses through predictive maintenance. In addition, the use of wireless data transmission technology reduces the need for physical connections and increases the flexibility and adaptability of the system.

[0120] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0121] The above is only a specific implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A roller temperature and load monitoring method, characterized in that: The following steps are involved: The temperature information and load information of the roller are collected by sensors respectively; The collected temperature information and load information are sent to a data processor, which is in communication connection with a pre-trained temperature and load multimodal data fusion prediction model; The prediction model estimates evaluation scores of a plurality of preset modal variables related to temperature information and load information; According to the evaluation scores of each modal variable and the weights of each modal variable, the comprehensive score of the current state of the roller is calculated; The temperature and load multimodal data fusion prediction model is trained in the following way: Collect sample data: simulate different working conditions for the roller, and use sensors to collect the temperature sample information and load sample information of the roller respectively, and record the correlation information between temperature, load and corresponding working conditions; Extracting one or more pieces of temperature characteristic information from the collected temperature sample information, and extracting one or more pieces of load characteristic information from the collected load sample information; The extracted temperature characteristic information and load characteristic information are spliced ​​and fused to form a comprehensive characteristic vector of multi-modal variables, and the weight of each modal variable is determined; Marking the comprehensive feature vector according to the corresponding associated working condition information; The support vector machine algorithm is used to train the labeled comprehensive feature vector, with the training purpose of minimizing the error between the predicted state comprehensive score of the sample data and the labeled label, to generate the temperature and load multimodal data fusion prediction model.

2. The roller temperature and load monitoring method according to claim 1, characterized in that: The current comprehensive score of the roller is calculated by the following formula: S = w 1× f 1+ w 2× f 2+…+ w n × f n ,in, S is the comprehensive score of status, f 1 is the evaluation score of the first modal variable estimated by the prediction model, w 1 is the weight of the first modal variable, f 2 is the evaluation score of the second modal variable estimated by the prediction model, w 2 is the weight of the second modal variable, f n Estimate the evaluation score of the nth modal variable for the prediction model, w n is the weight of the nth modal variable.

3. The roller temperature and load monitoring method according to claim 2, characterized in that: If the calculated comprehensive status score is lower than the preset alarm threshold, an early warning will be given that the roller is currently in poor working condition.

4. The roller temperature and load monitoring method according to claim 1, characterized in that: The training objective of the support vector machine algorithm is to minimize the following function: ， Its constraints are y i (w× x i + b )≥1- ξ i ,in ξ i ≥0; Among them, w is the weight vector, b is the deviation term, C A penalty factor to balance the relationship between model complexity and classification error. ξ i For the i The slack variable for samples, M is the number of samples to be predicted, x i For the i The comprehensive feature vector of samples, y i For the i The labels of samples.

5. The roller temperature and load monitoring method according to claim 1, characterized in that: The temperature characteristic information extracted from the collected temperature sample information includes one or more of the normalized real-time temperature value, the temperature change rate, and the maximum continuous operating temperature value; And / or, the load characteristic information extracted from the collected load sample information includes one or more of peak load, load change rate, maximum continuous working load, load fluctuation amplitude, and load fluctuation frequency.

6. The roller temperature and load monitoring method according to claim 5, characterized in that: Before extracting the temperature characteristic information and / or the load characteristic information, the following steps are also included: The temperature sample information and / or load sample information collected by the sensor is preprocessed, including one or more operations of data cleaning, normalization, and packaging into a preset standard data format.

7. The roller temperature and load monitoring method according to claim 1, characterized in that: The weights of the individual modal variables are determined in the following way: The weights of each modal variable are determined by the recursive feature elimination method of support vector machine; Alternatively, the weight of each modal variable is set through empirical values ​​or experiments.

8. The roller temperature and load monitoring method according to claim 2, characterized in that: The comprehensive feature vector is marked according to the working condition information in the following manner: The factors for presetting the working conditions include one or more of the ambient temperature, the load pressure, the ambient humidity, the continuous working time, and the roller speed; Different test conditions are set, and the value of at least one factor in different test conditions is different; After running various test conditions on the roller, obtaining roller performance results corresponding to the various test conditions; According to the roller performance results, the state scores are evaluated for various test conditions. The better the roller performance results, the higher the evaluation value of the state score. The evaluation value of the state score is associated with the test condition. In the stage of collecting sample data, different working conditions are simulated for the roller and executed as various test conditions set for operation, and then the evaluation value of its associated state score is used as the label value for marking.

9. The roller temperature and load monitoring method according to any one of claims 1 to 8, characterized in that: It also includes adjusting the equipment where the roller is located according to the calculated comprehensive score of the roller's current status: Multiple levels are pre-divided, and each level corresponds to a different range of comprehensive status scores; If the calculated comprehensive score of the current state of the roller belongs to the first level, the current motion parameters of the device where the roller is located are maintained; and / or, if the calculated comprehensive score of the current state of the roller belongs to the second level, reducing the operating speed of the equipment where the roller is located; and / or, if it is calculated that the current comprehensive score of the roller state belongs to the third level, the operating speed of the equipment where the roller is located is reduced and the refrigeration device is started; And / or, if it is calculated that the current comprehensive score of the roller status belongs to the fourth level, the device where the roller is located is forcibly shut down.

10. A roller conveying system, characterized in that: It includes a roller, a data processor, a temperature and load multimodal data fusion prediction model, a temperature sensor and a load sensor arranged on the roller, wherein the load sensor is configured to detect the weight of the material passing through the roller; The temperature sensor sends the collected temperature information to the data processor, and the load sensor sends the collected load information to the data processor; The temperature and load multimodal data fusion prediction model is trained in the following manner: Collecting sample data: simulating different working conditions for the roller, and using sensors to respectively collect the temperature sample information and load sample information of the roller, and recording the correlation information between the temperature, load and the corresponding working conditions; extracting one or more temperature feature information from the collected temperature sample information, and extracting one or more load feature information from the collected load sample information; splicing the extracted temperature feature information and load feature information to fuse them to form a comprehensive feature vector of multimodal variables, and determining the weight of each modal variable therein; marking the comprehensive feature vector according to the corresponding associated working condition information; using the support vector machine algorithm to train the marked comprehensive feature vector, with the training purpose of minimizing the error between the predicted state comprehensive score of the sample data and the marked label, to generate the temperature and load multimodal data fusion prediction model; The temperature and load multimodal data fusion prediction model is communicatively connected to the data processor, and the prediction model estimates the evaluation scores of a plurality of preset modal variables related to the temperature information and the load information, wherein the modal variables include a plurality of the real-time temperature value, the temperature change rate, the maximum continuous operating temperature value, the peak load, the load change rate, the maximum continuous operating load, the load fluctuation amplitude, and the load fluctuation frequency; According to the evaluation scores of each modal variable and the weights of each modal variable, the comprehensive score of the current state of the roller is calculated.

11. The roller conveyor system according to claim 10, characterized in that: The temperature sensor is arranged inside the roller shaft of the roller; and / or the load sensor is arranged on one side of the roller shaft.

12. The roller conveyor system according to claim 10, characterized in that: The roller conveying system has a plurality of rollers equipped with temperature sensors and load sensors; An average value is calculated for the comprehensive scores of the states of the multiple rollers, and the average value of the comprehensive scores is used as the monitoring result data.

13. The roller conveyor system according to claim 10, characterized in that: It also includes a communication module and a data receiving and processing module, wherein the data receiving and processing module is a host computer or a cloud server; The data receiving and processing module is configured to receive the temperature information and load information sent by the data processor through the communication module, pre-process the received information, extract the temperature characteristic information and the load characteristic information, and then input the extracted characteristic information into the pre-trained temperature and load multimodal data fusion prediction model; The temperature and load multimodal data fusion prediction model is integrated in the data processor, or is set in a cloud server.

14. The roller conveyor system according to claim 10, characterized in that: The data processor determines whether the comprehensive status score is lower than a preset threshold value. If so, the data processor triggers the early warning device of the roller conveying system or shuts down the roller conveying system.

Images