Multi-sensor detection method and platform based on Internet of Things

Abstract

The embodiment of the invention provides a multi-sensor detection method and platform based on the Internet of Things, and belongs to the technical field of detection processing. The multi-sensor detection platform based on the Internet of Things comprises a sensor group, a communication module group and an edge computing device, the sensor group is in communication connection with the edge computing device through the communication module group. The sensor group is used for collecting first vibration data and first gas concentration data of the target internet-of-things terminal equipment and uploading the data to the edge computing equipment through the communication module group; the edge computing device is used for receiving the first vibration data and the first gas concentration data, extracting a first vibration feature from the first vibration data, extracting a first gas concentration feature from the first gas concentration data, and inputting the first vibration feature and the first gas concentration feature into a pre-trained device fault prediction model, obtaining a fault prediction result of the target Internet of Things terminal equipment; and outputting a fault diagnosis report according to the fault prediction result of the target Internet of Things terminal device.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of detection technology, and in particular to a multi-sensor detection method and platform based on the Internet of Things. Background Art

[0002] With the development of industrial automation and intelligence, equipment reliability and maintenance efficiency have become the focus of enterprises. Traditional equipment maintenance methods mainly rely on regular manual inspections and post-maintenance, which has many shortcomings, such as high maintenance costs, slow fault response, and difficulty in predicting potential faults. Therefore, how to achieve real-time monitoring and fault prediction of equipment through advanced technical means has become an important research direction.

[0003] Existing equipment failure prediction methods mainly learn the normal and failure modes of equipment from a large amount of data to achieve failure prediction. However, existing fault prediction models often rely on a single data source and simple feature extraction methods, resulting in low prediction accuracy. Summary of the invention

[0004] The embodiments of the present invention provide a multi-sensor detection method and platform based on the Internet of Things, so as to achieve comprehensive detection and accurate fault prediction of target Internet of Things terminal devices by combining multiple sensor data and edge computing technology.

[0005] In a first aspect, an embodiment of the present invention provides a multi-sensor detection platform based on the Internet of Things, including a sensor group, a communication module group and an edge computing device; the sensor group and the edge computing device are communicatively connected through the communication module group;

[0006] The sensor group is used to collect the first vibration data and the first gas concentration data of the target IoT terminal device, and upload the first vibration data and the first gas concentration data to the edge computing device through the communication module group;

[0007] The edge computing device is used to receive the first vibration data and the first gas concentration data, extract the first vibration feature from the first vibration data, and extract the first gas concentration feature from the first gas concentration data, input the first vibration feature and the first gas concentration feature into a pre-trained device fault prediction model to obtain a fault prediction result of the target IoT terminal device; and output a fault diagnosis report based on the fault prediction result of the target IoT terminal device.

[0008] In a possible design of the first aspect, the edge computing device extracts a first vibration feature from the first vibration data, extracts a first gas concentration feature from the first gas concentration data, inputs the first vibration feature and the first gas concentration feature into a pre-trained device fault prediction model, and obtains a fault prediction result of the target IoT terminal device, specifically for:

[0009] using a low-pass filter to remove high-frequency noise in the first vibration data, using a moving average method to smooth the first vibration data after the high-frequency noise is removed, and normalizing the first vibration data after the smoothing process;

[0010] Extracting first vibration data time domain features and first vibration data frequency domain features from the first vibration data after normalization, wherein the first vibration data time domain features include mean, variance, peak, kurtosis and skewness; and the first vibration data frequency domain features include spectrum and power spectrum density;

[0011] Using a high-pass filter to remove low-frequency drift in the first gas concentration data, using an exponential smoothing method to smooth the first gas concentration data after the low-frequency drift is removed, and normalizing the first gas concentration data after the smoothing process;

[0012] extracting the first gas concentration feature from the first gas concentration data after normalization, wherein the first gas concentration feature includes a concentration change rate, a gas concentration peak value, and a gas concentration change trend;

[0013] combining the first vibration data time domain feature, the first vibration feature frequency domain feature, and the first gas concentration feature into a first multimodal feature vector;

[0014] The first multimodal feature vector is input into the device fault prediction model to obtain a fault prediction result of the target IoT terminal device.

[0015] In a possible design of the first aspect, when the sampling frequencies of the first vibration data time domain feature, the first vibration feature frequency domain feature, and the first gas concentration feature are different, the edge computing device is further used to:

[0016] The first vibration data time domain feature, the first vibration feature frequency domain feature and the first gas concentration feature are aligned on the same time axis.

[0017] In a possible design of the first aspect, when the edge computing module combines the first vibration data time domain feature, the first vibration feature frequency domain feature, and the first gas concentration feature into a first multimodal feature vector, it is specifically used to:

[0018] splicing the first vibration data time domain feature, the first vibration feature frequency domain feature and the first gas concentration feature in sequence into the first multimodal feature vector; or,

[0019] Using a feature selection model, the importance of each feature of the first vibration data time domain feature, the first vibration feature frequency domain feature and the first gas concentration feature is evaluated, and the least important features are gradually deleted until the number of remaining features reaches a preset number, and the remaining preset number of features are concatenated into the first multimodal feature vector; the first multimodal feature vector includes the most relevant features extracted from the first vibration data time domain feature, the first vibration feature frequency domain feature and the first gas concentration feature.

[0020] In a possible design of the first aspect, the platform further includes: a cloud server;

[0021] The cloud server is used to collect historical vibration data and historical gas concentration data of the target IoT terminal device from the edge computing device, the historical vibration data and the historical gas concentration data include data of the target IoT terminal device in a normal state and data in a faulty state, and use the collected historical vibration data and historical gas concentration data to train a fault prediction model for the device to be trained, and send the trained device fault prediction model to the edge computing device.

[0022] In a possible design of the first aspect, when training the to-be-trained equipment fault prediction model, the cloud server is specifically used to:

[0023] Judging whether the sample data has a fault according to a preset time window to generate labels for the historical vibration data and the historical gas concentration data; wherein each of the preset time windows corresponds to one of the sample data, and each of the sample data includes the historical vibration data and the historical gas concentration data;

[0024] Integrate the labeled sample data into a training data set, wherein each sample data in the training data set includes the historical vibration data, the historical gas concentration data and a corresponding label;

[0025] Dividing the training data set into a training set and a test set;

[0026] Using the sample data in the training set to train the constructed basic fault prediction model to obtain the trained basic fault prediction model;

[0027] Using the trained basic fault prediction model to predict the data in the test set, and calculating the performance index of the trained basic fault prediction model according to the prediction result;

[0028] Generating a ROC curve according to the calculated performance index to evaluate the performance of the trained basic fault prediction model;

[0029] When the performance evaluation of the basic fault prediction model after training is qualified, the equipment fault prediction model is obtained.

[0030] In a possible design of the first aspect, the target IoT terminal device includes a generator and a transformer;

[0031] The sensor group includes a first vibration sensor arranged on the generator rotor, a second vibration sensor arranged on the generator stator, a third vibration sensor arranged on the transformer core, a fourth vibration sensor arranged on the transformer winding, a first gas sensor arranged around the generator, and a second gas sensor arranged in the transformer oil;

[0032] The communication module group includes a first communication module integrated in the first vibration sensor, a second communication module integrated in the second vibration sensor, a third communication module integrated in the third vibration sensor, a fourth communication module integrated in the fourth vibration sensor, a fifth communication module integrated in the first gas sensor, and a sixth communication module integrated in the second gas sensor.

[0033] In a second aspect, an embodiment of the present invention provides a multi-sensor detection method based on the Internet of Things, which is applied to the multi-sensor detection platform based on the Internet of Things as described in the first aspect of the embodiment of the present invention, and the method includes:

[0034] The sensor group collects first vibration data and first gas concentration data of the target IoT terminal device;

[0035] The sensor group uploads the first vibration data and the first gas concentration data to the edge computing device through the communication module group;

[0036] The edge computing device receives the first vibration data and the first gas concentration data, extracts a first vibration feature from the first vibration data, extracts a first gas concentration feature from the first gas concentration data, inputs the first vibration feature and the first gas concentration feature into a pre-trained device fault prediction model, and obtains a fault prediction result of the target IoT terminal device;

[0037] The edge computing device outputs a fault diagnosis report according to the fault prediction result of the target IoT terminal device.

[0038] In a possible design of the first aspect, the edge computing device extracts a first vibration feature from the first vibration data, extracts a first gas concentration feature from the first gas concentration data, inputs the first vibration feature and the first gas concentration feature into a pre-trained device fault prediction model, and obtains a fault prediction result of the target IoT terminal device, including:

[0039] The edge computing device uses a low-pass filter to remove high-frequency noise in the first vibration data, uses a moving average method to smooth the first vibration data from which the high-frequency noise has been removed, and normalizes the first vibration data after the smoothing process;

[0040] The edge computing device extracts first vibration data time domain features and first vibration data frequency domain features from the first vibration data after normalization, wherein the first vibration data time domain features include mean, variance, peak, kurtosis and skewness; and the first vibration data frequency domain features include spectrum and power spectrum density;

[0041] The edge computing device uses a high-pass filter to remove low-frequency drift in the first gas concentration data, uses an exponential smoothing method to smooth the first gas concentration data after the low-frequency drift is removed, and normalizes the first gas concentration data after the smoothing process;

[0042] The edge computing device extracts the first gas concentration feature from the first gas concentration data after normalization, where the first gas concentration feature includes a concentration change rate, a gas concentration peak value, and a gas concentration change trend;

[0043] The edge computing device combines the first vibration data time domain feature, the first vibration feature frequency domain feature, and the first gas concentration feature into a first multimodal feature vector;

[0044] The edge computing device inputs the first multimodal feature vector into the device fault prediction model to obtain a fault prediction result of the target IoT terminal device.

[0045] In a possible design of the second aspect, the method further includes:

[0046] When the sampling frequencies of the first vibration data time domain feature, the first vibration feature frequency domain feature and the first gas concentration feature are different, the edge computing device aligns the first vibration data time domain feature, the first vibration feature frequency domain feature and the first gas concentration feature on the same time axis.

[0047] The embodiment of the present invention provides a multi-sensor detection method and platform based on the Internet of Things. The detection platform is provided with a sensor group, a communication module group and an edge computing device; the sensor group and the edge computing device are connected by communication module group. The sensor group is used to collect the first vibration data and the first gas concentration data of the target IoT terminal device and upload them to the edge computing device through the communication module group; the edge computing device is used to receive the first vibration data and the first gas concentration data, extract the first vibration feature from the first vibration data, and extract the first gas concentration feature from the first gas concentration data, input the first vibration feature and the first gas concentration feature into the pre-trained device fault prediction model, and obtain the fault prediction result of the target IoT terminal device; output the fault diagnosis report according to the fault prediction result of the target IoT terminal device. Thus, the first vibration data and the first gas concentration data of the target IoT terminal device are collected in real time by the sensor group, and uploaded to the edge computing device, so that the edge computing device uses the device fault prediction model obtained from the cloud server to perform fault prediction according to the first vibration data and the first gas concentration data, so as to timely determine whether the target IoT terminal device has a fault or is about to fail, and generate a fault diagnosis report to guide the maintenance personnel to perform timely maintenance and repair, thereby realizing comprehensive detection and accurate fault prediction of the target IoT terminal device. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present invention and, together with the description, serve to explain the principles of the embodiments of the present invention.

[0049] Figure 1 A schematic diagram of the structure of a multi-sensor detection platform based on the Internet of Things provided by an embodiment of the present invention;

[0050] Figure 2 A schematic diagram of the structure of another multi-sensor detection platform based on the Internet of Things provided by an embodiment of the present invention;

[0051] Figure 3 A flow chart of a multi-sensor detection method based on the Internet of Things provided by an embodiment of the present invention is shown;

[0052] Figure 4 A flowchart showing a training process of a device failure prediction model provided by an embodiment of the present invention is shown;

[0053] Figure 5 It is a schematic diagram of the structure of an electronic device according to an exemplary embodiment of the present invention.

[0054] The above drawings have shown specific embodiments of the present invention, which will be described in more detail below. These drawings and text descriptions are not intended to limit the scope of the embodiments of the present invention in any way, but to illustrate the concepts of the embodiments of the present invention for those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0055] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution 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 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 are within the scope of protection of the present invention.

[0056] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in the embodiments of the present invention are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data need to comply with the relevant laws, regulations and standards of relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0057] Some embodiments of the present invention are described in detail below in conjunction with the accompanying drawings. In the case where there is no conflict between the embodiments, the following embodiments and the features in the embodiments can be combined with each other. In addition, the step sequence in the following method embodiments is only an example and not a strict limitation.

[0058] There are many kinds of power transmission equipment, which often need to be inspected to ensure the normal operation of power transmission equipment. In practical applications, power equipment includes transformers, ice monitoring devices, transmission line monitoring devices, etc. With the development of industrial automation and intelligence, the reliability and maintenance efficiency of equipment have become the focus of enterprises. Traditional equipment maintenance methods mainly rely on regular manual inspections and post-maintenance. This method has many shortcomings, such as high maintenance costs, slow fault response, and difficulty in predicting potential faults. Therefore, how to achieve real-time monitoring and fault prediction of equipment through advanced technical means has become an important research direction.

[0059] The existing equipment fault prediction methods mainly include the following: one is based on regular inspections by operators, and faults are predicted through the experience and technical level of the operators during the inspection process; the other is based on machine learning, which trains machine learning models to learn the normal and fault modes of equipment from a large amount of data, thereby achieving fault prediction.

[0060] Although the above scheme can achieve equipment failure prediction to a certain extent, there are still some shortcomings, such as the subjectivity and large prediction errors of the method based on regular inspection by operators. The machine learning model trained based on machine learning often relies on a single data source and simple feature extraction methods, and it is difficult to process large-scale data in real time, resulting in low prediction accuracy and real-time performance.

[0061] In recent years, the rapid development of Internet of Things (IoT) technology has provided new solutions for equipment monitoring and fault prediction. In the embodiment of the present invention, various sensors and communication modules are deployed on target terminal devices, such as generators and transformers, through IoT technology, so that the operation data of the target terminal devices can be collected in real time, and the data can be processed and analyzed in real time through edge computing devices, thereby realizing real-time monitoring of the status of the target terminal devices and fault prediction.

[0062] The embodiments of the present invention are intended to solve the problems existing in the above-mentioned prior art. Specifically, the concepts of the embodiments of the present invention are developed around the following core points:

[0063] Multi-sensor data acquisition: Vibration data and gas concentration data of target equipment are collected in real time through the sensor group, providing comprehensive equipment status information.

[0064] Real-time data transmission: The collected data is transmitted to the edge computing device in real time through the communication module group to ensure the timeliness of the data.

[0065] Efficient data processing: Perform data preprocessing, feature extraction, and fault prediction on edge computing devices to fully utilize edge computing resources and improve data processing efficiency.

[0066] High-precision fault prediction: Through pre-trained equipment fault prediction models, features are extracted from multimodal data to achieve high-precision fault prediction.

[0067] Fault diagnosis report: Generate a fault diagnosis report based on the fault prediction results to guide maintenance personnel to perform timely maintenance and repairs.

[0068] In summary, the embodiment of the present invention constructs a multi-sensor detection platform based on the Internet of Things, which includes a sensor group, a communication module group and an edge computing device; the sensor group and the edge computing device are connected through the communication module group. The sensor group is used to collect the first vibration data and the first gas concentration data of the target IoT terminal device and upload them to the edge computing device through the communication module group; the edge computing device is used to receive the first vibration data and the first gas concentration data, extract the first vibration feature from the first vibration data, and extract the first gas concentration feature from the first gas concentration data, input the first vibration feature and the first gas concentration feature into a pre-trained device fault prediction model, and obtain the fault prediction result of the target IoT terminal device; output a fault diagnosis report according to the fault prediction result of the target IoT terminal device. Thus, the first vibration data and the first gas concentration data of the target IoT terminal device are collected in real time through the sensor group and uploaded to the edge computing device, so that the edge computing device uses the equipment fault prediction model obtained from the cloud server to perform fault prediction according to the first vibration data and the first gas concentration data, so as to timely determine whether the target IoT terminal device has a fault or is about to fail, and generate a fault diagnosis report to guide maintenance personnel to perform timely maintenance and repairs, thereby realizing comprehensive detection of the target IoT terminal device and accurate fault prediction.

[0069] The structure of the multi-sensor detection platform based on the Internet of Things involved in the embodiment of the present invention is briefly described below. Figure 1 The schematic diagram of the structure of a multi-sensor detection platform based on the Internet of Things provided by an embodiment of the present invention. Figure 1 As shown, the multi-sensor detection platform 100 based on the Internet of Things includes: a sensor group 110, a communication module group 120 and an edge computing device 130; the sensor group 110 and the edge computing device 130 are communicatively connected through the communication module group 120.

[0070] Among them, the sensor group 110 is used to collect the first vibration data and the first gas concentration data of the target IoT terminal device, and upload the first vibration data and the first gas concentration data to the edge computing device 130 through the communication module group 120.

[0071] In an embodiment of the present invention, the target IoT terminal device is a generator and a transformer. Specifically, the sensor group 110 includes a first vibration sensor disposed on the generator rotor, for collecting vibration data of the generator rotor. A second vibration sensor disposed on the generator stator, for collecting vibration data of the generator stator. A third vibration sensor disposed on the transformer core, for collecting vibration data of the transformer core. A fourth vibration sensor disposed on the transformer winding, for collecting vibration data on the transformer winding. A first gas sensor disposed around the generator, for collecting the SF6 gas concentration around the generator. And a second gas sensor disposed in the transformer oil, for collecting the concentration of dissolved gas.

[0072] In order to achieve the purpose of transmitting the collected data to the edge computing device 130 in real time, in an embodiment of the present invention, the communication module group 120 includes a first communication module integrated in the first vibration sensor, a second communication module integrated in the second vibration sensor, a third communication module integrated in the third vibration sensor, a fourth communication module integrated in the fourth vibration sensor, a fifth communication module integrated in the first gas sensor, and a sixth communication module integrated in the second gas sensor.

[0073] The first vibration data is composed of vibration data collected by the first vibration sensor, vibration data collected by the second vibration sensor, vibration data collected by the third vibration sensor, and vibration data collected by the fourth vibration sensor. The first gas concentration data is composed of gas concentration data collected by the first gas sensor and second gas concentration data collected by the second gas sensor.

[0074] The edge computing device 130 is used to receive the first vibration data and the first gas concentration data, extract the first vibration feature from the first vibration data, and extract the first gas concentration feature from the first gas concentration data, input the first vibration feature and the first gas concentration feature into a pre-trained device fault prediction model to obtain a fault prediction result of the target IoT terminal device; and output a fault diagnosis report based on the fault prediction result of the target IoT terminal device.

[0075] Specifically, when the edge computing device 130 extracts the first vibration feature from the first vibration data and the first gas concentration feature from the first gas concentration data, and inputs the first vibration feature and the first gas concentration feature into a pre-trained device fault prediction model to obtain a fault prediction result of the target IoT terminal device, a low-pass filter can be used to remove high-frequency noise from the first vibration data to retain low-frequency signal data. The first vibration data with high-frequency noise removed is smoothed using a moving average method to reduce short-term fluctuations in the data and make the data smoother. In another implementation method, the first vibration data with high-frequency noise removed can also be smoothed using an exponential smoothing method.

[0076] Next, the smoothed first vibration data is normalized. For example, by the formula The first vibration data after smoothing is normalized to be converted into a standard normal distribution with a mean of 0 and a standard deviation of 1.

[0077] The first vibration data time domain features and the first vibration data frequency domain features are extracted from the normalized first vibration data, wherein the first vibration data time domain features include mean, variance, peak, kurtosis and skewness; and the first vibration data frequency domain features include spectrum and power spectrum density.

[0078] When processing the first gas concentration data, the edge computing device 130 may use a high-pass filter to remove low-frequency drift in the first gas concentration data to retain high-frequency signals. Then, the first gas concentration data with low-frequency drift removed is smoothed using an exponential smoothing method to reduce short-term fluctuations in the data and make the data smoother. In another implementation method, the first vibration data with high-frequency noise removed may also be smoothed using a moving average method.

[0079] Next, the smoothed first gas concentration data is normalized. For example, by the formula The first vibration data after smoothing is normalized, and its data range is scaled to the range of [0, 1].

[0080] Extract the first gas concentration feature from the normalized first gas concentration data, the first gas concentration feature including the concentration change rate, the gas concentration peak value, and the gas concentration change trend. Combine the first vibration data time domain feature, the first vibration feature frequency domain feature, and the first gas concentration feature into a first multimodal feature vector. Input the first multimodal feature vector into the device fault prediction model to obtain the fault prediction result of the target IoT terminal device.

[0081] It should be noted that when the sampling frequencies of the first vibration data time domain feature, the first vibration feature frequency domain feature, and the first gas concentration feature are different, the edge computing device 130 can also be used to align the first vibration data time domain feature, the first vibration feature frequency domain feature, and the first gas concentration feature to the same time axis to ensure the time synchronization of the data. For example, assuming that the sampling frequency of the first vibration data is 1 Hz and the sampling frequency of the first gas concentration data is 1 / 60 Hz (once per minute). We can use linear interpolation to interpolate the gas concentration data to a sampling frequency of 1 Hz.

[0082] When the edge computing device 130 combines the first vibration data time domain features, the first vibration feature frequency domain features, and the first gas concentration features into a first multimodal feature vector, it can specifically splice the first vibration data time domain features, the first vibration feature frequency domain features, and the first gas concentration features in sequence into the first multimodal feature vector.

[0083] In another implementation, the edge computing device 130 can also use a feature selection model to evaluate the importance of each feature in the first vibration data time domain feature, the first vibration feature frequency domain feature, and the first gas concentration feature, and gradually delete the least important features until the number of remaining features is a preset number, and the remaining preset number of features are spliced ​​into a first multimodal feature vector; the first multimodal feature vector includes the most relevant features extracted from the first vibration data time domain feature, the first vibration feature frequency domain feature, and the first gas concentration feature.

[0084] In the embodiment of the present invention, the fault prediction results generally include multiple aspects, which can help maintenance personnel and engineers better understand and deal with potential faults. Exemplarily, the fault prediction results include: fault probability, predicted time, fault type, fault severity, fault cause, maintenance suggestions, information level, historical data comparison, and alarm information.

[0085] Among them, the failure probability refers to the probability that the target IoT terminal device will fail in a certain period of time in the future as predicted by the equipment failure prediction model. For example, the probability of a generator failing in the next week is 80%. It is used to help users assess risks and decide whether preventive measures need to be taken.

[0086] The prediction time refers to the specific time or time period when the equipment failure prediction model predicts that the failure may occur. For example, the equipment is expected to fail on November 10, 2024; it can also be a time period, for example, the equipment is expected to fail within the next 30 days. This helps users arrange maintenance plans and ensure that necessary inspections and repairs are carried out before the failure occurs.

[0087] Fault type refers to the fault type predicted by the equipment fault prediction model, such as mechanical fault, electrical fault, thermal fault, etc. This helps maintenance personnel prepare the corresponding tools and materials in advance and improves maintenance efficiency.

[0088] Fault severity refers to the severity of the fault predicted by the equipment fault prediction model, such as minor, moderate, or severe. It can be a classification label, such as "minor", "moderate", or "severe", or a numerical score, such as 1-5 points, with 5 points indicating the most severe. This helps users assess the urgency of the fault and decide whether immediate action is needed.

[0089] The cause of the fault refers to the cause of the fault predicted by the equipment fault prediction model, such as overload, wear, aging, etc. It is usually a text description, such as "overload", "wear", "aging". It helps maintenance personnel to conduct targeted inspections and repairs to prevent similar faults from happening again.

[0090] Maintenance suggestions refer to maintenance suggestions provided based on fault prediction results, such as replacing parts, cleaning equipment, adjusting parameters, etc. They are usually text descriptions, such as "recommended to replace bearings" and "recommended to clean the radiator." They provide specific maintenance operation guidelines to help maintenance personnel solve problems efficiently.

[0091] The confidence level refers to the confidence level of the equipment failure prediction model in the prediction results, indicating the reliability of the prediction. It is usually expressed as a percentage, for example, the confidence level of the model in the prediction results is 90%. This helps users evaluate the credibility of the prediction results and decide whether to rely on the prediction for decision making.

[0092] Historical data comparison refers to comparing the current prediction results with historical data to analyze trends and changes. It can be a chart or table showing the comparison between historical fault data and current prediction results. It helps identify long-term trends of equipment and evaluate the health of equipment.

[0093] Alarm information refers to the information that is triggered when the prediction results show potential failures. It can be email, SMS, system notification, etc. to notify relevant personnel in time to ensure a quick response.

[0094] The following is a specific example of a failure prediction report:

[0095] Failure prediction report:

[0096] Device number: 12345.

[0097] Forecast date: October 31, 2024.

[0098] Report generation date: October 31, 2024.

[0099] 1. Failure probability:

[0100] -Probability of failure within the next week: 80%

[0101] 2. Prediction time:

[0102] - Expected failure time: November 10, 2024

[0103] 3. Fault type:

[0104] -Predicted failure type: Mechanical failure

[0105] 4. Fault severity:

[0106] - Predicted failure severity: Medium (3 / 5)

[0107] 5. Cause of failure:

[0108] -Predicted failure cause: bearing wear

[0109] 6.Maintenance suggestions:

[0110] - Recommended actions:

[0111] -Replace worn bearings.

[0112] - Check the lubrication system to ensure that there is sufficient lubricating oil.

[0113] - Clean the radiator to ensure good heat dissipation.

[0114] 7. Confidence Level:

[0115] - Forecast confidence level: 90%

[0116] 8. Historical data comparison:

[0117] -Historical fault data:

[0118] - January 10, 2024: Mechanical failure (bearing wear).

[0119] - March 20, 2024: Electrical failure (circuit board aging).

[0120] -Current prediction results:

[0121] - November 10, 2024: Mechanical failure (bearing wear).

[0122] 9. Alarm information:

[0123] -Alarm level: High

[0124] -Alarm content: Equipment 12345 has an 80% probability of mechanical failure in the next week. Please arrange maintenance as soon as possible.

[0125] - Notification method: Email, SMS.

[0126] In the embodiment of the present invention, a sensor group 110, a communication module group 120 and an edge computing device 130 are provided; the sensor group 110 and the edge computing device 130 are connected in communication via the communication module group 120. The sensor group 110 is used to collect the first vibration data and the first gas concentration data of the target IoT terminal device and upload them to the edge computing device 130 via the communication module group 120; the edge computing device 130 is used to receive the first vibration data and the first gas concentration data, extract the first vibration feature from the first vibration data, and extract the first gas concentration feature from the first gas concentration data, input the first vibration feature and the first gas concentration feature into a pre-trained device fault prediction model, and obtain the fault prediction result of the target IoT terminal device; output a fault diagnosis report according to the fault prediction result of the target IoT terminal device. Thus, the first vibration data and the first gas concentration data of the target IoT terminal device are collected in real time through the sensor group 110 and uploaded to the edge computing device 130, so that the edge computing device 130 uses the equipment fault prediction model obtained from the cloud server 210 to perform fault prediction based on the first vibration data and the first gas concentration data, so as to timely determine whether the target IoT terminal device has a fault or is about to fail, and generate a fault diagnosis report to guide maintenance personnel to perform timely maintenance and repairs, thereby achieving comprehensive detection of the target IoT terminal device and accurate fault prediction.

[0127] Figure 2 A schematic diagram of another multi-sensor detection platform based on the Internet of Things provided by an embodiment of the present invention. Figure 2 As shown, based on the above embodiments, the multi-sensor detection platform based on the Internet of Things of this embodiment may also include: a cloud server 210; the cloud server 210 is used to collect historical vibration data and historical gas concentration data of the target Internet of Things terminal device from the edge computing device 130, the historical vibration data and the historical gas concentration data include data of the target Internet of Things terminal device in a normal state and data in a faulty state, and use the collected historical vibration data and historical gas concentration data to train the equipment fault prediction model to be trained, and send the trained equipment fault prediction model to the edge computing device 130.

[0128] Specifically, when training the fault prediction model of the device to be trained, the cloud server 210 first needs to collect the historical vibration data and historical gas concentration data of the target IoT terminal device. After that, it is judged whether the sample data has a fault according to the preset time window to generate labels for the historical vibration data and the historical gas concentration data; wherein each preset time window corresponds to a sample data, and each sample data includes historical vibration data and historical gas concentration data.

[0129] Among them, judging whether there is a fault in the time window based on the data in the time window can be specifically marked in the following ways: Classification label: Mark the window as "normal" or "faulty" if the data in the window corresponds to a fault event. Multi-category label: If there are multiple fault types, the window can be marked as different fault categories. Regression label: If you need to predict the severity of the fault, you can use a numerical label (for example, fault level).

[0130] Next, the labeled sample data is integrated into the training data set, where each sample data in the training data set contains historical vibration data, historical gas concentration data, and corresponding labels. Optionally, the edge computing device 130 also needs to check the ratio of normal and faulty samples in the training data set, and perform upsampling or downsampling when necessary to balance the data set to avoid the model being biased towards a certain type of label. The training data set is divided into a training set and a test set. The training set is used to train the basic fault prediction model, and the test set is used to test the performance of the trained basic fault prediction model.

[0131] Among them, the basic fault prediction model can be an integrated learning model composed of a random forest classifier and a support vector machine classifier. It combines the advantages of the random forest classifier in processing high-dimensional data and nonlinear relationships, and the support vector machine classifier in processing small samples and high-dimensional data, and has strong generalization capabilities, forming an integrated learning model with stronger performance.

[0132] In one implementation, the random forest classifier and support vector machine classifier of the basic fault prediction model can be implemented based on a voting mechanism, and the constructed basic fault prediction model is trained using sample data in the training set to obtain a trained basic fault prediction model. The trained basic fault prediction model is used to predict the data in the test set, and the performance index of the trained basic fault prediction model is calculated based on the prediction results; a ROC curve is generated based on the calculated performance index to evaluate the performance of the trained basic fault prediction model; when the performance evaluation of the trained basic fault prediction model is qualified, the equipment fault prediction model is obtained.

[0133] In another implementation, the random forest classifier and support vector machine classifier of the basic fault prediction model can be implemented based on a stacking mechanism. The basic fault prediction model built based on the stacking mechanism is trained using sample data in the training set, and the prediction results of the basic fault prediction model are generated using cross-validation. The prediction results of the basic fault prediction model are used as new features to train the meta-model (meta-model 1). After that, the basic fault prediction model is used to predict the data in the test set to obtain the corresponding prediction results. The prediction results obtained this time are used as new features to use the trained meta-model for prediction to generate the final prediction results. Based on the final prediction results, the performance indicators of the trained basic fault prediction model are calculated; the ROC curve is generated based on the calculated performance indicators to evaluate the performance of the trained basic fault prediction model; when the performance evaluation of the trained basic fault prediction model is qualified, the equipment fault prediction model is obtained.

[0134] It is understood that, in addition to training the equipment fault prediction model, the cloud server 210 can also store the first vibration data and the first gas concentration data of the target IoT terminal, as well as the corresponding target IoT terminal operation status, fault prediction report and other official information. No further details will be given here.

[0135] According to the detection platform of an embodiment of the present invention, the equipment fault prediction model can be pre-trained through a cloud server, and the trained equipment fault prediction model can be sent to an edge computing device, so that data processing, feature extraction and fault prediction can be performed on the edge computing device, making full use of the computing resources of edge computing and improving data processing efficiency.

[0136] The multi-sensor detection platform based on the Internet of Things provided by the embodiment of the present invention is described above. The detection method applied in the multi-sensor detection platform based on the Internet of Things is described in detail below with reference to the accompanying drawings.

[0137] Figure 3 FIG. 1 is a flow chart of a multi-sensor detection method based on the Internet of Things provided by an embodiment of the present invention. Figure 3 As shown, the method includes the following steps 301 to 304:

[0138] Step 301: The sensor group collects first vibration data and first gas concentration data of the target IoT terminal device.

[0139] Step 302: The sensor group uploads the first vibration data and the first gas concentration data to the edge computing device through the communication module group.

[0140] Step 303: The edge computing device receives the first vibration data and the first gas concentration data, extracts the first vibration feature from the first vibration data, and extracts the first gas concentration feature from the first gas concentration data, inputs the first vibration feature and the first gas concentration feature into a pre-trained device fault prediction model, and obtains a fault prediction result of the target IoT terminal device.

[0141] Step 304: The edge computing device outputs a fault diagnosis report based on the fault prediction result of the target IoT terminal device.

[0142] like Figure 4 As shown, in the above step 303, the edge computing device extracts the first vibration feature from the first vibration data, extracts the first gas concentration feature from the first gas concentration data, inputs the first vibration feature and the first gas concentration feature into a pre-trained device fault prediction model, and obtains the fault prediction result of the target IoT terminal device. Specifically, the process may include the following steps 401 to 406:

[0143] Step 401: The edge computing device uses a low-pass filter to remove high-frequency noise in the first vibration data, uses a moving average method to smooth the first vibration data from which the high-frequency noise has been removed, and normalizes the smoothed first vibration data.

[0144] Step 402: The edge computing device extracts time domain features of the first vibration data and frequency domain features of the first vibration data from the normalized first vibration data, where the time domain features of the first vibration data include mean, variance, peak, kurtosis, and skewness; and the frequency domain features of the first vibration data include spectrum and power spectrum density.

[0145] Step 403: The edge computing device uses a high-pass filter to remove low-frequency drift in the first gas concentration data, uses an exponential smoothing method to smooth the first gas concentration data from which the low-frequency drift is removed, and normalizes the smoothed first gas concentration data.

[0146] Step 404: The edge computing device extracts the first gas concentration characteristics from the normalized first gas concentration data, where the first gas concentration characteristics include the concentration change rate, the gas concentration peak value, and the gas concentration change trend.

[0147] Step 405: The edge computing device combines the first vibration data time domain features, the first vibration feature frequency domain features, and the first gas concentration features into a first multimodal feature vector.

[0148] It can be understood that when the sampling frequencies of the first vibration data time domain feature, the first vibration feature frequency domain feature, and the first gas concentration feature are different, the edge computing device aligns the first vibration data time domain feature, the first vibration feature frequency domain feature, and the first gas concentration feature to the same time axis to ensure the time synchronization of the data. For example, assuming that the sampling frequency of the first vibration data is 1Hz and the sampling frequency of the first gas concentration data is 1 / 60Hz (once per minute). We can use linear interpolation to interpolate the gas concentration data to a sampling frequency of 1Hz.

[0149] Step 406: The edge computing device inputs the first multimodal feature vector into the device fault prediction model to obtain a fault prediction result of the target IoT terminal device.

[0150] The multi-sensor detection method based on the Internet of Things of the embodiment of the present invention can be applied to the multi-sensor detection platform based on the Internet of Things, collects the first vibration data and the first gas concentration data of the target Internet of Things terminal device through the sensor group; uploads the first vibration data and the first gas concentration data to the edge computing device through the communication module group; receives the first vibration data and the first gas concentration data through the edge computing device, extracts the first vibration feature from the first vibration data, and extracts the first gas concentration feature from the first gas concentration data, inputs the first vibration feature and the first gas concentration feature into the pre-trained equipment fault prediction model, and obtains the fault prediction result of the target Internet of Things terminal device; outputs the fault diagnosis report according to the fault prediction result of the target Internet of Things terminal device through the edge computing device. Thus, the first vibration data and the first gas concentration data of the target Internet of Things terminal device are collected in real time through the sensor group, and uploaded to the edge computing device, so that the edge computing device uses the equipment fault prediction model obtained from the cloud server to perform fault prediction according to the first vibration data and the first gas concentration data, so as to timely determine whether the target Internet of Things terminal device has a fault or is about to fail, and generates a fault diagnosis report to guide the maintenance personnel to perform timely maintenance and repair, thereby realizing comprehensive detection and accurate fault prediction of the target Internet of Things terminal device.

[0151] Figure 5 FIG. 1 is a schematic diagram of the structure of an electronic device according to an exemplary embodiment of the present invention. Figure 5 As shown, an electronic device 500 provided in this embodiment includes: a processor 501 and a memory 502; wherein:

[0152] The memory 502 is used to store computer programs, and the memory may also be a flash memory.

[0153] The processor 501 is used to execute the execution instructions stored in the memory to implement the steps corresponding to each execution subject in the above method.

[0154] For example, when the electronic device 500 is a sensor group, the memory 502 is used to store computer programs, and the processor 501 is used to execute the execution instructions stored in the memory to collect the first vibration data and the first gas concentration data of the target IoT terminal device.

[0155] For example, when the electronic device 500 is a communication module group, the memory 502 is used to store computer programs, and the processor 501 is used to execute execution instructions stored in the memory to upload the first vibration data and the first gas concentration data to the edge computing device.

[0156] For another example, when the electronic device 500 is an edge computing device, the memory 502 is used to store computer programs, and the processor 501 is used to execute execution instructions stored in the memory, receive first vibration data and first gas concentration data, extract first vibration features from the first vibration data, and extract first gas concentration features from the first gas concentration data, input the first vibration features and the first gas concentration features into a pre-trained device fault prediction model to obtain a fault prediction result of the target IoT terminal device; and output a fault diagnosis report based on the fault prediction result of the target IoT terminal device.

[0157] Optionally, the memory 502 may be independent or integrated with the processor 501 .

[0158] When the memory 502 is a device independent of the processor 501, the electronic device 500 may further include:

[0159] The bus 503 is used to connect the memory 502 and the processor 501 .

[0160] This embodiment also provides a computer-readable storage medium, in which a computer program is stored. When at least one processor of an electronic device executes the computer program, the electronic device executes the methods provided in the above-mentioned various embodiments.

[0161] This embodiment also provides a program product, which includes a computer program stored in a readable storage medium. At least one processor of the electronic device can read the computer program from the readable storage medium, and at least one processor executes the computer program so that the electronic device implements the methods provided in the above various embodiments.

[0162] Those skilled in the art will readily appreciate other implementations of the embodiments of the present invention after considering the specification and practicing the invention disclosed herein. The embodiments of the present invention are intended to cover any variations, uses, or adaptations of the embodiments of the present invention that follow the general principles of the embodiments of the present invention and include common knowledge or customary technical means in the art that are not disclosed in the embodiments of the present invention. The specification and embodiments are to be regarded as exemplary only, and the true scope and spirit of the embodiments of the present invention are indicated by the claims.

[0163] It should be understood that the embodiments of the present invention are not limited to the precise structures described above and shown in the drawings, and various modifications and changes may be made without departing from the scope thereof. The scope of the embodiments of the present invention is limited only by the appended claims.

Claims

1. A multi-sensor detection platform based on the Internet of Things, characterized in that: It includes a sensor group, a communication module group and an edge computing device; the sensor group and the edge computing device are communicatively connected through the communication module group; The sensor group is used to collect the first vibration data and the first gas concentration data of the target IoT terminal device, and upload the first vibration data and the first gas concentration data to the edge computing device through the communication module group; The edge computing device is used to receive the first vibration data and the first gas concentration data, extract the first vibration feature from the first vibration data, and extract the first gas concentration feature from the first gas concentration data, input the first vibration feature and the first gas concentration feature into a pre-trained device fault prediction model to obtain a fault prediction result of the target IoT terminal device; and output a fault diagnosis report based on the fault prediction result of the target IoT terminal device.

2. The platform according to claim 1, characterized in that The edge computing device extracts a first vibration feature from the first vibration data, extracts a first gas concentration feature from the first gas concentration data, inputs the first vibration feature and the first gas concentration feature into a pre-trained device fault prediction model, and obtains a fault prediction result of the target IoT terminal device, specifically for: using a low-pass filter to remove high-frequency noise in the first vibration data, using a moving average method to smooth the first vibration data after the high-frequency noise is removed, and normalizing the first vibration data after the smoothing process; Extracting first vibration data time domain features and first vibration data frequency domain features from the first vibration data after normalization, wherein the first vibration data time domain features include mean, variance, peak, kurtosis and skewness; and the first vibration data frequency domain features include spectrum and power spectrum density; Using a high-pass filter to remove low-frequency drift in the first gas concentration data, using an exponential smoothing method to smooth the first gas concentration data after the low-frequency drift is removed, and normalizing the first gas concentration data after the smoothing process; extracting the first gas concentration feature from the first gas concentration data after normalization, wherein the first gas concentration feature includes a concentration change rate, a gas concentration peak value, and a gas concentration change trend; combining the first vibration data time domain feature, the first vibration feature frequency domain feature, and the first gas concentration feature into a first multimodal feature vector; The first multimodal feature vector is input into the device fault prediction model to obtain a fault prediction result of the target IoT terminal device.

3. The platform according to claim 2, characterized in that: When the sampling frequencies of the first vibration data time domain feature, the first vibration feature frequency domain feature, and the first gas concentration feature are different, the edge computing device is further used to: The first vibration data time domain feature, the first vibration feature frequency domain feature and the first gas concentration feature are aligned on the same time axis.

4. The platform according to claim 2, characterized in that: When the edge computing module combines the first vibration data time domain feature, the first vibration feature frequency domain feature, and the first gas concentration feature into a first multimodal feature vector, it is specifically used to: splicing the first vibration data time domain feature, the first vibration feature frequency domain feature and the first gas concentration feature in sequence into the first multimodal feature vector; or, Using a feature selection model, the importance of each feature of the first vibration data time domain feature, the first vibration feature frequency domain feature and the first gas concentration feature is evaluated, and the least important features are gradually deleted until the number of remaining features reaches a preset number, and the remaining preset number of features are concatenated into the first multimodal feature vector; the first multimodal feature vector includes the most relevant features extracted from the first vibration data time domain feature, the first vibration feature frequency domain feature and the first gas concentration feature.

5. The platform according to claim 1, characterized in that: The platform also includes: a cloud server; The cloud server is used to collect historical vibration data and historical gas concentration data of the target IoT terminal device from the edge computing device, the historical vibration data and the historical gas concentration data include data of the target IoT terminal device in a normal state and data in a faulty state, and use the collected historical vibration data and historical gas concentration data to train a fault prediction model for the device to be trained, and send the trained device fault prediction model to the edge computing device.

6. The platform according to claim 5, characterized in that When training the equipment failure prediction model to be trained, the cloud server is specifically used to: Judging whether the sample data has a fault according to a preset time window to generate labels for the historical vibration data and the historical gas concentration data; wherein each of the preset time windows corresponds to one of the sample data, and each of the sample data includes the historical vibration data and the historical gas concentration data; Integrate the labeled sample data into a training data set, wherein each sample data in the training data set includes the historical vibration data, the historical gas concentration data and a corresponding label; Dividing the training data set into a training set and a test set; Using the sample data in the training set to train the constructed basic fault prediction model to obtain the trained basic fault prediction model; Using the trained basic fault prediction model to predict the data in the test set, and calculating the performance index of the trained basic fault prediction model according to the prediction result; Generating a ROC curve according to the calculated performance index to evaluate the performance of the trained basic fault prediction model; When the performance evaluation of the basic fault prediction model after training is qualified, the equipment fault prediction model is obtained.

7. The platform according to claim 1, characterized in that The target IoT terminal device includes a generator and a transformer; The sensor group includes a first vibration sensor arranged on the generator rotor, a second vibration sensor arranged on the generator stator, a third vibration sensor arranged on the transformer core, a fourth vibration sensor arranged on the transformer winding, a first gas sensor arranged around the generator, and a second gas sensor arranged in the transformer oil; The communication module group includes a first communication module integrated in the first vibration sensor, a second communication module integrated in the second vibration sensor, a third communication module integrated in the third vibration sensor, a fourth communication module integrated in the fourth vibration sensor, a fifth communication module integrated in the first gas sensor, and a sixth communication module integrated in the second gas sensor.

8. A multi-sensor detection method based on the Internet of Things, applied to the multi-sensor detection platform based on the Internet of Things as described in any one of claims 1 to 7, characterized in that: The method comprises: The sensor group collects first vibration data and first gas concentration data of the target IoT terminal device; The sensor group uploads the first vibration data and the first gas concentration data to the edge computing device through the communication module group; The edge computing device receives the first vibration data and the first gas concentration data, extracts a first vibration feature from the first vibration data, extracts a first gas concentration feature from the first gas concentration data, inputs the first vibration feature and the first gas concentration feature into a pre-trained device fault prediction model, and obtains a fault prediction result of the target IoT terminal device; The edge computing device outputs a fault diagnosis report according to the fault prediction result of the target IoT terminal device.

9. The method according to claim 8, characterized in that The edge computing device extracts a first vibration feature from the first vibration data, and extracts a first gas concentration feature from the first gas concentration data, and inputs the first vibration feature and the first gas concentration feature into a pre-trained device fault prediction model to obtain a fault prediction result of the target IoT terminal device, including: The edge computing device uses a low-pass filter to remove high-frequency noise in the first vibration data, uses a moving average method to smooth the first vibration data from which the high-frequency noise has been removed, and normalizes the first vibration data after the smoothing process; The edge computing device extracts first vibration data time domain features and first vibration data frequency domain features from the first vibration data after normalization, wherein the first vibration data time domain features include mean, variance, peak, kurtosis and skewness; and the first vibration data frequency domain features include spectrum and power spectrum density; The edge computing device uses a high-pass filter to remove low-frequency drift in the first gas concentration data, uses an exponential smoothing method to smooth the first gas concentration data after the low-frequency drift is removed, and normalizes the first gas concentration data after the smoothing process; The edge computing device extracts the first gas concentration feature from the first gas concentration data after normalization, where the first gas concentration feature includes a concentration change rate, a gas concentration peak value, and a gas concentration change trend; The edge computing device combines the first vibration data time domain feature, the first vibration feature frequency domain feature, and the first gas concentration feature into a first multimodal feature vector; The edge computing device inputs the first multimodal feature vector into the device fault prediction model to obtain a fault prediction result of the target IoT terminal device.

10. The method according to claim 9, characterized in that The method further comprises: When the sampling frequencies of the first vibration data time domain feature, the first vibration feature frequency domain feature and the first gas concentration feature are different, the edge computing device aligns the first vibration data time domain feature, the first vibration feature frequency domain feature and the first gas concentration feature on the same time axis.

Images