A smart monitoring method and system for motor rotor production based on the Internet of Things

Abstract

This invention belongs to the field of electric vehicle device and component manufacturing technology, specifically relating to an intelligent monitoring method and system for motor rotor production based on the Internet of Things. The method includes: setting benchmark values ​​for vibration and current load using the third and fourth quartiles of historical data; calculating the vibration risk factor per second in real time; and weighting the current load ratio accordingly to obtain the instantaneous operating intensity; subsequently, fusing the average intensity values ​​of short, medium, and long-term windows, dynamically determining the weights through the window change rate, calculating the cumulative operating intensity, and comparing this intensity with historical cumulative normal intensity to obtain an abnormal risk factor; finally, dynamically scaling the benchmark alarm threshold of the LSTM self-encoder using the abnormal risk factor, and achieving intelligent monitoring of electronic rotor production by comparing the reconstruction error with the dynamic threshold in real time.

Description

Technical Field

[0001] This invention relates to the field of electric vehicle device and component manufacturing technology, and in particular to an intelligent monitoring method and system for motor rotor production based on the Internet of Things. Background Technology

[0002] In the production process of motor rotors, the stable operation of key equipment such as winding machines, die-casting machines, and dynamic balancing machines is the core to ensuring the product quality and production efficiency of motor rotors. By deploying various types of sensors on these devices and using Internet of Things (IoT) technology for data collection and transmission, remote monitoring of the equipment's operating status can be achieved.

[0003] Long Short-Term Memory (LSTM) autoencoders are a type of deep learning model widely used for anomaly detection in time-series data. An encoder network compresses the input time-series data into a low-dimensional feature vector, and a decoder network reconstructs the original input sequence from this feature vector. By training on a large amount of data from equipment operating without anomalies, the model learns the inherent patterns of normal operating modes. During monitoring, the reconstruction error between the input production time-series data and the model's reconstructed data is calculated to determine whether the equipment's current behavior is abnormal, thus enabling the monitoring of abnormal conditions in the motor rotor production process.

[0004] Existing LSTM self-encoders typically use a globally fixed reconstruction error threshold to determine whether to trigger an alarm. However, in real motor rotor production scenarios, the operating conditions of the equipment are dynamically changing. When the equipment is running under high load, the normal fluctuation range of the sensor signal itself will increase, which may cause the reconstruction error to temporarily increase and trigger false alarms. When the equipment is running under low load, a small, early abnormal signal may be missed because its absolute value is small and fails to break through the fixed high threshold. Summary of the Invention

[0005] To address the technical problem that existing LSTM self-encoders typically use a globally fixed reconstruction error threshold to determine whether to alarm, the fixed threshold cannot adapt to the dynamic changes of the equipment under different loads, resulting in a temporary increase in reconstruction error and triggering false alarms when operating under high loads, and the failure to break through the fixed high threshold and thus missing alarms when operating under low loads, this invention provides an intelligent monitoring method and system for motor rotor production based on the Internet of Things.

[0006] In a first aspect, the present invention provides an intelligent monitoring method for motor rotor production based on the Internet of Things, comprising: obtaining a vibration reference value and a current load reference value based on the third and fourth quartiles of historical operating condition data; obtaining a vibration risk factor per second based on the difference between the vibration data per second and the vibration reference value in the operating condition data; weighting the ratio of the current load data per second to the current load reference value using the vibration risk factor per second to obtain the instantaneous operating intensity of the equipment per second; obtaining a window change rate based on the difference in the average instantaneous operating intensity between adjacent moments within different ranges of windows, obtaining the dynamic weight corresponding to each window, and weighting the average instantaneous operating intensity of the equipment within each window to obtain the cumulative operating intensity of the equipment per second; obtaining the cumulative normal intensity based on the third and fourth quartiles of historical operating condition data; obtaining an abnormal risk factor per second based on the difference between the cumulative operating intensity and the cumulative normal intensity of the equipment per second; obtaining a dynamic threshold per second based on the abnormal risk factor and the reference alarm threshold of the equipment per second; calculating the mean square error between the input instantaneous operating intensity sequence and the output reconstructed sequence per second, comparing the reconstructed error with the dynamic threshold in real time, and obtaining the intelligent monitoring result of motor rotor production.

[0007] This invention provides a more stable reference by using the third and fourth quartiles of historical data as benchmark values. It innovatively introduces a vibration risk factor-weighted current load ratio to generate an instantaneous operating intensity index that comprehensively reflects the overall pressure of the equipment. It uses multi-time window change rate dynamic weighting to calculate the cumulative operating intensity, achieving intelligent trend tracking. By comparing the cumulative intensity with the normal level, it obtains the abnormal risk factor, and finally achieves dynamic adaptive adjustment of the threshold: when the risk is high, the threshold is relaxed to reduce false alarms, and when the risk is low, the threshold is tightened to avoid missed alarms, significantly improving the accuracy of motor rotor production monitoring.

[0008] Preferably, obtaining the vibration risk factor per second based on the difference between the vibration data per second in the operating condition data and the vibration reference value includes: Where, ω a (t) is the vibration data a corresponding to the t-th second. t The vibration risk factor; s represents the vibration sensitivity threshold of the vibration risk factor; exp is an exponential function with the natural constant e as its base; a t This is the vibration data corresponding to second t; a T It is the vibration reference value.

[0009] This method introduces a smooth transition mechanism based on the Sigmoid function, mapping the relative deviation between vibration data and the benchmark value to a range of 1 to 2. This effectively avoids the instability caused by sudden changes near the critical value in traditional hard threshold judgment methods. The vibration sensitivity threshold precisely controls the critical point of risk growth, enabling the system to adaptively adjust its risk sensitivity according to changes in vibration intensity: when the vibration intensity is close to the benchmark value, the risk factor tends to 1, and the system maintains a basic monitoring state; while when the vibration significantly exceeds the benchmark value, the risk factor smoothly approaches 2, reflecting the upward trend of risk in a timely and continuous manner. This provides a stable and reasonable quantitative basis for subsequent dynamic threshold adjustments, thereby significantly enhancing the monitoring system's ability to perceive vibration anomalies and its response smoothness.

[0010] Preferably, the step of weighting the ratio of the current load data to the current load reference value using a vibration risk factor per second to obtain the instantaneous operating intensity of the equipment per second includes: Among them, P t b is the instantaneous running intensity corresponding to second t; t This is the current load data corresponding to second t; b T It is the current load reference value; ω a (t) is the vibration data a corresponding to the t-th second. t Vibration risk factors.

[0011] This method achieves coordinated evaluation of two key parameters, vibration and current, by using the vibration risk factor as the weight of the current load ratio, thus overcoming the limitations of traditional single-parameter monitoring. When the vibration level is normal, the operating intensity is mainly determined by the current load, and the system maintains a basic monitoring state. However, when the vibration level rises abnormally, the vibration risk factor increases, which amplifies the contribution of the current load, thereby significantly increasing the value of the instantaneous operating intensity. This reflects that under high vibration risk, even a normal current load may mean higher overall pressure, enabling the system to more comprehensively capture the overall operating intensity of the equipment.

[0012] Preferably, the step of weighting the average instantaneous operating intensity of the device within each window to obtain the cumulative operating intensity of the device per second includes: Among them, Q t It is the cumulative running intensity corresponding to second t; It is the average value of all instantaneous running intensities within the short-term window; It is the average value of all instantaneous running intensities within the medium-term window; It is the average of all instantaneous running intensities within the long-term window; β i It represents the dynamic weight; i is the index of the dynamic weight, which can be 1, 2, or 3.

[0013] This method integrates the average equipment operating intensity within short-term, medium-term, and long-term time windows and introduces dynamic weighting coefficients for weighted calculation, thereby achieving multi-scale dynamic evaluation of equipment operating status. This processing can both keenly capture instantaneous abnormal fluctuations and smooth out random disturbances, thus significantly improving the comprehensiveness and reliability of operating intensity characterization.

[0014] Preferably, obtaining the abnormal risk factor of the equipment per second based on the difference between the cumulative operating intensity and the cumulative normal intensity of the equipment per second includes: Among them, Y t This represents the abnormal risk factor of the device at second t; max() indicates taking the maximum value; Q t Q is the cumulative running intensity corresponding to second t; T ε is the cumulative normal intensity; ε is the sensitivity coefficient.

[0015] This method calculates the deviation between the cumulative operating intensity and the preset normal intensity, and uses the max() function to ensure that only the abnormal intensity is sensitively monitored, effectively avoiding false alarms. The introduction of an adjustable sensitivity coefficient makes the abnormal risk assessment mechanism not only flexibly adapt to the early warning needs of different production environments, but also significantly enhances the ability to identify potential faults and its real-time performance.

[0016] Secondly, the present invention provides an intelligent monitoring system for motor rotor production based on the Internet of Things, comprising: a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned intelligent monitoring method for motor rotor production based on the Internet of Things is implemented.

[0017] By adopting the above technical solution, a computer program is generated from the above-mentioned intelligent monitoring method for motor rotor production based on the Internet of Things, and stored in the memory so that it can be loaded and executed by the processor. In this way, a terminal device can be made based on the memory and the processor for convenient use.

[0018] The beneficial effects of this invention are as follows: Using the third and fourth quartiles of historical operating data as the benchmark values ​​for vibration and current load avoids interference from extreme outliers and provides a stable and reliable reference for all subsequent calculations; innovatively introducing a vibration risk factor and weighting the current load ratio yields a more comprehensive reflection of the instantaneous operating intensity, effectively integrating multi-source heterogeneous data; dynamically determining the weight of each window by analyzing the rate of change of instantaneous operating intensity within different time windows, and then weighted averaging to obtain the cumulative operating intensity, allows the system to rely more on long-term trends to maintain stability when operating conditions are stable, and to rely more on short-term windows to quickly capture anomalies when operating conditions change drastically, thus achieving better monitoring of equipment performance. The system adaptively captures cumulative losses and trend changes; it compares cumulative operating intensity with cumulative normal intensity to obtain an anomaly risk factor, reflecting the degree to which the current cumulative operating intensity deviates from the upper limit of the historical normal level, making the system evaluation consistent and comparable; it uses the anomaly risk factor to dynamically scale the baseline alarm threshold of the LSTM autoencoder and compares the reconstruction error with the dynamic threshold in real time. When the anomaly risk factor increases, the dynamic threshold automatically widens, effectively preventing normal high load fluctuations from being misjudged as anomalies and reducing the false alarm rate; when the anomaly risk factor decreases, the dynamic threshold automatically tightens, making the system more sensitive to minor anomalies, reducing the false alarm rate, and thus obtaining more accurate intelligent monitoring results for electronic rotor production. Attached Figure Description

[0019] Figure 1 This is a schematic flowchart illustrating an intelligent monitoring method for motor rotor production based on the Internet of Things (IoT) according to the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0022] This invention discloses an intelligent monitoring method and system for motor rotor production based on the Internet of Things (IoT), referring to... Figure 1 This includes steps S1 to S4:

[0023] S1. Real-time data collection of operating conditions of key production equipment for motor rotors via IoT gateway, followed by preprocessing.

[0024] It should be noted that various types of sensors, including vibration sensors and current sensors, are deployed on the key production equipment of the motor rotor to collect the equipment's operating condition data in real time through an IoT gateway. The vibration sensors mainly monitor the vibration changes during equipment operation, while the current sensors are used to monitor the load changes of the equipment. Since the collected operating condition data has problems such as noise interference, non-normalization, and asynchronous sampling, it is necessary to preprocess the operating condition data to improve its quality.

[0025] Based on this, vibration sensors are installed at key vibration points in the critical production equipment of the motor rotor to collect vibration signals at a high-frequency sampling rate of 10kHz, and the root mean square value of all vibration amplitude values ​​in each second is used as the vibration data for each second; current sensors are installed at the motor power line in the critical production equipment to collect the current load data of the equipment every second.

[0026] Specifically, the collected vibration data and current load data are used as the operating condition data of the equipment. The collected operating condition data is transmitted to the IoT gateway via industrial Ethernet, and the edge computing gateway is used for preliminary data processing and caching.

[0027] Furthermore, a Gaussian filtering algorithm is used to process the operating condition data, retaining effective features while removing high-frequency noise; timestamp alignment is performed on data streams from different sensors to ensure that data at the same time point corresponds to the equipment status at the same moment; a maximum and minimum value normalization method is used to transform the operating condition data into the range of [0,1]; missing data is filled with the average of its adjacent data.

[0028] After the above processing, the preprocessed operating condition data is obtained and used as input for subsequent steps. The above methods are all existing technologies and will not be described in detail here.

[0029] S2. Based on the vibration data per second in the operating condition data, construct the vibration risk factor per second, and weight the current load data per second to obtain the instantaneous operating intensity of the equipment per second.

[0030] It should be noted that the instantaneous operating intensity of the equipment is used to comprehensively characterize the level of mechanical stress and energy consumption that the equipment experiences within any given second. The instantaneous operating intensity of the equipment per second is characterized by current load data, as it directly reflects the work done by the equipment to overcome processing resistance during the production of the motor rotor. Vibration data serves as an auxiliary feature, used to verify and supplement the current load from a dynamic response perspective: under healthy conditions, high loads are usually only accompanied by a slight increase in vibration; however, when early faults exist, the vibration response will be abnormally amplified even under normal loads. Therefore, vibration data can provide important supplementary information for operating intensity.

[0031] Specifically, taking the vibration data a corresponding to the t-th second as an example. t and current load data b t Taking this as an example for analysis: The third and fourth quartiles of all vibration data collected during the production of motor rotors in a healthy state of the equipment are used as the vibration reference value 'a'. T The third quartile of all current load data is used as the current load reference value b. T Vibration reference value a T and current load reference value b T The third and fourth quartiles represent a relatively high load level during normal equipment operation; they can effectively eliminate the interference of occasional extreme values.

[0032] Furthermore, calculate the vibration data a corresponding to the t-th second. t Vibration risk factor ω a (t), for details please refer to the following relation:

[0033]

[0034] Where, ω a (t) is the vibration data a corresponding to the t-th second. t The vibration risk factor; 1 represents the base value of the vibration risk factor; s represents the vibration sensitivity threshold of the vibration risk factor; exp is an exponential function with the natural constant e as its base; a t This is the vibration data corresponding to second t; a T It is the vibration reference value; when the vibration data a corresponds to the t-th second... t Far below the vibration reference value a T At that time, vibration risk factor ω a (t) approaches 1, indicating the vibration data a corresponding to the t-th second. t Without generating additional risk gain, the instantaneous operating intensity at second t is determined solely by the current load; when the ratio When the vibration sensitivity threshold is exceeded, the vibration risk factor begins to increase rapidly and non-linearly.

[0035] The specific value of the vibration sensitivity threshold of the vibration risk factor can be set according to the actual application scenario and requirements, and the value range of the vibration sensitivity threshold of the vibration risk factor is [1,2]. In this invention, the vibration sensitivity threshold of the vibration risk factor is set to s = 1.2, which means that when the vibration data a corresponds to the t-th second... t Greater than the vibration reference value a T When it reaches 20% of its maximum value, the vibration data a corresponding to second t is... t The corresponding vibration risk factor began to grow rapidly and non-linearly.

[0036] Furthermore, calculate the instantaneous running intensity P corresponding to the t-th second. t For details, please refer to the following relation:

[0037]

[0038] Among them, P t b is the instantaneous running intensity corresponding to second t; t This is the current load data corresponding to second t; b T It is the current load reference value; This is normalized current load data, reflecting the energy consumption level of the equipment at second t during the production of the motor rotor; ω a (t) is the vibration data a corresponding to the t-th second. t The vibration risk factor is used to quantify the risk impact of vibration on the equipment at second t.

[0039] Furthermore, repeat the above steps to obtain the vibration risk factor of the equipment's vibration data per second, and further obtain the instantaneous operating intensity per second.

[0040] S3. Based on the window change rate within different windows, obtain the dynamic weight corresponding to each window; based on the dynamic weight corresponding to each window and the average value of the instantaneous operating intensity, obtain the cumulative operating intensity of the equipment per second; combined with the cumulative normal intensity, obtain the abnormal risk factor of the equipment per second.

[0041] It should be noted that the abnormal risks of the equipment stem from continuous overload operation, rather than instantaneous fluctuations. Therefore, it is necessary to further calculate the cumulative operating intensity of the equipment: the cumulative operating intensity is obtained by calculating the average instantaneous operating intensity of the equipment in different windows. The cumulative operating intensity reflects the cumulative load level borne by the equipment at second t, and is more representative of the characteristics that lead to mechanical fatigue and performance degradation during the production of motor rotors. The third and fourth quartiles of the cumulative operating intensity corresponding to historical operating condition data are used as the cumulative normal intensity, which represents the upper limit of the load that the equipment can reach under most normal operating conditions. It is a robust reference value that can exclude the interference of occasional extreme operating conditions.

[0042] Specifically, we define short-term, medium-term, and long-term windows with the t-th second as their right endpoint. T is the initial length of the short-term window, representing the number of time points within it. The range of the short-term window is [t-T+1, t], the range of the medium-term window is [t-2T+1, t], and the range of the long-term window is [t-4T+1, t], all in seconds. We denote the indices of the short-term, medium-term, and long-term windows as k, corresponding to k = 1, 2, and 3 respectively. The dynamic weight β corresponding to window k is then... k for:

[0043]

[0044] Where, β k It is the dynamic weight corresponding to window k; It is the average value of all instantaneous running intensities within window k; It is the average value of all instantaneous running intensities within window k after the right endpoint of window k is moved one moment to the left from second t to second t-1. λ is the window change rate of window k; e is the natural constant; λ is the sensitivity coefficient, which ranges from [1,2]. The larger λ is, the more concentrated the weight distribution is on the window with the larger change rate. Therefore, in this invention, the sensitivity coefficient is set to λ = 1.3.

[0045] Furthermore, based on the above operations, the dynamic weights β1, β2, and β3 corresponding to the short-term window, medium-term window, and long-term window are obtained.

[0046] It should be noted that short-term windows can capture instantaneous load fluctuations and reflect the rapid response characteristics of equipment more quickly, such as the start-up and shutdown status of equipment; medium-term windows can smooth short-term load fluctuations and capture load trends over a slightly longer period, such as the status of continuous production cycles; long-term windows can reflect slowly changing load baselines, such as long-term load drift caused by equipment performance degradation; the introduction of exponential functions can amplify the impact of load changes within different windows.

[0047] Specifically, the cumulative operating intensity Q of the equipment at second t. t The relationship is as follows:

[0048]

[0049] Among them, Q t It represents the cumulative operating intensity at second t, reflecting the cumulative load level borne by the equipment at second t. It is the average value of all instantaneous running intensities within the short-term window; It is the average value of all instantaneous running intensities within the medium-term window; It is the average of all instantaneous running intensities within the long-term window; β i It represents the dynamic weight; i is the index of the dynamic weight, which can be 1, 2, or 3. It is the sum of the dynamic weights corresponding to i = 1, 2, and 3.

[0050] The specific value of the initial length can be set according to the actual characteristics of the equipment, such as the mechanical response time. In this invention, it is set to T=60. Based on the above operation, the cumulative operating intensity of the equipment per second is obtained.

[0051] It should be noted that the abnormal risk factor is calculated based on the cumulative excess intensity, which is the cumulative operating intensity at the current moment that exceeds the cumulative normal intensity. When the equipment continues to operate at a level higher than its normal historical mode, the cumulative excess intensity will significantly increase the risk of failure, and the abnormal risk factor of the equipment per second will increase significantly. The increase of the abnormal risk factor of the equipment per second is non-linear, reflecting the actual situation that the risk of small overload increases slowly, while the risk of large overload increases sharply.

[0052] Furthermore, following the same procedure, the third quartile of all cumulative operating intensities from a large amount of historical operating condition data collected when the equipment was in a healthy state and producing motor rotors was used as the cumulative normal intensity Q. T The relationship between the abnormal risk factors of the equipment at second t is as follows:

[0053]

[0054] Among them, Y t Q is the abnormal risk factor of the equipment at second t; t -Q T It is the cumulative excess intensity corresponding to second t, reflecting the cumulative operating intensity Q corresponding to second t. t Exceeding the cumulative normal intensity Q T Part of Q; t Q is the cumulative running intensity corresponding to second t; T It is the cumulative normal intensity; max(0,Q) t -Q T () indicates the maximum value between 0 and the cumulative excess strength; ε is the normalized excess ratio, which measures the relative magnitude of the current cumulative operating intensity to the cumulative normal intensity, making the calculation of abnormal risk factors more comparable; ε is the sensitivity coefficient, and the value range of the sensitivity coefficient is [1,2]. In this invention, the sensitivity coefficient is set to ε = 1.5, which means that the abnormal risk factor of the equipment at second t will increase exponentially by 1.5 as the cumulative excess intensity increases, which can amplify the abnormal situation of the equipment.

[0055] It should be noted that the cumulative running intensity Q corresponding to second t is... t Not exceeding the cumulative normal intensity Q T When the equipment is considered to be operating within the normal range and the cumulative excess intensity is 0, the abnormal risk factor of the equipment at second t is equal to its baseline value of 1, indicating that the equipment has no additional risk at second t.

[0056] Repeat the above steps to obtain the abnormal risk factors of the device per second.

[0057] S4. Based on the abnormal risk factors and baseline alarm thresholds corresponding to the equipment per second, dynamically adjust the dynamic thresholds used to judge abnormalities in the LSTM self-encoder, compare the reconstruction error with the dynamic thresholds in real time, and obtain intelligent monitoring results of motor rotor production.

[0058] It should be noted that LSTM autoencoders learn the normal operating mode of a device by reconstructing the input sequence, and their reconstruction error is a key indicator for identifying anomalies. However, a fixed threshold cannot adapt to the dynamic changes of the device under different loads: when the anomaly risk factor is large, it indicates that the device is under high load or in a potentially abnormal state, and the volatility of its operating data will increase. In this case, the threshold should be appropriately relaxed to avoid misjudging normal high load fluctuations as anomalies and reduce false alarms. When the anomaly risk factor is small, it indicates that the device is in a stable or low-load normal state. In this case, the threshold should be tightened to improve detection sensitivity and ensure that subtle anomalies can be captured in time to prevent missed alarms. By introducing anomaly risk factors to dynamically scale the baseline alarm threshold, the anomaly judgment criteria can adapt to the current operating intensity of the device, improving the accuracy and reliability of the monitoring system.

[0059] Specifically, according to step S1, the corresponding instantaneous operating intensity sequence is calculated using the historical operating condition data of the equipment in a healthy state according to step S2, and used as the training set. With the goal of minimizing the reconstruction mean square error, an LSTM autoencoder is trained to learn the data features of normal operating conditions.

[0060] Furthermore, a sequence of instantaneous running intensity in a healthy state is used as a validation dataset. This validation dataset is input into the trained LSTM autoencoder model to calculate its reconstruction error sequence; the mean μ of this set of reconstruction error sequences is then calculated. e and standard deviation σ e Based on the normal distribution assumption, the baseline alarm threshold θ is set according to the 3σ principle. e θ e =μ e +3σ e ; Baseline alarm threshold θ t This represents the maximum normal fluctuation of the reconstruction error generated by the LSTM autoencoder when the device is in good health.

[0061] Specifically, based on the baseline alarm threshold θ e And the abnormal risk factor of the device at second t, to obtain the dynamic threshold of the device at second t, as detailed in the following formula:

[0062] θ t =θ e ×(Y t ) η ;

[0063] Where, θ t θ is the dynamic threshold of the device at second t; e It is the baseline alarm threshold; Y t η is the abnormal risk factor corresponding to the device in second t; η is the scaling factor; when the abnormal risk factor corresponding to the device in second t is larger, the dynamic threshold corresponding to the device in second t is larger, thus relaxing the alarm conditions; conversely, it is tightened.

[0064] The specific value of the scaling factor can be set according to the actual application scenario and needs, and the value range of the scaling factor is [1,2]. In this invention, the sensitivity coefficient is set to η=1.5 to adjust the sensitivity of the dynamic threshold to abnormal risk factors.

[0065] Based on the above operations, the dynamic threshold corresponding to the device per second is obtained.

[0066] Furthermore, for the real-time input operating condition data, the instantaneous operating intensity at the current moment is first calculated in step S2. The length of the short-term window in step S3 is recorded as the time window size. For each second, an instantaneous operating intensity sequence containing the size of the most recent time window needs to be constructed and input into the trained LSTM autoencoder model. The trained LSTM autoencoder model will output a corresponding reconstruction sequence. The mean square error between the input instantaneous operating intensity sequence and the output reconstruction sequence for each second is calculated as the reconstruction error for each second.

[0067] Simultaneously, step S3 calculates the cumulative operating intensity and abnormal risk factor at the current moment, and further calculates the dynamic threshold based on the baseline threshold. Finally, the reconstruction error corresponding to each second is compared with the corresponding dynamic threshold in real time to generate the monitoring result of motor rotor production: if the reconstruction error at the current moment is less than or equal to the dynamic threshold at the current moment, it indicates that the equipment is normal and the motor rotor production is normal, so no alarm is triggered; otherwise, it indicates that the equipment is abnormal and the motor rotor production is abnormal, so an alarm operation is triggered.

[0068] An embodiment of the present invention also discloses an intelligent monitoring system for motor rotor production based on the Internet of Things, including a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, an intelligent monitoring method for motor rotor production based on the Internet of Things according to the present invention is implemented.

[0069] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

Claims

1. An intelligent monitoring method for motor rotor production based on the Internet of Things, characterized in that, include: Vibration reference values ​​and current load reference values ​​are obtained based on the third and fourth quartiles of historical operating condition data; Based on the difference between the vibration data per second and the vibration reference value in the operating condition data, the vibration risk factor per second is obtained; by weighting the ratio of the current load data to the current load reference value using the vibration risk factor per second, the instantaneous operating intensity of the equipment per second is obtained. The window change rate is obtained by considering the difference in the average instantaneous running intensity between adjacent time points within different window ranges. The method for obtaining this rate is as follows: set the window change rate based on the difference in the average instantaneous running intensity between adjacent time points within different window ranges. Short-term window, medium-term window, and long-term window, with seconds as the right endpoint. It is the initial length of the short-run window, indicating the number of elements contained within the short-run window. At any given moment, the range of the short-term window is The range of the medium-term window is The range of the long-term window is The indices of the short-term window, medium-term window, and long-term window are denoted as... , respectively corresponding , It is a window The average value of all instantaneous running intensities within the range; It is to open the window The right endpoint from the first Move the second one hour to the left to the next hour. Seconds later, the corresponding window The average value of all instantaneous running intensities within the range; It is a window The rate of change of the window; Obtain the dynamic weights corresponding to each window, including: , It is a window The corresponding dynamic weights, It is a natural constant. It is the sensitivity coefficient. ; The average instantaneous operating intensity of the equipment in each window is weighted to obtain the cumulative operating intensity of the equipment per second. The cumulative normal intensity is obtained from the third and fourth quartiles of historical operating data; the abnormal risk factor of the equipment per second is obtained from the difference between the cumulative operating intensity and the cumulative normal intensity of the equipment per second. Based on the abnormal risk factors and baseline alarm thresholds corresponding to the equipment per second, the dynamic thresholds per second are obtained; the reconstruction error between the input instantaneous operating intensity sequence and the output reconstruction sequence per second is calculated, and the reconstruction error is compared with the dynamic thresholds in real time to obtain the intelligent monitoring results of motor rotor production.

2. The intelligent monitoring method for motor rotor production based on the Internet of Things according to claim 1, characterized in that, The method of obtaining the vibration risk factor per second based on the difference between the vibration data per second in the operating condition data and the vibration reference value includes: ； in, It is the first Vibration data corresponding to seconds Vibration risk factors; The vibration sensitivity threshold represents the vibration risk factor; Based on the natural constant An exponential function with base 0; It is the first Vibration data corresponding to seconds; It is the vibration reference value.

3. The intelligent monitoring method for motor rotor production based on the Internet of Things according to claim 1, characterized in that, The method of weighting the ratio of the current load data to the current load reference value using a vibration risk factor per second to obtain the instantaneous operating intensity of the equipment per second includes: ； in, It is the first The instantaneous operational intensity corresponding to a second; It is the first Current load data corresponding to each second; It is the current load reference value; It is the first Vibration data corresponding to seconds Vibration risk factors.

4. The intelligent monitoring method for motor rotor production based on the Internet of Things according to claim 1, characterized in that, The average instantaneous operating intensity of the device within each window is weighted to obtain the cumulative operating intensity of the device per second. include: ； in, It is the first The cumulative operating intensity corresponding to each second; It is the average value of all instantaneous running intensities within the short-term window; It is the average value of all instantaneous running intensities within the medium-term window; It is the average value of all instantaneous running intensities within the long-term window; It is a dynamic weight; It is the index of the dynamic weight, taking the values ​​1, 2, and 3.

5. The intelligent monitoring method for motor rotor production based on the Internet of Things according to claim 1, characterized in that, The method of obtaining the abnormal risk factor of the equipment per second based on the difference between the cumulative operating intensity and the cumulative normal intensity of the equipment per second includes: ； in, The equipment is in the The abnormal risk factor corresponding to the second; This indicates taking the maximum value; It is the first The cumulative operating intensity corresponding to each second; It is the cumulative normal intensity; It is the sensitivity coefficient.

6. The intelligent monitoring method for motor rotor production based on the Internet of Things according to claim 1, characterized in that, The process of obtaining the dynamic threshold per second based on the abnormal risk factor and baseline alarm threshold corresponding to the device per second includes: ； in, The equipment is in the The dynamic threshold corresponding to seconds; It is the baseline alarm threshold; The equipment is in the The abnormal risk factor corresponding to the second; It is the scaling factor.

7. The intelligent monitoring method for motor rotor production based on the Internet of Things according to claim 1, characterized in that, The real-time comparison of reconstruction error with dynamic thresholds to obtain intelligent monitoring results for motor rotor production includes: If the reconstruction error at the current moment is less than or equal to the dynamic threshold at the current moment, it indicates that the equipment is normal and the motor rotor is producing normally, so no alarm is triggered; otherwise, it indicates that the equipment is abnormal and the motor rotor is producing abnormally, so an alarm is triggered.

8. An intelligent monitoring system for motor rotor production based on the Internet of Things, characterized in that, include: A processor and a memory, the memory storing computer program instructions, which, when executed by the processor, implement an intelligent monitoring method and system for motor rotor production based on the Internet of Things, as described in any one of claims 1-7.

Images