Diagnostic system for predicting falling of rotating blades of intelligent factory based on Internet of Things

Abstract

The invention relates to a diagnosis system for predicting falling of rotating blades of an intelligent factory based on the Internet of Things, and aims to solve the problems that the existing monitoring dimension is single and only end-stage faults can be identified. The system obtains environmental parameters (liquid level height and motor frequency) and dynamic parameters (vibration signals and current signals) of stirring equipment through acquisition equipment, the vibration signals and the current signals are preprocessed to construct a current feature set and a vibration feature set, and the current feature set and the vibration feature set are normalized to form a fusion feature vector; and calculating the correlation degree deviation of the current characteristics and the vibration characteristics, aggregating the correlation degree deviation and the characteristic average deviation of the fusion characteristic vector to obtain a comprehensive deviation degree, mapping the comprehensive deviation degree into a fault risk index, and hierarchically outputting normal, slight loosening, serious loosening and falling fault states and corresponding operation and maintenance suggestions. The method achieves the full-stage early warning of the blade fault, improves the recognition accuracy, shortens the operation and maintenance time, reduces the production loss, and is suitable for multiple working conditions.

Description

Technical Field

[0001] This invention relates to the field of mixing equipment technology, and more specifically to a diagnostic system for predicting the fall of rotating blades in an IoT-based smart factory. Background Technology

[0002] In modern industrial fields such as chemical engineering, water treatment, food processing, and biomedicine, mixing equipment is widely used as a core process equipment in key production processes such as reactor stirring, submersible liquid mixing, and material mass and heat transfer. It is a fundamental piece of equipment to ensure continuous production processes and stable product quality.

[0003] In existing factories, mixing equipment encompasses various types, including reactor agitators, submersible agitators, and vertical mixing agitators. Their core structure typically includes a drive motor, a main shaft, and evenly distributed rotating blades (the blades are perpendicularly connected to the main shaft at 90° angles). Most of these devices require long-term immersion in liquid media (such as chemical raw materials, wastewater, and food slurries). This type of equipment usually needs to meet the continuous operation requirements of the factory. Especially in the context of large-scale smart factory production, the deployment of mixing equipment in a single plant can reach dozens to hundreds of units, distributed across different workshops or outdoor treatment areas (such as submersible agitators in wastewater treatment ponds), forming a cross-regional, multi-condition equipment cluster.

[0004] The stable operation of mixing equipment directly determines the continuity of the production process. For example, in chemical reactors, uneven mixing by the blades can lead to insufficient mixing of raw materials, causing violent or incomplete local reactions and affecting product purity. In water treatment, if the blades of a submersible mixer fail, it can cause wastewater sedimentation, reduced aeration efficiency, and disrupt the water treatment cycle. Therefore, factories have extremely high requirements for the reliability of mixing equipment. However, due to the operating environment (liquid immersion, long-term vibration, media corrosion), rotating blades, as vulnerable parts, are prone to fatigue wear and vibration loosening at their connection points (such as screws and flanges), eventually leading to detachment, which is the most common type of failure.

[0005] The current monitoring of the condition of mixing equipment blades in factories relies on the traditional model of manual inspection and periodic shutdown checks. On the one hand, since the blades are immersed in liquid, it is impossible to directly observe their looseness or wear by human eyes. It can only be judged by indirect phenomena such as equipment operating noise and overall vibration, which is easily affected by environmental interference (such as background noise in the workshop and vibration of other equipment) and leads to misjudgment. On the other hand, the inspection has a fixed cycle, which cannot capture the risk of slight loosening, severe loosening, or imminent falling of the blades in real time. It is also impossible to monitor the gradual failure process in real time. Often, the problem is only discovered after the blades have fallen off or the failure has expanded, missing the opportunity for early intervention.

[0006] Prior art 1, such as utility model patent No. 202321181994.2, discloses a stirring shaft rotation detection and alarm system for a stirring device, including a stirring shaft and a transmission component for driving the stirring shaft to rotate; it also includes: a magnet, a speed detection device, an instrument control device, and an alarm device. The magnet is mounted on the transmission component and is positioned away from the axis of the stirring shaft; a predetermined distance is provided between the speed detection device and the transmission component, and the speed detection device is used to generate magnetic induction with the magnet. By sensing the magnet mounted on the transmission component through the speed detection device, the induced signal is converted into a speed signal, and then the instrument control device visually displays the operating status of the stirring device. When the speed exceeds a limit value, the alarm device emits an audible and visual alarm signal, improving the monitoring capability of the stirring device's operating status, reducing the risk of the stirring shaft being buried by clumps of material in the container when it unexpectedly stops rotating, and reducing the impact of equipment failure on production.

[0007] The solution disclosed in prior art 2 is essentially as follows: a magnet mounted on the stirring transmission component rotates with the shaft. A magnetic sensor in the speed detection device generates an induction signal by cutting magnetic field lines with the magnet, and converts the signal into a speed signal, which is then transmitted to the instrument control device for display. The instrument control device presets a speed limit. When the detected speed signal exceeds the limit, an alarm device is triggered to issue an audible and visual alarm. The visual alarm is displayed on the instrument control device's screen, and the audible alarm is emitted by an audible and visual alarm mounted on the top of the instrument control device, thereby monitoring the actual operating status of the stirring shaft and preventing it from unexpectedly stopping without being detected. The optimal sensing distance between the speed detection device and the rotating magnet is 20mm, and the device itself is also equipped with an indicator light to display the sensor's operating status.

[0008] Prior art 2, such as utility model patent with patent number 202120640087.4, discloses a stirring device for a reaction vessel, including a control module, a stirring drive module, a stirrer, and a stirring detection module. The reaction vessel includes a reaction vessel body with a mechanical sealing layer at the center of the upper part of the reaction vessel body, and the stirrer is installed inside the reaction vessel body. The stirring drive module includes a motor and a reducer, with the motor shaft connected to the reducer and the reducer's output shaft passing through the mechanical sealing layer and connected to the stirrer. The stirring detection module includes a speed sensor, which is installed at the output shaft of the reducer. The control module is connected to the motor, reducer, and speed sensor. By obtaining the speed of the reducer's output shaft through the speed sensor, the control module can accurately control the motor and reducer, preventing excessively rapid acceleration or deceleration, or even sudden stopping of the stirrer. A photoelectric sensor can promptly detect stirrer malfunctions, and a current transformer can promptly detect motor malfunctions.

[0009] The solution disclosed in prior art 2 is essentially as follows: a speed sensor monitors the speed of the reducer output shaft and transmits the signal to the control module; when the speed exceeds the normal range set by the control module, the control module determines that the agitator is operating abnormally and intervenes to prevent excessive acceleration or deceleration. A photoelectric detector monitors the status of the agitator inside the reactor through an explosion-proof transparent window. If the agitator stops rotating or the impeller breaks, the signal processor transmits the photoelectric signal through a wireless transmitting unit, which is received by the control module's wireless receiving unit and triggers an alarm. Three current transformers collect the three-phase operating current of the motor in real time. When the control module detects an imbalance in the three-phase current, it determines that the motor has malfunctioned and triggers a corresponding alarm to ensure continuous and stable operation of the stirring process.

[0010] While existing technologies incorporate automated monitoring, their core monitoring dimensions focus on abnormal speed and motor current imbalance. They can only identify terminal faults such as stoppages and motor failures caused by blade detachment, failing to capture early characteristics like symmetrical blade structural damage and load coupling imbalance. This results in a lack of early warning for the most common fault, blade loosening and detachment, leading to a reactive response only after the fault has escalated. Traditional manual inspections, with their fixed cycles, cannot capture subtle anomalies in the early stages of loosening in real time, often only discovering the problem when the blade has become severely loose or detached.

[0011] Therefore, it is necessary to study a diagnostic system for predicting the fall of rotating blades in an IoT-based smart factory. Summary of the Invention

[0012] Therefore, the purpose of this invention is to provide a diagnostic system for predicting the fall of rotating blades in an IoT-based smart factory. This system effectively solves the problem that existing condition detection equipment has a single monitoring dimension, focusing only on abnormal speed and unbalanced motor current, and cannot capture the progressive fault process of slight loosening, severe loosening, and imminent fall of blades. The warning is only triggered in the terminal stage of the fault.

[0013] To achieve the above objectives, the technical solution adopted by this invention is: a diagnostic system for predicting the fall of rotating blades in an IoT-based smart factory, comprising... Based on the acquisition device deployed on the mixing equipment, environmental parameters and dynamic parameters of the mixing equipment are collected, including vibration information and current signals; After preprocessing the vibration and current signals, current feature sets and vibration feature sets are constructed. The current feature set and vibration feature set are normalized to construct a fusion feature vector with a unified dimension; Calculate the correlation deviation between current characteristics and vibration characteristics; The overall deviation is obtained by combining the aggregate correlation deviation with the average deviation of the fused feature vectors. The overall deviation is mapped to a failure risk index, and the failure status of the rotating blades of the mixing equipment is output based on the failure risk index.

[0014] Furthermore, the environmental parameters include the liquid level and the motor operating frequency.

[0015] Furthermore, the acquisition device includes: a vibration sensor, installed at the root of the main shaft or blades of the mixing equipment, for acquiring vibration acceleration signals of the main shaft and blades; a Hall current sensor, connected in series in the motor power supply circuit of the mixing equipment, for acquiring motor operating current signals; and a liquid level sensor, installed in the mixing chamber of the mixing equipment, for acquiring real-time liquid level signals of the liquid in the chamber.

[0016] Furthermore, a combination of fundamental notch filtering and Kalman filtering is used to filter out high-frequency noise and voltage jump spurious fluctuations in the current signal; the harmonic content of the current signal is extracted based on fast Fourier transform, and the current coefficient of variation and current cumulative deviation of the current signal are calculated using a preset duration sliding window to form a current feature set containing the specific harmonic content, current coefficient of variation, and current cumulative deviation.

[0017] Furthermore, wavelet threshold filtering is used to remove noise interference from the vibration signal and retain the vibration signal associated with blade faults; the blade frequency amplitude value of the vibration signal is extracted based on power spectral density analysis, the peak factor of the vibration signal is calculated, and a vibration feature set containing the blade frequency amplitude value and peak factor is formed.

[0018] Furthermore, the blade frequency amplitude value includes a single frequency amplitude value and a blade multiple frequency amplitude value; the single frequency amplitude value is the vibration amplitude value with the same frequency as the motor operating frequency, which is used to reflect the unbalanced state of the main shaft; the frequency of the blade multiple frequency amplitude value is the product of the number of blades of the mixing equipment and the motor operating frequency, which is used to reflect the degree of symmetrical damage to the blade structure.

[0019] Furthermore, the implementation process for the correlation deviation includes: The first correlation pair is constructed based on the amplitude of the blade's several harmonic frequencies of the vibration signal and the content of specific harmonics in the current signal; A second correlation pair is constructed based on the single-frequency amplitude of the vibration signal and the current variation coefficient of the current signal; Based on the current environmental parameters, the operating conditions are matched, and the feature association benchmark ratio of the first and second association pairs under the corresponding operating conditions is obtained from the database. The feature association benchmark ratio is the standard ratio of the feature values ​​of the association pairs when there is no fault in the corresponding operating conditions. The correlation deviation between the first and second association pairs is calculated using the following procedure: Calculate the real-time association ratio of the two normalized feature values ​​in each association pair; The real-time correlation ratio of each pair of correlations is compared with the feature correlation benchmark ratio of the corresponding working condition to obtain the single correlation degree deviation of each pair of correlations. The average correlation deviation is obtained by averaging the two individual correlation deviations. The average correlation deviation is used to reflect whether the coupling relationship between current and vibration characteristics is disrupted due to blade failure.

[0020] Furthermore, normalized feature values ​​directly related to blade failures are extracted from the fused feature vector of a unified dimension, and the feature average deviation value is obtained by arithmetically averaging the effective feature values.

[0021] Furthermore, the calculation method of the comprehensive deviation includes: multiplying the average deviation of the correlation degree by the average deviation of the feature, and taking the square root of the product to obtain the comprehensive deviation.

[0022] Furthermore, the fault risk index of the fault decision module is the product of the comprehensive deviation and a preset coefficient; based on the fault risk index, the module outputs the fault status and maintenance suggestions for the rotating blades of the mixing equipment, including: When the fault risk index is less than the first risk threshold, it is determined to be in a normal state, no reminder is output, and a suggestion to update the current working condition normal feature database is provided. When the fault risk index is greater than or equal to the first risk threshold and less than the second risk threshold, it is determined to be in a slightly loose state, and an inspection reminder is output, suggesting that the blade connection screws be checked in particular. When the fault risk index is greater than or equal to the second risk threshold and less than the third risk threshold, it is determined to be in a serious loose state, and a shutdown warning is output, and a suggestion to arrange maintenance within the preset maintenance time is given. When the fault risk index is greater than or equal to the third risk threshold, it is determined to be in an imminent fall state, and an emergency stop command is output, suggesting that the mixing equipment be stopped immediately.

[0023] The beneficial effects of the above technical solution are as follows: For proactive early warning and intelligent diagnosis of blade failures in mixing equipment in chemical, water treatment, and other fields, this invention is not limited to traditional single parameters such as rotational speed and current. Instead, it integrates environmental and dynamic parameters. The former covers liquid level and motor frequency, while the latter covers vibration acceleration and motor current, covering all dimensions of operating conditions and fault characteristics, thus avoiding misjudgments caused by non-fault interference. For the progressive fault characteristics of blades, from slight loosening to severe loosening and then to imminent fall, it achieves quantitative analysis of fault severity and full-stage early warning by fusing correlation deviation and characteristic average deviation. The former reflects coupling imbalance, while the latter reflects overall deviation, replacing fixed thresholds. Simultaneously, by combining sensor deployment, remote data transmission, and equipment positioning functions, it establishes a closed loop for data collection, analysis and early warning, and operation and maintenance, solving the problems of low efficiency and difficulty in fault location associated with traditional manual inspections.

[0024] The core advantage of this invention lies in: From passive maintenance to proactive prediction: By integrating multi-dimensional data and using quantitative algorithms, blade fault monitoring is upgraded from terminal alarm to full-stage prediction, filling the technological gap in early warning of slight blade loosening. From single-point monitoring to IoT closed-loop: Leveraging IoT technology to achieve unified monitoring and collaborative operation and maintenance of cross-regional equipment clusters, solving the problems of scattered and difficult-to-manage mixing equipment and inefficient operation and maintenance in large-scale deployments of smart factories. Shifting from experience-based judgment to data-driven approaches: By iteratively optimizing benchmarks and thresholds through a self-learning database, we can break free from reliance on human experience, ensure monitoring accuracy during long-term operation, provide technical support for the continuous and stable operation of mixing equipment, and ultimately reduce factory production risks and maintenance costs.

[0025] This invention effectively distinguishes between non-fault interferences such as power grid fluctuations and pipeline resonance and actual blade faults through dual correlation pair collaborative verification and coupling correlation coefficient verification. The fault identification accuracy is significantly improved, and the misjudgment rate under normal conditions is greatly reduced. Compared with traditional manual inspection, it can detect slight blade loosening in advance, avoiding serious accidents such as blade detachment and main shaft jamming due to delayed warnings, and significantly reducing the loss of a single fault. Combined with equipment location push, it pushes the location of faulty equipment and outputs hierarchical operation and maintenance suggestions at the same time, including spare parts list, maintenance priority, etc., which greatly shortens the time for operation and maintenance personnel to locate equipment, significantly compresses the fault handling cycle, and reduces production interruption losses. Attached Figure Description

[0026] Figure 1 This is a flowchart illustrating the implementation process of the present invention; Figure 2 This is a system implementation block diagram of the present invention; Figure 3 This is a schematic diagram of the data acquisition process; Figure 4 This is a schematic diagram of the data processing flow; Figure 5 Implementation flowchart for correlation deviation; Figure 6 A flowchart for implementing comprehensive deviation. Detailed Implementation

[0027] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments: Example 1: This example aims to provide a diagnostic system for predicting the fall of rotating blades in an IoT-based smart factory. It is primarily used for predicting the fall of rotating blades in mixing equipment within the IoT-based smart factory. Mixing equipment includes reactor agitators, submersible agitators, etc. This system is particularly suitable for intelligent control of submersible agitator control systems. Based on existing submersible agitator control systems, improvements and performance optimizations are made, and a diagnostic device for predicting the fall of rotating blades based on IoT technology is provided. This device can quickly locate abnormalities in the rotating blades of specific equipment and diagnose the predicted fall of rotating blades.

[0028] like Figure 1-6 As shown, a diagnostic system for predicting the fall of rotating blades in an IoT-based smart factory is not limited to single-parameter monitoring. It focuses on the core correlation between structural symmetry damage (vibration characteristics) and load coupling imbalance (current characteristics) of blade faults, avoiding misjudgments due to non-fault interference. It introduces operating condition adaptation logic and progressive quantization to replace fixed thresholds, achieving accurate early warning at all fault stages. At the same time, it combines IoT to realize remote data transmission and equipment positioning, realizing a monitoring and diagnostic system for monitoring, early warning, and operation and maintenance.

[0029] In industrial fields such as chemical engineering and water treatment, mixing equipment is a core component, and the stable operation of its rotating blades directly determines the continuity of production. Traditional methods rely on manual inspections, which cannot distinguish between non-blade faults and abnormalities caused by blade loosening or falling off. The core concept of this embodiment is to break through the bottlenecks of existing mixing equipment blade monitoring, such as single-dimensional monitoring, delayed early warning, and poor adaptability to operating conditions. By integrating multi-source data of environmental parameters (liquid level, motor frequency) and dynamic parameters (current, vibration) through Internet of Things (IoT) technology, and through signal preprocessing, feature fusion, and quantitative calculation, a progressive fault early warning model is constructed to detect slight blade loosening, severe loosening, and impending blade falling off. This enables early identification, accurate positioning, and graded handling, ultimately ensuring the continuous operation of the mixing equipment and reducing failure losses.

[0030] In specific implementation, such as Figure 1 As shown, based on the acquisition equipment deployed on the mixing equipment, environmental and dynamic parameters of the mixing equipment are collected. The dynamic parameters include vibration information and current signals; the environmental parameters include liquid level and motor operating frequency. In implementation, for example, this embodiment targets a submersible mixer (model: QJB4 / 12-620 / 3-480, core parameters: 4 rotating blades, motor rated frequency 50Hz, mixing tank liquid level range 1.5-3m) in a municipal water treatment plant. This equipment is immersed in sewage for a long time, and the blade connecting screws are prone to corrosion from sewage and loosening due to equipment vibration. Traditional manual inspections are conducted once every 24 hours, but the delayed warnings often lead to blade detachment, requiring shutdown and replacement.

[0031] The data acquisition device includes a vibration sensor, which is installed at the root of the main shaft or blades of the mixing equipment to collect vibration acceleration signals of the main shaft and blades. Specifically, in this embodiment, the vibration sensor is installed at the top of the main shaft of the submersible mixer (10cm away from the blades). The vibration of the main shaft directly transmits the state of the blades, avoiding interference from sewage.

[0032] The Hall current sensor, connected in series in the motor power supply circuit of the mixing equipment and installed in the control cabinet, is used to collect the motor operating current signal and reflect the change in blade resistance.

[0033] The liquid level sensor is installed inside the stirring chamber of the mixing equipment to collect real-time liquid level signals of the liquid inside the chamber, obtain the liquid level in real time, and match the normal parameters at different immersion depths.

[0034] The motor frequency acquisition module is integrated into the control cabinet and installed at the output of the motor frequency converter to obtain the actual operating frequency and adapt to non-rated frequency operating conditions.

[0035] The Beidou positioning module is installed on the top of the control cabinet next to the mixing tank to locate the faulty equipment.

[0036] After preprocessing the vibration and current signals, current and vibration feature sets are constructed; in specific implementations, such as... Figure 4 As shown, a combination of fundamental notch filtering and Kalman filtering is used to filter out high-frequency noise and voltage jump spurious fluctuations in the current signal. A 50Hz fundamental notch filter (to remove power grid interference) and a Kalman filter (to remove voltage jump spurious fluctuations, with a process noise covariance Q=0.01 and a measurement noise covariance R=0.1) are used. The third harmonic content is extracted based on FFT, and the current coefficient of variation (CV_I) and current cumulative deviation (ΔI_sum, the sum of deviations between the current and the normal mean within one minute) are calculated using a 1-minute sliding window. The harmonic content of the current signal is extracted based on Fast Fourier Transform, and the current coefficient of variation and current cumulative deviation are calculated using a preset duration sliding window, forming a current feature set containing the specific harmonic content, current coefficient of variation, and current cumulative deviation.

[0037] In this embodiment, wavelet threshold filtering is used to remove noise interference from the vibration signal, retaining the vibration signal associated with blade failure. Based on power spectral density analysis, the blade frequency amplitude value of the vibration signal is extracted, and the peak factor of the vibration signal is calculated, forming a vibration feature set containing the blade frequency amplitude value and the peak factor. The blade frequency amplitude value includes single harmonic amplitude and several harmonic amplitude values. The single harmonic amplitude value is the vibration amplitude at the same frequency as the motor's operating frequency, used to reflect the unbalanced state of the main shaft. The frequency of the several harmonic amplitude values ​​is the product of the number of blades in the mixing equipment and the motor's operating frequency, used to reflect the degree of symmetrical damage to the blade structure.

[0038] In specific implementation, wavelet threshold filtering based on db4 wavelet base with 3-level decomposition is used to filter out pool vibration interference less than 1Hz and motor electromagnetic noise greater than 500Hz, while retaining the blade fault frequency band of 1-500Hz. Based on power spectral density analysis, the first harmonic amplitude (A1, 50Hz, reflecting main shaft imbalance) and the fourth harmonic amplitude (A4, 200Hz=4×50Hz, reflecting symmetrical damage to the blade structure) are extracted, and the peak factor (peak factor = vibration peak value / effective value, reflecting impact fault) is calculated.

[0039] The current feature set and vibration feature set are normalized to construct a fusion feature vector with a unified dimension. In view of the large differences in the magnitude of different features and the scattered fault information, this embodiment eliminates the magnitude interference of parameters such as vibration amplitude (unit: mm / s) and current variation coefficient by unifying the feature scale. At the same time, the scattered fault-related features are integrated into a structured vector, providing a unified data foundation for subsequent correlation calculation and comprehensive deviation measurement.

[0040] In practical implementation, this embodiment adopts the working condition adaptive Min-Max normalization method (different from the traditional Min-Max method with a fixed range). That is, based on the current environmental parameters (liquid level, motor frequency), it retrieves the minimum normal value (Min) and maximum normal value (Max) of the feature under the corresponding working condition from the self-learning database, and maps the feature values ​​to the [0,1] interval; the specific formula is as follows:

[0041] in These are the original collected values ​​of the features. , These are the normal maximum and minimum values ​​of this feature under the current operating conditions; After normalization, the eigenvalues, in specific implementation, if , Take 0, if , Set the value to 1 to avoid outliers interfering with subsequent calculations.

[0042] The core of feature vector fusion is to integrate normalized current features and vibration features in a fixed order to ensure that the vector structure is unique and the feature positions are fixed, providing a unified feature index for subsequent correlation calculation and comprehensive deviation quantification. In implementation, the vector dimension is the sum of the number of current features and the number of vibration features. In this embodiment, it is a 6-dimensional vector. The first 3 dimensions are the normalized current features (3rd harmonic content, current variation coefficient, and current cumulative deviation, in order); the last 3 dimensions are the normalized vibration features (first harmonic amplitude, fourth harmonic amplitude, and peak factor, in order).

[0043] The role of fused feature vectors is to provide a data foundation for calculating the average deviation of features. For example, by directly summing and averaging the six elements in the vector, we can obtain the overall deviation of all fault features. At the same time, it provides feature indexes for correlation calculation: for example, when calculating the correlation pair between the fourth harmonic amplitude (the fifth dimension of the vector) and the third harmonic content (the first dimension of the vector), the feature value can be located directly through the index, avoiding feature matching confusion.

[0044] This embodiment solves the problem of interference of different features by normalization and fusion vector construction, and realizes the structured integration of fault information. It provides accurate and unified data support for subsequent correlation deviation calculation and comprehensive deviation measurement. At the same time, the fusion vector is easy to transmit and store in the Internet of Things. Compared with the scattered original data, the structured vector transmits fewer bytes, which is suitable for the concurrent data transmission needs of multiple devices in smart factories.

[0045] The core concept of this embodiment for calculating the correlation deviation between current characteristics and vibration characteristics is to overcome the limitations of single-feature monitoring, which is susceptible to interference and unable to locate core faults. Based on the strong coupling between the structural state of the mixing equipment blades and the motor load response, two sets of targeted feature correlation pairs are constructed to quantify the degree of imbalance in the vibration and current coordination caused by blade faults. The core logic is: under normal operating conditions, the blade structure is symmetrical, the main shaft is under balanced stress, and the motor load is stable, resulting in a fixed ratio between vibration and current characteristics; when the blades loosen or detach, this coupling relationship is broken, and the ratio deviates from the baseline value. By fusing the deviations of the two sets of correlation pairs, the accurate distinction between non-fault interference (power grid fluctuations, pipeline resonance) and actual blade faults is achieved, providing a core quantitative basis for early warning.

[0046] The methods for calculating correlation deviation include: The first correlation pair is constructed based on the amplitude of the blade's several harmonic frequencies and the specific harmonic content of the current signal. In practice, the amplitude of the blade's several harmonic frequencies (such as the 4th harmonic of a 4-blade device and the 6th harmonic of a 6-blade device): When the blades are evenly distributed in a circle, the amplitude of the blade's several harmonic frequencies of the vibration signal is extremely low (the structure is symmetrical and the force is balanced). When the blades loosen or fall off, the structural symmetry is destroyed, and periodic impacts will be generated during rotation, resulting in a significant increase in the amplitude of the blade's several harmonic frequencies. This feature directly reflects the degree of damage to the symmetry of the blade structure.

[0047] During implementation, the core parameters of the mixing equipment (number of blades) are first obtained through the equipment parameter acquisition module. Motor rated frequency (Motor type) The key frequency characteristics are determined based on the following formula. Leaf frequency ; single frequency multiplication ; in This refers to the real-time operating frequency of the motor.

[0048] The specific harmonic content (usually the 3rd or 5th harmonic, with the 3rd harmonic preferred in this embodiment) is important because the blades are the core load of the motor. Loose blades can lead to uneven load (such as increased force on one blade while decreased force on others), causing magnetomotive force distortion in the motor stator windings and increasing the specific harmonic content in the current signal. This characteristic directly reflects the degree of imbalance in blade load coupling.

[0049] In practice, a harmonic contribution analysis algorithm is used to determine the specific harmonics of the current signal. The specific calculation formula is as follows: Harmonic contribution ; in For the first Total harmonic distortion of the second harmonics The upper limit for harmonic analysis is usually set to 15th order; [selection / selection] The two largest harmonics are selected; the third and fifth harmonics are given priority because the magnetomotive force distortion of the stator winding of the asynchronous motor is most sensitive to this type of harmonic, and they are selected as candidates. Combined with the equipment load characteristics, such as prioritizing the fifth harmonic under heavy load conditions and prioritizing the third harmonic under light load conditions, a specific set of harmonics is finally determined.

[0050] Blade structural symmetry failure (amplitude increase at multiple harmonics) and load coupling imbalance (current harmonic increase) are twin characteristics of blade failure. Under normal operating conditions, the two are in a fixed ratio (structural symmetry, load balance, low harmonics). During a failure, this ratio is broken. Therefore, this correlation can be used to detect structural and load-related anomalies in blade failure.

[0051] A second correlation pair is constructed based on the single-frequency amplitude of the vibration signal and the current variation coefficient of the current signal; Single octave amplitude (the same frequency as the motor's operating frequency, such as the 50Hz vibration amplitude of a 50Hz motor): The main shaft is the mounting carrier for the blades. Loose blades will cause uneven force on the main shaft (such as the eccentric inertial force generated by loose blades), which will cause the single octave amplitude of the main shaft vibration to increase. This characteristic directly reflects the degree of imbalance of the main shaft-blade system.

[0052] The coefficient of variation of current (the ratio of the standard deviation to the mean of the current signal) is a characteristic that directly reflects the degree of fluctuation in the motor load due to imbalance in the spindle blade system (the load changes once per revolution). This causes fluctuations in the operating current and increases the coefficient of variation of current.

[0053] The imbalance between the main shaft and blades (increased amplitude of single octave) and the fluctuation of motor load (increased coefficient of variation of current) are strongly coupled. Under normal operating conditions, the ratio between the two is stable (main shaft is balanced, load is stable, and coefficient of variation is low). When the blades become loose, this ratio becomes unbalanced. Therefore, this correlation can be used to detect the mechanical and electrical synergistic anomalies of blade faults.

[0054] Based on the current environmental parameters, the operating conditions are matched, and the feature association benchmark ratio of the first and second association pairs under the corresponding operating conditions is obtained from the database. The feature association benchmark ratio is the standard ratio of the feature values ​​of the association pairs when there is no fault in the corresponding operating conditions. This embodiment uses environmental parameters (liquid level) Motor operating frequency Using as the core dimension, the K-means clustering algorithm is employed to classify historical fault-free operation data into different operating conditions. The clustering objective function is:

[0055] To preset the number of working condition categories, Let i be the sample set of the i-th type of working condition. The cluster centers for the i-th type of working condition are ultimately formed. The working conditions are further subdivided; the K-Nearest Neighbors (KNN) algorithm is used to match the current working condition, calculate the Euclidean distance between the real-time environmental parameter vector and the cluster center of each working condition, and select the working condition corresponding to the smallest cluster center as the current matching working condition.

[0056] Finally, the correlation degree deviation between the first and second correlation pairs is calculated. This embodiment is based on the structural damage caused by blade failure leading to load imbalance, which in turn triggers a chain reaction of mechanical and electrical synergistic anomalies. Two sets of core feature correlation pairs are selected to ensure the coverage of the key coupling relationship of blade failure.

[0057] The specific calculation method for correlation deviation, such as Figure 5 As shown, it includes: Calculate the real-time correlation ratio of the two normalized feature values ​​in each correlation pair. In implementation, this embodiment first normalizes the feature values ​​(vibration feature V, current feature I) of each correlation pair and calculates the real-time correlation ratio of each correlation pair. The specific formula is as follows: The ratio of first association to real-time association:

[0058] The ratio of the second association to the real-time association:

[0059] In the formula, This represents the normalized value of the harmonic amplitude of the leaf blade. This is the normalized value for the content of a specific subharmonic; This is the normalized value of the single octave amplitude; This is the normalized value of the current variation coefficient. The vibration-to-current ratio, rather than the current-to-vibration ratio, is used because blade failure is essentially a structural-mechanical anomaly (vibration characteristics precede) followed by load anomaly (current characteristics follow). This ratio more directly reflects the changes in the coupling relationship caused by the failure. Normalization ensures the comparability of the proportions of the two sets of correlation pairs and avoids interference from differences in characteristic magnitudes. The real-time correlation ratio of each correlation pair is compared with the characteristic correlation benchmark ratio of the corresponding working condition to obtain the single correlation degree deviation of each correlation pair; the relative deviation is used, the core of which is to eliminate the influence of the difference in the benchmark ratio on the deviation judgment. The specific calculation formula is as follows: The first association's bias towards the degree of single association: ; The second association's bias towards the degree of single association:

[0060] and The current operating condition characteristics are associated with a benchmark ratio. Then, based on a preset deviation level threshold, the deviation levels are divided to initially determine the degree of imbalance in a single set of associated pairs.

[0061] Since the two sets of correlation pairs correspond to the direct dimension (structure-load) and indirect dimension (main shaft-load) of blade failure respectively, with no distinction between primary and secondary, an arithmetic mean method is used for fusion. The average correlation deviation is obtained by averaging the deviations of the two individual correlation degrees. This average correlation deviation reflects whether the coupling relationship between current and vibration characteristics is disrupted due to blade failure. The specific formula is as follows:

[0062] In further implementation, for special equipment (such as large reactor agitators, where the main shaft and blade system are highly rigid and the second correlation pair is more sensitive), a weighted average method can be used. The specific calculation formula is as follows:

[0063] and The weight is determined through offline training (based on the fault identification accuracy of two sets of associated pairs in the fault simulation experiment; the higher the accuracy, the greater the weight).

[0064] like Figure 6 As shown in the figure, the aggregate correlation deviation and the average deviation of the fused feature vectors are used to obtain the comprehensive deviation. The overall deviation is mapped to a failure risk index, and the failure status of the rotating blades of the mixing equipment is output based on the failure risk index. The calculation method for the overall deviation includes: obtaining the feature-averaged deviation value based on the normalized failure correlation feature values ​​in the fused feature vector; the feature-averaged deviation value... The arithmetic mean is used for calculation, and the formula is as follows:

[0065] In the formula, The number of effective features; For the first One effective normalized fault association feature value.

[0066] For agitated equipment with large liquid level fluctuations, dynamic weights can be assigned to liquid level-related features (such as cumulative current deviation) when calculating the characteristic average deviation value. The specific correction formula is as follows:

[0067] Weights are applied to the liquid level.

[0068] All features in the fused feature vector are core features associated with the fault. Current features reflect abnormal loads, and vibration features reflect abnormal structural mechanics. Together, they constitute a complete feature profile of blade faults and are equally important.

[0069] Average correlation deviation The imbalance in the coupling relationship of the focusing current vibration characteristics reflects abnormal coordination between characteristics (such as blade loosening leading to an imbalance in the ratio of structural characteristics to load characteristics). Characteristic mean deviation Focus on the overall deviation of all features to reflect the absolute abnormality of the features themselves (e.g., if all features are outside the normal range, it indicates that the fault has spread). The two complement each other: only A high (coupling imbalance) could indicate a minor fault (such as a slightly loose screw); only A high (overall deviation) may be due to multi-dimensional interference; both high values ​​indicate a serious fault (such as a blade about to fall off), providing coordinated, overall dual-dimensional support for the comprehensive deviation.

[0070] Multiply the average correlation deviation by the average characteristic deviation, and take the square root of the product to obtain the overall deviation.

[0071] The specific calculation formula is as follows;

[0072] The fault risk index of the fault decision module is the product of the comprehensive deviation and the preset coefficient; the specific mapping formula is as follows:

[0073] in This is the failure risk index. These are preset coefficients.

[0074] Based on the failure risk index, the failure status and maintenance recommendations for the rotating blades of the mixing equipment are output. The failure status and maintenance recommendations include: When the fault risk index is less than the first risk threshold, it is determined to be in a normal state, no warning is output, and a suggestion to update the current operating condition normal characteristic database is provided; for example... Based on fault-free operating conditions The system calculates the maximum statistical value, provides maintenance-free reminders, automatically incorporates the current data into the normal feature database, and iteratively optimizes the baseline ratio.

[0075] When the fault risk index is greater than or equal to the first risk threshold and less than the second risk threshold, it is determined to be in a slightly loose state, and an inspection reminder is output, suggesting a focus on checking the blade connection screws; for example... Based on minor loosening fault Minimum value, output inspection reminder, mark the key inspection blade connecting screws, and push the equipment's Beidou positioning coordinates.

[0076] When the fault risk index is greater than or equal to the second risk threshold and less than the third risk threshold, it is determined to be in a severely loose state, a shutdown warning is output, and a suggestion to arrange maintenance within a preset maintenance period is made; for example... Based on severe loosening fault The system will issue a shutdown warning and recommend that maintenance be arranged within 24 hours, while simultaneously sending a spare parts list (such as blade connecting screws).

[0077] When the fault risk index is greater than or equal to the third risk threshold, it is determined to be in an imminent fall state, and an emergency stop command is output, suggesting that the mixing equipment be stopped immediately. For example... Based on the hour before the leaf fell At the minimum value, an emergency stop command is output, forcibly cutting off the motor power supply and simultaneously notifying maintenance personnel to immediately arrive at the site for repair.

[0078] In summary, this embodiment provides a diagnostic system for predicting the fall of rotating blades in an IoT-based smart factory, aiming to solve the problem of existing monitoring systems having only a single dimension and only being able to identify terminal faults. It acquires environmental parameters (liquid level, motor frequency) and dynamic parameters (vibration signals, current signals) of the mixing equipment through data acquisition devices. After preprocessing, the vibration and current signals are used to construct current and vibration feature sets, which are then normalized to form a fused feature vector. The correlation deviation between the current and vibration features is calculated, and the correlation deviation is aggregated with the average deviation of the fused feature vector to obtain a comprehensive deviation degree. This comprehensive deviation degree is mapped to a fault risk index, and the system outputs graded fault states—normal, slightly loose, severely loose, and about to fall—along with corresponding maintenance suggestions. This invention achieves full-stage early warning of blade faults, improves identification accuracy, shortens maintenance time, reduces production losses, and is adaptable to multiple operating conditions.

Claims

1. A diagnostic system for predicting the fall of rotating blades in an IoT-based smart factory, characterized in that: include Based on the acquisition device deployed on the mixing equipment, environmental parameters and dynamic parameters of the mixing equipment are collected, including vibration information and current signals; After preprocessing the vibration and current signals, current feature sets and vibration feature sets are constructed. The current feature set and vibration feature set are normalized to construct a fusion feature vector with a unified dimension; Calculate the correlation deviation between current characteristics and vibration characteristics; The overall deviation is obtained by combining the aggregate correlation deviation with the average deviation of the fused feature vectors. The overall deviation is mapped to a failure risk index, and the failure status of the rotating blades of the mixing equipment is output based on the failure risk index.

2. The diagnostic system for predicting the fall of rotating blades in an IoT-based smart factory according to claim 1, characterized in that: The environmental parameters include the liquid level and the motor operating frequency.

3. The diagnostic system for predicting the fall of rotating blades in an IoT-based smart factory according to claim 2, characterized in that: The acquisition equipment includes: a vibration sensor installed at the root of the main shaft or blades of the mixing equipment to acquire vibration acceleration signals of the main shaft and blades; a Hall current sensor connected in series in the motor power supply circuit of the mixing equipment to acquire motor operating current signals; and a liquid level sensor installed in the mixing chamber of the mixing equipment to acquire real-time liquid level signals of the liquid in the chamber.

4. The diagnostic system for predicting the fall of rotating blades in an IoT-based smart factory according to claim 1, characterized in that: A combination of fundamental notch filtering and Kalman filtering is used to filter out high-frequency noise and voltage jump spurious fluctuations in the current signal. The harmonic content of the current signal is extracted based on fast Fourier transform, and the current coefficient of variation and current cumulative deviation are calculated using a preset time sliding window to form a current feature set containing specific harmonic content, current coefficient of variation, and current cumulative deviation.

5. The diagnostic system for predicting the fall of rotating blades in an IoT-based smart factory according to claim 1, characterized in that: Wavelet threshold filtering is used to remove noise interference from the vibration signal and retain the vibration signal associated with blade faults. The blade frequency amplitude value of the vibration signal is extracted based on power spectral density analysis, and the peak factor of the vibration signal is calculated to form a vibration feature set containing the blade frequency amplitude value and the peak factor.

6. The diagnostic system for predicting the fall of rotating blades in an IoT-based smart factory according to claim 5, characterized in that: The blade frequency amplitude includes single frequency amplitude and blade multiple frequency amplitude; the single frequency amplitude is the vibration amplitude at the same frequency as the motor operating frequency, used to reflect the unbalanced state of the main shaft; the frequency of the blade multiple frequency amplitude is the product of the number of blades of the mixing equipment and the motor operating frequency, used to reflect the degree of symmetrical damage to the blade structure.

7. The diagnostic system for predicting the fall of rotating blades in an IoT-based smart factory according to claim 6, characterized in that: The implementation process for the correlation deviation includes: The first correlation pair is constructed based on the amplitude of the blade's several harmonic frequencies of the vibration signal and the content of specific harmonics in the current signal; A second correlation pair is constructed based on the single-frequency amplitude of the vibration signal and the current variation coefficient of the current signal; Based on the current environmental parameters, the operating conditions are matched, and the feature association benchmark ratio of the first and second association pairs under the corresponding operating conditions is obtained from the database. The feature association benchmark ratio is the standard ratio of the feature values ​​of the association pairs when there is no fault in the corresponding operating conditions. The correlation deviation between the first and second association pairs is calculated using the following procedure: Calculate the real-time association ratio of the two normalized feature values ​​in each association pair; The real-time correlation ratio of each pair of correlations is compared with the feature correlation benchmark ratio of the corresponding working condition to obtain the single correlation degree deviation of each pair of correlations. The average correlation deviation is obtained by averaging the two individual correlation deviations. The average correlation deviation is used to reflect whether the coupling relationship between current and vibration characteristics is disrupted due to blade failure.

8. The diagnostic system for predicting the fall of rotating blades in an IoT-based smart factory according to claim 7, characterized in that: From the fused feature vector of a unified dimension, normalized feature values ​​directly related to blade failure are extracted, and the feature average deviation value is obtained by arithmetically averaging the effective feature values.

9. The diagnostic system for predicting the fall of rotating blades in an IoT-based smart factory according to claim 8, characterized in that: The calculation method of the comprehensive deviation includes: multiplying the average deviation of the correlation degree by the average deviation of the feature, and taking the square root of the product to obtain the comprehensive deviation.

10. The diagnostic system for predicting the fall of rotating blades in an IoT-based smart factory according to any one of claims 1-9, characterized in that: The fault risk index of the fault decision module is the product of the comprehensive deviation degree and the preset coefficient. Based on the failure risk index, the failure status and maintenance recommendations for the rotating blades of the mixing equipment are output. The failure status and maintenance recommendations include: When the fault risk index is less than the first risk threshold, it is determined to be in a normal state, no reminder is output, and a suggestion to update the current working condition normal feature database is provided. When the fault risk index is greater than or equal to the first risk threshold and less than the second risk threshold, it is determined to be in a slightly loose state, and an inspection reminder is output, suggesting that the blade connection screws be checked in particular. When the fault risk index is greater than or equal to the second risk threshold and less than the third risk threshold, it is determined to be in a serious loose state, and a shutdown warning is output, and a suggestion to arrange maintenance within the preset maintenance time is given. When the fault risk index is greater than or equal to the third risk threshold, it is determined to be in an imminent fall state, and an emergency stop command is output, suggesting that the mixing equipment be stopped immediately.

Images