Method and system for online monitoring of a mechanical installation process

Abstract

The application relates to the technical field of electromechanical equipment assembly and monitoring, and discloses an online monitoring method and system for an electromechanical equipment installation process, wherein the method comprises the following steps: acquiring multi-source sensor sampling data; adaptively dividing a tightening stage; performing online identification of tightening abnormalities; performing pre-tightening force estimation by adopting a LeNet type one-dimensional convolutional neural network and a long short-term memory network; performing comprehensive evaluation of tightening quality and matching of installation decisions; and outputting installation decision results. Compared with the prior art which relies on artificial offline sampling inspection and judgment, especially under the condition that the tightening sequence of key bolt groups of large electromechanical equipment is complex, the technical problems of online identification of the bolt tightening process abnormalities and overall evaluation of the bolt group pre-tightening quality cannot be solved. Due to the synchronous construction of multi-source tightening time sequence data and the adaptive labeling of the tightening stage, continuous monitoring of the slip, torque attenuation, tightening sequence deviation and pre-tightening force imbalance state is realized.

Description

Technical Field

[0001] This invention relates to the field of electromechanical equipment assembly and monitoring technology, and in particular to an online monitoring method and system for the installation process of electromechanical equipment. Background Technology

[0002] In the installation of large electromechanical equipment, such as wind turbine generators, heavy machine tools, and aerospace structures, bolted connections are crucial for ensuring structural integrity, load transfer, and operational safety. Traditional bolt tightening quality control primarily relies on preset torque or torque-angle methods, where tightening tools stop working once the target value is reached. However, this method only focuses on the final torque value and fails to reflect the dynamic behavior during the tightening process, such as poor initial fit, stripping during the linear preload stage, material yielding during the plastic deformation stage, and stress relaxation during the final holding stage—all critical stages.

[0003] The shortcomings of existing technologies are mainly reflected in the following aspects: First, most monitoring methods are based on a single torque sensor, resulting in limited data dimensions and difficulty in distinguishing between abnormal angle increases (slippage) and normal plastic deformation; second, the division of tightening stages often relies on fixed thresholds or empirical formulas, failing to adapt to different tightening curve characteristics; third, anomaly identification is mostly post-analysis, lacking online real-time discrimination capabilities, especially for detecting hidden problems such as torque decay and incorrect tightening sequence; finally, direct measurement of preload is costly and difficult to implement online, while model-based indirect estimation methods lack accuracy and robustness, failing to meet the requirements of high-consistency assembly. Therefore, there is an urgent need for a method and system that can adaptively monitor the entire tightening process in complex installation environments and accurately assess the final preload quality. Summary of the Invention

[0004] To address the aforementioned technical shortcomings, the present invention aims to propose an online monitoring method for the installation process of electromechanical equipment. This method addresses the technical problem that existing technologies rely on manual offline sampling inspections, particularly in situations where the tightening sequence of critical bolt groups in large electromechanical equipment is complex. This makes it impossible to achieve online identification of abnormalities in the bolt tightening process and overall assessment of the pre-tightening quality of bolt groups.

[0005] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: The present invention provides an online monitoring method for the installation process of electromechanical equipment.

[0006] The online monitoring method for the installation process of electromechanical equipment includes: Step S10: Acquire multi-source sensor sampling data during the bolt tightening process, perform a tightening time series data construction task based on the multi-source sensor sampling data, and output a standardized tightening time series dataset; Step S20: Based on the standardized tightening time series dataset, the Savitzky-Golay smoothing and first-order difference slope discrimination method are used to perform the adaptive partitioning task of the tightening stage, and the tightening stage annotation results are output. Step S30: Based on the annotation results of the tightening stage, the online identification task of tightening anomalies is performed by using the joint discrimination method of CUSUM cumulative sum test and Pearson correlation degradation discrimination, and the anomaly identification result set is output. Step S40: Based on the standardized tightening time series dataset and the anomaly identification result set, the preload estimation task is performed by combining a LeNet-style one-dimensional convolutional neural network with a long short-term memory network, and the preload quality evaluation result is output. Step S50: Based on the pre-tightening quality assessment results, the TOPSIS comprehensive evaluation method is used to classify the tightening quality level and match the installation decision, and the installation decision results are output.

[0007] Preferably, step S10, which involves acquiring multi-source sensor sampling data during the bolt tightening process, performing a tightening time-series data construction task based on the multi-source sensor sampling data, and outputting a standardized tightening time-series dataset, specifically includes: Step S101: During the same bolt tightening process, the original torque sequence output by the high-precision torque sensor, the original angle sequence output by the angle encoder, and the original length measurement sequence output by the embedded ultrasonic length measurement module are synchronously collected according to the preset sampling frequency, and the hardware timestamps corresponding to each sampling point are recorded respectively. Step S102: Construct a unified time axis based on the hardware timestamp recorded in step S101, and map the original torque sequence, original angle sequence and original length measurement sequence to the unified time axis through linear interpolation to obtain the time-aligned torque sequence, angle sequence and elongation sequence. Step S103: Perform Donoho wavelet soft thresholding on the time-aligned torque sequence, angle sequence, and elongation sequence obtained in step S102, and normalize them according to the upper and lower limits of the corresponding sensor range to obtain a standardized tightening time series dataset containing the normalized torque sequence, normalized angle sequence, and normalized elongation sequence. The normalization process is performed according to the following formula:

[0008] in, This represents any sequence of physical quantities after filtering. This indicates the lower limit of the corresponding sensor's measurement range. This indicates the upper limit of the corresponding sensor's measurement range. This represents the normalized sequence value.

[0009] Preferably, step S20, which involves performing an adaptive partitioning task of the tightening stage based on a standardized tightening time-series dataset using Savitzky-Golay smoothing and a first-order difference slope discrimination method, and outputting the tightening stage annotation results, specifically includes: Step S201: Extract the normalized torque sequence and normalized angle sequence from the standardized tightening time series dataset output in step S10, and use the Savitzky-Golay smoothing method to smooth the normalized torque sequence and normalized angle sequence respectively to obtain the smoothed torque sequence and smoothed angle sequence. Step S202: Based on the smoothed torque sequence and smoothed angle sequence obtained in step S201, the slope of the torque angle curve is obtained by using the first-order difference slope calculation method, and the moving average of the slope is calculated by using a sliding window based on the slope of the torque angle curve. Step S203: Compare the slope moving average obtained in step S202 with the preset derivative threshold, and combine the target torque, torque fluctuation threshold and stable holding time to make point-by-point stage judgment. Merge continuous sampling points with the same stage type into the corresponding tightening stage interval and output the tightening stage labeling result.

[0010] Preferably, in step S203, the moving average slope obtained in step S202 is compared with a preset derivative threshold, and a point-by-point stage determination is made in conjunction with the target torque, torque fluctuation threshold, and stable holding time, specifically including: Preset the first derivative threshold and the second derivative threshold; When the moving average slope is less than the first derivative threshold and the current normalized torque value is less than the target torque, the corresponding sampling point is determined as the sampling point of the initial bonding stage. When the moving average slope is greater than or equal to the first derivative threshold and less than the second derivative threshold, and the current normalized torque value is less than the target torque, the corresponding sampling point is determined to be a sampling point in the linear preload stage. When the average slope moving average is greater than or equal to the second derivative threshold, and the torque fluctuation amplitude within the sliding window is less than the torque fluctuation threshold, the corresponding sampling point is determined to be a sampling point in the plastic deformation stage. When the current normalized torque value reaches the target torque, and within a continuous time interval with a duration not less than the stable holding time, the deviation between the normalized torque value and the target torque does not exceed the torque fluctuation threshold, the corresponding interval is determined as the final holding stage. The determination criteria for the final holding stage are as follows:

[0011] in, Indicates sampling point The normalized torque value at that point, Indicates the target torque. Indicates the torque fluctuation threshold. Indicates the duration of stability; This represents the normalized torque value at any time t within the continuous holding time interval.

[0012] Preferably, step S30, based on the tightening stage annotation results, employs a joint discrimination method combining CUSUM cumulative sum test and Pearson correlation degradation discrimination to perform the online identification task of tightening anomalies and output the anomaly identification result set, specifically includes: Step S301: Based on the start and end times of the linear pre-tightening stage and the plastic deformation stage output in step S20, extract the corresponding stage data segments from the standardized tightening timing dataset output in step S10, use the CUSUM cumulative sum test to detect the deviation between the angle increment and the torque increment, and calculate the slippage index based on the ratio of the cumulative angle increment to the cumulative torque increment. Step S302: Based on the start and end times of the final holding phase output in step S20, extract the holding start torque value and holding end torque value from the standardized tightening timing dataset output in step S10. Use Pearson correlation degradation discrimination to confirm the torque decrease trend in the final holding phase and calculate the torque decay rate. When the torque decay rate is greater than the decay alarm threshold, output the torque decay abnormal mark of the corresponding bolt. Step S303: Based on the bolt numbering order in the preset process sequence and the actual tightening timestamps of each bolt recorded during the acquisition process in step S10, generate the actual tightening sequence. Use Hamming distance to calculate the sequence deviation between the actual tightening sequence and the preset process sequence. When the sequence deviation is greater than the sequence deviation threshold, output a tightening sequence abnormality mark. Then, summarize the tightening sequence abnormality mark, torque attenuation abnormality mark, and tightening sequence abnormality mark into an abnormality identification result set.

[0013] Preferably, step S40, which involves performing the preload estimation task and outputting the preload quality assessment result based on a standardized tightening time-series dataset and anomaly identification result set, using a combination of a LeNet-style one-dimensional convolutional neural network and a long short-term memory network, specifically includes: Step S401: Based on the standardized tightening time series dataset output in step S10, extract the normalized torque sequence, normalized angle sequence, and normalized elongation sequence corresponding to each bolt, and set an abnormal additional channel for bolt samples with abnormal markers according to the abnormal identification result set output in step S30, forming a four-channel time series input sample. Step S402: Input the four-channel time-series input samples obtained in step S401 into the pre-trained LeNet-style one-dimensional convolutional neural network and long short-term memory network combined model. Extract local tightening waveform features through the one-dimensional convolutional layer, extract cross-stage time-series correlation features through the long short-term memory network, and output the estimated value of the preload of each bolt through a fully connected regression layer. Step S403: Based on the estimated preload of each bolt output in step S402, calculate the average preload, standard deviation of preload, and consistency score of preload within the same bolt group, and use the estimated preload of each bolt, the consistency score of preload of the bolt group, and the statistical results of abnormal bolts as the preload quality assessment results.

[0014] Preferably, in step S30, the torque attenuation rate is calculated according to the following formula:

[0015] in, Indicates the start time of the final hold phase. This indicates the end time of the final hold phase. This represents the torque attenuation rate corresponding to the i-th bolt; This represents a correction constant to prevent the denominator from being zero; This represents the normalized torque value corresponding to the start time of the final hold phase; This represents the normalized torque value at the end of the final hold phase.

[0016] The present invention also provides an online monitoring system for the installation process of electromechanical equipment, comprising: The data acquisition and preprocessing module is used to acquire multi-source sensor sampling data during the bolt tightening process, perform a tightening time series data construction task based on the multi-source sensor sampling data, and output a standardized tightening time series dataset. The tightening stage segmentation module is used to perform adaptive segmentation of tightening stages based on a standardized tightening time series dataset, using Savitzky-Golay smoothing and first-order difference slope discrimination method, and outputs tightening stage annotation results. The online anomaly identification module is used to perform online anomaly identification tasks based on the tightening stage annotation results, using a joint discrimination method of CUSUM cumulative sum test and Pearson correlation degradation discrimination, and outputs anomaly identification result set. The consistency evaluation module is used to perform the preload estimation task based on the standardized tightening time series dataset and the anomaly identification result set, using a combination of LeNet-style one-dimensional convolutional neural network and long short-term memory network, and outputs the preload quality evaluation result. The comprehensive evaluation and decision-making module is used to classify tightening quality levels and match installation decisions based on the pre-tightening quality assessment results using the TOPSIS comprehensive evaluation method, and output the installation decision results.

[0017] The present invention also provides an online monitoring device for the installation process of electromechanical equipment, the online monitoring device for the installation process of electromechanical equipment includes: a memory, a processor, and an online monitoring program for the installation process of electromechanical equipment stored in the memory and executable on the processor. When the online monitoring program for the installation process of electromechanical equipment is executed by the processor, it implements an online monitoring method for the installation process of electromechanical equipment.

[0018] The present invention also provides a computer program product, the computer program product including an online monitoring program for the installation process of electromechanical equipment, wherein when the online monitoring program for the installation process of electromechanical equipment is executed by a processor, it implements any one of the methods for online monitoring of the installation process of electromechanical equipment.

[0019] The beneficial effects of this invention are as follows: 1. This invention constructs a standardized time-series dataset of multi-source sensors (torque, angle, elongation) and uses Savitzky-Golay smoothing and first-order differential slope discrimination for adaptive stage division. This enables accurate identification of key tightening stages such as initial bonding, linear pre-tightening, plastic deformation, and final holding, providing an accurate time-series framework for subsequent refined analysis and overcoming the shortcomings of poor adaptability of fixed threshold division.

[0020] 2. This invention achieves online real-time identification of abnormal thread slippage in the linear preload stage and abnormal torque decay in the final holding stage through the combined application of CUSUM cumulative sum test and Pearson correlation degradation judgment. Combined with process sequence comparison to detect tightening sequence errors, it forms a multi-dimensional and highly sensitive anomaly detection system that can promptly detect potential installation defects and prevent operation with defects. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating the first embodiment of an online monitoring method for the installation process of electromechanical equipment according to the present invention. Detailed Implementation

[0022] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0023] 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.

[0024] Example 1: As Figure 1 The diagram shown is a flowchart of the first embodiment of the online monitoring method for the installation process of electromechanical equipment according to the present invention, which presents the first embodiment of the online monitoring method for the installation process of electromechanical equipment according to the present invention.

[0025] In the first embodiment, the online monitoring method for the installation process of electromechanical equipment includes: Step S10: Acquire multi-source sensor sampling data during the bolt tightening process, perform a tightening time series data construction task based on the multi-source sensor sampling data, and output a standardized tightening time series dataset; The "multi-source sensor sampling data" in this step specifically refers to the raw torque, angle, and bolt elongation sequences synchronously collected by a high-precision torque sensor, angle encoder, and embedded ultrasonic length measurement module during the same bolt tightening process. The "tightening time-series data construction task" includes multi-series time alignment based on hardware timestamps, noise reduction using Donoho wavelet soft threshold filtering, and normalization according to the measurement range of each sensor. The output "standardized tightening time-series dataset" is a collection of time-aligned and dimensionlessly standardized torque, angle, and elongation sequences, providing a unified and clean input data foundation for all subsequent analysis steps. This step, as the data cornerstone of the entire monitoring method, transforms the raw, potentially asynchronous, and noisy physical signals into standardized time-series data that can be directly processed by subsequent algorithms through multi-source information fusion and standardized preprocessing. This ensures the accuracy and reliability of advanced analysis functions such as stage division, anomaly identification, and preload estimation, achieving a reliable conversion from physical signals to information features.

[0026] Compared to traditional methods that only use a single torque signal or simple filtering, this invention simultaneously collects three key physical quantities—torque, angle, and elongation—and employs a preprocessing method that combines time alignment and wavelet filtering. This allows for a more comprehensive reflection of the mechanical state changes during the tightening process and effectively suppresses common industrial field pulse noise and random interference.

[0027] For example, during the installation of flange bolts on wind turbine towers, the on-site environment experiences significant vibration, making sensor signals susceptible to interference. This step unifies the fluctuating torque signal output by the torque wrench, the rotation angle signal read by the encoder, and the bolt micro-elongation signal measured by ultrasound onto a time axis with millisecond-level precision. It also filters out high-frequency noise caused by vibration, ultimately yielding a smooth, synchronized standardized curve showing the change of "torque-angle-elongation" over time. This curve fully depicts the dynamic process of a single bolt from the start of tightening to the end of torque holding.

[0028] Step S20: Based on the standardized tightening time series dataset, the Savitzky-Golay smoothing and first-order difference slope discrimination method are used to perform the adaptive partitioning task of the tightening stage, and the tightening stage annotation results are output. The "Savitzky-Golay smoothing" in this step is a data processing method that effectively smooths noise while preserving the original shape characteristics of the curve, and is particularly suitable for processing trend-based signals such as tightening curves. "First-order differential slope discrimination" refers to calculating the first derivative of torque with respect to angle (i.e., dT / dθ), which physically represents the tightening stiffness. The core of "adaptive task partitioning for tightening stages" is based on the smoothed torque-angle curve. By analyzing the trend of its slope (stiffness) change and combining it with process parameters such as target torque and fluctuation threshold, it automatically identifies different mechanical stages of the tightening process, such as the initial contact stage with low stiffness, the linear stage with basically constant stiffness, the plastic stage with a sharp increase in stiffness, and the torque holding stage.

[0029] This step enables intelligent and precise segmentation of the continuous tightening process. It decomposes a continuous tightening action into several stages with clear physical meaning, providing precise time windows and contextual information for subsequent specific analyses of different stages (such as detecting slippage in the linear segment and detecting attenuation in the holding segment). It serves as a bridge connecting raw data with advanced anomaly identification and quality assessment.

[0030] Compared to traditional methods that rely on fixed torque or angle thresholds for stage division, this invention employs an adaptive division method based on curve slope variations. This method can automatically adapt to differences in curve morphology under different bolt specifications, friction coefficients, and tightening strategies. For example, the slopes of the linear segments differ significantly between well-lubricated and dry-friction bolts. The fixed threshold method may misjudge these differences, while this method, by comparing the moving average of the slope with the threshold, can robustly identify the start and end points of each stage, improving the universality and accuracy of stage division.

[0031] For example, when tightening a set of M36 high-strength bolts, the initial contact phase may be prolonged and the torque increase slow due to slight unevenness of the contact surface. This step accurately identifies the initial contact phase by calculating the moving average of the slope of the smoothed curve. The range where the slope is less than a first threshold (e.g., 0.1). When the slope rises and stabilizes within a certain range (e.g., between the first and second thresholds), the linear preload phase is identified. When the slope exceeds the second threshold (e.g., indicating the material is beginning to yield), the plastic deformation phase is marked. This classification does not rely on absolute torque values ​​but is based on the shape characteristics of the curve itself.

[0032] Step S30: Based on the annotation results of the tightening stage, the online identification task of tightening anomalies is performed by using the joint discrimination method of CUSUM cumulative sum test and Pearson correlation degradation discrimination, and the anomaly identification result set is output. The "CUSUM cumulative sum test" in this step is a statistical process control method used to detect small deviations in the process mean. Here, it is used in real time during the linear preload and plastic deformation stages to accumulate the deviation between the angle increment and the torque increment, thereby sensitively detecting the sign of slippage, namely "abnormally large angle increase with slow torque increase". The "Pearson correlation degradation discrimination" is used to analyze the correlation between the torque sequence and the time sequence during the final holding stage. If a significant negative correlation is shown (i.e., torque decreases over time), it indicates stress relaxation or torque decay. The "online tightening anomaly identification task" combines the above two methods and adds a comparison with the preset process sequence to identify three typical anomalies: slippage, torque decay, and incorrect tightening sequence.

[0033] This step constitutes the core early warning module of the online monitoring system of this invention. It can issue an anomaly alarm in real-time or near real-time, even before or immediately after the tightening process is complete. Through a phased, multi-indicator joint discrimination mechanism, the coverage and timeliness of anomaly detection are significantly improved, shifting quality issues from "post-event discovery" to "in-process intervention," providing a decision-making basis for real-time quality control during the installation process. For example, when installing a bolt on a device base, impurities in the threads cause slight stripping during the linear pre-tightening stage. The CUSUM test continuously calculates the cumulative sum of angle-torque increments; when this sum exceeds a preset control limit, the bolt is immediately marked as having a "stripping risk." Similarly, after tightening to the target torque and holding for 30 seconds, the torque values ​​at the start and end of the holding period are extracted, and the attenuation rate is calculated. If a bolt's torque is found to have decreased from 1000 Nm to 950 Nm (attenuation rate 5%, exceeding the alarm threshold of 3%), the Pearson discrimination will output a "torque attenuation anomaly" flag. At the same time, if it is recorded that the installer did not tighten the screws in the diagonal order required by the process, but instead skipped around, the Hamming distance between the actual order and the process order will be calculated, and a "tightening sequence abnormal" flag will be output.

[0034] Step S40: Based on the standardized tightening time series dataset and the anomaly identification result set, the preload estimation task is performed by combining a LeNet-style one-dimensional convolutional neural network with a long short-term memory network, and the preload quality evaluation result is output. The "LeNet-style one-dimensional convolutional neural network" in this step excels at extracting features from local waveforms, such as inflection points and fluctuation patterns in the torque curve; the "Long Short-Term Memory Network" excels at capturing long-term dependencies in time series, such as how the state of the previous stage affects the state of the next stage. The combined model can simultaneously learn the local details of the tightening process and the global temporal correlations across stages. The "preload estimation task" utilizes this combined deep learning model, taking a standardized time-series dataset (torque, angle, elongation) and anomaly markers as input, to directly map the final preload estimate of the bolt. The output "preload quality assessment result" includes not only the preload of a single bolt but also statistical indicators within the bolt group, such as average preload, standard deviation, and consistency score.

[0035] This step enables indirect, high-precision online estimation of preload, a key quality indicator. It overcomes the challenges of high cost and difficulty in integrating direct preload measurement (such as using strain gauges or hydraulic tensioners) into online systems. By using a deep learning model to mine the complex nonlinear relationship between preload and final preload from rich tightening process data, it provides a quantitative basis for comprehensively evaluating the fastening quality of connection points, and is a key step in realizing intelligent quality decision-making.

[0036] Compared to preload estimation based on simplified mechanical formulas (such as the torque coefficient method), traditional methods are greatly affected by fluctuations in the friction coefficient and have limited accuracy. The deep learning model used in this invention, through training, can automatically learn the deep correlation between the dynamic interaction patterns of torque, angle, and elongation throughout the tightening process and the final preload. This effectively compensates for the influence of factors such as friction variations and material differences, thus providing a more accurate and stable preload estimate, making it particularly suitable for critical connection parts where high consistency of preload is required.

[0037] For example, consider a set of bolts used to connect a large gearbox housing and a machine base. The standardized tightening curve (three-channel) of each bolt and its anomaly markers (an additional channel is set to 1 if there is a stripped bolt, otherwise 0) are input into a pre-trained ensemble neural network model. The model's one-dimensional convolutional layer first identifies the "hardening" characteristics of the curve during the plastic deformation stage, while the LSTM layer incorporates the stiffness variation trend from the linear segment to the plastic segment. Ultimately, the model outputs an estimated preload of 85 kN for the bolt. After calculating for all eight bolts, the average preload is 83.5 kN, with a standard deviation of 1.2 kN, and a good consistency score. This more accurately reflects the actual clamping state of the connection pair than simply checking whether the final torque of all bolts is 1200 Nm.

[0038] Step S50: Based on the pre-tightening quality assessment results, the TOPSIS comprehensive evaluation method is used to classify the tightening quality level and match the installation decision, and the installation decision results are output.

[0039] The "TOPSIS comprehensive evaluation method" in this step is a multi-attribute decision-making method that ranks the evaluated objects by calculating their relative proximity to the ideal and negative ideal solutions. In this invention, the evaluation attributes include the estimated preload of a single bolt output in step S40, the bolt group preload consistency score, and the severity of various anomalies output in step S30 (such as the stripping index, attenuation rate, and sequence deviation). "Tightening quality grade classification" categorizes the installation results into levels such as "Excellent, Good, Medium, Poor" or "Qualified, Requires Re-inspection, Unqualified" based on the comprehensive score calculated by TOPSIS. "Installation decision matching" automatically provides subsequent operation suggestions based on the quality grade, such as "Acceptance passed," "Mark and focus on monitoring," "Immediately stop and re-tighten," or "Replace bolts."

[0040] This step is the final output and decision-making stage of the monitoring process. It integrates the multi-dimensional and heterogeneous quality indicators (process data, anomaly markers, preload estimation) generated in the preceding steps into an intuitive, operable comprehensive evaluation result and clear action instructions. This greatly reduces the difficulty and subjectivity of manual data interpretation, realizes automated and intelligent decision-making output of monitoring results, and completes a closed loop from data collection to action guidance. For example, after the installation of turbine cylinder block bolts, the summarized data shows: bolt group preload consistency score of 85 points (good), but one bolt was found to have slight torque decay (decay rate 2.5%), and the tightening sequence of another bolt deviated from the process requirements. The TOPSIS method calculates based on preset weights (e.g., preload consistency has the highest weight, slight decay has the second highest weight, and sequence error has the lowest weight), and concludes that the comprehensive score of this installation is in the "good" range. Therefore, the output installation decision result is: "Quality level: Good. Recommendation: Mark the decayed bolt and pay special attention to it in subsequent inspections; record the sequence deviation and strengthen process discipline. This installation is acceptable." This provides the field engineer with a clear and quantifiable acceptance basis.

[0041] Example 2: Furthermore, the present invention provides an online monitoring system for the installation process of electromechanical equipment, employing an online monitoring method for the installation process of electromechanical equipment as described in the above embodiments, which can solve a technical problem related to online monitoring of the installation process of electromechanical equipment. The beneficial effects of the online monitoring system for the installation process of electromechanical equipment provided by the present invention are the same as those of the online monitoring method for the installation process of electromechanical equipment provided in the above embodiments, and other technical features of the online monitoring system for the installation process of electromechanical equipment are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0042] Example 3: This invention provides an online monitoring device for the installation process of electromechanical equipment. The device includes: at least one processor; and a memory communicatively connected to the at least one processor. The memory stores instructions executable by the at least one processor, which are then executed to enable the at least one processor to perform the online monitoring method for the installation process of electromechanical equipment described in Example 1. The online monitoring device for the installation process of electromechanical equipment in this embodiment may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital radio receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), and vehicle terminals (e.g., vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers. This online monitoring device for the installation process of electromechanical equipment is merely an example and should not limit the functionality or scope of the embodiments of this invention. The online monitoring device for the installation process of electromechanical equipment may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory or a program loaded from a storage device into a random access memory. The random access memory also stores various programs and data required for the operation of an online monitoring device for the installation process of electromechanical equipment. The processing unit, read-only memory, and random access memory are interconnected via a bus. The I / O interface is also connected to the bus. Typically, the following systems can be connected to the I / O interface: input devices including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices including, for example, magnetic tapes, hard disks, etc.; and communication devices. The communication device allows the online monitoring device for the installation process of electromechanical equipment to communicate wirelessly or wiredly with other devices to exchange data. Although the figure shows an online monitoring device for the installation process of electromechanical equipment with various systems, it should be understood that it is not required to implement or possess all the systems shown. More or fewer systems can be implemented alternatively.

[0043] Example 4: This invention also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the online monitoring method for the installation process of electromechanical equipment as described above. The computer program product provided by this invention can solve the technical problem of online monitoring of the installation process of electromechanical equipment. Compared with the prior art, the beneficial effects of the computer program product provided by this invention are the same as the beneficial effects of the online monitoring method for the installation process of electromechanical equipment provided in the above embodiments, and will not be repeated here.

[0044] In particular, according to the embodiments disclosed in this invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from a storage device, or installed from a read-only memory. When the computer program is executed by a processing device, it performs the functions defined in the methods of the embodiments disclosed in this invention.

[0045] It should be understood that the various parts disclosed in this invention can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics may be combined in any suitable manner in one or more embodiments or examples.

[0046] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of this invention and its equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for online monitoring of the installation process of electromechanical equipment, characterized in that, The methods include: Step S10: Acquire multi-source sensor sampling data during the bolt tightening process, perform a tightening time series data construction task based on the multi-source sensor sampling data, and output a standardized tightening time series dataset; Step S20: Based on the standardized tightening time series dataset, the Savitzky-Golay smoothing and first-order difference slope discrimination method are used to perform the adaptive partitioning task of the tightening stage, and the tightening stage annotation results are output. Step S20, based on the standardized tightening time-series dataset, employs Savitzky-Golay smoothing and first-order difference slope discrimination to perform an adaptive partitioning task for the tightening stage, and outputs the tightening stage annotation results. Specifically, this includes: Step S201: Extract the normalized torque sequence and normalized angle sequence from the standardized tightening time series dataset output in step S10, and use the Savitzky-Golay smoothing method to smooth the normalized torque sequence and normalized angle sequence respectively to obtain the smoothed torque sequence and smoothed angle sequence. Step S202: Based on the smoothed torque sequence and smoothed angle sequence obtained in step S201, the slope of the torque angle curve is obtained by using the first-order difference slope calculation method, and the moving average of the slope is calculated by using a sliding window based on the slope of the torque angle curve. Step S203: Compare the slope moving average obtained in step S202 with the preset derivative threshold, and combine the target torque, torque fluctuation threshold and stable holding time to make point-by-point stage judgment. Merge continuous sampling points with the same stage type into the corresponding tightening stage interval and output the tightening stage labeling result. In step S203, the moving average slope obtained in step S202 is compared with a preset derivative threshold, and a point-by-point stage determination is made in conjunction with the target torque, torque fluctuation threshold, and stable holding time. Specifically, this includes: Preset the first derivative threshold and the second derivative threshold; When the moving average slope is less than the first derivative threshold and the current normalized torque value is less than the target torque, the corresponding sampling point is determined as the sampling point of the initial bonding stage. When the moving average slope is greater than or equal to the first derivative threshold and less than the second derivative threshold, and the current normalized torque value is less than the target torque, the corresponding sampling point is determined to be a sampling point in the linear preload stage. When the average slope moving average is greater than or equal to the second derivative threshold, and the torque fluctuation amplitude within the sliding window is less than the torque fluctuation threshold, the corresponding sampling point is determined to be a sampling point in the plastic deformation stage. When the current normalized torque value reaches the target torque, and within a continuous time interval with a duration not less than the stable holding time, the deviation between the normalized torque value and the target torque does not exceed the torque fluctuation threshold, the corresponding interval is determined as the final holding stage. The determination criteria for the final holding stage are as follows: in, Indicates sampling point The normalized torque value at that point, Indicates the target torque. Indicates the torque fluctuation threshold. Indicates the duration of stability; This represents the normalized torque value at any time t within the continuous holding time interval; Step S30: Based on the annotation results of the tightening stage, the online identification task of tightening anomalies is performed by using the joint discrimination method of CUSUM cumulative sum test and Pearson correlation degradation discrimination, and the anomaly identification result set is output. Step S30, based on the annotation results of the tightening stage, employs a joint discrimination method of CUSUM cumulative sum test and Pearson correlation degradation discrimination to perform the online identification task of tightening anomalies and output the anomaly identification result set. This step specifically includes: Step S301: Based on the start and end times of the linear pre-tightening stage and the plastic deformation stage output in step S20, extract the corresponding stage data segments from the standardized tightening timing dataset output in step S10, use the CUSUM cumulative sum test to detect the deviation between the angle increment and the torque increment, and calculate the slippage index based on the ratio of the cumulative angle increment to the cumulative torque increment. Step S302: Based on the start and end times of the final holding phase output in step S20, extract the holding start torque value and holding end torque value from the standardized tightening timing dataset output in step S10. Use Pearson correlation degradation discrimination to confirm the torque decrease trend in the final holding phase and calculate the torque decay rate. When the torque decay rate is greater than the decay alarm threshold, output the torque decay abnormal mark of the corresponding bolt. Step S303: Based on the bolt numbering order in the preset process sequence and the actual tightening timestamps of each bolt recorded during the acquisition process in step S10, generate the actual tightening sequence. Use Hamming distance to calculate the sequence deviation between the actual tightening sequence and the preset process sequence. When the sequence deviation is greater than the sequence deviation threshold, output a tightening sequence abnormality mark. Then, summarize the tightening sequence abnormality mark, torque attenuation abnormality mark, and tightening sequence abnormality mark into an abnormality identification result set. Step S40: Based on the standardized tightening time series dataset and the anomaly identification result set, the preload estimation task is performed by combining a LeNet-style one-dimensional convolutional neural network with a long short-term memory network, and the preload quality evaluation result is output. Step S40, based on the standardized tightening time-series dataset and the anomaly identification result set, employs a combination of a LeNet-style one-dimensional convolutional neural network and a long short-term memory network to perform the preload estimation task and output the preload quality assessment result. This step specifically includes: Step S401: Based on the standardized tightening time series dataset output in step S10, extract the normalized torque sequence, normalized angle sequence, and normalized elongation sequence corresponding to each bolt, and set an abnormal additional channel for bolt samples with abnormal markers according to the abnormal identification result set output in step S30, forming a four-channel time series input sample. Step S402: Input the four-channel time-series input samples obtained in step S401 into the pre-trained LeNet-style one-dimensional convolutional neural network and long short-term memory network combined model. The one-dimensional convolutional layer of the model first identifies the "hardening" feature of the curve in the plastic deformation stage, while the LSTM layer combines the stiffness change trend from the linear segment to the plastic segment and outputs the estimated value of the preload of each bolt through a fully connected regression layer. Step S403: Based on the estimated preload of each bolt output in step S402, calculate the average preload, standard deviation of preload, and preload consistency score of the same bolt group, and use the estimated preload of each bolt, the preload consistency score of the bolt group, and the statistical results of abnormal bolts as the preload quality assessment results; Step S50: Based on the pre-tightening quality assessment results, the TOPSIS comprehensive evaluation method is used to classify the tightening quality level and match the installation decision, and the installation decision results are output.

2. The online monitoring method for the installation process of electromechanical equipment as described in claim 1, characterized in that, Step S10 involves acquiring multi-source sensor sampling data during the bolt tightening process, performing a tightening time-series data construction task based on the multi-source sensor sampling data, and outputting a standardized tightening time-series dataset. Specifically, this includes: Step S101: During the same bolt tightening process, the original torque sequence output by the high-precision torque sensor, the original angle sequence output by the angle encoder, and the original length measurement sequence output by the embedded ultrasonic length measurement module are synchronously collected according to the preset sampling frequency, and the hardware timestamps corresponding to each sampling point are recorded respectively. Step S102: Construct a unified time axis based on the hardware timestamp recorded in step S101, and map the original torque sequence, original angle sequence and original length measurement sequence to the unified time axis through linear interpolation to obtain the time-aligned torque sequence, angle sequence and elongation sequence. Step S103: Perform Donoho wavelet soft thresholding on the time-aligned torque sequence, angle sequence, and elongation sequence obtained in step S102, and normalize them according to the upper and lower limits of the corresponding sensor range to obtain a standardized tightening time series dataset containing the normalized torque sequence, normalized angle sequence, and normalized elongation sequence. The normalization process is performed according to the following formula: in, This represents any sequence of physical quantities after filtering. This indicates the lower limit of the corresponding sensor's measurement range. This indicates the upper limit of the corresponding sensor's measurement range. This represents the normalized sequence value.

3. The online monitoring method for the installation process of electromechanical equipment as described in claim 1, characterized in that, In step S30, the torque attenuation rate is calculated according to the following formula: in, Indicates the start time of the final hold phase. This indicates the end time of the final hold phase. This represents the torque attenuation rate corresponding to the i-th bolt; This represents a correction constant to prevent the denominator from being zero; This represents the normalized torque value corresponding to the start time of the final hold phase; This represents the normalized torque value at the end of the final hold phase.

4. An online monitoring system for the installation process of electromechanical equipment, applied to the online monitoring method for the installation process of electromechanical equipment as described in any one of claims 1 to 3, characterized in that, The online monitoring system for the installation process of the electromechanical equipment includes: The data acquisition and preprocessing module is used to acquire multi-source sensor sampling data during the bolt tightening process, perform a tightening time series data construction task based on the multi-source sensor sampling data, and output a standardized tightening time series dataset. The tightening stage segmentation module is used to perform adaptive segmentation of tightening stages based on a standardized tightening time series dataset, using Savitzky-Golay smoothing and first-order difference slope discrimination method, and outputs tightening stage annotation results. The online anomaly identification module is used to perform online anomaly identification tasks based on the tightening stage annotation results, using a joint discrimination method of CUSUM cumulative sum test and Pearson correlation degradation discrimination, and outputs anomaly identification result set. The consistency evaluation module is used to perform the preload estimation task based on the standardized tightening time series dataset and the anomaly identification result set, using a combination of LeNet-style one-dimensional convolutional neural network and long short-term memory network, and outputs the preload quality evaluation result. The comprehensive evaluation and decision-making module is used to classify tightening quality levels and match installation decisions based on the pre-tightening quality assessment results using the TOPSIS comprehensive evaluation method, and output the installation decision results.

5. An online monitoring device for the installation process of electromechanical equipment, characterized in that, The online monitoring device for the installation process of electromechanical equipment includes: a memory, a processor, and an online monitoring program for the installation process of electromechanical equipment stored in the memory and executable on the processor. When the online monitoring program for the installation process of electromechanical equipment is executed by the processor, it implements the online monitoring method for the installation process of electromechanical equipment according to any one of claims 1 to 3.

6. A computer program product, characterized in that, The computer program product includes an online monitoring program for the installation process of electromechanical equipment, which, when executed by a processor, implements an online monitoring method for the installation process of electromechanical equipment as described in any one of claims 1 to 3.

Images