Correlation analysis method and system for rotation flexibility test data and assembly parameters of self-aligning roller bearing

Abstract

The invention discloses a self-aligning roller bearing rotation flexibility test data and assembly parameter correlation analysis method and a self-aligning roller bearing rotation flexibility test data and assembly parameter correlation analysis system. According to the method, bearing assembly parameters and multi-dimensional time sequence data of a rotation test are acquired through a system, and a quantitative correlation model between the bearing assembly parameters and the multi-dimensional time sequence data is constructed by using a machine learning algorithm. On the basis of the model, reverse analysis can be carried out on abnormal test data, and an assembly deviation source which causes poor flexibility is accurately positioned; target parameters can be input in the process design stage, and the expected performance of the bearing can be predicted in the forward direction. The corresponding system integrates a data acquisition module, a model analysis module, an application output module and a process feedback control module, and is integrated with a production line control system. According to the method, traditional qualitative judgment depending on experience is converted into quantitative analysis driven by data and models, intelligent closed loop from quality detection and root diagnosis to process optimization is achieved, and the consistency of bearing assembly quality and the intelligent level of process control are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of industrial intelligence and predictive quality control, and in particular to a method and system for correlation analysis of test data and assembly parameters of self-aligning roller bearing rotational flexibility. Background Technology

[0002] As a critical basic component, the rotational flexibility of self-aligning roller bearings directly affects the operating accuracy, energy consumption, and reliability of main equipment. Therefore, rotational flexibility testing after assembly is a crucial quality inspection step before the bearing leaves the factory. Currently, the industry commonly uses manual rotation for sensory evaluation or simple instrument measurement of frictional torque. Test results are often judged as pass or fail or based on a single value. This process is highly dependent on operator experience, and the test data is independent and scattered, failing to establish an effective correlation with bearing assembly parameters (such as clearance, cage compression, and grease injection). This traditional method makes it difficult to achieve quantitative analysis and data traceability, and it cannot provide data support for precise optimization of the assembly process.

[0003] With the development of intelligent manufacturing and digital testing technologies, the bearing industry urgently needs to establish a deep connection between test data and the assembly process. Current technologies lack a systematic method to correlate and analyze the multi-dimensional dynamic data (such as starting torque, rotational uniformity, and abnormal noise spectrum) collected during rotational flexibility testing with key bearing assembly parameters. This makes it difficult to trace the root causes of assembly quality problems, and process adjustments lack data support, hindering further improvements in bearing consistency and reliability. Therefore, developing a method and system that can deeply integrate test data and assembly parameters, achieving quantitative correlation and intelligent analysis, is of great significance for achieving precise assembly and improving product quality. Summary of the Invention

[0004] This invention specifically relates to a method and system for correlating test data and assembly parameters of self-aligning roller bearings with rotational flexibility. The aim is to collect bearing assembly parameters and multi-dimensional test data through the system, and to construct a quantitative correlation model using machine learning algorithms. This achieves a leap from subjective experience-based judgment to objective data-driven analysis of rotational flexibility, significantly improving the accuracy and consistency of quality assessment. To achieve the above objectives, the specific technical solution of this invention's method and system for correlating test data and assembly parameters with rotational flexibility of self-aligning roller bearings is as follows: A method for correlation analysis between test data and assembly parameters of self-aligning roller bearing rotational flexibility includes the following steps: S1. Data acquisition steps: Obtain the assembly parameter set of the batch of self-aligning roller bearings to be analyzed, as well as the test data sequence collected by each bearing in the rotational flexibility test; the assembly parameter set includes at least any two of the following: clearance, cage compression amount, and grease injection amount; S2. Association Model Construction Steps: Using the assembly parameter set as input features and the quantitative index representing rotational flexibility extracted from the test data sequence as the output target, a machine learning algorithm is used to train and construct an association model that can quantitatively reflect the impact of assembly parameters on rotational flexibility. S3. Analysis and Application Steps: Apply the correlation model to new bearing test data or target assembly parameters to generate analysis results for locating assembly deviation sources or evaluating the performance of assembly schemes, so as to guide the adjustment of assembly processes.

[0005] Furthermore, in step S1, the test data sequence includes torque values, vibration signals, and / or sound signals that vary with time or angle.

[0006] Furthermore, step S1 also includes feature extraction of the test data sequence to obtain a feature dataset; the features include at least the average starting torque, rotational stability index, and characteristic frequency amplitude.

[0007] Furthermore, in step S2, the machine learning algorithm is a supervised learning algorithm.

[0008] Furthermore, step S3 includes at least one of the following application modes: S31. Reverse Diagnosis Mode: Input the new bearing test data into the correlation model, reverse deduce the corresponding predicted assembly parameters, and compare the predicted assembly parameters with the process design standard values ​​to locate the assembly deviation source that causes abnormal rotational flexibility. And / or, S32. Forward prediction mode: Input the target assembly parameters into the correlation model and forward predict the corresponding rotational flexibility index to evaluate whether the expected performance of the bearing under the assembly scheme meets the standard.

[0009] Furthermore, after step S3, the method further includes: S4. Process Iteration and Optimization Steps: Based on the analysis results, dynamically adjust the control parameters of the corresponding workstations in the assembly line.

[0010] A system for correlating and analyzing test data and assembly parameters of self-aligning roller bearings for rotational flexibility includes: a data acquisition and management module for acquiring and storing the assembly parameter set and the test data sequence from the bearing assembly line and rotational flexibility test bench; The model building and analysis module is used to build, store, and retrieve the associated model based on the data in the data acquisition and management module. The application and output module is used to receive new bearing test data or target assembly parameters, call the associated model to perform calculations, and output analysis results to guide process adjustments.

[0011] Furthermore, it also includes a process feedback control module, which is connected to the application and output module, for converting the analysis results into specific process parameter adjustment instructions and sending them to the assembly line control system.

[0012] Furthermore, it also includes a human-computer interaction interface module, which is used to configure model parameters, visualize the relationship between assembly parameters and test data, analyze results, and provide suggestions for process adjustments.

[0013] Furthermore, the system is deployed as an integrated hardware platform or industrial software; the data acquisition and management module communicates with the programmable logic controller of the bearing assembly line and the sensors and data acquisition cards of the rotational flexibility test bench through an industrial communication interface; the application and output module interacts with the manufacturing execution system or database server through the workshop network.

[0014] Compared with existing technologies, this invention systematically collects bearing assembly parameters and multi-dimensional test data, and utilizes machine learning algorithms to construct a quantitative correlation model. This achieves a leap from subjective experience-based judgment to objective data-driven analysis of rotational flexibility, significantly improving the accuracy and consistency of quality assessment. Specifically, the model-based reverse diagnostic function can quickly locate specific assembly deviations causing abnormal flexibility based on test data, enabling precise tracing of quality problems. The forward prediction function can simulate performance under different assembly parameters during the process design stage, providing predictions for optimization solutions and effectively reducing trial production costs and timelines. Furthermore, the system can form a closed-loop feedback loop of analysis results, automatically guiding assembly line parameter adjustments, thereby achieving continuous self-optimization of the production process. The entire method is implemented in highly integrated hardware or software, seamlessly integrating with existing production line control systems and management platforms. It not only solves the problems of traditional methods relying on manual labor and lacking traceability, but also forms a practical "test-analysis-optimization" intelligent closed loop, demonstrating significant benefits in improving product consistency, reducing defect rates, and achieving intelligent manufacturing. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating the working principle of the present invention. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of the present invention clearer, the following description, in conjunction with preferred embodiments and appendices, provides further details. Figure 1 This invention will be described in detail. This embodiment is only for explaining the invention and is not intended to limit the scope of protection of the invention.

[0017] Example 1 This embodiment details a method for correlating test data and assembly parameters of self-aligning roller bearings with rotational flexibility, including the following steps: S1. Data acquisition step: Acquire the assembly parameter set of the batch of self-aligning roller bearings to be analyzed, and the test data sequence collected by each bearing in the rotational flexibility test; the assembly parameter set includes at least two of clearance, cage compression amount, and grease injection amount; S2. Correlation model construction step: Using the assembly parameter set as input features and the quantitative index representing rotational flexibility extracted from the test data sequence as the output target, train the model using a machine learning algorithm to construct a correlation model that can quantitatively reflect the influence of assembly parameters on rotational flexibility; S3. Analysis and application step: Apply the correlation model to new bearing test data or target assembly parameters to generate analysis results for locating assembly deviation sources or evaluating the performance of assembly schemes, so as to guide the adjustment of assembly processes. In step S1, the test data sequence includes torque values ​​and vibration signals that change with time or angle. Step S1 also includes feature extraction of the test data sequence to obtain a feature dataset; the features include at least the average starting torque, rotational stability index, and characteristic frequency amplitude. In step S2, the machine learning algorithm is a supervised learning algorithm. Step S3 includes at least one of the following application modes: S32. Forward prediction mode: The target assembly parameters are input into the correlation model to forward predict the corresponding rotational flexibility index, so as to evaluate whether the expected performance of the bearing under the assembly scheme meets the standard. After step S3, it also includes: S4. Process iteration optimization step: Based on the analysis results, the control parameters of the corresponding station in the assembly line are dynamically adjusted.

[0018] A system for correlating test data and assembly parameters of self-aligning roller bearings with rotational flexibility includes: a data acquisition and management module for acquiring and storing the assembly parameter set and the test data sequence from the bearing assembly line and rotational flexibility test bench; a model construction and analysis module for constructing, storing, and calling the correlation model based on the data in the data acquisition and management module; and an application and output module for receiving new bearing test data or target assembly parameters, calling the correlation model for calculation, and outputting analysis results to guide process adjustments. It also includes a process feedback control module connected to the application and output module for converting the analysis results into specific process parameter adjustment instructions and sending them to the assembly line control system. Finally, it includes a human-machine interface module for configuring model parameters, visually displaying the correlation between assembly parameters and test data, analysis results, and process adjustment suggestions.

[0019] Example 2 The contents that are the same as in Example 1 will not be repeated here; the different aspects of this embodiment compared to Example 1 are as follows: In this embodiment, for the mass production of a certain model of self-aligning roller bearing, the specific implementation steps are as follows: On the assembly line, key assembly parameters of each bearing are recorded in real time during the assembly process, forming a structured parameter set. For example, the radial clearance value is 0.055 mm, the cage compression interference is 0.12 mm, and the grease injection amount is 5.2 g, obtained and recorded by high-precision sensors and an industrial control system. After the bearing comes off the production line, it is transferred to a dedicated high-precision rotational flexibility testing bench for full inspection. The testing bench drives the bearing inner ring to rotate at a standard speed in a constant temperature and humidity environment, and simultaneously collects the torque curve, three-dimensional vibration signal waveform, and sound signal that change continuously over time during the rotation process through a high-response torque sensor, vibration accelerometer, and microphone. Subsequently, the system automatically and uniquely binds the "assembly parameter set" and "test data sequence" of the same bearing and stores them in the central database, forming the basic data unit for subsequent analysis. In the model building phase, algorithm engineers extract the "assembly parameter set" as input features (X) from tens of thousands of historically accumulated data units. From the corresponding test data sequences, they use signal processing algorithms to calculate quantitative indicators such as "average starting torque," "rotational uniformity coefficient," and "energy at specific abnormal noise frequencies" as output labels (Y). Using these (X, Y) data pairs, a Gradient Boosting Decision Tree (GBDT) model is trained on a server cluster. This model can accurately quantify causal relationships such as "how much the average starting torque is expected to decrease when the clearance increases by 0.01 mm." In the application phase, process engineers can input test data of a new bearing (such as an abnormal torque curve) into the model. The model can then calculate the predicted combination of assembly parameters, providing feedback such as "the predicted clearance is only 0.038 mm, significantly lower than the standard lower limit of 0.045 mm," thus directly guiding adjustments to the grinding dimensions. This embodiment achieves, for the first time, a digital and model-based connection between the bearing assembly process and the final performance test. It deeply links the traditionally scattered and isolated "assembly records" and "test results," replacing the experience-based model that relies entirely on experienced workers to "diagnose by listening" with a data-driven model. This allows the analysis and optimization of rotational flexibility to move from qualitative to quantitative, and from vague to precise, fundamentally improving the objectivity and scientific nature of quality analysis and laying a solid data and model foundation for subsequent precise process control.

[0020] In this embodiment, the acquisition of test data sequences is further refined. The torque sensor on the test bench has a sampling frequency of 10kHz, recording the millinewton values ​​of torque over time (t) during the start-up and uniform rotation phases in real time. A triaxial vibration accelerometer (installed at a specific location on the bearing housing) synchronously acquires vibration acceleration signals (a_x(t), a_y(t), a_z(t)) in the X, Y, and Z directions at a sampling rate of 20kHz. A highly directional microphone acquires the broadband sound signal s(t) generated during bearing rotation at a sampling rate of 44.1kHz. These raw, high-fidelity time-series signals can be completely recorded, forming a multi-dimensional, synchronous test data sequence. For example, the torque sequence T(t) can reveal the transient characteristics of rotational resistance, the vibration sequence a(t) contains information such as the contact state between the rolling elements and the raceway, and the cage dynamics, while the sound sequence s(t) is extremely sensitive to surface defects such as scratches and impacts, or lubrication abnormalities. These rich raw signals provide a comprehensive data source for subsequent comprehensive evaluation of the bearing's rotational state from different physical dimensions. This embodiment constructs a comprehensive and multi-dimensional sensing system for bearing rotational flexibility by collecting multi-dimensional physical signals such as torque, vibration, and sound. Compared to methods that only measure a single torque average value, multi-dimensional time-series signals can more sensitively and earlier detect minute defects or improper assembly within the bearing. For example, it can identify slight cage jamming through vibration signal spectrum or microscopic scratches on the raceway through sound signals. This greatly enriches the dimensions and depth of the test information, enabling subsequent correlation analysis to uncover deeper and more complex quality influencing factors, resulting in more precise diagnosis.

[0021] In this embodiment, after acquiring the original test data sequence, it is not directly used for model training, but rather a series of feature extraction and engineering processing are performed first. For the torque sequence T(t), the algorithm calculates its "starting peak torque", "average torque within 0.5 seconds to 1 second", and "torque fluctuation coefficient (standard deviation / mean)" for the entire constant speed range as features. For the vibration signal, Fast Fourier Transform (FFT) is performed on a_x(t), a_y(t), and a_z(t) respectively to extract the bearing characteristic frequencies (such as the cage passing frequency and the rolling element passing the inner ring frequency) and their harmonic amplitudes, and the total effective vibration value (RMS) is calculated. For the sound signal s(t), wavelet packet transform is performed to extract the energy proportion of a specific frequency band (such as 2kHz-8kHz, which often corresponds to abnormal friction sound). Finally, the original high-dimensional data sequence of a bearing is transformed into a sequence containing "starting torque mean = 15.6mNm", "rotational stability index (torque fluctuation coefficient) = 0.032", and "cage passing frequency amplitude = 0.15 m / s". 2The feature vector contains dozens of specific values, such as "high-frequency sound energy percentage = 0.08%". This feature vector, together with the assembly parameter set, constitutes the standardized input of the machine learning model. This embodiment achieves dimensionality reduction and knowledge aggregation by selectively extracting features from the original high-dimensional time-series signal. It transforms massive amounts of unstructured waveform data into a set of low-dimensional, structured quantitative indicators that can represent the specific physical state of the bearing. This greatly reduces the complexity and computational burden of subsequent machine learning model training, avoids the "curse of dimensionality", and makes the model's learning objectives more explicit and physically meaningful (such as "torque fluctuation coefficient" directly related to rotational stability), improving the model's training efficiency, interpretability, and final prediction accuracy.

[0022] In this embodiment, the association model construction step explicitly adopts a supervised learning paradigm. Specifically, N bearing sample data that have completed assembly and testing are selected from the database. The input feature (X_i) of each sample i is its assembly parameter set (such as clearance, compression amount, grease injection amount) and / or a feature set extracted from the test data. The output label (Y_i) of each sample i is a human-defined quantitative index reflecting the rotational flexibility, for example, it can be a comprehensive score (0-100 points) given by the test bench, or a "flexibility index" calculated based on torque and vibration characteristics. Thus, a labeled training dataset containing N (X_i, Y_i) samples is constructed. Subsequently, the random forest regression algorithm is used to train this dataset. The algorithm learns the complex mapping relationship from the input feature X to the output target Y by iteratively constructing multiple decision trees. The random forest model obtained after training is an association model that can predict the expected flexibility index based on the input assembly parameters, or infer the assembly parameter state based on the flexibility index. This embodiment employs a supervised learning algorithm, whose core advantage lies in its ability to fully utilize historically accumulated, known "experience data" (i.e., labeled data) to train the model. This allows the model to not only learn the mathematical relationship between input and output but also to learn the "knowledge" implicit in the data that conforms to actual physical laws. The trained model possesses strong generalization ability, and can make accurate predictions or diagnoses based on learned patterns even for new and unseen bearing data. This method digitizes and intelligently reuses the long-term production and testing experience accumulated by enterprises, representing a core aspect of intelligent manufacturing.

[0023] In this embodiment, the analysis and application steps include two flexibly switchable operating modes. In reverse diagnostic mode, a bearing test fails (e.g., excessive starting torque). The system automatically inputs the test data of its high torque fluctuation characteristics into a pre-trained correlation model. After the model runs, it outputs a set of predicted assembly parameter values: "Predicted clearance = 0.041mm, predicted pressing amount = 0.15mm, predicted grease amount = 5.8g". The system then compares these values ​​with the process standards (clearance 0.05±0.01mm, pressing amount 0.12±0.02mm, grease amount 5.0±0.5g), highlighting "predicted clearance (0.041mm) is lower than the standard lower limit (0.044mm)" as the most likely cause, guiding maintenance personnel to focus on checking and adjusting the clearance matching process. In forward prediction mode, the process department plans to try a new solution: increasing the grease amount from 5.0g to 5.3g to improve lubrication, but is concerned about increased resistance. The engineer inputs the target parameter set {clearance = 0.05mm, pressing amount = 0.12mm, grease injection amount = 5.3g} into the system interface. The model instantly predicts its "flexibility index" to be 92 points and provides the following prompt: "Increasing the grease injection amount is expected to increase the starting torque by about 5%, but the rotational smoothness index will improve by 8%, and the overall score is within an acceptable range," thus supporting the decision on this process scheme. The two application modes provided in this embodiment correspond to the two core industrial scenarios of "problem solving" and "process design," respectively. The reverse diagnostic mode realizes a "CT scan" of quality problems, which can quickly and accurately locate the root cause of the abnormality, shortening the troubleshooting time from hours or even days to minutes, greatly improving the efficiency of one-time problem solving. The forward prediction mode acts as a "digital twin testbed," allowing engineers to test and evaluate various process schemes in a virtual space at extremely low cost and zero risk, significantly shortening the process development cycle, reducing trial and error costs, and realizing the transformation of process design from "experience-driven" to "model prediction-driven."

[0024] In this embodiment, after completing the analysis and application steps, the system further executes a process iteration and optimization step. When the reverse diagnostic mode locates a common deviation of "predicted clearance being too small" in multiple bearings in the current batch, the system not only issues an alarm but also automatically generates a process adjustment suggestion based on the statistical regularity of the deviation (e.g., an average deviation of 0.008mm): "It is recommended to increase the median tolerance zone of the inner ring grouping and matching machine by 0.008mm." After confirmation by the engineer on the interface, this adjustment instruction is automatically sent to the programmable logic controller (PLC) of the assembly line through the system interface. The PLC adjusts the operating parameters of the matching machine accordingly. The clearance value of subsequent assembled bearings can be automatically corrected. The system continuously monitors the test data of the new batch of bearings and compares the new "predicted clearance" with the target value, forming a closed loop of "detection-analysis-adjustment-re-detection". Through multiple iterations, the output parameters (actual clearance) of the assembly process can automatically and dynamically converge to the target value set by the process. This embodiment extends a single quality analysis into a continuous and automated process optimization closed loop. It breaks away from the traditional, lagging, and fragmented model of "problem discovery - manual machine adjustment - re-production verification," enabling adaptive fine-tuning of process parameters based on real-time data feedback. This not only quickly corrects existing systemic deviations but also gives the entire assembly process "self-healing" and "self-optimizing" capabilities, maintaining long-term process stability and elevating product quality consistency to a new level. It is a key step towards achieving intelligent and adaptive production.

[0025] In this embodiment, the system implementing the above method exists as an integrated hardware and software platform. The data acquisition and management module is deployed on the workshop data server. It communicates with the assembly line MES system via the OPC UA protocol to acquire the work order number, serial number, and corresponding assembly parameters (clearance, compression amount, grease injection amount) of each bearing in real time. Simultaneously, this module connects to the NI data acquisition card on the test bench via Ethernet to receive and store the complete test data sequence (torque, vibration, sound waveform) corresponding to each bearing in real time. The core function of this module is to accurately associate and store the "assembly parameter set" and "test data sequence" from two independent data streams in the database based on the unique serial number of the bearing. The model building and analysis module is deployed on the company's cloud computing platform. It periodically extracts historical data from the database of the data acquisition and management module, providing graphical tools for data scientists to perform feature engineering, model training, validation, and deployment. The trained model is published as a callable API service. The application and output module is deployed on the engineer workstations of the process quality department, providing a graphical interface. Engineers can select "Diagnostic Mode" to upload a test file for a new bearing, or select "Prediction Mode" to input a set of target parameters. The module calls the cloud-based model API through the internal network, and the calculation results (such as prediction parameters, performance indicators, and deviation analysis) are presented on the interface in graphical and report form. The system described in this embodiment clearly divides the data flow, calculation flow, and application flow through modular design. It achieves end-to-end connectivity from production site data acquisition to cloud-based intelligent model calculation and then to the engineer's desktop application. This architecture gives the system good scalability (the model can be upgraded independently), flexibility (the application can be deployed at multiple points), and maintainability. It encapsulates the complex machine learning analysis process into a tool that is easy for process and quality personnel to use, truly embedding artificial intelligence technology seamlessly into the actual production quality control process.

[0026] In this embodiment, the system further includes a process feedback control module. This module, deployed as a standalone software service within the workshop network, resides in the same security domain as the MES system and PLC control system. When the analysis results generated by the application and output modules contain explicit process adjustment instructions (such as "It is recommended to adjust the pulse count of the grease injection machine from 105 to 108"), these instructions are sent to the process feedback control module. This module has an embedded configurable instruction-signal conversion rule base, converting the process instruction "grease injection amount + 0.3g" into a control signal recognizable by the specific device, "grease injection machine #3, pulse count setpoint = 108". Subsequently, it establishes communication with the corresponding grease injection machine PLC via a secure industrial Ethernet protocol (such as Profinet), and after obtaining secondary confirmation from the operator on the MES terminal, issues the new setpoint. Simultaneously, this module records the time, content, and target device of each instruction issuance, forming a complete traceable log. In this embodiment, the process feedback control module is a crucial bridge between "analysis and decision-making" and "physical execution". It achieves a closed loop from digital diagnosis to automated execution, directly and automatically applying the conclusions of quality analysis to production equipment. This significantly shortens the time cycle from anomaly detection to corrective action, truly achieving "second-level" process adjustment response. This not only improves efficiency but, more importantly, reduces delays and operational errors that may result from manual intervention, making data-driven real-time process optimization possible. It is a core component for building adaptive intelligent manufacturing systems.

[0027] In this embodiment, the system includes a human-computer interaction interface module developed based on web technology. This interface provides a unified access portal for process engineers, quality engineers, and equipment maintenance personnel. The main dashboard displays real-time trends of key indicators for production batches using visual charts, such as "predicted clearance pass rate" and "flexibility index distribution." Users can select the data time range, feature engineering methods, and algorithm hyperparameters for model training through the configuration panel. When applying the model, the interface provides an intuitive operation guide: users can drag and drop to upload test data files, and the system automatically generates a diagnostic report, visually comparing the differences between "predicted parameters" and "standard parameters" in the form of a "radar chart." During forward prediction, users can interactively adjust parameters such as "clearance" and "grease injection volume" using sliders, and the "predictive performance dashboard" on the right dynamically displays the real-time changes in predicted torque, vibration, and other indicators. All analysis results, historical records, and adjustment suggestions are presented in structured reports and visual charts, supporting one-click export. The human-computer interaction interface module in this embodiment greatly reduces the technical threshold for system use, encapsulating the complex underlying data processing and model calculations behind a user-friendly graphical interface. It offers powerful data visualization capabilities, transforming abstract model predictions into easily understandable graphs and reports for engineers, enhancing the intuitiveness and efficiency of decision-making. Meanwhile, its rich configuration features provide professional users with sufficient flexibility, making it a powerful and easy-to-use production analysis decision support center that effectively promotes cross-departmental (process, quality, production) collaboration.

[0028] In this embodiment, the entire system is deployed in two forms. The first is an integrated hardware platform, consisting of a ruggedized industrial PC integrating a data acquisition card, switch, and analysis software. This PC is installed in a cabinet near the test bench, connected to the assembly line PLC via a PROFIBUS-DP bus, and to the test bench sensors via analog / digital I / O modules, forming a physically compact local analysis station. The second form is an industrial software architecture. The software is installed on a workshop server and collects data through edge computing gateways deployed at assembly stations and the test bench. The gateway communicates with the PLC via an RS-485 bus and with the NI acquisition card on the test bench via Ethernet, pre-processing the data before uploading it to the server. The data acquisition and management module reads and writes the assembly parameter registers stored in the PLC in real time through the aforementioned industrial communication interfaces and simultaneously acquires data from the test bench's data acquisition card buffer. The application and output modules interact with the database of the upper-level Manufacturing Execution System (MES) via the workshop industrial Ethernet, writing analysis results (such as bearing quality status and process adjustment suggestions) into the corresponding work order data table of the MES. This data is then used by the MES for production scheduling, quality traceability, and report generation. Simultaneously, the modules retrieve production plans and material batch information from the MES to enrich the analysis dimensions. This embodiment clarifies the specific methods for integrating the system with the existing industrial automation system. Whether as an integrated hardware device or a distributed software system, it achieves seamless integration with core production systems such as PLCs, sensors, and MES through standard industrial protocols. This ensures that the system can acquire high-fidelity data in real time without interfering with existing production processes and seamlessly feed analytical decisions back to the production control and management system. This "plug-and-play" integration capability solves the "last mile" problem of implementing industrial data intelligence applications, enabling advanced correlation analysis methods to be quickly deployed on existing production lines, fully utilizing existing automation investments, and achieving a smooth upgrade from informatization to intelligentization.

[0029] It is understood that the present invention has been described through some embodiments, and those skilled in the art will recognize that various changes or equivalent substitutions can be made to these features and embodiments without departing from the spirit and scope of the invention. Furthermore, under the teachings of the present invention, these features and embodiments can be modified to adapt to specific situations and materials without departing from the spirit and scope of the invention. Therefore, the present invention is not limited to the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are within the protection scope of the present invention.

Claims

1. A method for correlation analysis between test data and assembly parameters of self-aligning roller bearing rotational flexibility, characterized in that, Includes the following steps: S1. Data acquisition steps: Obtain the assembly parameter set of the batch of self-aligning roller bearings to be analyzed, as well as the test data sequence collected by each bearing in the rotational flexibility test; the assembly parameter set includes at least any two of the following: clearance, cage compression amount, and grease injection amount; S2. Association Model Construction Steps: Using the assembly parameter set as input features and the quantitative index representing rotational flexibility extracted from the test data sequence as the output target, a machine learning algorithm is used to train and construct an association model that can quantitatively reflect the impact of assembly parameters on rotational flexibility. S3. Analysis and Application Steps: Apply the correlation model to new bearing test data or target assembly parameters to generate analysis results for locating assembly deviation sources or evaluating the performance of assembly schemes, so as to guide the adjustment of assembly processes.

2. The method for correlation analysis of self-aligning roller bearing rotational flexibility test data and assembly parameters according to claim 1, characterized in that, In step S1, the test data sequence includes torque values, vibration signals, and / or sound signals that vary with time or angle.

3. The method for correlation analysis of self-aligning roller bearing rotational flexibility test data and assembly parameters according to claim 2, characterized in that, Step S1 further includes feature extraction of the test data sequence to obtain a feature dataset; the features include at least the average starting torque, rotational stability index and characteristic frequency amplitude.

4. The method for correlation analysis of self-aligning roller bearing rotational flexibility test data and assembly parameters according to claim 1, characterized in that, In step S2, the machine learning algorithm is a supervised learning algorithm.

5. The method for correlation analysis of self-aligning roller bearing rotational flexibility test data and assembly parameters according to claim 1, characterized in that, Step S3 includes at least one of the following application modes: S31. Reverse Diagnosis Mode: Input the new bearing test data into the correlation model, reverse deduce the corresponding predicted assembly parameters, and compare the predicted assembly parameters with the process design standard values ​​to locate the assembly deviation source that causes abnormal rotational flexibility. And / or, S32. Forward prediction mode: Input the target assembly parameters into the correlation model and forward predict the corresponding rotational flexibility index to evaluate whether the expected performance of the bearing under the assembly scheme meets the standard.

6. The method for correlation analysis of self-aligning roller bearing rotational flexibility test data and assembly parameters according to claim 1, characterized in that, The step S3 is followed by: S4. Process Iteration and Optimization Steps: Based on the analysis results, dynamically adjust the control parameters of the corresponding workstations in the assembly line.

7. A system for correlating test data and assembly parameters of self-aligning roller bearings with rotational flexibility, used to implement the method for correlating test data and assembly parameters of self-aligning roller bearings as described in any one of claims 1-6, characterized in that, include: The data acquisition and management module is used to acquire and associate the assembly parameter set and the test data sequence from the bearing assembly line and the rotational flexibility test bench; The model building and analysis module is used to build, store, and call the associated model based on the data in the data acquisition and management module; The application and output module is used to receive new bearing test data or target assembly parameters, call the associated model to perform calculations, and output analysis results to guide process adjustments.

8. The system for correlating test data and assembly parameters of self-aligning roller bearings according to claim 7, characterized in that, Also includes: The process feedback control module, connected to the application and output module, is used to convert the analysis results into specific process parameter adjustment instructions and send them to the assembly line control system.

9. The system for correlating test data and assembly parameters of self-aligning roller bearings according to claim 7, characterized in that, It also includes a human-computer interaction interface module, which is used to configure model parameters, visualize the relationship between assembly parameters and test data, analyze results and provide suggestions for process adjustment.

10. The system for correlating test data and assembly parameters of self-aligning roller bearings according to claim 7, characterized in that, The system is deployed in the form of an integrated hardware platform or industrial software; the data acquisition and management module communicates with the programmable logic controller of the bearing assembly line and the sensors and data acquisition cards of the rotational flexibility test bench through an industrial communication interface; the application and output module interacts with the manufacturing execution system or database server through the workshop network.

Images