A method and system for optimizing process parameters of a generator oversize retainer ring bulging

Abstract

The application provides a generator oversize retainer ring bulging process parameter optimization method and system, relates to the plastic processing technical field, and the method comprises the steps of constructing a multi-source sensor array, synchronously collecting trace gas concentration field information and sealing gap change information, forming a leakage original data set and a wear original data set respectively, obtaining a leakage characteristic parameter and a wear amount parameter, constructing a high-dimensional feature vector based on the leakage characteristic parameter and the wear amount parameter, calculating bulging uniformity, wear unevenness, wear leakage coupling abnormality index and instability risk index according to the high-dimensional feature vector, determining a bulging process state category according to a multi-threshold-based judgment logic tree, and generating a corresponding process parameter adjustment strategy according to the bulging process state category. The technical problems of isolated multi-source sensing information, incomplete state representation, lagging abnormality identification and lack of closed-loop control mechanism are solved.

Description

Technical Field

[0001] This application relates to the field of plastic processing technology, and in particular to a method and system for optimizing the process parameters of bulging forming of ultra-large retaining rings for generators. Background Technology

[0002] With the rapid development of my country's power equipment manufacturing industry, the retaining ring of ultra-large generators, as a key component of the fixed rotor end winding, directly affects the safe operation of generator sets due to the stability and quality of its bulging process. The retaining ring bulging process uses high-pressure liquid to expand the blank from the inside, combined with axial extrusion, to achieve the required strength and dimensional accuracy from difficult-to-deform materials such as high-nitrogen austenitic stainless steel. However, with the increase in retaining ring diameter and the increasing complexity of bulging process parameters, the sealing cone surface, as the core interface between the mold and the retaining ring end, has a decisive impact on bulging quality, mold life, and process stability due to its leakage behavior and wear state. Therefore, there is an urgent need for a technology that can monitor the sealing interface state in real time and perform adaptive control.

[0003] In existing technologies, monitoring of the retaining ring bulging process mainly relies on the detection of a single physical quantity, such as monitoring the internal pressure of bulging through pressure sensors, measuring the dimensional changes of the retaining ring through displacement sensors, or assessing the wear degree of the mold through offline detection. However, these methods have significant shortcomings: on the one hand, tracer gas leak detection and sealing gap measurement are usually performed independently, lacking a unified spatial coordinate system and time synchronization mechanism, making it difficult to achieve spatial correlation analysis between leak distribution and wear evolution; on the other hand, existing methods mostly focus on post-event detection or single-feature alarms, failing to integrate leakage intensity, leakage fluctuations, wear accumulation, and their coupling relationships into a unified state representation vector, resulting in the inability to identify complex abnormal patterns such as local damage, uneven wear, and pressure fluctuations in real time, and lacking the ability to adaptively adjust closed-loop process parameters based on multi-index collaborative criteria. In addition, traditional bulging process control relies on manual experience to set fixed process parameters, and cannot dynamically optimize the bulging rate, internal pressure curve, and contact pressure distribution according to the real-time state of the sealing interface, thus restricting further improvement of bulging quality and extension of mold life.

[0004] In summary, existing monitoring technologies for retaining ring bulging processes suffer from problems such as isolated multi-source sensor information, incomplete state characterization, delayed anomaly identification, and a lack of closed-loop control mechanisms. Therefore, a method is urgently needed to address these issues. Summary of the Invention

[0005] This disclosure provides a method and system for optimizing the process parameters of the expansion forming of ultra-large retaining rings for generators, in order to solve the technical problems in the prior art, such as isolated multi-source sensor information, incomplete state characterization, delayed anomaly identification, and lack of closed-loop control mechanism.

[0006] According to a first aspect of this disclosure, a method for optimizing the process parameters of the bulging process for an ultra-large retaining ring of a generator is provided, comprising: A multi-source sensor array is constructed to simultaneously collect tracer gas concentration field information and sealing gap change information, forming original datasets of leakage and wear respectively. The multi-source sensor array includes a gas concentration sensor array and a displacement sensor array. Feature extraction and physical quantity decoupling calculation were performed on the original leakage dataset and the original wear dataset to obtain leakage feature parameters and wear parameters. The leakage feature parameters include the full-time leakage intensity sequence, average leakage intensity, maximum leakage location, spectral main frequency and fluctuation amplitude. The wear parameters include the cumulative wear amount and the estimated value of instantaneous cumulative wear. A high-dimensional feature vector is constructed based on leakage characteristic parameters and wear parameters. The high-dimensional feature vector includes leakage intensity feature vector, wear accumulation feature vector, leakage fluctuation feature vector, and wear-leak coupling feature vector. The bulging uniformity, wear non-uniformity, wear-leakage coupling anomaly index, and instability risk index are calculated based on the high-dimensional feature vectors. The bulging process state category is determined based on a multi-threshold-based judgment logic tree, and a corresponding process parameter adjustment strategy is generated based on the bulging process state category.

[0007] According to a second aspect of this disclosure, a system for optimizing the process parameters of an ultra-large retaining ring for generators is provided, comprising: A multi-source sensor data acquisition module is used to construct a multi-source sensor array, synchronously acquire tracer gas concentration field information and sealing gap change information, and form leakage raw dataset and wear raw dataset respectively. The multi-source sensor array includes a gas concentration sensor array and a displacement sensor array. The feature extraction and parameter calculation module is used to extract features and decouple physical quantities from the original leakage dataset and the original wear dataset respectively, to obtain leakage feature parameters and wear parameters. The leakage feature parameters include the full-time leakage intensity sequence, average leakage intensity, maximum leakage location, spectral main frequency and fluctuation amplitude. The wear parameters include the cumulative wear amount and the estimated value of instantaneous cumulative wear amount. A high-dimensional feature vector construction module is used to construct high-dimensional feature vectors based on leakage feature parameters and wear parameters. The high-dimensional feature vectors include leakage intensity feature vectors, wear accumulation feature vectors, leakage fluctuation feature vectors, and wear-leak coupling feature vectors. The key indicator calculation module is used to calculate the bulging uniformity, wear non-uniformity, wear leakage coupling anomaly index and instability risk index based on high-dimensional feature vectors. The state determination and strategy generation module is used to determine the state category of the bulging process based on a multi-threshold-based judgment logic tree, and generate corresponding process parameter adjustment strategies based on the state category of the bulging process.

[0008] One or more technical solutions provided in this disclosure have at least the following technical effects or advantages: Constructing a multi-source sensor array to simultaneously collect tracer gas concentration field information and sealing gap change information, forming original leakage datasets and original wear datasets respectively, wherein the multi-source sensor array includes a gas concentration sensor array and a displacement sensor array; performing feature extraction and physical quantity decoupling calculations on the original leakage dataset and the original wear dataset respectively to obtain leakage characteristic parameters and wear parameters, wherein the leakage characteristic parameters include a full-time leakage intensity sequence, average leakage intensity, maximum leakage location, spectral dominant frequency, and fluctuation amplitude, and the wear parameters include cumulative wear and estimated instantaneous cumulative wear; constructing a high-dimensional feature vector based on the leakage characteristic parameters and wear parameters, wherein the high-dimensional feature vector includes a leakage intensity feature vector, a wear cumulative feature vector, a leakage fluctuation feature vector, and a wear-leak coupling feature vector; calculating bulging uniformity, wear non-uniformity, wear-leak coupling anomaly index, and instability risk index based on the high-dimensional feature vector; determining the bulging process state category based on a multi-threshold-based judgment logic tree, and generating corresponding process parameter adjustment strategies based on the bulging process state category. This invention solves the technical problems of isolated multi-source sensor information, incomplete state representation, delayed anomaly identification, and lack of closed-loop control mechanisms in existing technologies. It achieves the technical effects of improving the consistency of bulging quality, delaying mold wear, and reducing the risk of instability.

[0009] The above description is merely an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, specific embodiments of this application are given below. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in this disclosure or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0011] Figure 1 A flowchart illustrating a method for optimizing process parameters of an ultra-large retaining ring for generators, provided in an embodiment of this application; Figure 2This is a schematic diagram of a system for optimizing the process parameters of an ultra-large retaining ring for generators, provided in an embodiment of this application.

[0012] Figure labeling: 11 Multi-source sensor data acquisition module, 12 Feature extraction and parameter calculation module, 13 High-dimensional feature vector construction module, 14 Key index calculation module, 15 State determination and strategy generation module. Detailed Implementation

[0013] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0014] Example 1: This disclosure provides a method for optimizing the process parameters of the bulging process for an ultra-large retaining ring of a generator. (Refer to...) Figure 1 The methods include: S1: Construct a multi-source sensor array to synchronously collect tracer gas concentration field information and sealing gap change information to form original leakage dataset and original wear dataset, respectively. The multi-source sensor array includes a gas concentration sensor array and a displacement sensor array. Furthermore, step S1 also includes: The bulging working medium is marked using a tracer gas-assisted detection method; The number of circumferential gas concentration sensors is determined based on the nominal diameter of the retaining ring, and the number of axial gas concentration sensor layers is determined based on the height of the sealing cone. A two-dimensional spatial measuring point grid marked with circumferential index i and axial index j is constructed. Miniature gas concentration sensors are installed based on the two-dimensional spatial measuring point grid to form a gas concentration sensor array. The tracer gas concentration values ​​at each measuring point (i,j) are collected by a gas concentration sensor array, and the concentration time series of each measuring point is constructed in chronological order. The concentration time series of all measuring points are then summarized to form the original dataset of the leak. An eddy current displacement sensor array is formed by installing eddy current displacement sensors on a two-dimensional spatial measuring point grid. By using the real-time gap values ​​of each measuring point in the displacement sensor array and taking the initial gap value before the bulging begins as a benchmark, the gap change is calculated. A time series of gap changes for each measuring point is constructed in chronological order, and the time series of gap changes for all measuring points are summarized to form the original wear dataset.

[0015] Specifically, in step S1, a multi-source sensor array is first constructed to acquire the leakage behavior and wear state of the sealing cone surface. By synchronously collecting tracer gas concentration field information and sealing gap change information, original leakage dataset and original wear dataset are formed respectively, providing a unified data input basis for subsequent state decoupling and feature calculation.

[0016] In the process of acquiring tracer gas concentration field information, the expanded working medium is first marked using a tracer gas-assisted detection method. Before the ultra-high pressure liquid is injected into the high-pressure pump, a micro-gas mixer is installed in its inlet pipeline. This mixer adopts an online static mixer structure, installed on a horizontal pipe section 0.5~1.0m before the high-pressure pump inlet, and is sealed and fixed to the pipeline by a flange connection. Its interior has multiple layers of staggered baffles to promote turbulent mixing of gas and liquid. Then, a mass flow controller is used to inject tracer gas with a purity of 99.99%. A mass flow controller is installed on a bypass branch pipe upstream of the tracer gas mixer, injecting a constant flow rate into the working medium. Its inlet is connected to the tracer gas source, and its outlet is connected to the mixer's inlet via a one-way valve. The controller incorporates a thermal mass flow sensor and a proportional control valve, automatically adjusting its opening according to a set value to maintain the tracer gas concentration in the medium stably within the range of 5–20 ppm. Simultaneously, an online gas concentration analyzer is installed at the outlet of the tracer gas mixer to monitor the current tracer gas concentration in real time and feeds the monitoring results back to the mass flow controller for closed-loop adjustment of its opening, ensuring the tracer gas concentration remains stable throughout the expansion process.

[0017] After adding the tracer gas, a spatially distributed gas concentration sensor array is constructed on the outer side of the sealing cone surface of the expansion mold. With the retaining ring axis as the reference center, M miniature gas concentration sensors are evenly distributed circumferentially on the fixed part of the mold. The specific value of M is determined according to the nominal diameter of the retaining ring. When the retaining ring diameter is less than or equal to 1500 mm, M=8; when the diameter is greater than 1500 mm but less than or equal to 2500 mm, M=12; and when the diameter is greater than 2500 mm, M=16. In this embodiment, the retaining ring diameter is 1800 mm, so M=12 is chosen. The circumferential miniature gas concentration sensors are numbered and indexed as i=1,2,…,M, with corresponding circumferential angles of… This creates a complete circumferential coverage. Simultaneously, L=2 to 3 layers of miniature gas concentration sensors are arranged axially along the generatrix direction on the sealing cone surface. The specific value of L is determined based on the height of the sealing cone surface. When the cone surface height is less than or equal to 100mm, L=2 is used, covering the root and opening. When the cone surface height is greater than 100mm, L=3 is used, covering the root, middle, and opening. In this example, the cone surface height is 150mm, so L=3 is chosen. The index of the axial miniature gas concentration sensors is set as j=1,2,…,L, corresponding to the root, middle, and opening regions of the sealing cone surface, respectively. This constructs a two-dimensional spatial measurement point grid indexed by (i,j), where each measurement point uniquely corresponds to a local region on the sealing interface. The miniature gas concentration sensor is installed using a threaded bushing fixing method. Internal threads are machined into the pre-drilled mounting hole on the outer side of the mold's sealing cone surface, and external threads are provided at the front end of the sensor housing. After being screwed into the mounting hole, it is positioned by tightening a lock nut. The sensor probe end face is flush with the outer surface of the mold. A 2mm diameter gas guide hole is opened in front of the probe, connecting to the working surface of the sealing cone surface to capture the escaping tracer gas. The miniature gas concentration sensor uses the non-dispersive infrared absorption principle, with a detection limit not exceeding 1 ppm, a response time not exceeding 0.1 s, and a sampling frequency not less than 50 Hz, thus enabling real-time capture of concentration disturbances caused by weak leaks.

[0018] During the bulging process, at each sampling time Where k = 1, 2, ..., K, K is the total number of sampling points, and the tracer gas concentration value at each measuring point (i, j) is collected simultaneously. And construct the concentration time series of a single measurement point in chronological order, the specific expression is as follows: Furthermore, the concentration time series data from all measuring points are compiled to form a complete original dataset of the leak, which characterizes the tracer gas concentration field information during the bulging process. The specific expression for the original leak dataset is as follows: ,in, This indicates that at the i-th circumferential position and the j-th axial layer position, at time... The collected tracer gas concentration values ​​represent the degree of gas enrichment in the local area of ​​the sealed interface. This dataset serves as the sole original input for subsequent leak feature extraction and leak location identification.

[0019] To achieve high-precision online measurement of the sealing cone surface gap change during the acquisition of sealing gap variation information, a high-precision displacement sensor array is embedded in the back region of the mold's sealing cone surface. Specifically, the displacement sensor employs an eddy current displacement sensor. The sensor's probe penetrates the mold substrate, and its measuring end face is flush with or slightly below the working surface of the sealing cone surface (0.05 mm below the surface) to ensure measurement stability and avoid interfering with the sealing contact state. This type of sensor directly measures the change in distance between the probe end face and the outer surface of the retaining ring, thus indirectly reflecting the evolution of the sealing interface gap.

[0020] In terms of spatial deployment, the eddy current displacement sensor and the miniature gas concentration sensor adopt the same grid system, also numbered based on a two-dimensional spatial grid of measurement points. This ensures that each gas measurement point has a matching displacement measurement point below or in the vicinity of its corresponding location, thereby achieving spatial alignment between leakage information and gap change information. Each eddy current displacement sensor has a range of 0–5 mm, a resolution no higher than 1 μm, and a sampling frequency no lower than 50 Hz to meet the high-precision capture requirements of dynamic gap changes.

[0021] In actual operation, at each sampling time Obtain the real-time gap value of each measuring point. and the initial gap value before the bulging begins. Based on this, the gap change is calculated using the following formula: ,in, The initial clearance value obtained through no-load calibration under unloaded conditions. Indicates the current moment The real-time interval values ​​obtained from the acquisition, This indicates that the measuring point is at time [time]. The amount of gap change relative to the initial state is obtained directly from the sensor output signal after calibration, without the need for additional calculation or processing.

[0022] Furthermore, a time series of gap changes at each measuring point is constructed in chronological order. These time series of gap changes at all measuring points are then summarized to form the original wear dataset. The specific expression is as follows: It should be noted that, It also includes the elastic or plastic deformation component generated at the end of the retaining ring under high pressure and the gap increase caused by the wear of the sealing cone material. Therefore, this data is only used as the comprehensive response input. In the subsequent step S2, it is separated by constructing a decoupling model to obtain the real wear evolution information.

[0023] Through the above methods, a unified spatiotemporal coordinate system is established. This enables the synchronous acquisition and unified representation of leakage behavior data and structural gap response data, thus providing a complete and consistent data foundation for subsequent multiphysics coupling analysis, state decoupling, and feature extraction.

[0024] S2: Perform feature extraction and physical quantity decoupling calculation on the original leakage dataset and the original wear dataset respectively to obtain leakage feature parameters and wear parameters. The leakage feature parameters include the full-time leakage intensity sequence, average leakage intensity, maximum leakage location, spectrum main frequency and fluctuation amplitude. The wear parameters include the cumulative wear amount and the estimated value of instantaneous cumulative wear. Furthermore, step S2 also includes: Obtain the original dataset of the leak, and calculate the leak intensity based on the concentration time series at each monitoring point. The specific calculation formula is as follows: ; in, This represents the leakage intensity at each measuring point at each time point. and Representing time respectively and The tracer gas concentration value below, The time interval between adjacent sampling times; The leakage intensity at each time point is integrated to generate a full-time leakage intensity sequence. The average leakage intensity is then obtained from the full-time leakage intensity sequence, using the following specific calculation formula: ; in, This represents the average leakage intensity at measurement point (i,j) during the entire bulging process, where K represents the total number of sampling points. This represents the leakage intensity at each measuring point at each time point; The location of the maximum leakage is determined by comparing the average leakage intensity at each measuring point throughout the bulging process; Frequency domain analysis was performed on the full-time leakage intensity sequence. The corresponding power spectral density function was obtained through Fast Fourier Transform, and the dominant frequency of the spectrum was defined. The specific calculation formula is as follows: ; in, Represents the main frequency of the spectrum. Represents the power spectral density function; Meanwhile, the fluctuation range of leakage intensity is quantified by the ratio of the standard deviation to the mean of the leakage intensity, specifically expressed as: ; in, Represents the fluctuation range. This represents the leakage intensity at each measuring point at each time point. This represents the average leakage intensity at measuring point (i,j) throughout the entire bulging process. Represents the smallest positive number; The original wear dataset was obtained, and combined with the mechanical response mechanism of the sealing interface, the deformation component and wear amount were decoupled using the benchmark calibration method. The specific calculation formula is as follows: ; in, This represents the amount of wear during the Nth bulging process. This represents the change in the gap at the final moment of the measurement point. The plastic deformation benchmark represents the first measurement of bulging. The cumulative wear amount under multiple bulging cycles is calculated based on the wear amount. The specific calculation formula is as follows: ; in, Represents cumulative wear and tear. This represents the cumulative wear amount after the previous bulging process, satisfying the initial conditions. This represents the amount of wear during the Nth bulging process; An empirical model based on gap variation is used to approximate the estimate of instantaneous wear accumulation. The specific calculation formula is as follows: ; in, This represents the estimated value of instantaneous cumulative wear. This represents the cumulative wear amount after the previous bulging process ended. Indicates the time of measurement point. The change in gap relative to the initial state. This serves as the benchmark for the first bulging measurement of plastic deformation. This is the proportionality coefficient.

[0025] Specifically, the unified spatiotemporal coordinate system constructed in step S1 The obtained original datasets of leakage and wear are systematically feature extracted and physical quantity decoupled to obtain leakage characteristic parameters for characterizing the leakage behavior of the sealing interface and wear parameters for characterizing the material degradation state. The above calculation results are used as input data for state determination and correlation analysis in the subsequent step S3.

[0026] First, during the calculation of leakage characteristic parameters, the original leakage dataset is used as input to perform joint time-domain and frequency-domain analysis on the concentration time series of each measuring point. Specifically, for any measuring point (i,j), at time... The leakage intensity is defined as the rate of change of concentration per unit time, and the specific calculation formula is as follows: ; in, This represents the leakage intensity at each measuring point at each time point. and Representing time respectively and The tracer gas concentration value below, The time interval between adjacent sampling times is a known constant, specifically a value of When k=1, the definition is... The physical meaning of this parameter is the enrichment rate of the tracer gas in a local area per unit time, and its value directly reflects the instantaneous severity of the leakage behavior at the corresponding location.

[0027] Integrate leakage intensity data at various time points to generate a full-time leakage intensity sequence. Furthermore, spatial statistical analysis was performed on the full-time leakage intensity sequence to extract the leakage location distribution characteristics. Specifically, on a fixed axial layer j, the leakage intensity at each circumferential measuring point was averaged over time to obtain the average leakage intensity. The specific calculation formula is as follows: ; in, This represents the average leakage intensity at measurement point (i,j) during the entire bulging process, where K represents the total number of sampling points. This represents the leakage intensity at each measuring point at each time point.

[0028] Based on this, the characteristics of the maximum leakage location are solved, through... This method identifies the circumferential location where leakage is most concentrated on the j-th axial layer, thus achieving a quantitative characterization of circumferential leakage concentration. Simultaneously, by comparing the average leakage intensity of different axial layers j under a fixed circumferential position i, it can be determined that leakage mainly occurs in the root, middle, or opening region of the sealing cone, thereby identifying the axial leakage distribution.

[0029] Furthermore, to characterize the fluctuation patterns of leakage behavior, frequency domain analysis was performed on the full-time leakage intensity sequence. The corresponding power spectral density function was obtained through Fast Fourier Transform, and the dominant frequency of the spectrum was defined. The specific calculation formula is as follows: ; in, Represents the main frequency of the spectrum. The power spectral density function represents the dominant frequency of the spectrum, which reflects the dominant fluctuation rhythm during the leakage process.

[0030] Meanwhile, the fluctuation range of leakage intensity is quantified by the ratio of the standard deviation to the mean of the leakage intensity, and its expression is as follows: ; in, Represents the fluctuation range. This represents the leakage intensity at each measuring point at each time point. This represents the average leakage intensity at measuring point (i,j) throughout the entire bulging process. This represents a very small positive number and is used to avoid the denominator being zero.

[0031] The aforementioned frequency domain dominant frequency and fluctuation amplitude together constitute the characteristics of the leakage fluctuation mode.

[0032] After extracting the leakage characteristic parameters, the physical quantity decoupling calculations were further performed on the original wear dataset to obtain the true wear amount of the sealing cone surface. Based on the mechanical response mechanism of the sealing interface, the gap change satisfies the following relationship: ; in, Indicates the time of measurement point. The change in gap relative to the initial state. This indicates the deformation component caused by the elastic or plastic deformation of the retaining ring end under high pressure. This indicates the amount of irreversible wear caused by the wear of the sealing cone material.

[0033] A benchmark calibration method is used to decouple the deformation component from the wear amount. Specifically, when the new mold is used for the first bulging process, the wear amount is approximately zero, and the end time of bulging is recorded. The amount of gap change, which is entirely caused by the plastic deformation at the end of the retaining ring, is defined as the plastic deformation reference. Because the plastic deformation behavior of the same retaining ring material under the same bulging process parameters has good repeatability, the deformation component at the end of the retaining ring is basically consistent with this reference value in each subsequent bulging. Therefore, for the end time of the Nth bulging (N≥2), the newly generated wear amount in this bulging can be calculated according to the following formula: ; in, This represents the amount of wear during the Nth bulging process. This represents the change in the gap at the final moment of the measurement point. Representing the plastic deformation benchmark obtained from the first bulging test, this decoupling method avoids complex finite element calculations and is easy to implement in engineering.

[0034] Under multiple bulging cycles, to describe the cumulative effect of wear, the cumulative wear amount is defined. Let be the total wear thickness at measurement point (i,j) after the Nth bulging process, and satisfy the initial conditions. And update according to the following recursive relationship, the specific expression is: ; in, Represents cumulative wear and tear. This represents the cumulative wear amount after the previous bulging process ended. This represents the amount of new wear generated during the Nth bulging process, and its value is the same as described above. Equivalent, obtained by decoupling from the above benchmark calibration method.

[0035] In the real-time stage before the bulging process is completed, an estimate of the instantaneous wear accumulation is introduced to achieve online estimation of the wear state. Considering that the wear process is related to contact pressure and relative sliding behavior, but these physical quantities are difficult to measure directly, an empirical model based on gap variation is used for approximate expression. The specific calculation formula is as follows: ; in, This represents an estimated value of instantaneous cumulative wear; it is a real-time value. This represents the cumulative wear amount after the previous bulging process ended. Indicates the time of measurement point. The change in gap relative to the initial state. The plastic deformation datum measured during the first bulging is the same as the datum in the aforementioned decoupling formula. This is the proportionality coefficient.

[0036] It should be noted that the proportionality coefficient The physical meaning is the proportion of the gap change exceeding the plastic deformation reference that is converted into wear. This is determined through offline wear calibration tests. Specifically, a set of identical molds was used for continuous bulging tests, and the actual wear amount after each bulging test was recorded. and with the corresponding Performing linear regression, the regression slope is... The typical value range is 0.6 to 0.8; in this example, we take... =0.7. This expression is activated only when the gap change exceeds the reference plastic deformation datum, thus avoiding misjudgment in the purely elastic stage.

[0037] Through the above calculation process, two key outputs are ultimately obtained: one is leakage characteristic parameters describing leakage behavior, including the full-time leakage intensity sequence, average leakage intensity, maximum leakage location, dominant frequency of the spectrum, and fluctuation amplitude. The other is wear parameters describing the structural degradation state, including cumulative wear and estimated instantaneous wear accumulation. These results form structured feature data under a unified spatial indexing system, providing direct input for state identification and correlation analysis in subsequent step S3.

[0038] S3: Construct a high-dimensional feature vector based on leakage characteristic parameters and wear parameters. The high-dimensional feature vector includes leakage intensity feature vector, wear accumulation feature vector, leakage fluctuation feature vector, and wear-leak coupling feature vector. Furthermore, step S3 also includes: The leakage intensity of each measuring point is arranged in the order of the two-dimensional spatial measuring point grid to construct a leakage intensity feature vector with dimension M×L; The estimated instantaneous wear accumulation values ​​of each measuring point are also arranged in the order of the two-dimensional spatial measuring point grid to construct a wear accumulation feature vector with dimension M×L; The full-time leakage intensity sequence is obtained, and the real-time fluctuation amplitude of each measuring point is dynamically calculated using a sliding time window mechanism. The real-time fluctuation amplitude is arranged in the order of the two-dimensional spatial measuring point grid to form a leakage fluctuation feature vector with dimension M×L. Based on the estimated instantaneous wear accumulation and leakage intensity, the wear-leakage correlation coefficient at each circumferential location is calculated. The specific calculation formula is as follows: ; in, The wear-leakage correlation coefficient represents the circumferential position i. This represents the estimated value of instantaneous cumulative wear. Represents the leakage intensity. This represents the estimated average instantaneous cumulative wear of all axial layers at circumferential position i. This represents the average leakage intensity of all axial layers at circumferential position i; By integrating the wear and leakage correlation coefficients of all circumferential positions, a wear and leakage coupled feature vector of dimension M×1 is constructed; The leakage intensity feature vector, wear accumulation feature vector, leakage fluctuation feature vector, and wear-leak coupling feature vector are concatenated into a high-dimensional feature vector.

[0039] Specifically, based on the leakage characteristic parameters and wear parameters obtained in step S2, the above parameters are uniformly organized and fused to construct a high-dimensional feature vector F(t) that can comprehensively characterize the operating state of the sealing interface. This feature vector F(t) is then used as the sole input data source for the state identification and risk assessment model in the subsequent step S4, thereby realizing the transformation from multi-parameter discrete representation to a unified state vector-driven modeling approach.

[0040] At a unified time t, the high-dimensional feature vector F(t) consists of four parts: leakage intensity feature vector, wear accumulation feature vector, leakage fluctuation feature vector, and wear-leakage coupling feature vector. Each component is derived from the calculation results of step S2 and is arranged in a structured manner in the two-dimensional space measurement point grid (i,j).

[0041] First, a leakage intensity feature vector Q(t) is constructed. This vector is composed of the leakage intensity at each spatial measuring point at the current moment. Arranged according to a predetermined spatial index order, among which This is derived from the result obtained by differentiating the concentration time series in step S2. By uniformly expanding the leakage intensity of all M×L measuring points, a leakage intensity feature vector of dimension M×L is formed, which is used to characterize the instantaneous leakage distribution state of the sealing interface at the current moment.

[0042] Secondly, a wear accumulation feature vector W(t) is constructed. This vector is composed of the estimated instantaneous wear accumulation values ​​at each measuring point. Arranged according to the same spatial indexing rules, where This vector originates from the real-time wear estimation model in step S2. It also has an M×L dimension, and its physical meaning represents the distribution of material degradation in various local areas of the sealing cone surface under the current operating conditions.

[0043] To further reflect the dynamic instability of leakage behavior, a leakage fluctuation characteristic vector R(t) is introduced. This is taken into account the fluctuation amplitude calculated in step S2. Since statistics based on the entire time series cannot directly reflect real-time fluctuation characteristics, a sliding time window mechanism is used to dynamically update them in this step. Specifically, at the current time t, the full-time series leakage intensity sequence within the most recent time range of ΔT = 0.5s is selected, and the fluctuation amplitude of the data within this local time period is recalculated to obtain the real-time fluctuation amplitude. The leakage fluctuation feature vector R(t) is formed by arranging the vectors according to their spatial indices (i,j) to form a vector of dimension M×L. This vector is used to characterize the stability and disturbance intensity of the leakage process over a short timescale.

[0044] After constructing the single physical quantity features, a wear-leakage coupling feature vector ρ(t) is further constructed to characterize the coupling relationship between wear and leakage. Specifically, at each circumferential position i, the estimated instantaneous wear accumulation and leakage intensity of different axial layers j=1..L at that position are selected, and their linear correlation in the axial direction is calculated. The specific calculation formula is as follows: ; in, The wear-leakage correlation coefficient represents the circumferential position i. This represents the estimated value of instantaneous cumulative wear. Represents the leakage intensity. The value represents the estimated average instantaneous cumulative wear of all axial layers at circumferential position i, specifically... , This represents the average leakage intensity of all axial layers at circumferential position i, specifically the value... Correlation coefficient The value range of is [-1, 1], and its physical meaning lies in characterizing the degree of consistency between the wear distribution and leakage distribution along the axial direction at the same circumferential position. When the value is positive and close to 1, it indicates that the area with greater wear also corresponds to a higher leakage intensity, which is consistent with the physical mechanism of normal wear driving enhanced leakage. When the value is close to 0 or negative, it indicates a deviation or inverse relationship between the leakage distribution and the wear distribution, which may indicate local damage, changes in the sealing failure mode, or other abnormal conditions.

[0045] By calculating for all circumferential positions i=1..M, the wear-leakage coupling characteristic vector is finally obtained. The vector has a dimension of M×1 and is used to characterize the cooperative degradation behavior of the sealed interface from the perspective of spatial coupling.

[0046] After constructing the above-mentioned features, they are finally concatenated to form a high-dimensional feature vector. This feature vector simultaneously contains information on leakage intensity distribution, wear accumulation status, leakage dynamic fluctuations, and the spatial coupling relationship between wear and leakage, thus constructing a complete multi-dimensional state representation space for the sealing interface. It is important to emphasize that the various state recognition models constructed in subsequent step S4, including bulge uniformity assessment, sealing cone surface health assessment, joint state inference, and instability risk determination, all use the aforementioned feature vector F(t) as the sole input data source, no longer directly calling the original data or intermediate variables from steps S1 and S2, thereby achieving centralized management of the data flow and consistency of model input.

[0047] S4: Calculate the bulging uniformity, wear non-uniformity, wear-leakage coupling anomaly index, and instability risk index based on the high-dimensional feature vector; Furthermore, step S4 also includes: The leakage intensity feature vector is extracted from the high-dimensional feature vector, and its circumferential distribution is analyzed under the axial layered structure. The circumferential leakage non-uniformity of the leakage intensity at each circumferential measuring point on each axial layer is calculated. The specific calculation formula is as follows: ; in, denoted by , representing the circumferential leakage non-uniformity of the j-th axial layer, and M representing the total number of circumferential locations. Represents the leakage intensity. This represents the average leakage intensity of all circumferential measuring points in the j-th layer; Based on the circumferential leakage non-uniformity, the overall bulging uniformity index is further obtained through a weighted summation method. The specific calculation formula is as follows: ; in, Represents the uniformity of bulging. This represents the circumferential leakage non-uniformity of the j-th axial layer. Represents the weighting coefficient; The wear accumulation feature vector is extracted from the high-dimensional feature vector, and the wear distribution of all spatial measuring points is statistically analyzed to calculate the maximum, minimum and average values ​​of the instantaneous wear accumulation estimate of each measuring point at the current moment. Wear unevenness is calculated based on the maximum, minimum, and average values ​​of the estimated instantaneous wear accumulation. The specific calculation formula is as follows: ; in, This represents uneven wear. This represents the maximum value of the estimated instantaneous cumulative wear. This represents the minimum value of the estimated instantaneous cumulative wear. This represents the average value of the estimated instantaneous cumulative wear. Represents the smallest positive number; The wear-leakage coupling feature vector is extracted from the high-dimensional feature vector, and a local anomaly index is defined based on the wear-leakage correlation coefficient. The specific calculation formula is as follows: ; in, Represents a local anomaly index. Represents the correlation coefficient between wear and leakage; By combining the wear accumulation feature vector and the leakage intensity feature vector in the high-dimensional feature vector, the wear leakage deviation is calculated. The specific calculation formula is as follows: ; in, This represents the wear leakage deviation at position i in the circumferential direction. This represents the estimated value of instantaneous cumulative wear. Representing leakage intensity, a and b are regression coefficients; The overall wear-leakage coupling anomaly index is defined by combining the local anomaly index and the wear-leakage deviation. The specific calculation formula is as follows: ; in, Represents an abnormal coupling index of wear and leakage. Represents a local anomaly index. This represents wear and leakage deviation. This represents the average value of the estimated instantaneous cumulative wear. and These are the weighting coefficients; The above-mentioned bulging uniformity, wear non-uniformity, and wear-leakage coupling anomaly index are unified and integrated, and combined with the global average value of the fluctuation amplitude obtained from the leakage fluctuation characteristic vector, the instability risk index is calculated. The specific calculation formula is as follows: ; in, Represents the risk index of instability. Represents the uniformity of bulging, Represents uneven wear, Represents an abnormal coupling index of wear and leakage. The global average value representing the fluctuation range. and These are the threshold parameters for each indicator under normal operating conditions. and These are the weighting coefficients.

[0048] Specifically, this step uses the previously constructed high-dimensional feature vector F(t) as the sole input data source, and constructs a multi-level state recognition model under a unified data interface. Through layer-by-layer analysis of leakage characteristics, wear state, and their coupling relationships, the bulging uniformity is calculated. Uneven wear Wear and leakage coupling anomaly index and instability risk index This allows for a quantitative assessment of the stability of the bulging process.

[0049] First, a bulge uniformity evaluation model based on leakage characteristics is constructed in the first layer. This model extracts the leakage intensity feature vector Q(t) from the high-dimensional feature vector F(t) and analyzes its circumferential distribution under the axial layered structure. Specifically, for each axial layer j, the circumferential leakage non-uniformity of the leakage intensity at each circumferential measuring point on that layer is calculated using the following formula: ; in, denoted by , representing the circumferential leakage non-uniformity of the j-th axial layer, and M representing the total number of circumferential locations. Represents the leakage intensity. This represents the average leakage intensity of all circumferential measuring points in the j-th layer, specifically the value... .

[0050] After obtaining the circumferential leakage non-uniformity of each axial layer, the overall bulging uniformity index is further obtained by weighted summation. The specific calculation formula is as follows: ; in, Represents the uniformity of bulging. This represents the circumferential leakage non-uniformity of the j-th axial layer. Represents the weighting coefficient, satisfying This index is used to reflect the difference in importance of different axial layers in sealing performance. It is a dimensionless positive value; when the sealing contact pressure distribution is uniform, the difference in leakage intensity across the circumferential directions is small, thus... A lower value indicates a higher value, while a higher value indicates a significant difference in leakage intensity in the circumferential direction when non-uniform deformations such as local bulges or depressions occur. Increase.

[0051] It should be noted that the weighting coefficients The setting is based on the degree of influence of different axial regions of the sealing cone surface on the bulging quality. The root region (j=1) directly constrains the deformation of the end of the retaining ring and has the greatest impact on the uniformity of bulging, therefore, it is selected as... The central region (j=2) is the next best option. The oral region (j=3) has the least impact, so we take... If L=2, then take , .

[0052] Secondly, a health assessment model for the sealing cone surface based on wear accumulation state is constructed in the second layer. This model extracts the wear accumulation feature vector W(t) from the high-dimensional feature vector F(t) and performs statistical analysis on the wear distribution of all spatial measuring points. Specifically, the maximum, minimum, and average values ​​of the instantaneous wear accumulation estimates at each measuring point at the current moment are first calculated, with the specific calculation formulas as follows: Maximum value: ; Minimum value: ; average value: ; in, This represents the maximum value of the estimated instantaneous cumulative wear. This represents the minimum value of the estimated instantaneous cumulative wear. This represents the average value of the estimated instantaneous cumulative wear. This represents an estimated value of the instantaneous cumulative wear.

[0053] Based on this, the wear unevenness index is defined as follows: ; in, This represents uneven wear. Representing a very small positive number to prevent the denominator from being zero, this index is used to characterize the degree of concentration of wear in spatial distribution. When the wear on the sealing cone surface is uniform, the wear difference between each measuring point is small. When the value is close to 0, and localized uneven wear or concentrated wear occurs, A significant increase reflects a deteriorating trend in the health of the sealing surface. Generally, based on historical experience, To ensure uniform wear, Slight uneven wear is indicated by a wear level between 0.2 and 0.4. This indicates severe uneven wear.

[0054] Furthermore, a joint inference model of leakage and wear is constructed in the third layer to identify abnormal damage behavior beyond the conventional wear mechanism. This model simultaneously extracts the wear accumulation feature vector W(t), the leakage intensity feature vector Q(t), and the wear-leakage coupling feature vector ρ(t) from the high-dimensional feature vector F(t). For each circumferential position i, the wear-leakage correlation coefficient is first used as the basis for... The local anomaly index is defined, and its specific calculation formula is as follows: ; in, This represents a local anomaly index, used to measure the degree of deviation between wear distribution and leakage distribution. This represents the correlation coefficient between wear and leakage. The value 0.2 in the formula is a preset threshold for the correlation coefficient between wear and leakage. When the value is less than a preset threshold of 0.2, the positive correlation between wear and leakage is considered to be significantly weakened, which may indicate an anomaly. This threshold is set based on data from 20 groups under normal wear conditions. Statistical analysis showed that the lower limit of the 95% confidence interval was 0.23, so 0.2 was taken as the dividing line between normal and abnormal.

[0055] Meanwhile, a linear regression model based on historical normal data calibration is introduced to establish the expected relationship between leakage intensity and wear accumulation. Specifically, in historical bulging operations, the wear accumulation and leakage intensity of the first 50 normal bulging operations are collected. A univariate linear regression is performed with the average leakage intensity as the independent variable and the corresponding cumulative wear amount as the dependent variable, yielding regression coefficients a and b. In this example, the calibrated values ​​are a = 0.32 μm·s / ppm and b = 5.2 μm. Based on this, the wear leakage deviation between the actual observed value and the predicted value is calculated using the following formula: ; in, This represents the wear leakage deviation at position i in the circumferential direction. This represents the estimated value of instantaneous cumulative wear. The value represents the leakage intensity, and a and b are regression coefficients. This deviation measures the difference between the actual wear and the wear predicted based on the leakage intensity; a larger deviation indicates that the current wear behavior deviates from the normal pattern.

[0056] Combining the two types of anomaly metrics mentioned above, an overall wear-leakage coupling anomaly index is defined, and the specific calculation formula is as follows: ; in, Represents an abnormal coupling index of wear and leakage. Represents a local anomaly index. This represents wear and leakage deviation. This represents the average value of the estimated instantaneous cumulative wear. and Let be the weighting coefficient, satisfying This indicator is used to comprehensively reflect whether there is an abnormal wear mechanism at the sealing interface, such as local scratches, indentations or foreign object embedding. The larger the value, the higher the degree of abnormality.

[0057] It should be noted that the weighting coefficients were set based on sensitivity analysis and the local anomaly index. It is sensitive to scratch-type damage, and the deviation term is sensitive to the aggravation of uniform wear. The two are equally important, so we take... .

[0058] Finally, a quantitative model for the instability risk of the bulging process is constructed in the fourth layer, and the above-mentioned multiple state indicators are unified and integrated. Specifically, the bulging uniformity U(t) and wear non-uniformity are extracted. Wear and leakage coupling anomaly index Based on this, the leakage fluctuation feature vector R(t) is further obtained from the high-dimensional feature vector F(t), and the global average value of the fluctuation amplitude is calculated. .

[0059] Based on this, an instability risk index is defined, and the specific calculation formula is as follows: ; Where U(t) represents the uniformity of bulging, Represents uneven wear, Represents an abnormal coupling index of wear and leakage. The global average value representing the fluctuation range. , and These are the threshold parameters for each indicator under normal operating conditions. and The weighting coefficients satisfy the normalization constraint; the recommended value is [value]. =0.3, =0.2, =0.3, =0.2, the instability risk index quantifies system stability from a multidimensional perspective. When the value is at a low level, it indicates that the bulging process is stable. When it gradually increases and exceeds the preset threshold, it indicates that the system has entered a warning or high-risk state, and process parameter intervention is required.

[0060] It should be noted that the specific values ​​and setting methods for the above thresholds are as follows: By statistically analyzing U(t) during 50 groups of normal expansion forming processes (finished products are qualified and have no defects), the 90th quantile is 0.13 and the 95th quantile is 0.16. The values ​​are then rounded down to determine the value. ; The mold life test determined that when When the value exceeds 0.2, the probability of visible wear marks appearing at the end of the retaining ring increases significantly with continued bulging. Therefore, 0.2 is taken as the threshold for mild wear. By statistically analyzing the real-time fluctuation amplitude during the normal expansion process, the mean is 0.12 and the standard deviation is 0.06. The mean + 3σ ≈ 0.30 is taken as the threshold. The above threshold is exactly the same as the threshold used in the subsequent step S5, which can maintain the consistency of the whole scheme.

[0061] Furthermore, the weighting coefficients were primarily determined through pairwise comparisons of the importance of the four indicators using the Analytic Hierarchy Process (AHP). Bulging uniformity directly reflects the current bulging quality and is the most important; the wear-leakage coupling anomaly index indicates potential damage and is the second most important; wear unevenness reflects long-term degradation and is the next most important; leakage fluctuation serves as an auxiliary indicator. After consistency testing, the recommended weights are: .

[0062] Through the above multi-level computing structure, the transformation from single physical quantity analysis to multi-source information fusion judgment is realized, so that the stability assessment of the bulging process not only depends on local feature changes, but also can comprehensively reflect leakage behavior, wear state and their coupling relationship, thereby significantly improving the accuracy and robustness of state identification.

[0063] S5: Determine the state category of the bulging process based on the multi-threshold judgment logic tree, and generate the corresponding process parameter adjustment strategy based on the state category of the bulging process.

[0064] Furthermore, step S5 also includes: Obtain the average values ​​of bulging uniformity, wear non-uniformity, wear-leakage coupling anomaly index, instability risk index, and instantaneous wear accumulation estimate; The thresholds for bulging uniformity, wear non-uniformity, coupling anomaly, instability risk, and ultimate wear amount were determined through historical data statistics and experimental calibration. Based on the above parameters and thresholds, the bulging process state categories are determined sequentially according to a preset priority order. The bulging process state categories include local damage alarm state, mold life termination state, high instability risk state, circumferential non-uniformity dominant state, wear dominant state, coupling abnormality dominant state, fluctuation abnormality dominant state, and normal working condition state. According to different bulging process state categories, corresponding process parameter adjustment strategies are implemented. The process parameter adjustment strategies include damage marking and early warning strategies, mold replacement strategies, emergency load reduction strategies, adaptive pressure compensation strategies, wear mitigation strategies, damage location and enhanced monitoring strategies, and pressure smoothing and damping enhancement strategies.

[0065] Specifically, based on the obtained bulging uniformity, wear non-uniformity, wear leakage coupling anomaly index, and instability risk index, a judgment logic tree based on multiple thresholds is constructed. By setting preset thresholds and preset priority order, the state of the bulging process is determined, and corresponding process parameter adjustment strategies are generated accordingly, thus forming a complete closed-loop control mechanism.

[0066] First, a decision logic tree based on multiple thresholds is constructed. The logic tree is based on... , , , and the average value of the estimated instantaneous wear accumulation. The input criteria are determined by statistical analysis of historical data and experimental calibration, specifically: Bulging uniformity threshold The value is set to 0.15. This value is derived from the 90th percentile of U(t) in 50 sets of normal bulging data. When U(t) exceeds 0.15, the probability of the circumferential dimension deviation of the retaining ring exceeding 0.5mm increases to more than 15%.

[0067] Wear unevenness threshold The value was set to 0.2, determined through accelerated mold life testing. Recording was performed during 100 consecutive bulging cycles. With the change in surface roughness at the end of the retaining ring, when When the value exceeds 0.2, the probability of visible wear marks at the end increases significantly, reaching the severe wear threshold. The value is set to 0.4. When the wear unevenness exceeds 0.4, it indicates that the wear has been highly concentrated in a certain local area. According to past engineering experience, continued use at this time will cause unacceptable stress concentration at the end of the retaining ring, so it is determined that the mold life has ended.

[0068] Coupling anomaly threshold The value was set to 0.3, obtained through artificial damage experiments. Scratches with depths of 0.1 mm, 0.2 mm, and 0.3 mm were created on the sealing cone surface, and the corresponding values ​​were measured. Value, when the scratch depth is greater than or equal to 0.2 mm. Therefore, 0.3 is chosen as the alarm threshold.

[0069] The instability risk threshold includes a normal upper limit threshold and a severe deviation threshold, which are set as follows: and The setting of these two thresholds is based on statistical process control. Based on past engineering experience, under normal operating conditions, the mean of S(t) is about 0.2 and the standard deviation is about 0.1. 0.5 corresponds to the mean +3σ, which is the upper limit of normal, and 1.0 corresponds to the mean +8σ, which is a serious deviation.

[0070] Limit wear The depth is determined based on the mold material and the maximum allowable wear depth of the sealing cone surface. In this embodiment, the sealing cone surface is made of 5CrNiMo mold steel, and the maximum allowable wear depth is 0.3mm. Exceeding this value will cause the sealing cone angle to change by more than 0.5°, affecting the sealing performance. Therefore, the maximum allowable wear depth is determined by the mold material and the maximum allowable wear depth of the sealing cone surface. .

[0071] During the determination process, the bulging process state category is determined sequentially according to the preset thresholds mentioned above and the preset priority order: First priority, when Furthermore, if the damage persists for more than 3 sampling periods, it is preferentially determined that there is local damage on the sealing cone surface. At this time, the bulging process status is classified as a local damage alarm state. The threshold is used instead of the U(t) threshold because coupling anomalies are a direct indicator of damage; Second priority, when wear average Exceeding the limit of wear or When the mold life ends, the state of the bulging process is classified as the mold life end state. Third priority, when At this time, the expansion process is classified as a high-risk instability state; Fourth priority, when At this point, the bulging process carries a risk of instability, requiring further identification of the source of the anomaly through a dominant factor identification mechanism. The dominant factor is determined by normalizing various indicators. as well as By comparing and determining the index corresponding to the maximum value, the dominant anomaly source is selected, thereby identifying the dominant state of circumferential non-uniformity. Maximum), wear-dominant state ( Maximum), coupled abnormal dominant state ( ) or the dominant state of fluctuation abnormality ( maximum); Fifth priority: If none of the above conditions are met, the expansion process status is determined to be normal operating condition. In this case, there is no need to generate a process parameter adjustment strategy; the system can simply continue to operate normally.

[0072] After determining the state category of the bulging process, the corresponding process parameter adjustment strategy is further implemented according to the different bulging process state categories: When the bulging process status is in the local damage alarm state, the system executes the damage marking and early warning strategy. The system triggers an alarm for local damage to the sealing cone surface and automatically performs the following operations: First, it tightens the product quality tolerance, temporarily reducing the bulging dimension tolerance of subsequent retaining rings from ±1.0 mm to ±0.6 mm, and tightening the leakage intensity fluctuation tolerance from ±20% to ±10%, so as to detect quality deviations caused by seal damage earlier; Second, it prompts the operator to check, and displays the possible location of seal cone surface damage on the HMI interface based on the location of the maximum leakage, suggesting that the area be checked using an endoscope or the rework method during the next mold change; Finally, it automatically marks abnormal batches, marking the production data of the current and the next 5 retaining rings as batches suspected of being affected by damage, facilitating quality traceability.

[0073] When the bulging process status is in the mold life end state, the system will execute the mold lock replacement strategy, implement mold life end control, lock the bulging program, refuse to execute the next bulging command, and display "The sealing cone surface has reached the end of its life, please replace the mold" on the HMI interface. After the operator completes the mold replacement, he / she must confirm and reset the wear counter in the system before the system can be unlocked.

[0074] When the expansion process is classified as a high instability risk state, the system will execute an emergency load reduction strategy, immediately stop the downward movement of the hydraulic press, and linearly release the pressure of the ultra-high pressure pump at a rate not exceeding 50 MPa / s, with a depressurization time of greater than or equal to 3 seconds to prevent impact caused by a sudden pressure drop. At the same time, the contact pressure of the sealing cone surface will be maintained unchanged to avoid instability and warping of the end of the retaining ring due to the sudden disappearance of the constraint during the depressurization process. After the system stabilizes, the HMI interface will display a high instability risk alarm and record the current expansion parameters. The operator will then decide whether to re-expand or scrap the retaining ring based on the alarm information.

[0075] When the bulging process is classified as circumferentially non-uniform, the system executes an adaptive pressure compensation strategy. Specifically, it first calculates the leakage deviation ratio at each circumferential position: ; in, The leakage deviation ratio at circumferential position i. The leakage deviation ratio represents the leakage intensity. A positive value indicates that the leakage intensity in this area is higher than the circumferential average, indicating insufficient contact pressure. A negative value indicates that the leakage intensity is lower than the average, indicating excessive contact pressure.

[0076] Then, the sealing pressure of each circumferential sector is adjusted in a closed loop. The contact pressure of the sealing cone surface is controlled by M independent hydraulic servo cylinders, each servo cylinder corresponding to one circumferential sector. Initial sealing pressure Given the preset process value, which is 25 MPa in this example, the adjustment formula is: ; in, Represents the new sealing pressure. Represents the original sealing pressure. This is the proportional adjustment coefficient. This adjustment strategy uses leakage deviation to drive pressure compensation, thereby enabling areas with larger leaks to obtain higher contact pressure and thus restoring circumferential contact uniformity.

[0077] It should be noted that the proportional adjustment coefficient The determination is mainly achieved through step response tests, when When the coefficient of performance is 0.8, the system reduces the leakage deviation ratio from 0.5 to below 0.1 within 3 sampling periods, exhibiting fast response and no overshoot oscillation. Therefore, the coefficient of performance is chosen. =0.8.

[0078] When the bulging process is in a wear-dominated state, the system implements a wear mitigation strategy, specifically by adjusting the following three parameters to reduce localized wear: First, reduce the bulging rate by decreasing the target value of the hydraulic press's downward speed from the current value vcurrent to [a lower value]. The maximum single pressure reduction should not exceed 20%. The reduction rate decreases the relative sliding speed between the sealing cone and the end of the retaining ring, thereby reducing frictional work and wear rate. Secondly, the internal pressure loading curve is optimized, changing the original steep-rise internal pressure curve to a gradual rise-plateau-gradient curve. Specifically, the time for the internal pressure to rise from 0 to the target pressure is extended to 1.2 times the original time, and a period of [duration missing] is inserted simultaneously when the target pressure is reached (0.8 times). The pressure holding platform lasts for seconds. During the unloading phase, the pressure release rate is reduced from 5 MPa / s to 3 MPa / s. This curve can reduce the peak contact stress, allowing the retaining ring material sufficient time for stress relaxation, thereby reducing wear concentration. Finally, micro-lubrication is introduced. If the mold is equipped with a lubrication system, the micro-spraying device on the sealing cone surface is activated to spray molybdenum disulfide dry film lubricant. The spray amount is controlled at 0.1 g / m² each time, sprayed once every 10 seconds, for a total of 3 sprays. The lubricant can reduce the coefficient of friction and significantly reduce wear.

[0079] When the bulging process is in a state dominated by coupling anomalies, the system executes a damage localization and enhanced monitoring strategy. At this time, the sealing interface exhibits mild or intermittent wear and leakage coupling anomalies, which have not yet triggered an immediate alarm, but require proactive intervention to prevent further deterioration. First, live local damage tracking is performed, and the suspected damage area is highlighted on the HMI based on the circumferential location of the maximum leakage in each axial layer. Then, the coupling anomaly threshold is dynamically adjusted; if this state persists for more than 5 sampling cycles, the system is temporarily... The threshold for damage alarm was lowered from 0.3 to 0.25, enabling early warning. Finally, process parameters were fine-tuned. Without triggering overall pressure adjustments, a small pressure increment was applied to the independent hydraulic servo cylinder corresponding to the suspected damage area to attempt to suppress leakage through local clamping. The response effect was recorded for subsequent damage evolution analysis.

[0080] When the expansion process is in a state dominated by abnormal fluctuations, the system implements a pressure smoothing and damping enhancement strategy. At this time, the leakage intensity fluctuates drastically within a short period, usually caused by pressure pulsation, contact chatter, or lubrication instability; if not suppressed, it may develop into instability. First, pressure smoothing filtering is activated. In the ultra-high pressure pump control loop, the pressure command signal is passed through a first-order low-pass filter, reducing the cutoff frequency from 20Hz to 5Hz to suppress high-frequency pressure pulsation. Then, hydraulic damping is increased. For the sealing cone contact pressure control loop, the proportional adjustment coefficient is adjusted... Temporarily reduce the coefficient of friction from 0.8 to 0.5. Suspend micro-lubrication. If the lubrication system is operating, immediately stop spraying, as uneven lubricant distribution can also cause fluctuations in the coefficient of friction. After the fluctuation subsides for 2 seconds, try again with half the spray volume. Finally, perform fluctuation source diagnosis, marking the location of the measurement point with the largest fluctuation amplitude on the HMI, prompting the operator to check the sealing cone surface in that area for foreign objects or early scratches.

[0081] Through the above-mentioned judgment and adjustment mechanism, a closed-loop control process from state recognition to active intervention is realized, enabling the system to dynamically adjust process parameters according to the real-time operating status of the sealing interface, thereby extending the service life of the mold and reducing the risk of instability while ensuring the bulging quality.

[0082] Example 2: Based on the same inventive concept as the generator ultra-large retaining ring bulging process parameter optimization method in the foregoing examples, this application also provides a generator ultra-large retaining ring bulging process parameter optimization system. Please refer to the appendix. Figure 2 The system includes: A multi-source sensor data acquisition module 11 is used to construct a multi-source sensor array, synchronously acquire tracer gas concentration field information and sealing gap change information, and respectively form a leak raw dataset and a wear raw dataset. The multi-source sensor array includes a gas concentration sensor array and a displacement sensor array. The feature extraction and parameter calculation module 12 is used to perform feature extraction and physical quantity decoupling calculation on the original leakage dataset and the original wear dataset respectively, to obtain leakage feature parameters and wear parameters. The leakage feature parameters include the full-time leakage intensity sequence, average leakage intensity, maximum leakage location, spectrum main frequency and fluctuation amplitude. The wear parameters include the cumulative wear amount and the estimated value of instantaneous cumulative wear. A high-dimensional feature vector construction module 13 is used to construct a high-dimensional feature vector based on leakage feature parameters and wear parameters. The high-dimensional feature vector includes a leakage intensity feature vector, a wear accumulation feature vector, a leakage fluctuation feature vector, and a wear-leak coupling feature vector. Key indicator calculation module 14 is used to calculate bulging uniformity, wear non-uniformity, wear leakage coupling anomaly index and instability risk index based on high-dimensional feature vectors. The state determination and strategy generation module 15 is used to determine the state category of the bulging process according to the judgment logic tree based on multiple thresholds, and generate the corresponding process parameter adjustment strategy according to the state category of the bulging process.

[0083] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0084] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for optimizing the process parameters of the bulging forming process for an ultra-large retaining ring of a generator, characterized in that, The method includes: A multi-source sensor array is constructed to simultaneously collect tracer gas concentration field information and sealing gap change information, forming original datasets of leakage and wear respectively. The multi-source sensor array includes a gas concentration sensor array and a displacement sensor array. Feature extraction and physical quantity decoupling calculation were performed on the original leakage dataset and the original wear dataset to obtain leakage feature parameters and wear parameters. The leakage feature parameters include the full-time leakage intensity sequence, average leakage intensity, maximum leakage location, spectral main frequency and fluctuation amplitude. The wear parameters include the cumulative wear amount and the estimated value of instantaneous cumulative wear. A high-dimensional feature vector is constructed based on leakage characteristic parameters and wear parameters. The high-dimensional feature vector includes leakage intensity feature vector, wear accumulation feature vector, leakage fluctuation feature vector, and wear-leak coupling feature vector. The bulging uniformity, wear non-uniformity, wear-leakage coupling anomaly index, and instability risk index are calculated based on the high-dimensional feature vectors. The bulging process state category is determined based on a multi-threshold-based judgment logic tree, and a corresponding process parameter adjustment strategy is generated based on the bulging process state category.

2. The method for optimizing the process parameters of the bulging forming process of an ultra-large retaining ring for a generator as described in claim 1, characterized in that, The original datasets for leakage and wear were generated separately, including: The bulging working medium is marked using a tracer gas-assisted detection method; The number of circumferential gas concentration sensors is determined based on the nominal diameter of the retaining ring, and the number of axial gas concentration sensor layers is determined based on the height of the sealing cone. A two-dimensional spatial measuring point grid marked with circumferential index i and axial index j is constructed. Miniature gas concentration sensors are installed based on the two-dimensional spatial measuring point grid to form a gas concentration sensor array. The tracer gas concentration values ​​at each measuring point (i,j) are collected by a gas concentration sensor array, and the concentration time series of each measuring point is constructed in chronological order. The concentration time series of all measuring points are then summarized to form the original dataset of the leak. An eddy current displacement sensor array is formed by installing eddy current displacement sensors on a two-dimensional spatial measuring point grid. By using the real-time gap values ​​of each measuring point in the displacement sensor array and taking the initial gap value before the bulging begins as a benchmark, the gap change is calculated. A time series of gap changes for each measuring point is constructed in chronological order, and the time series of gap changes for all measuring points are summarized to form the original wear dataset.

3. The method for optimizing the process parameters of the bulging forming process of an ultra-large retaining ring for a generator as described in claim 1, characterized in that, Obtain leakage characteristic parameters and wear parameters, including: Obtain the original dataset of the leak, and calculate the leak intensity based on the concentration time series at each monitoring point. The specific calculation formula is as follows: ； in, This represents the leakage intensity at each measuring point at each time point. and Representing time respectively and The tracer gas concentration value below, The time interval between adjacent sampling times; The leakage intensity at each time point is integrated to generate a full-time leakage intensity sequence. The average leakage intensity is then obtained from the full-time leakage intensity sequence, using the following specific calculation formula: ； in, Indicates the measuring points throughout the entire bulging process. The average leakage intensity, where K represents the total number of sampling points. This represents the leakage intensity at each measuring point at each time point; The location of the maximum leakage is determined by comparing the average leakage intensity at each measuring point throughout the bulging process; Frequency domain analysis was performed on the full-time leakage intensity sequence. The corresponding power spectral density function was obtained through Fast Fourier Transform, and the dominant frequency of the spectrum was defined. The specific calculation formula is as follows: ； in, Represents the main frequency of the spectrum. Represents the power spectral density function; Meanwhile, the fluctuation range of leakage intensity is quantified by the ratio of the standard deviation to the mean of the leakage intensity, specifically expressed as: ； in, Represents the fluctuation range. This represents the leakage intensity at each measuring point at each time point. Indicates the measuring points throughout the entire bulging process. Average leakage intensity, Represents the smallest positive number; The original wear dataset was obtained, and combined with the mechanical response mechanism of the sealing interface, the deformation component and wear amount were decoupled using the benchmark calibration method. The specific calculation formula is as follows: ; in, This represents the amount of wear during the Nth bulging process. This represents the change in the gap at the final moment of the measurement point. The plastic deformation benchmark represents the first measurement of bulging. The cumulative wear amount under multiple bulging cycles is calculated based on the wear amount. The specific calculation formula is as follows: ； in, Represents cumulative wear and tear. This represents the cumulative wear amount after the previous bulging process, satisfying the initial conditions. , This represents the amount of wear during the Nth bulging process; An empirical model based on gap variation is used to approximate the estimate of instantaneous wear accumulation. The specific calculation formula is as follows: ; in, This represents the estimated value of instantaneous cumulative wear. This represents the cumulative wear amount after the previous bulging process ended. Indicates the time of measurement point. The change in gap relative to the initial state. This serves as the benchmark for the first bulging measurement of plastic deformation. This is the proportionality coefficient.

4. The method for optimizing the process parameters of the bulging forming process for an ultra-large retaining ring of a generator as described in claim 1, characterized in that, A high-dimensional feature vector is constructed based on leakage characteristic parameters and wear parameters, including: The leakage intensity of each measuring point is arranged in the order of the two-dimensional spatial measuring point grid to construct a leakage intensity feature vector with dimension M×L; The estimated instantaneous wear accumulation values ​​of each measuring point are also arranged in the order of the two-dimensional spatial measuring point grid to construct a wear accumulation feature vector with dimension M×L; The full-time leakage intensity sequence is obtained, and the real-time fluctuation amplitude of each measuring point is dynamically calculated using a sliding time window mechanism. The real-time fluctuation amplitude is arranged in the order of the two-dimensional spatial measuring point grid to form a leakage fluctuation feature vector with dimension M×L. Based on the estimated instantaneous wear accumulation and leakage intensity, the wear-leakage correlation coefficient at each circumferential location is calculated. The specific calculation formula is as follows: ； in, The wear-leakage correlation coefficient represents the circumferential position i. This represents the estimated value of instantaneous cumulative wear. Represents the leakage intensity. This represents the estimated average instantaneous cumulative wear of all axial layers at circumferential position i. This represents the average leakage intensity of all axial layers at circumferential position i; By integrating the wear and leakage correlation coefficients of all circumferential positions, a wear and leakage coupled feature vector of dimension M×1 is constructed; The leakage intensity feature vector, wear accumulation feature vector, leakage fluctuation feature vector, and wear-leak coupling feature vector are concatenated into a high-dimensional feature vector.

5. The method for optimizing the process parameters of the bulging forming process of an ultra-large retaining ring for a generator as described in claim 1, characterized in that, The following parameters are calculated based on high-dimensional eigenvectors: bulging uniformity, wear non-uniformity, wear-leakage coupling anomaly index, and instability risk index, including: The leakage intensity feature vector is extracted from the high-dimensional feature vector, and its circumferential distribution is analyzed under the axial layered structure. The circumferential leakage non-uniformity of the leakage intensity at each circumferential measuring point on each axial layer is calculated. The specific calculation formula is as follows: ； in, denoted by , representing the circumferential leakage non-uniformity of the j-th axial layer, and M representing the total number of circumferential locations. Represents the leakage intensity. This represents the average leakage intensity of all circumferential measuring points in the j-th layer; Based on the circumferential leakage non-uniformity, the overall bulging uniformity index is further obtained through a weighted summation method. The specific calculation formula is as follows: ； in, Represents the uniformity of bulging. This represents the circumferential leakage non-uniformity of the j-th axial layer. Represents the weighting coefficient; The wear accumulation feature vector is extracted from the high-dimensional feature vector, and the wear distribution of all spatial measuring points is statistically analyzed to calculate the maximum, minimum and average values ​​of the instantaneous wear accumulation estimate of each measuring point at the current moment. Wear unevenness is calculated based on the maximum, minimum, and average values ​​of the estimated instantaneous wear accumulation. The specific calculation formula is as follows: ； in, This represents uneven wear. This represents the maximum value of the estimated instantaneous cumulative wear. This represents the minimum value of the estimated instantaneous cumulative wear. This represents the average value of the estimated instantaneous cumulative wear. Represents the smallest positive number; The wear-leakage coupling feature vector is extracted from the high-dimensional feature vector, and a local anomaly index is defined based on the wear-leakage correlation coefficient. The specific calculation formula is as follows: ； in, Represents a local anomaly index. Represents the correlation coefficient between wear and leakage; By combining the wear accumulation feature vector and the leakage intensity feature vector in the high-dimensional feature vector, the wear leakage deviation is calculated. The specific calculation formula is as follows: ； in, This represents the wear leakage deviation at position i in the circumferential direction. This represents the estimated value of instantaneous cumulative wear. Representing leakage intensity, a and b are regression coefficients; The overall wear-leakage coupling anomaly index is defined by combining the local anomaly index and the wear-leakage deviation. The specific calculation formula is as follows: ; in, Represents an abnormal coupling index of wear and leakage. Represents a local anomaly index. This represents wear and leakage deviation. This represents the average value of the estimated instantaneous cumulative wear. and These are the weighting coefficients; The above-mentioned bulging uniformity, wear non-uniformity, and wear-leakage coupling anomaly index are unified and integrated, and combined with the global average value of the fluctuation amplitude obtained from the leakage fluctuation characteristic vector, the instability risk index is calculated. The specific calculation formula is as follows: ； in, Represents the risk index of instability. Represents the uniformity of bulging, Represents uneven wear, Represents an abnormal coupling index of wear and leakage. The global average value representing the fluctuation range. and These are the threshold parameters for each indicator under normal operating conditions. and These are the weighting coefficients.

6. The method for optimizing the process parameters of the bulging forming process of an ultra-large retaining ring for a generator as described in claim 1, characterized in that, The bulging process state category is determined based on a multi-threshold-based judgment logic tree, and a corresponding process parameter adjustment strategy is generated based on the bulging process state category, including: Obtain the average values ​​of bulging uniformity, wear non-uniformity, wear-leakage coupling anomaly index, instability risk index, and instantaneous wear accumulation estimate; The thresholds for bulging uniformity, wear non-uniformity, coupling anomaly, instability risk, and ultimate wear amount were determined through historical data statistics and experimental calibration. Based on the above parameters and thresholds, the bulging process state categories are determined sequentially according to a preset priority order. The bulging process state categories include local damage alarm state, mold life termination state, high instability risk state, circumferential non-uniformity dominant state, wear dominant state, coupling abnormality dominant state, fluctuation abnormality dominant state, and normal working condition state. According to different bulging process state categories, corresponding process parameter adjustment strategies are implemented. The process parameter adjustment strategies include damage marking and early warning strategies, mold replacement strategies, emergency load reduction strategies, adaptive pressure compensation strategies, wear mitigation strategies, damage location and enhanced monitoring strategies, and pressure smoothing and damping enhancement strategies.

7. A system for optimizing process parameters of ultra-large retaining ring bulging for generators, characterized in that, The system is used to implement the method for optimizing the process parameters of the generator ultra-large retaining ring expansion forming process as described in any one of claims 1 to 6, and the system includes: A multi-source sensor data acquisition module is used to construct a multi-source sensor array, synchronously acquire tracer gas concentration field information and sealing gap change information, and form leakage raw dataset and wear raw dataset respectively. The multi-source sensor array includes a gas concentration sensor array and a displacement sensor array. The feature extraction and parameter calculation module is used to extract features and decouple physical quantities from the original leakage dataset and the original wear dataset respectively, to obtain leakage feature parameters and wear parameters. The leakage feature parameters include the full-time leakage intensity sequence, average leakage intensity, maximum leakage location, spectral main frequency and fluctuation amplitude. The wear parameters include the cumulative wear amount and the estimated value of instantaneous cumulative wear amount. A high-dimensional feature vector construction module is used to construct high-dimensional feature vectors based on leakage feature parameters and wear parameters. The high-dimensional feature vectors include leakage intensity feature vectors, wear accumulation feature vectors, leakage fluctuation feature vectors, and wear-leak coupling feature vectors. The key indicator calculation module is used to calculate the bulging uniformity, wear non-uniformity, wear leakage coupling anomaly index and instability risk index based on high-dimensional feature vectors. The state determination and strategy generation module is used to determine the state category of the bulging process based on a multi-threshold-based judgment logic tree, and generate corresponding process parameter adjustment strategies based on the state category of the bulging process.

Images