A slurry pump flow part life prediction system and optimization method based on reverse engineering

Abstract

The application provides a slurry pump overcurrent component life prediction system and optimization method based on reverse engineering, belonging to the technical field of slurry pumps, comprising a data acquisition layer, a model construction layer, a decision execution layer and a closed-loop iteration layer. The data acquisition layer acquires global point cloud data of the overcurrent component by using multi-view splicing scanning through a non-contact three-dimensional scanning node, and acquires actual wear data and operating condition data; the model construction layer constructs a corresponding digital twin model, and establishes an adaptive wear rate model based on the above data; the decision execution layer predicts the remaining life according to the model, and outputs operation and maintenance optimization instructions in stages; the closed-loop iteration layer feeds back the execution result to the first two layers to form a whole life cycle data closed loop. The prediction system combines reverse engineering and digital twin technology to realize data acquisition, model construction, life prediction and operation and maintenance decision-making whole-process quality improvement and efficiency increase, solve the problems of prediction distortion and operation and maintenance without basis, reduce costs, reduce unplanned shutdowns and improve equipment operation stability.

Description

Technical Field

[0001] This invention belongs to the field of slurry pump technology, and more specifically, it relates to a slurry pump flow component life prediction system and optimization method based on reverse engineering. Background Technology

[0002] Slurry pumps, as core fluid transport equipment in mining, metallurgy, power, and chemical industries, transport media containing solid particles. Their impellers, sleeves, and wear plates are subjected to harsh conditions of high-speed scouring, abrasive wear, and corrosion coupling, making them highly susceptible to localized wear, spalling, and erosion failure. Industry statistics show that under high-concentration, high-hardness slurry conditions, the average service life of slurry pump components is only a few months. Frequent downtime for replacement and spare parts wear not only significantly increases equipment maintenance costs but also causes unplanned production line interruptions, severely restricting continuous production efficiency and enterprise economic benefits. Furthermore, uneven wear of flow components can trigger a chain reaction of problems, such as increased pump vibration, efficiency decline, and escalating energy consumption, further exacerbating equipment operation risks. Therefore, achieving accurate life prediction and structural optimization of slurry pump flow components is a key technical challenge that the industry urgently needs to solve.

[0003] Currently, in assessing the lifespan of slurry pump flow components, insufficient accuracy in wear data acquisition and a lack of fidelity in digital models lead to distorted lifespan predictions and a lack of basis for optimization decisions. Traditional techniques either rely on contact measurements, which cannot adapt to the complex curved surface structure of flow components and are difficult to achieve high-precision wear morphology acquisition across the entire area; or reverse modeling only reaches the level of theoretical geometric reconstruction, and the wear rate model cannot match the actual working conditions and wear evolution laws, ultimately resulting in a large deviation in remaining lifespan predictions. Subsequent operation and maintenance optimization and structural improvements also lack accurate data support, failing to meet the long-term management needs of industrial sites. Summary of the Invention

[0004] The purpose of this invention is to provide a slurry pump flow component life prediction system based on reverse engineering, which aims to solve the problems of low accuracy of wear data acquisition and wear rate model that cannot fit the actual working conditions and wear evolution law in slurry pump flow component life assessment, resulting in distorted life prediction and lack of accurate basis for operation and maintenance optimization.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is: to provide a slurry pump flow component life prediction system based on reverse engineering, comprising: The data acquisition layer is used to collect actual wear data and operating condition data of the slurry pump's flow-through components. The model building layer is used to build digital twin models that correspond one-to-one with the flow parts of the slurry pump, and to establish a wear rate model adapted to the operating conditions based on the actual wear data and operating condition data. The decision execution layer is used to predict the remaining lifespan based on the wear rate model and output operation and maintenance optimization instructions in a hierarchical manner. The closed-loop iteration layer is used to feed back the execution results of the operation and maintenance optimization instructions to the data acquisition layer and the model building layer, forming a data closed loop throughout the entire lifecycle; The data acquisition layer includes a non-contact 3D scanning node, which uses a multi-view splicing scanning method to collect point cloud data of the entire area of ​​the slurry pump flow component.

[0006] In one possible implementation, the model building layer includes a point cloud processing unit and a registration and fusion unit; The point cloud processing unit is used to preprocess the point cloud data by denoising, simplifying and filling holes, and to generate a wear three-dimensional solid model. The registration and fusion unit is used to register the wear 3D solid model with the original design theoretical model, generate a wear depth cloud map, quantify and extract wear feature parameters.

[0007] In one possible implementation, the model building layer further includes a model training and validation unit; The model training and verification unit is used to construct the wear rate model by training a neural network, using the actual wear data as the dependent variable and the operating condition data and operating time as independent variables. The model training and verification unit is also used to verify and iteratively optimize the wear rate model using historical maintenance data.

[0008] In one possible implementation, the decision execution layer includes a lifetime prediction unit and a scheme generation unit; The life prediction unit is used to predict the remaining service life of each high-risk area based on the iteratively optimized wear rate model, and to determine the overall remaining service life of the flow component and the recommended replacement time window. The scheme generation unit is used to generate maintenance suggestion schemes, including maintenance timing, specific measures, and implementation costs, based on the remaining service life.

[0009] In one possible implementation, the closed-loop iteration layer includes a data synchronization and feedback unit; The data synchronization and feedback unit is used to synchronously transmit the execution results of the operation and maintenance suggestion plan back to the data acquisition layer; The data synchronization and feedback unit is also used to drive the model building layer to update the wear rate model and the digital twin model.

[0010] The beneficial effects of the reverse engineering-based slurry pump flow component life prediction system provided by this invention are as follows: Compared with the prior art, this prediction system adopts a four-layer collaborative architecture of data acquisition layer, model building layer, decision execution layer and closed-loop iteration layer, combined with non-contact three-dimensional scanning core acquisition method, to form a full-chain closed-loop solution from data source, model building, decision output to iterative optimization.

[0011] The data acquisition layer is equipped with non-contact 3D scanning nodes, employing a multi-view stitching scanning method to collect point cloud data across the entire domain. This avoids secondary damage to the flow components caused by measurement tools and can adapt to the complex curved surface structure of the flow components, achieving seamless data acquisition and completely restoring physical data such as wear morphology, depth, and area. Simultaneously, it couples operational condition data to construct a multi-source dataset. The model construction layer, based on the aforementioned multi-source dataset, builds a digital twin model corresponding one-to-one with the slurry pump flow components, achieving a mapping between physical entities and virtual models. Through deep fusion of actual wear data and real-time operating parameters, the model is iteratively trained, ensuring that the wear rate model fully adapts to the operating characteristics of a single piece of equipment and the actual wear evolution pattern. The execution layer, relying on a wear rate model adapted to operating conditions, can dynamically quantify and predict the remaining lifespan of flow components, accurately locate critical wear nodes, and simultaneously output hierarchical operation and maintenance optimization instructions based on the prediction results. It formulates maintenance, repair, replacement, and operating condition parameter adjustment plans for different wear levels and remaining lifespan ranges, providing a basis for operation and maintenance work. The closed-loop iteration layer further synchronously feeds back the operation and maintenance instruction execution results, subsequent operating data, and secondary wear data to the data acquisition layer and model building layer, forming a full lifecycle data closed loop. As data accumulates, the digital twin model and wear rate model are continuously corrected and iterated, constantly improving the accuracy of lifespan prediction and operation and maintenance optimization, and realizing long-term dynamic management of equipment.

[0012] Overall, this solution addresses the technical shortcomings in slurry pump flow component life assessment by integrating reverse engineering and digital twin technologies. It improves the efficiency and effectiveness of the entire process, from data acquisition and model building to life prediction and operation and maintenance decision-making. This solution effectively solves the practical problems of inaccurate predictions and lack of data for operation and maintenance, while also reducing operation and maintenance costs, minimizing unplanned downtime, and improving equipment operational stability.

[0013] This invention also provides a method for predicting and optimizing the lifespan of slurry pump flow components based on reverse engineering, comprising the following steps: S1: Use non-contact 3D scanning equipment to acquire point cloud data of the slurry pump flow parts, and simultaneously collect on-site operating data of the slurry pump in real time. S2: Preprocess and register the point cloud data to construct a three-dimensional solid model of wear, and establish a digital twin model bound to the slurry pump flow component; S3: Based on the digital twin model and on-site working data, construct and verify the wear rate model to predict the remaining service life of the slurry pump flow components; S4: Output an optimization scheme based on the prediction results, and feed back the new data after the optimization scheme is executed to step S1.

[0014] In one possible implementation, in step S1, the field operating data includes the particle size distribution of the conveying medium, the medium concentration, the pump operating speed, and the inlet and outlet pressures. The data acquisition process is achieved by deploying multiple sensor monitoring nodes at the slurry pump operation site, and the acquired data is processed by outlier removal and time series interpolation.

[0015] In one possible implementation, in step S2, the registration fusion adopts a combination of coarse registration and fine registration; The coarse registration is achieved through an iterative nearest-point algorithm, while the fine registration employs an algorithm based on surface features to control the registration error between the wear model and the original design theoretical model within a preset range.

[0016] In one possible implementation, the prediction step of the remaining useful life in step S3 is as follows: S11. Calculate the wear development trend of each high-risk area based on the wear rate model, and set the failure judgment threshold for the slurry pump flow parts. S12. Combining the current wear status and failure judgment threshold, calculate the remaining life of each region using the life prediction algorithm, take the minimum value as the overall remaining life of the flow component, and determine the recommended replacement time window.

[0017] In one possible implementation, step S4, feeding back the new data after the optimization scheme is executed to step S1, includes: The newly designed, repaired, or replaced slurry pump flow components are again incorporated into the 3D scanning and life prediction process during shutdown maintenance. Wear data is continuously accumulated and the wear rate model and flow component design scheme are iteratively optimized to form a closed-loop management system.

[0018] The beneficial effects of the reverse engineering-based method for optimizing the lifespan of slurry pump flow components provided by this invention are as follows: Compared with existing technologies, this optimization method addresses the pain points in slurry pump flow component lifespan assessment, such as low accuracy of wear data acquisition, wear rate models failing to match real operating conditions and wear evolution patterns, distorted lifespan predictions, and lack of precise basis for operation and maintenance optimization. It achieves high-precision, full-dimensional data acquisition of the wear morphology of flow components through non-contact 3D scanning, effectively compensating for the accuracy deficiencies of traditional measurement methods. By combining a digital twin model with real-time field operating data to construct a wear rate model, it breaks through the limitations of traditional models that are detached from actual operating conditions and statically fitted, accurately restoring the wear evolution pattern and significantly improving the realism and accuracy of lifespan prediction. Simultaneously, relying on reliable prediction results, it outputs adaptive optimization schemes and forms a data closed-loop feedback mechanism, providing precise decision-making basis for equipment operation and maintenance, promoting slurry pump operation and maintenance from passive emergency repairs to proactive prediction and precise control, effectively improving equipment operational stability and reducing operation and maintenance costs. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 A schematic diagram of a slurry pump flow component life prediction system based on reverse engineering provided in an embodiment of the present invention; Figure 2 The flowchart illustrates a method for predicting and optimizing the lifespan of slurry pump flow components based on reverse engineering, as provided in this embodiment of the invention. Detailed Implementation

[0021] To make the technical problems to be solved, the technical solutions, and the beneficial effects of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present invention and are not intended to limit the present invention.

[0022] Unless otherwise explicitly specified, the use of terms such as "first," "second," or "third" is intended to distinguish different objects, not to describe a specific order.

[0023] Unless otherwise expressly defined, the use of directional terms such as “center,” “lateral,” “longitudinal,” “horizontal,” “vertical,” “top,” “bottom,” “inner,” “outer,” “upper,” “lower,” “front,” “rear,” “left,” “right,” “clockwise,” “counterclockwise,” “high,” and “low” to indicate orientation or positional relationships is based on the orientation and positional relationships shown in the accompanying drawings and is only for the convenience of describing the invention and simplifying the description, and is not intended to indicate or imply that the device or element referred to must have a specific orientation or be constructed and operated in a specific orientation, and therefore should not be construed as limiting the specific scope of protection of the invention.

[0024] Please see Figure 1 This invention provides a system and optimization method for predicting the lifespan of slurry pump flow components based on reverse engineering. The system comprises a data acquisition layer, a model building layer, a decision execution layer, and a closed-loop iteration layer.

[0025] The data acquisition layer is used to collect actual wear data and operating condition data of the slurry pump's flow components; the model building layer is used to build digital twin models corresponding one-to-one with the slurry pump's flow components, and establish a wear rate model adapted to the operating conditions based on the actual wear data and operating condition data; the decision execution layer is used to predict the remaining lifespan according to the wear rate model, and output maintenance optimization instructions in stages; the closed-loop iteration layer is used to feed back the execution results of the maintenance optimization instructions to the data acquisition layer and the model building layer, forming a data closed loop throughout the entire life cycle; The data acquisition layer includes a non-contact 3D scanning node, which uses a multi-view splicing scanning method to collect point cloud data of the entire area of ​​the slurry pump flow component.

[0026] This invention provides a slurry pump flow component life prediction system based on reverse engineering. Compared with the prior art, this prediction system adopts a four-layer collaborative architecture of data acquisition layer, model building layer, decision execution layer and closed-loop iteration layer, combined with non-contact three-dimensional scanning core acquisition method, to form a full-chain closed-loop solution from data source, model building, decision output to iterative optimization.

[0027] The data acquisition layer is equipped with non-contact 3D scanning nodes, employing a multi-view stitching scanning method to collect point cloud data across the entire domain. This avoids secondary damage to the flow components caused by measurement tools and can adapt to the complex curved surface structure of the flow components, achieving seamless data acquisition and completely restoring physical data such as wear morphology, depth, and area. Simultaneously, it couples operational condition data to construct a multi-source dataset. The model construction layer, based on the aforementioned multi-source dataset, builds a digital twin model corresponding one-to-one with the slurry pump flow components, achieving a mapping between physical entities and virtual models. Through deep fusion of actual wear data and real-time operating parameters, the model is iteratively trained, ensuring that the wear rate model fully adapts to the operating characteristics of a single piece of equipment and the actual wear evolution pattern. The execution layer, relying on a wear rate model adapted to operating conditions, can dynamically quantify and predict the remaining lifespan of flow components, accurately locate critical wear nodes, and simultaneously output hierarchical operation and maintenance optimization instructions based on the prediction results. It formulates maintenance, repair, replacement, and operating condition parameter adjustment plans for different wear levels and remaining lifespan ranges, providing a basis for operation and maintenance work. The closed-loop iteration layer further synchronously feeds back the operation and maintenance instruction execution results, subsequent operating data, and secondary wear data to the data acquisition layer and model building layer, forming a full lifecycle data closed loop. As data accumulates, the digital twin model and wear rate model are continuously corrected and iterated, constantly improving the accuracy of lifespan prediction and operation and maintenance optimization, and realizing long-term dynamic management of equipment.

[0028] Overall, this solution addresses the technical shortcomings in slurry pump flow component life assessment by integrating reverse engineering and digital twin technologies. It improves the efficiency and effectiveness of the entire process, from data acquisition and model building to life prediction and operation and maintenance decision-making. This solution effectively solves the practical problems of inaccurate predictions and lack of data for operation and maintenance, while also reducing operation and maintenance costs, minimizing unplanned downtime, and improving equipment operational stability.

[0029] Specifically, the data acquisition layer is responsible for collecting actual wear data and operating condition data of the slurry pump flow components. It consists of two parts: a non-contact 3D scanning node and a multi-sensor operating condition monitoring node. The two types of nodes work together to achieve spatiotemporal synchronous acquisition of wear data and operating condition data, providing a high-quality data source for subsequent modeling.

[0030] The non-contact 3D scanning node employs a multi-view stitching scanning method to collect point cloud data across the entire flow area of ​​the slurry pump. A high-precision laser 3D scanner (Model Track Scan-Sharp 49) is used, with a scanning accuracy of 0.025mm and adjustable point cloud density, adaptable to scanning requirements of flow components with different structures such as impeller curved surfaces, volute irregular surfaces, and guard plate flat surfaces. This scanning node is deployed at the slurry pump shutdown and maintenance station using a mobile deployment scheme, requiring no structural modifications to the pump body. Scanning operations are only conducted during shutdown and maintenance windows, without affecting normal production operations.

[0031] The specific implementation process of multi-view stitching scanning is as follows: First, mark positioning points: evenly paste 3-5 reflective positioning points on the surface of the flow-through component, with the spacing between the marking points controlled at 100mm-150mm to ensure accurate stitching of scan data from different perspectives. Second, perform multi-view scanning: set up 6-8 scanning stations around the flow-through component, collecting point cloud data from different angles such as the front, side, top, bottom, and inner cavity. The scanning time for each station is controlled at 3-5 minutes to ensure coverage of all wear areas of the flow-through component, with no blind spots or missed scans. Third, automatically stitch and fuse: the scanner's built-in algorithm performs preliminary stitching of the multi-view point cloud data, removing duplicate point clouds, and generating a full-domain original point cloud dataset for the flow-through component. The amount of point cloud data in a single set is controlled at 5 million-10 million points, balancing data accuracy and processing efficiency.

[0032] For areas prone to wear and with complex structures, such as the twisted surfaces of impeller blades and the volute tongue, a localized high-density scanning mode is employed. This increases the point cloud density in these areas to twice that of conventional areas, accurately capturing microscopic wear defects (such as microcracks, pits, and wall thinning). During the scanning process, the ambient light intensity is controlled between 200 lx and 500 lx to avoid interference from strong light and dust, and the scanning environment temperature is controlled between 10℃ and 30℃, with relative humidity ≤80% to ensure the stability of point cloud data acquisition.

[0033] The multi-sensor operational condition monitoring nodes adopt an embedded deployment scheme, with various sensors deployed on the slurry pump body, inlet and outlet pipelines, and media delivery pipelines to collect operational condition data in real time. The collection frequency is set to 1 time / second, and the data is uploaded to the edge computing gateway in real time. After preliminary preprocessing, it is transmitted to the model construction layer. The collected operational condition data specifically includes core parameters such as the particle size distribution of the conveyed medium, medium concentration, pump operating speed, inlet and outlet pressure, medium temperature, flow rate, vibration amplitude, and motor power.

[0034] Sensor selection and deployment location: ① Particle size sensor: installed in the pump inlet pipeline, using laser scattering method for detection, with a detection range of 0.01mm-30mm; ② Concentration sensor: An ultrasonic concentration meter is used, installed on the straight pipe section at the pump inlet and outlet, away from bends and valves, with a detection range of 5%-50%; ③Speed ​​sensor: magnetically mounted on the pump shaft end, non-contact speed detection, accuracy ±1r / min; ④ Pressure sensor: installed at the pump inlet and outlet flanges, measuring range 0-10MPa, accuracy class 0.1; ⑤ Temperature sensor: Surface mount type, installed on the outer wall of the flow-through component and the medium pipeline, detection range -20℃ to 200℃; ⑥ Vibration sensor: Installed on the pump body bearing housing and the outer wall of the volute, it detects triaxial vibration signals with a frequency range of 0-1000Hz.

[0035] The data acquisition layer incorporates a data preprocessing module to perform preliminary cleaning of the collected raw data, removing outliers and missing values, and using linear interpolation to fill in time series gaps, ensuring data continuity and validity. Simultaneously, a data timestamp matching mechanism is established to bind the wear data acquired from 3D scanning with the corresponding time period's operating condition data, forming a correlated dataset that provides standardized input data for the model building layer.

[0036] The model building layer is the core processing unit of the system. It constructs digital twin models corresponding one-to-one with the flow components of the slurry pump, and simultaneously establishes a wear rate model adapted to the operating conditions based on the collected actual wear data and operating condition data. The model building layer mainly includes a point cloud processing unit, a registration and fusion unit, and a model training and verification unit. Each unit works together to complete the entire process of building a high-precision wear model from the original point cloud.

[0037] The point cloud processing unit is responsible for denoising, simplifying and filling holes in the raw point cloud data transmitted from the data acquisition layer, eliminating noise points and redundant points, repairing the missing areas of the point cloud, and generating a high-precision wear 3D solid model.

[0038] Raw point cloud data is affected by environmental dust, scanning jitter, and surface contaminants, resulting in a large number of isolated noise points and discrete artifacts, directly impacting model accuracy. The point cloud processing unit employs a combined statistical filtering and radius filtering denoising algorithm: First, statistical filtering calculates the average distance between points in the neighborhood of each point cloud, eliminating isolated noise points whose distance exceeds a threshold set to twice the standard deviation of the average distance. Second, radius filtering filters out discrete artifacts with insufficient neighborhood points, with a radius set to 0.5 mm and a minimum neighborhood size of 5. After denoising, the noise point removal rate is ≥98%, preserving the integrity of the valid point cloud data.

[0039] The original global point cloud data is massive, and direct modeling would lead to excessive computational consumption and low processing efficiency. Therefore, it is necessary to simplify the point cloud in non-critical areas while retaining the point cloud density of critical wear areas (blades, tongues, and working surfaces of protective plates). A curvature-based simplification algorithm is used to significantly simplify areas with gentle curvature changes (such as non-working surfaces of flow components and outer walls), with a point cloud retention rate of 30%-50%. For high-risk wear areas with drastic curvature changes, the original point cloud density is maintained without simplification. After simplification, the point cloud data volume is compressed by more than 60%, modeling efficiency is improved by 50%, and the accuracy of wear features is not lost.

[0040] To address point cloud defects and holes (especially in the impeller cavity and volute dead corners) caused by occlusion and angle limitations during scanning, a hole-filling algorithm based on Poisson surface reconstruction is employed. This algorithm automatically fits the missing surface based on the curvature and normal vector characteristics of the point cloud surrounding the defective area, achieving a hole-filling accuracy within 0.03mm. After hole filling, reverse engineering software (GeomagicDesign X) is used to reconstruct the processed point cloud data into a 3D solid model. The model format is the common STEP and IGS format, allowing for seamless integration with subsequent digital twin models. The generated 3D solid model of the wear component fully reproduces the current wear morphology, wall thickness reduction, surface pits, and other features.

[0041] The core function of the registration and fusion unit is to accurately register the 3D physical model of the wear model with the original design theoretical model, generate a wear depth cloud map, quantify and extract wear characteristic parameters, and realize the digital and visual presentation of the wear state. The registration process adopts a two-level registration strategy of coarse registration and fine registration to completely eliminate model position deviation and angle deviation, and ensure that the registration error is controlled within a preset range (≤0.3mm).

[0042] Coarse registration employs the Iterative Closest Point (ICP) algorithm to initially align the worn 3D solid model with the original design theoretical model. First, feature points (such as corners, centers, and surface vertices) are extracted from both models to establish a correspondence between them. Second, the rotation matrix and translation vector between the models are iteratively optimized to gradually reduce model deviations. This iteration is repeated 100-200 times until the average distance between the models converges, completing the coarse registration. After coarse registration, the model position deviation is controlled within 0.5mm, laying the foundation for fine registration.

[0043] The precise registration employs a surface feature-based matching algorithm, focusing on local surface details of the model to further improve registration accuracy. Features such as surface normals, curvature, and contours of the two models are extracted to construct local feature descriptors. Feature matching is then used to accurately align detailed areas of the model. A weighted least squares method is introduced to optimize registration error. The weighting rules are as follows: critical wear areas (impeller blades, volute tongue, and wear plate working surface) are assigned a weight of 0.7, while non-critical areas are assigned a weight of 0.3. The weight calculation formula is ω. i =0.7×C i +0.3×S i (C) i S represents the regional wear risk coefficient. i The system optimizes the registration objective function by weighting the surface feature complexity coefficients to ensure that the registration error in critical areas is ≤0.3mm and the registration error in non-critical areas is ≤0.5mm. After fine registration is completed, the system automatically checks the registration accuracy. If the error exceeds the preset range, the registration process is restarted until the standard is met.

[0044] After registration, the system performs three-dimensional differential calculations on the two models to obtain data such as wall thickness reduction, wear depth, and wear volume at various points on the surface of the flow component. The wear distribution is then visualized using a color cloud map: red areas represent severe wear (70mm ≥ wear depth > 60mm), yellow areas represent moderate wear (20mm ≤ wear depth ≤ 60mm), and green areas represent mild wear (wear depth < 20mm). High-risk wear areas are clearly marked. See Table 1 below.

[0045] Table 1 Simultaneously, the system automatically quantifies and extracts core wear characteristic parameters, including: maximum wear depth, average wear depth, wear area percentage, wear volume, remaining wall thickness, and dimensional deviations of key components, with an extraction accuracy of ≤0.1mm. The extracted wear characteristic parameters are integrated with corresponding working condition data to form a model training sample library, providing data support for subsequent wear rate model construction.

[0046] The model training and validation unit uses actual wear data as the dependent variable and operating condition data and operating time as independent variables. It constructs a wear rate model through neural network training and uses historical maintenance data to validate and iteratively optimize the model to ensure that the model can adapt to the prediction needs of different operating conditions and different wear stages.

[0047] The model training and validation unit uses a Long Short-Term Memory (LSTM) neural network model, which is adapted to the nonlinear fitting characteristics of time-series data and accurately depicts the dynamic relationship between operating parameters and wear rate. The model input layer variables include 12 core operating parameters such as medium particle size, concentration, pump speed, inlet and outlet pressure, running time, temperature, and vibration amplitude; the output layer variables are the wear rate and wear depth increment of each region of the flow component.

[0048] Model training process: ① Sample splitting: The collected “working condition / wear” related dataset is divided into a training set and a test set in a 7:3 ratio. The training set is used for model parameter learning, and the test set is used for model accuracy verification. ② Data standardization: The Min-Max normalization method is used to map the input data to the [0,1] interval to eliminate the difference in units; ③ Network initialization: Set the LSTM network to 3 hidden layers, with 64, 32 and 16 neurons respectively, learning rate to 0.001 and batch size to 32; ④ Iterative training: Use the Adam optimizer for gradient descent training, with 500 iterations, until the loss function converges (loss value < 0.1). ⑤ Model Output: After training, a wear rate model adapted to the current slurry pump flow components is generated, which can predict future wear trends based on real-time operating data.

[0049] A validation dataset was constructed using historical on-site maintenance data, actual wear measurement data, and failure record data to validate the accuracy of the trained wear rate model. Key validation metrics included: Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Coefficient of Determination (R²). The model was required to have a prediction error ≤ ±10% and an R² ≥ 0.95. If the accuracy requirements were not met, the source of error was analyzed: if insufficient sample data was the cause, additional on-site data was collected to expand the sample library; if unreasonable network parameters were the cause, the number of neurons in the hidden layer, the learning rate, and other parameters were adjusted for retraining.

[0050] Meanwhile, a regular model iteration mechanism is established. After each shutdown maintenance scan and acquisition of new wear data, the new data is automatically added to the training set for incremental training of the wear rate model, updating the model parameters to ensure that the model always reflects changes in on-site operating conditions and the aging patterns of flow components, maintaining long-term prediction accuracy. In addition, physical wear mechanisms (coupling mechanisms of erosion wear and corrosion wear) are integrated with the data-driven model to construct a mechanism-data hybrid model, further improving the prediction stability under extreme operating conditions.

[0051] The decision-making and execution layer relies on the iteratively optimized wear rate model to predict remaining lifespan and outputs operation and maintenance optimization instructions in a hierarchical manner. The core includes a lifespan prediction unit and a solution generation unit, realizing the transformation from wear prediction to operation and maintenance decision-making.

[0052] The life prediction unit is based on the iteratively optimized wear rate model. It combines the current wear status and operating trend to predict the remaining service life of each high-risk area, determine the overall remaining service life of the flow components and the recommended replacement time window. The prediction process takes into account both local wear and overall failure, so as to avoid sudden failure caused by local severe wear.

[0053] The system automatically identifies high-risk wear areas such as the volute tongue, impeller blade inlet edge, and guard plate working surface. It retrieves the current wear depth and remaining wall thickness data for these areas and calculates the wear development curves under different operating conditions using a wear rate model. Considering the impact of operating condition fluctuations, three simulation scenarios are set up: normal operating condition, overload operating condition, and low load operating condition. The system predicts the wear increment under each scenario and outputs the trend of wear depth change with operating time, accurately predicting the time point when wear exceeds the standard.

[0054] Based on flow component design standards and industry specifications, failure thresholds are set as follows: remaining wall thickness ≤ 20% of design wall thickness, maximum wear depth ≥ 70mm, and critical dimension deviation ≥ 0.8mm. Failure is deemed complete if any one of these thresholds is met. The remaining life calculation formula is as follows:

[0055] In the formula: RUL is the remaining service life (h); The failure wear threshold (mm); Current wear depth (mm); The wear rate is expressed in mm / h.

[0056] Calculate the remaining lifespan of each high-risk area separately, and take the minimum value as the overall remaining lifespan of the current-carrying components to avoid localized failures leading to overall system failure. Simultaneously, combine production plans and maintenance windows to determine recommended replacement time windows and provide optimal maintenance periods, avoiding peak production periods and minimizing downtime losses. Current-carrying components with a remaining lifespan <300 hours are marked as red alert; those with a remaining lifespan between 300 and 1000 hours are marked as yellow alert; and those with a remaining lifespan >1000 hours are marked as green normal.

[0057] The solution generation unit generates tiered maintenance (O&M) recommendations based on remaining service life, warning level, and wear status. These recommendations include maintenance timing, specific measures, and implementation costs, enabling precise O&M and on-demand maintenance, replacing the traditional scheduled maintenance model. The O&M solutions are divided into three levels, corresponding to different warning levels: Level 1 Operation and Maintenance Plan (Red Alert, Emergency Response) This solution is applicable to flow-through components with a remaining life of <300 hours, severe wear, and a risk of sudden failure. The solution includes: immediately arranging a shutdown for maintenance, prioritizing the replacement of flow-through components; controlling maintenance time to 4-8 hours, and selecting replacement parts of the same model made of high-wear-resistant material; simultaneously checking pump body installation clearances and operating parameters to prevent accelerated wear of new spare parts; implementing cost accounting including spare parts procurement costs, labor costs for maintenance, and downtime losses, and providing cost optimization suggestions.

[0058] Level 2 Maintenance Plan (Yellow Alert, Planned Overhaul) This solution is suitable for flow components with a remaining life of 300-1000 hours and moderate wear. The solution includes: including the component in the next planned maintenance schedule; determining the maintenance window based on production shifts; prioritizing repair processes such as laser cladding and welding, eliminating the need for complete replacement; adjusting operating conditions before maintenance to reduce medium concentration and speed, thus slowing the wear rate; and providing repair process parameters, quality acceptance standards, a cost-benefit analysis comparing repair and replacement costs, and recommending the most cost-effective solution.

[0059] Level 3 Operation and Maintenance Solution (Green Normal, Status Monitoring) This solution is applicable to current-carrying components with a remaining life of >1000 hours and slight wear. The solution includes: maintaining current operating conditions, conducting regular condition monitoring, collecting additional operating data daily, and performing a 3D scan re-inspection every 7 days; optimizing operating parameters to avoid overload and unbalanced operating conditions; providing long-term maintenance recommendations to extend the service life of current-carrying components without requiring immediate maintenance. Refer to Table 2 below.

[0060] Table 2 The core of the closed-loop iteration layer is the data synchronization and feedback unit, which is responsible for feeding back the execution results of operation and maintenance optimization instructions to the data acquisition layer and model building layer, forming a data closed loop throughout the entire life cycle, enabling the system to self-optimize and continuously improve accuracy.

[0061] After the maintenance plan is executed, maintenance personnel enter the execution results through the system terminal, including: whether the current-carrying components were replaced, the implementation status of the repair process, the effect of the operating condition adjustment, the maintenance time, the actual cost, and the wear status after repair. The data synchronization and feedback unit binds the above execution results with the corresponding life prediction records and wear data to form a closed-loop dataset, which is then transmitted back to the data acquisition layer in real time to supplement the system's historical database and expand the data sample size.

[0062] Simultaneously, for repaired or replaced flow components, non-contact 3D scanning nodes and operating condition monitoring nodes are immediately activated to collect a new round of wear and operating condition data, achieving seamless integration of old and new data and ensuring the integrity of the data chain. For prediction deviations occurring during execution (such as discrepancies between actual remaining lifespan and predicted values), the reasons for the deviations are noted to provide a basis for model optimization.

[0063] All operation and maintenance solutions are output in the form of visual reports and electronic instructions, which can be viewed simultaneously on mobile devices and industrial control terminals. They also include wear cloud maps, life prediction curves, and cost analysis tables to facilitate quick decision-making and implementation by operation and maintenance personnel.

[0064] The closed-loop iteration layer automatically drives the model building layer to update the model based on the returned execution results and newly collected data. ① Digital twin model update: Replace the original model with the 3D solid model of the repaired or replaced flow component, and update the operating parameters and wear status simultaneously to ensure real-time virtual-real mapping between the digital twin model and the physical flow component; ② Wear rate model iteration: The newly added "operating condition-wear-execution result" data is included in the training set, incremental training is started, the model parameters are fine-tuned, the prediction bias is corrected, and the accuracy decline caused by model aging and changes in operating conditions is eliminated.

[0065] In addition, a closed-loop management and control mechanism is established to regularly generate system iteration reports, analyze and evaluate indicators such as changes in prediction accuracy, execution rate of operation and maintenance plans, and cost savings rate, evaluate the system's operating effect, optimize the data collection plan, modeling parameters, and decision rules for weak links, and form a complete closed-loop management system to achieve continuous improvement in the accuracy of flow component life prediction and operation and maintenance efficiency.

[0066] Please see Figure 2 The present invention also discloses a method for predicting and optimizing the lifespan of slurry pump flow components based on reverse engineering, comprising the following steps: S1: Use non-contact 3D scanning equipment to acquire point cloud data of the slurry pump flow parts, and simultaneously collect on-site operating data of the slurry pump in real time to achieve synchronous acquisition, binding and storage of wear data and operating data. Detailed implementation method: ① Point cloud data acquisition: After the slurry pump is shut down, the flow-through components are disassembled and scanned using a laser 3D scanner in a multi-view stitching manner to obtain full-area point cloud data. After the scan is completed, the data is uploaded to the system. ② Operating data acquisition: Through the deployment of multi-sensor monitoring nodes, the particle size distribution, medium concentration, pump operating speed and inlet and outlet pressure of the conveyed medium are collected in real time, while auxiliary parameters such as flow rate, temperature and vibration are also collected. ③ Data preprocessing: Outlier removal and time series interpolation are performed on the collected data to fill in missing data, remove abnormal fluctuation data, and generate a standardized dataset. In this embodiment, a total of 120 valid samples were collected, covering data from different operating conditions and different wear stages.

[0068] S2: Preprocess and register the point cloud data to construct a three-dimensional solid model of wear, and establish a digital twin model bound to the slurry pump flow component. Detailed implementation method: ① Point cloud preprocessing: Denoising, simplification, and hole filling operations are carried out in sequence to eliminate noise and redundant data, repair missing areas, and reconstruct the worn 3D solid model; ② Registration and fusion: A combination of coarse registration (ICP algorithm) and fine registration (surface feature algorithm) is used to accurately align the wear model with the original design theoretical model, with the registration error controlled within 0.05mm; ③ Digital twin modeling: Import the wear 3D solid model, working condition data, material parameters, and operating parameters into the digital twin platform to build a 1:1 high-precision digital twin model, realize real-time visualization mapping of the operating status and wear morphology of the flow component, and support virtual simulation and wear source analysis.

[0070] S3: Based on the digital twin model and on-site operating data, construct and verify the wear rate model to predict the remaining service life of the slurry pump's flow components. The prediction steps for the remaining service life are as follows: S11. Calculate the wear development trend of each high-risk area based on the wear rate model, and set the failure judgment threshold for the slurry pump flow components. Input the standardized operating condition data and wear characteristic parameters into the LSTM neural network to train and generate the wear rate model; set the failure judgment threshold for the slurry pump flow components, combining industry standards and field experience.

[0071] This embodiment sets the following conditions for failure: remaining wall thickness < 25mm (design wall thickness 100mm), maximum wear depth > 70mm, and flow component efficiency decrease ≥ 15%. Failure is determined when any of these conditions is met. The wear development trend of each high-risk area is calculated using a wear rate model, and predicted wear depth values ​​for the next 300h, 1000h, and 3000h are output.

[0072] S12. Combining the current wear status and failure judgment threshold, calculate the remaining life of each region using the life prediction algorithm, take the minimum value as the overall remaining life of the flow component, and determine the recommended replacement time window.

[0073] In this embodiment, the wear is most severe in the impeller blade inlet edge area, with a current wear depth of 60mm, a wear rate of 0.08mm / h, a failure threshold of 75mm, and a calculated remaining lifespan of 187.5h. This is considered a red alert, and the recommended replacement time window is within 7 days. The specific downtime will be determined in conjunction with the production plan.

[0074] S4: Output an optimization plan based on the prediction results, and feed back the new data after the optimization plan is implemented to step S1 to form a closed-loop management system. Detailed implementation method: ① Solution Output: Based on the remaining life prediction results, a Level 1 emergency operation and maintenance plan is generated, specifying measures such as immediately shutting down the machine to replace the impeller, adjusting the medium concentration to below 15%, and re-inspecting the pump body clearance; ② Execution feedback: After the operation and maintenance personnel execute the plan, they enter the replacement spare parts model, maintenance time, cost, and post-repair operating conditions. At the same time, they conduct a 3D scan of the new impeller to collect new point cloud data and operating condition data. ③ Data feedback: New data and execution results are fed back to the data acquisition layer to drive the update of the wear rate model and digital twin model, continuously accumulate wear data, and iteratively optimize the model accuracy.

[0076] For newly designed, repaired, or replaced slurry pump flow components, they are again incorporated into the 3D scanning and life prediction process during shutdown maintenance, forming a closed loop of prediction, maintenance, retesting, and optimization. This continuously improves the wear database, enhances the accuracy of life prediction, and provides data support for the optimization of flow component structures and material improvements, thereby achieving lean management throughout the entire life cycle.

[0077] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A slurry pump flow component life prediction system based on reverse engineering, characterized in that, include: The data acquisition layer is used to collect actual wear data and operating condition data of the slurry pump's flow-through components. The model building layer is used to build digital twin models that correspond one-to-one with the flow parts of the slurry pump, and to establish a wear rate model adapted to the operating conditions based on the actual wear data and operating condition data. The decision execution layer is used to predict the remaining lifespan based on the wear rate model and output operation and maintenance optimization instructions in a hierarchical manner. The closed-loop iteration layer is used to feed back the execution results of the operation and maintenance optimization instructions to the data acquisition layer and the model building layer, forming a data closed loop throughout the entire lifecycle; The data acquisition layer includes a non-contact 3D scanning node, which uses a multi-view splicing scanning method to collect point cloud data of the entire area of ​​the slurry pump flow component.

2. The slurry pump flow component life prediction system based on reverse engineering as described in claim 1, characterized in that, The model building layer includes a point cloud processing unit and a registration and fusion unit; The point cloud processing unit is used to preprocess the point cloud data by denoising, simplifying and filling holes, and to generate a wear three-dimensional solid model. The registration and fusion unit is used to register the wear 3D solid model with the original design theoretical model, generate a wear depth cloud map, quantify and extract wear feature parameters.

3. The slurry pump flow component life prediction system based on reverse engineering as described in claim 2, characterized in that, The model building layer also includes a model training and validation unit; The model training and verification unit is used to construct the wear rate model by training a neural network, using the actual wear data as the dependent variable and the operating condition data and operating time as independent variables. The model training and verification unit is also used to verify and iteratively optimize the wear rate model using historical maintenance data.

4. The slurry pump flow component life prediction system based on reverse engineering as described in claim 3, characterized in that, The decision execution layer includes a lifetime prediction unit and a scheme generation unit; The life prediction unit is used to predict the remaining service life of each high-risk area based on the iteratively optimized wear rate model, and to determine the overall remaining service life of the flow component and the recommended replacement time window. The scheme generation unit is used to generate maintenance suggestion schemes, including maintenance timing, specific measures, and implementation costs, based on the remaining service life.

5. The slurry pump flow component life prediction system based on reverse engineering as described in claim 4, characterized in that, The closed-loop iteration layer includes a data synchronization and feedback unit; The data synchronization and feedback unit is used to synchronously transmit the execution results of the operation and maintenance suggestion plan back to the data acquisition layer; The data synchronization and feedback unit is also used to drive the model building layer to update the wear rate model and the digital twin model.

6. A method for predicting and optimizing the lifespan of slurry pump flow components based on reverse engineering, characterized in that, Includes the following steps: S1: Use non-contact 3D scanning equipment to acquire point cloud data of the slurry pump flow parts, and simultaneously collect on-site operating data of the slurry pump in real time. S2: Preprocess and register the point cloud data to construct a three-dimensional solid model of wear, and establish a digital twin model bound to the slurry pump flow component; S3: Based on the digital twin model and on-site working data, construct and verify the wear rate model to predict the remaining service life of the slurry pump flow components; S4: Output an optimization scheme based on the prediction results, and feed back the new data after the optimization scheme is executed to step S1.

7. The method for predicting and optimizing the lifespan of slurry pump flow components based on reverse engineering as described in claim 6, characterized in that, In step S1, the field operating data includes the particle size distribution of the conveying medium, the medium concentration, the pump operating speed, and the inlet and outlet pressures. The data acquisition process is achieved by deploying multiple sensor monitoring nodes at the slurry pump operation site, and the acquired data is processed by outlier removal and time series interpolation.

8. The method for predicting and optimizing the lifespan of slurry pump flow components based on reverse engineering as described in claim 6, characterized in that, In step S2, the registration fusion adopts a combination of coarse registration and fine registration; The coarse registration is achieved through an iterative nearest-point algorithm, while the fine registration employs an algorithm based on surface features to control the registration error between the wear model and the original design theoretical model within a preset range.

9. The method for predicting and optimizing the lifespan of slurry pump flow components based on reverse engineering as described in claim 6, characterized in that, In step S3, the prediction step for the remaining useful life is as follows: S11. Calculate the wear development trend of each high-risk area based on the wear rate model, and set the failure judgment threshold for the slurry pump flow parts. S12. Combining the current wear status and failure judgment threshold, calculate the remaining life of each region using the life prediction algorithm, take the minimum value as the overall remaining life of the flow component, and determine the recommended replacement time window.

10. The method for predicting and optimizing the lifespan of slurry pump flow components based on reverse engineering as described in claim 6, characterized in that, In step S4, feeding back the new data after the optimization scheme is executed to step S1 includes: The newly designed, repaired, or replaced slurry pump flow components are again incorporated into the 3D scanning and life prediction process during shutdown maintenance. Wear data is continuously accumulated and the wear rate model and flow component design scheme are iteratively optimized to form a closed-loop management system.

Images