A method for predicting stress distribution in crusher liners based on multi-field coupling

Abstract

This invention discloses a method for predicting stress distribution in crusher liners based on multi-field coupling, comprising the following steps: Step 1: Collecting dynamic information and constructing a disturbance behavior map; Step 2: Constructing a thermal-grinding cross-analysis model and extracting thermal-grinding cross-factors; Step 3: Generating a geometric latent field using a deformation trend generation network and fusing it with a wear evolution trend vector to update the liner's geometric parameters; Step 4: Obtaining a set of hot spot induction points; Step 5: Constructing a multi-field coupling model and using the set of hot spot induction points as stress propagation constraints to obtain the initial predicted stress distribution; Step 6: Calculating the stress residual and introducing a geometric backtracking feedback mechanism for backtracking correction; Step 7: Updating the multi-field coupling model to obtain the final predicted stress distribution. This invention achieves stress distribution prediction based on liner geometric evolution perception by constructing a disturbance behavior map, thermal-grinding cross-analysis, and multi-field coupling modeling.

Description

Technical Field

[0001] This invention relates to the field of intelligent structural mechanics analysis technology, and in particular to a method for predicting stress distribution in crusher liners based on multi-field coupling. Background Technology

[0002] With the increasing demand for intelligent mining machinery and efficient operation of large crushing equipment, the monitoring of stress state and life prediction of crusher liners under complex working conditions has attracted widespread attention. As a key vulnerable component in the crusher that directly bears the impact, friction, and heat of materials, the stress distribution of the liner directly affects the equipment's operational safety, maintenance cycle, and overall production efficiency. However, in practical engineering applications, obtaining and predicting the stress distribution of crusher liners still faces many technical challenges.

[0003] In existing technologies, stress analysis of liner plates mostly relies on finite element simulation or simplified calculation methods based on empirical formulas. These methods typically assume that the liner plate geometry and operating conditions are constant, making it difficult to reflect the dynamic influence of random material movement, time-varying impact loads, and the coupling effect of thermal wear on the stress distribution of the liner plate during crushing. Some studies have attempted to introduce sensor data for stress inversion, but these are mostly concentrated on single measuring points or local areas, lacking systematic modeling of the disturbance propagation relationship across the entire liner plate area. This results in poor spatial consistency of stress prediction results and a lag in response to changes in operating conditions. Furthermore, during long-term service, the geometric parameters of the liner plate continuously evolve with wear, and these geometric changes, in turn, affect local stress concentration and heat accumulation characteristics. Existing methods often treat the geometric structure as a static input, ignoring the feedback relationship between geometric evolution and the stress field. At the same time, the phenomenon of alternating dominance of thermal and wear effects is common during crushing, but existing technologies struggle to accurately characterize the alternating thermal-wear features and their impact on liner plate deformation and stress distribution in both time and space dimensions.

[0004] Therefore, how to provide a method for predicting the stress distribution of crusher liners based on multi-field coupling is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] One objective of this invention is to propose a method for predicting the stress distribution of crusher liners based on multi-field coupling. This invention integrates multi-source dynamic information generated during crusher operation with the geometric evolution characteristics of the liner by constructing a disturbance behavior map, thermal grinding alternation analysis, and multi-field coupling modeling. Under a unified spatiotemporal framework, it characterizes the comprehensive influence of disturbance propagation, thermal grinding alternation, and geometric changes on the stress state of the liner, thereby achieving a refined prediction of the stress distribution in the entire crusher liner area and improving the accuracy, stability, and adaptability to complex working conditions.

[0006] A method for predicting stress distribution in crusher liners based on multi-field coupling according to an embodiment of the present invention includes the following steps:

[0007] Step 1: Collect dynamic information on the interaction between material and liner during crusher operation, and construct a disturbance behavior map based on the dynamic information;

[0008] Step 2: Using the disturbance behavior map as the driving input, collect information on temperature rise, impact frequency, and friction work during the operation of the liner, construct a thermal-grind cross-analysis model, and extract the thermal-grind cross-factor;

[0009] Step 3: Based on the aforementioned thermal wear crossover factor, a deformation trend generation network is used to generate a geometric latent field describing the inherent deformation trend of the liner under the current operating state. The geometric latent field is then fused with the wear evolution trend vector to update the geometric structure parameters of the liner.

[0010] Step 4: Based on the updated liner geometry parameters, and combined with the disturbance behavior map and thermal wear crossover factor, obtain the set of hot spot induction points on the liner;

[0011] Step 5: Construct a multi-field coupling model and use the set of hot spot induction points as stress propagation constraints. In the multi-field coupling model, realize the layer-by-layer propagation of the perturbation field and evolution field to the stress field to obtain the initial predicted stress distribution.

[0012] Step 6: Calculate the stress residual between the initial predicted stress distribution and the actual monitored stress data, and introduce a geometric backtracking feedback mechanism to backtrack and correct the updated liner geometric parameters.

[0013] Step 7: Update the multi-field coupling model based on the back-corrected geometric parameters of the liner to obtain the final predicted stress distribution.

[0014] Optionally, the dynamic information specifically includes multi-point acceleration signals, impact signals, and material motion image data.

[0015] Optionally, the construction of the disturbance behavior map based on dynamic information specifically includes:

[0016] Based on the multi-point acceleration and impact signals in the dynamic information, the average acceleration, impact frequency and impact intensity of each measuring point per unit time are extracted.

[0017] The motion image data of the material is processed frame by frame to identify the measurement point area where the material contacts the liner and obtain the material's motion trajectory.

[0018] Using the measurement points of the liner as the nodes of the graph, the average acceleration, impact frequency and impact intensity of each measurement point per unit time are used as the node attributes.

[0019] The correlation frequency between nodes is determined by calculating the number of times the material movement trajectory passes through two measurement point areas in a continuous frame. The correlation frequency is then normalized to establish edge connections and obtain the corresponding edge weights.

[0020] A perturbation behavior graph is constructed based on the node attributes, edge connections, and corresponding edge weights.

[0021] Optionally, step two specifically includes:

[0022] Taking the liner measuring point area corresponding to each node in the disturbance behavior map as the analysis object, the temperature rise change information, impact frequency information and friction work information of the corresponding measuring point area are collected in the same time window, and the thermal grinding characteristic time series of the corresponding measuring point area is formed in time order.

[0023] Based on the edge connection relationships in the disturbance behavior map, adjacent measurement point regions with edge connections are selected as combined units;

[0024] Within the combined unit, the temperature rise change information of adjacent measuring point areas is multiplied by the corresponding edge weights in the disturbance behavior map, and the results of the multiplication are accumulated within the combined unit to obtain the combined temperature rise characteristic value of the corresponding time window.

[0025] Within the combined unit, the impact frequency information of adjacent measurement point areas is multiplied by the corresponding edge weights in the disturbance behavior spectrum, and the results of the multiplication are accumulated within the combined unit to obtain the combined impact frequency characteristic value of the corresponding time window.

[0026] Within the combined unit, the friction work information of adjacent measuring point areas is multiplied by the corresponding edge weights in the disturbance behavior spectrum, and the results of the multiplication are accumulated within the combined unit to obtain the combined characteristic value of friction work for the corresponding time window.

[0027] Using the time series formed by the combined characteristic values ​​of temperature rise, impact frequency, and friction work within a continuous time window as input, a thermal grinding cross-analysis model constrained by the edge connection relationship of the perturbation behavior graph is constructed.

[0028] In the thermal-grinding cross-analysis model, the variation range of the temperature rise combination characteristic value and the variation range of the impact frequency combination characteristic value and the friction work combination characteristic value within a continuous time window are compared window by window. When the variation range of the temperature rise combination characteristic value is greater than the variation range of the impact frequency combination characteristic value and greater than the variation range of the friction work combination characteristic value, the time window is determined to be a thermal-dominant window; otherwise, the time window is determined to be a grinding-dominant window.

[0029] The number of state transitions between the heat-dominated window and the grinding-dominated window is counted sequentially over time and used as the heat-grind interleaving factor.

[0030] Optionally, based on the thermal grinding interleaving factor, the step of generating a geometric latent field describing the inherent deformation tendency of the liner under the current operating state using a deformation tendency generation network specifically involves:

[0031] Using the measurement point area of ​​the liner plate as a network processing unit, the deformation tendency generation network includes a geometric state input layer, a thermal grinding interleaving factor modulation layer, a deformation tendency calculation layer, and a geometric latent output layer.

[0032] In the geometric state input layer, the current liner geometric parameters of each measuring point area are received and organized so that the liner geometric parameters are aligned in the time window and measuring point area dimensions. The liner geometric parameters include the thickness parameters, curvature parameters and surface roughness parameters of the measuring point area.

[0033] In the thermally ground interlacing factor modulation layer, the thermally ground interlacing factor of the corresponding measurement point area is used as the modulation factor, and multiplied by the thickness parameter, curvature parameter and surface roughness parameter respectively to obtain the modulation geometric feature vector of the corresponding measurement point area.

[0034] The modulation geometric feature vector is input to the deformation tendency calculation layer, and the modulation geometric feature vector is subjected to vector difference operation within adjacent time windows to obtain the geometric change vector.

[0035] The sign of each component of the geometric change vector is determined to obtain the deformation change direction component vector of the corresponding measurement point area;

[0036] The absolute values ​​of each component of the geometric change vector are taken and normalized to obtain the deformation change amplitude component vector of the corresponding measurement point area.

[0037] The deformation change direction component vector and the deformation change amplitude component vector are concatenated along the feature dimension to form the deformation trend feature vector of the corresponding measurement point area.

[0038] The deformation tendency feature vector is input to the geometric latent variable output layer. Each component of the deformation tendency feature vector is multiplied by the corresponding set geometric mapping coefficient, and the results of the multiplication are summed to obtain the geometric latent variable corresponding to the measurement point area in the current operating state. The geometric latent variables of all measurement point areas together constitute the geometric latent variable field.

[0039] Optionally, the step of fusing the geometric latent field with the wear evolution trend vector to update the liner's geometric parameters specifically involves:

[0040] Taking each measuring point area of ​​the liner as the processing object, historical wear monitoring data of the corresponding measuring point area is collected within multiple consecutive historical time windows. The historical wear monitoring data includes liner thickness data, curvature data and surface roughness data.

[0041] Within adjacent historical time windows, the thickness data, curvature data, and surface roughness data are respectively subjected to difference calculation to obtain the thickness wear change, curvature wear change, and surface roughness wear change of the corresponding measuring point area under each historical time window;

[0042] The thickness wear variation obtained within multiple consecutive historical time windows is arranged into a thickness wear variation sequence in chronological order; the curvature variation obtained within multiple consecutive historical time windows is arranged into a curvature variation sequence in chronological order; and the roughness wear variation obtained within multiple consecutive historical time windows is arranged into a roughness wear variation sequence in chronological order.

[0043] The thickness wear change sequence, curvature wear change sequence, and roughness wear change sequence are respectively subjected to least squares linear fitting operation to obtain the thickness wear change slope, curvature wear change slope, and roughness wear change slope of the corresponding measuring point area, and the thickness wear change slope, curvature wear change slope, and roughness wear change slope are combined to form a wear evolution trend vector;

[0044] The geometric latent variables of the corresponding measurement point area are multiplied by each component of the wear evolution trend vector to obtain the fused modulation vector;

[0045] Each component in the fused modulation vector is used as a correction increment for the thickness parameter, curvature parameter, and surface roughness parameter of the corresponding measurement point area. These components are then added one by one to the corresponding current thickness parameter, curvature parameter, and surface roughness parameter to obtain the updated geometric structure parameters.

[0046] Optionally, step four specifically involves:

[0047] Using the measurement point regions corresponding to the updated liner geometric parameters as the analysis objects, in the disturbance behavior graph, the graph nodes corresponding to each measurement point region are determined, and the edge connection relationships and corresponding edge weights connected to the nodes are obtained.

[0048] The thermal grinding interleaving factor of the corresponding measuring point area is multiplied with the updated liner geometric structure parameters one by one to obtain the updated modulation geometric feature vector of the corresponding measuring point area.

[0049] Based on the edge connection relationship of the perturbation behavior map, the vector magnitude of the updated modulation geometric feature vector of all adjacent measurement point regions that are connected to the measurement point region is calculated respectively. The obtained vector magnitude is multiplied by the corresponding edge weight, and the multiplication results are accumulated to obtain the perturbation weighted feature value of the target measurement point region.

[0050] The perturbation weighted feature value is compared with the set hot spot determination threshold, and combined with the dominant state of the thermal wear interleaving factor of the corresponding measurement point area in the current time window, when the perturbation weighted feature value is greater than the set hot spot determination threshold and the corresponding current time window is a thermal dominant window, the corresponding measurement point area is determined as a hot spot induction point.

[0051] By collecting all the measurement points that have been identified as hot spot induction points, a set of hot spot induction points on the liner is obtained.

[0052] Optionally, step five specifically includes:

[0053] A discrete model of the liner is constructed using the measurement point regions corresponding to the updated geometric parameters of the liner as the basic units. The surface of the liner is discretized in plane according to the measurement point regions, and the volume elements corresponding to each measurement point region are divided into layers along the thickness direction of the liner to form a set of volume elements consisting of surface volume elements, middle volume elements and bottom volume elements.

[0054] A multi-field coupling model is constructed, which includes a perturbation field sub-model, an evolution field sub-model, and a stress field sub-model. The perturbation field sub-model, the evolution field sub-model, and the stress field sub-model are established on the same set of volume elements and share the spatial adjacency relationship between volume elements.

[0055] In the disturbance field sub-model, the node attributes corresponding to each measurement point region in the disturbance behavior map are used as the disturbance input vector of the corresponding surface volume unit, and the edge weights between measurement point regions in the disturbance behavior map are used as the disturbance transmission coefficients between adjacent surface volume units to describe the lateral propagation relationship of disturbance between volume units.

[0056] In the evolution field sub-model, for the surface volume unit corresponding to each measurement point region, the updated thickness parameters, curvature parameters and surface roughness parameters of the corresponding measurement point region are read respectively, and a weighted summation operation is performed to obtain the evolution modulation coefficient of the surface volume unit.

[0057] The set of hot spot induction points is used as a stress propagation constraint. In the multi-field coupling model, propagation constraint markers are set for the volume elements corresponding to the measurement point regions belonging to the set of hot spot induction points. The perturbation transmission coefficient and evolution modulation coefficient corresponding to the volume elements with propagation constraint markers are amplified and updated, while the perturbation transmission coefficient and evolution modulation coefficient corresponding to the volume elements without propagation constraint markers are attenuated and updated.

[0058] In the stress field sub-model, layer-by-layer propagation calculations are performed in the order of surface volume elements to middle volume elements and then to bottom volume elements. The perturbation input vector of the middle volume element is obtained by multiplying the perturbation input vector of the corresponding surface volume element with the updated perturbation transfer coefficient component by component.

[0059] The perturbation input vector of the bottom volume element is obtained by multiplying the perturbation input vector of the corresponding middle volume element with the updated perturbation transfer coefficient component by component.

[0060] In each layer of volume element, the perturbation input vector of the corresponding layer of volume element is multiplied component by component with the updated evolution modulation coefficient, and the multiplication results are summed to obtain the equivalent stress value of the corresponding layer of volume element.

[0061] After completing the layer-by-layer propagation calculations for all body elements, the equivalent stress values ​​of the surface body elements, middle body elements, and bottom body elements corresponding to the same body element area are summed and averaged according to the spatial positional relationship of the measuring point area on the liner surface. This average value is then mapped back to the corresponding position on the liner surface to form the initial predicted stress distribution covering the entire area of ​​the liner.

[0062] Optionally, the geometric backtracking feedback mechanism specifically includes:

[0063] The absolute value of the stress residual is taken to obtain the stress deviation amplitude, and the sign of the stress residual is determined.

[0064] The stress deviation amplitude is multiplied by the updated thickness parameter, curvature parameter and surface roughness parameter of the corresponding measuring point area, and the product is divided by the absolute value of the initial predicted stress value of the corresponding measuring point area in the initial predicted stress distribution to obtain the corresponding thickness correction amount, curvature correction amount and roughness correction amount.

[0065] When the stress residual is positive, the thickness correction, curvature correction, and roughness correction are subtracted from the updated thickness parameters, curvature parameters, and surface roughness parameters of the corresponding measurement point area, respectively.

[0066] When the stress residual is negative, the thickness correction amount, curvature correction amount, and roughness correction amount are added to the updated thickness parameter, curvature parameter, and surface roughness parameter of the corresponding measuring point area, respectively.

[0067] After correcting the geometric parameters of all measuring point areas, the corrected geometric structural parameters of the liner are obtained.

[0068] Optionally, step seven specifically includes:

[0069] Based on the back-corrected liner geometry parameters, the evolution modulation coefficients in the multi-field coupling model are recalculated to obtain an updated multi-field coupling model.

[0070] In the updated multi-field coupling model, following the layer-by-layer propagation calculation process from the disturbance field and evolution field to the stress field, the corrected equivalent stress values ​​of the surface, middle and bottom layers of the corresponding volume elements in each measuring point area are obtained. The three corrected equivalent stress values ​​are summed again and averaged, and then mapped back to the corresponding position on the liner surface to obtain the final predicted stress distribution of the liner.

[0071] The beneficial effects of this invention are:

[0072] This invention introduces a disturbance behavior map to structurally model the dynamic interaction between materials and liners during crusher operation. It maps multi-point acceleration signals, impact signals, and material motion image data into a graph structure with spatial correlation, effectively depicting the propagation path and intensity differences of disturbances in different measurement areas of the liner. Based on this, a thermal-grinding cross-analysis model is constructed using the disturbance behavior map as the driving input. By combining and modeling temperature rise change information, impact frequency information, and friction work information within a unified time window, the thermal-grinding cross-factor is extracted, realizing a quantitative characterization of the alternating dominance of thermal and wear effects. This avoids the operational condition identification bias caused by relying on only a single thermal or wear index in existing methods.

[0073] Furthermore, this invention constructs a deformation trend generation network based on the thermal-wear interleaving factor, modulates and differentially analyzes the liner geometric parameters in the temporal and spatial dimensions to generate a geometric latent field characterizing the intrinsic deformation trend of the liner. This geometric latent field is then fused with a wear evolution trend vector obtained by fitting historical wear monitoring data, achieving a collaborative geometric structure update mechanism driven by current operating condition response and long-term wear patterns. This ensures the evolution of the liner geometric parameters possesses temporal consistency and physical interpretability. Building upon this, the invention obtains a set of hot spot induction points based on the updated liner geometric parameters, perturbation behavior maps, and the thermal-wear interleaving factor, and integrates these points into a multi-field coupling model. The set of spot-induced points serves as a stress propagation constraint, differentially modulating the layer-by-layer propagation process of the disturbance field and evolution field to the stress field, thereby effectively enhancing the stress transfer characteristics of high-risk areas. Simultaneously, a geometric backtracking feedback mechanism is introduced to backtrack and correct the stress residual between the initial predicted stress distribution and the actual monitored stress data, forming a closed-loop iterative relationship between the liner geometric parameters and the stress prediction results. Finally, a stable and reliable final predicted stress distribution is obtained in the updated multi-field coupling model. This invention can achieve high-precision prediction of the stress distribution of the entire crusher liner area under complex working conditions, significantly improving the response capability of the stress prediction results to changes in working conditions and geometric evolution, as well as the reliability of engineering applications. Attached Figure Description

[0074] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0075] Figure 1 This is an overall flowchart of a method for predicting stress distribution in crusher liners based on multi-field coupling proposed in this invention.

[0076] Figure 2 This is a flowchart illustrating the generation of a geometric latent field based on a deformation tendency generation network in a method for predicting stress distribution in crusher liners based on multi-field coupling, as proposed in this invention.

[0077] Figure 3 This is a flowchart illustrating the initial predicted stress distribution generation process for a multi-field coupling-based stress distribution prediction method for crusher liners proposed in this invention. Detailed Implementation

[0078] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0079] refer to Figures 1-3 A method for predicting stress distribution in crusher liners based on multi-field coupling includes the following steps:

[0080] Step 1: Collect dynamic information on the interaction between material and liner during crusher operation, and construct a disturbance behavior map based on the dynamic information;

[0081] Step 2: Using the disturbance behavior map as the driving input, collect information on temperature rise, impact frequency, and friction work during the operation of the liner, construct a thermal-grind cross-analysis model, and extract the thermal-grind cross-factor;

[0082] Step 3: Based on the aforementioned thermal wear crossover factor, a deformation trend generation network is used to generate a geometric latent field describing the inherent deformation trend of the liner under the current operating state. The geometric latent field is then fused with the wear evolution trend vector to update the geometric structure parameters of the liner.

[0083] Step 4: Based on the updated liner geometry parameters, and combined with the disturbance behavior map and thermal wear crossover factor, obtain the set of hot spot induction points on the liner;

[0084] Step 5: Construct a multi-field coupling model and use the set of hot spot induction points as stress propagation constraints. In the multi-field coupling model, realize the layer-by-layer propagation of the perturbation field and evolution field to the stress field to obtain the initial predicted stress distribution.

[0085] Step 6: Calculate the stress residual between the initial predicted stress distribution and the actual monitored stress data, and introduce a geometric backtracking feedback mechanism to backtrack and correct the updated liner geometric parameters.

[0086] Step 7: Update the multi-field coupling model based on the back-corrected geometric parameters of the liner to obtain the final predicted stress distribution.

[0087] In this embodiment, the dynamic information specifically includes multi-point acceleration signals, impact signals, and material motion image data;

[0088] In this invention, the dynamic information refers to data reflecting the interaction behavior between the material and the liner during the operation of the crusher. Multi-point acceleration signals are acquired by deploying triaxial MEMS accelerometers at multiple key structural locations inside the crusher cavity, enabling real-time acquisition of the instantaneous vibration intensity and direction changes experienced by the liner in different areas. Impact signals are acquired by piezoelectric impact sensors to monitor the instantaneous impact load changes generated by the material impacting the liner. Material motion image data is acquired by a high-speed industrial camera with a set viewing angle outside the transparent window of the crushing cavity. This dynamic information constitutes the original data basis for the disturbance behavior map, possessing high temporal resolution and multimodal complementary characteristics, and can comprehensively describe the dynamic coupling behavior between the material and the liner.

[0089] In this embodiment, the construction of the disturbance behavior map based on dynamic information specifically includes:

[0090] Based on the multi-point acceleration and impact signals in the dynamic information, the average acceleration, impact frequency and impact intensity of each measuring point per unit time are extracted.

[0091] The motion image data of the material is processed frame by frame to identify the measurement point area where the material contacts the liner and obtain the material's motion trajectory.

[0092] Using the measurement points of the liner as the nodes of the graph, the average acceleration, impact frequency and impact intensity of each measurement point per unit time are used as the node attributes.

[0093] The correlation frequency between nodes is determined by calculating the number of times the material movement trajectory passes through two measurement point areas in a continuous frame. The correlation frequency is then normalized to establish edge connections and obtain the corresponding edge weights.

[0094] In constructing the disturbance behavior map, to reflect the correlation between disturbance behaviors in different regions of the liner, it is necessary to establish edge weights between nodes in the graph structure. The material motion image data is extracted frame by frame, and the motion trajectory sequence of the material in the crushing chamber is generated based on the pixel position change of the material centroid in the continuous image frames. The trajectory is spatially matched with multiple preset measurement point areas of the liner to determine whether the material centroid falls into the spatial range of a certain measurement point in each frame. If the trajectory passes through two different measurement point areas in two adjacent frames, it is recorded as a valid association of the measurement point pair. All consecutive frames are traversed and counted to obtain the total number of times each measurement point pair is traversed by the material trajectory in the whole time period, which is the association frequency of the node pair. Edge connections are established between node pairs with valid associations. At the same time, in order to avoid the large difference in values ​​between different measurement point pairs leading to unbalanced graph construction, the min-max normalization method is used to uniformly map all association frequencies to the [0,1] interval. The normalized result is used as the edge weight of the corresponding edge in the graph to characterize the potential intensity of disturbance propagation between different measurement point areas.

[0095] A perturbation behavior graph is constructed based on the node attributes, edge connections, and corresponding edge weights.

[0096] The purpose of constructing a disturbance behavior map is to accurately characterize the dynamic relationship between the material and various measuring points on the crusher liner. By integrating acceleration, impact response and material motion trajectory information, the multi-source disturbance data is structured into a disturbance behavior map, which can effectively depict the propagation path and intensity of disturbances in space. This provides a data foundation and correlation constraints for thermal grinding coupling analysis and stress field prediction, thereby improving the accuracy and physical consistency of the prediction results.

[0097] In this embodiment, step two specifically includes:

[0098] Taking the liner measuring point area corresponding to each node in the disturbance behavior map as the analysis object, the temperature rise change information, impact frequency information and friction work information of the corresponding measuring point area are collected in the same time window, and the thermal grinding characteristic time series of the corresponding measuring point area is formed in time order.

[0099] Based on the edge connection relationships in the disturbance behavior map, adjacent measurement point regions with edge connections are selected as combined units;

[0100] Within the combined unit, the temperature rise change information of adjacent measuring point areas is multiplied by the corresponding edge weights in the disturbance behavior map, and the results of the multiplication are accumulated within the combined unit to obtain the combined temperature rise characteristic value of the corresponding time window.

[0101] Within the combined unit, the impact frequency information of adjacent measurement point areas is multiplied by the corresponding edge weights in the disturbance behavior spectrum, and the results of the multiplication are accumulated within the combined unit to obtain the combined impact frequency characteristic value of the corresponding time window.

[0102] Within the combined unit, the friction work information of adjacent measuring point areas is multiplied by the corresponding edge weights in the disturbance behavior spectrum, and the results of the multiplication are accumulated within the combined unit to obtain the combined characteristic value of friction work for the corresponding time window.

[0103] Using the time series formed by the combined characteristic values ​​of temperature rise, impact frequency, and friction work within a continuous time window as input, a thermal grinding cross-analysis model constrained by the edge connection relationship of the perturbation behavior graph is constructed.

[0104] In the thermal-grinding cross-analysis model, the variation range of the temperature rise combination characteristic value and the variation range of the impact frequency combination characteristic value and the friction work combination characteristic value within a continuous time window are compared window by window. When the variation range of the temperature rise combination characteristic value is greater than the variation range of the impact frequency combination characteristic value and greater than the variation range of the friction work combination characteristic value, the time window is determined to be a thermal-dominant window; otherwise, the time window is determined to be a grinding-dominant window.

[0105] The number of state transitions between the heat-dominated window and the grinding-dominated window is counted sequentially over time and used as the heat-grind interleaving factor.

[0106] During the operation of the crusher, the present invention deploys force sensors in the corresponding measuring point area of ​​the liner to obtain the normal load and tangential friction force between the material and the liner. At the same time, it combines the relative sliding velocity and contact duration extracted from the material motion image data to perform time integration on the friction force and relative displacement, thereby calculating the friction work information of the corresponding measuring point area.

[0107] This invention introduces a disturbance behavior map to drive the analysis of the liner's operating state, enabling thermal-wear cross-analysis to move beyond relying on single measuring points or isolated time series. Instead, it uses joint modeling based on the spatial correlation between measuring points. By treating adjacent measuring point regions with edge connections as combined units and using edge weights to weight and accumulate temperature rise, impact frequency, and friction work information, it effectively reflects the propagation path and intensity distribution of disturbances on the liner surface, thus avoiding the amplified impact of local measuring point anomalies on the analysis results. Simultaneously, by constructing combined characteristic values ​​of temperature rise, impact frequency, and friction work within a unified time window, the thermal and wear effects are comparable on the same time scale, facilitating accurate identification of the dominant relationship between thermal and wear effects under different operating conditions. Furthermore, by statistically analyzing the number of state transitions between the thermal-dominant and wear-dominant windows within a continuous time window, it can intuitively depict the frequency of alternation between thermal and wear effects in local areas of the liner, providing stable and traceable input for geometric creep modeling and stress distribution prediction, thereby improving the responsiveness and overall reliability of stress prediction results to changes in actual operating conditions.

[0108] In this embodiment, the step of generating a geometric latent field describing the inherent deformation tendency of the liner under the current operating state using a deformation tendency generation network based on the thermal grinding crossover factor is as follows:

[0109] Using the measurement point area of ​​the liner plate as a network processing unit, the deformation tendency generation network includes a geometric state input layer, a thermal grinding interleaving factor modulation layer, a deformation tendency calculation layer, and a geometric latent output layer.

[0110] In the geometric state input layer, the current liner geometric parameters of each measuring point area are received and organized so that the liner geometric parameters are aligned in the time window and measuring point area dimensions. The liner geometric parameters include the thickness parameters, curvature parameters and surface roughness parameters of the measuring point area.

[0111] In the thermally ground interlacing factor modulation layer, the thermally ground interlacing factor of the corresponding measurement point area is used as the modulation factor, and multiplied by the thickness parameter, curvature parameter and surface roughness parameter respectively to obtain the modulation geometric feature vector of the corresponding measurement point area.

[0112] The modulation geometric feature vector is input to the deformation tendency calculation layer, and the modulation geometric feature vector is subjected to vector difference operation within adjacent time windows to obtain the geometric change vector.

[0113] The sign of each component of the geometric change vector is determined to obtain the deformation change direction component vector of the corresponding measurement point area;

[0114] The absolute values ​​of each component of the geometric change vector are taken and normalized to obtain the deformation change amplitude component vector of the corresponding measurement point area.

[0115] The deformation change direction component vector and the deformation change amplitude component vector are concatenated along the feature dimension to form the deformation trend feature vector of the corresponding measurement point area.

[0116] The deformation tendency feature vector is input to the geometric latent variable output layer. Each component of the deformation tendency feature vector is multiplied by the corresponding set geometric mapping coefficient, and the results of the multiplication are summed to obtain the geometric latent variable corresponding to the measuring point area in the current operating state. The geometric latent variables of all measuring point areas together constitute the geometric latent variable field, which is used to describe the deformation tendency of the liner in the current operating state.

[0117] In this invention, a deformation tendency generation network is used to transform the thermal grinding crossover factor into a geometric latent field that can directly participate in geometric evolution calculations. Specifically, the crusher liner is divided into multiple measurement point regions, each corresponding to a network processing unit. The geometric state input layer is only used to receive and organize the liner geometric structure parameters of the measurement point region. The thickness parameter is taken as the current effective liner thickness of the measurement point region, the curvature parameter is taken as the average value of the principal curvatures of the surface of the measurement point region, and the surface roughness parameter is taken as the arithmetic mean roughness value. The time window length is set to 10 seconds, with no overlap between adjacent time windows, and is used to characterize the geometric change trend under the current operating state.

[0118] In the thermal-grinding interleaving factor modulation layer, the thermal-grinding interleaving factor is directly used as the modulation factor to perform component-wise multiplication operations on the thickness parameter, curvature parameter, and surface roughness parameter, thereby numerically reflecting the influence of alternating thermal-grinding on the evolution of geometric state. In the deformation trend calculation layer, only the modulation geometric feature vectors corresponding to the current time window and the previous time window are selected for vector difference calculation to ensure that only one geometric change vector is generated for each measurement point area in one prediction process, avoiding multi-value ambiguity. The sign of each component of the geometric change vector is judged, with a positive sign indicating the direction of geometric parameter increase and a negative sign indicating the direction of geometric parameter decrease. At the same time, the absolute value of the change is taken and processed using the maximum value normalization method. The normalization benchmark value is set to 1.0 times the maximum change in historical observation, thereby obtaining a stable deformation change amplitude component vector.

[0119] In the geometric latent variable output layer, the deformation direction component vector and the deformation amplitude component vector are concatenated along the feature dimension to form a unified deformation trend feature vector. This vector is then multiplied component by component with preset geometric mapping coefficients and summed. The geometric mapping coefficients are set to 0.5, 0.3, and 0.2 to distinguish the contribution weight of different geometric features to the latent variable result. The resulting geometric latent variable is represented in scalar form as the inherent deformation trend of the corresponding measurement point area under the current operating state. The geometric latent variables of all measurement point areas together constitute the geometric latent variable field, providing a definite and traceable input basis for updating the liner geometry and predicting stress distribution.

[0120] In this embodiment, the process of fusing the geometric latent field with the wear evolution trend vector to update the liner's geometric parameters specifically involves:

[0121] Taking each measuring point area of ​​the liner as the processing object, historical wear monitoring data of the corresponding measuring point area is collected within multiple consecutive historical time windows. The historical wear monitoring data includes liner thickness data, curvature data and surface roughness data.

[0122] Within adjacent historical time windows, the thickness data, curvature data, and surface roughness data are respectively subjected to difference calculation to obtain the thickness wear change, curvature wear change, and surface roughness wear change of the corresponding measuring point area under each historical time window;

[0123] The thickness wear variation obtained within multiple consecutive historical time windows is arranged into a thickness wear variation sequence in chronological order; the curvature variation obtained within multiple consecutive historical time windows is arranged into a curvature variation sequence in chronological order; and the roughness wear variation obtained within multiple consecutive historical time windows is arranged into a roughness wear variation sequence in chronological order.

[0124] The thickness wear change sequence, curvature wear change sequence, and roughness wear change sequence are respectively subjected to least squares linear fitting operation to obtain the thickness wear change slope, curvature wear change slope, and roughness wear change slope of the corresponding measuring point area, and the thickness wear change slope, curvature wear change slope, and roughness wear change slope are combined to form a wear evolution trend vector;

[0125] The geometric latent variables of the corresponding measurement point area are multiplied by each component of the wear evolution trend vector to obtain the fused modulation vector;

[0126] Each component in the fused modulation vector is used as a correction increment for the thickness parameter, curvature parameter, and surface roughness parameter of the corresponding measurement point area. These are then added one by one to the corresponding current thickness parameter, curvature parameter, and surface roughness parameter to obtain the updated geometric structure parameters.

[0127] In this invention, a geometric structure update mechanism that takes into account both current operating condition response and historical wear patterns is constructed by dividing the liner geometric evolution process into two interconnected stages: deformation trend modeling and wear trend fusion update. First, in the deformation trend modeling stage, the liner geometry is not directly modified. Instead, based on the thermal wear crossover factor, the thickness, curvature, and surface roughness parameters of each measuring point region under the current operating condition are modulated and subjected to time-difference analysis to obtain geometric latent variables reflecting the direction and intensity of geometric changes. The geometric latent variable field, composed of the geometric latent variables of each measuring point region, is used to characterize the spatial distribution of the liner's inherent deformation trend under current operating conditions. Essentially, it is a potential representation of geometric change, rather than a direct geometric update result.

[0128] Based on this, wear monitoring data within historical time windows are further introduced to statistically model the historical changes in liner thickness, curvature, and surface roughness. The wear evolution trend vector is extracted by least-squares linear fitting to characterize the stable wear direction and rate of change of the liner over a longer time scale. This wear evolution trend comes from historical data and can effectively suppress the interference of unstable factors such as instantaneous impact and short-term load fluctuations on geometric updates, so that the geometric evolution process has time consistency and engineering reliability.

[0129] By fusing geometric latent variables with the wear evolution trend vector component by component, a coupled-driven geometric update mechanism based on current deformation potential and historical wear patterns is realized. This fusion method ensures that the update magnitude of the liner geometry is not only affected by the deformation driving force under the current working condition, but also constrained by the long-term wear evolution trend. This avoids the instability caused by relying solely on instantaneous information for geometric updates. Compared with existing methods that predict geometric evolution based solely on load response or a single wear model, this invention achieves physical interpretability and numerical stability of the geometric structure update process through time scale separation and multi-source information fusion, providing a reliable geometric basis for stress distribution prediction and geometric backtracking feedback.

[0130] In this embodiment, step four specifically includes:

[0131] Using the measurement point regions corresponding to the updated liner geometric parameters as the analysis objects, in the disturbance behavior graph, the graph nodes corresponding to each measurement point region are determined, and the edge connection relationships and corresponding edge weights connected to the nodes are obtained.

[0132] The thermal grinding interleaving factor of the corresponding measuring point area is multiplied with the updated liner geometric structure parameters one by one to obtain the updated modulation geometric feature vector of the corresponding measuring point area.

[0133] Based on the edge connection relationship of the perturbation behavior map, the vector magnitude of the updated modulation geometric feature vector of all adjacent measurement point regions that are connected to the measurement point region is calculated respectively. The obtained vector magnitude is multiplied by the corresponding edge weight, and the multiplication results are accumulated to obtain the perturbation weighted feature value of the target measurement point region.

[0134] The perturbation weighted feature value is compared with the set hot spot determination threshold, and combined with the dominant state of the thermal wear interleaving factor of the corresponding measurement point area in the current time window, when the perturbation weighted feature value is greater than the set hot spot determination threshold and the corresponding current time window is a thermal dominant window, the corresponding measurement point area is determined as a hot spot induction point.

[0135] By collecting all the measurement points identified as hot spot induction points, a set of hot spot induction points on the liner is obtained;

[0136] In this invention, a spatial weighted determination mechanism for locating hot spot induction points on the liner is established by introducing a disturbance behavior map and a thermal grinding cross factor based on the updated liner geometry parameters. In specific implementation, the crusher liner is divided into several measurement point regions, each measurement point region corresponds to a node in the disturbance behavior map, and the nodes are connected by edges to characterize the propagation path of the disturbance on the liner surface. The edge weight is used to reflect the relative intensity of the disturbance propagation between adjacent measurement point regions.

[0137] In the process of obtaining hot spot induction points, the updated thickness, curvature, and surface roughness parameters are modulated component by component using the thermal-grinding interleaving factor, so that the measurement point area with significant thermal dominance can obtain amplified response at the numerical level. For adjacent measurement point areas that have edge connections with the target measurement point area, the Euclidean modulus of the corresponding updated modulated geometric feature vector is calculated, and weighted and accumulated according to the edge weight in the perturbation behavior spectrum, so as to obtain the perturbation weighted feature value that characterizes the degree of neighborhood perturbation aggregation. This feature value can comprehensively reflect the superposition of geometric perturbation and thermal effect on the target measurement point area in the spatial neighborhood.

[0138] During the determination phase, the perturbation weighted feature value is compared with a pre-set hot spot determination threshold, wherein the hot spot determination threshold is set to 0.75, and the measurement point area is marked as a hot spot induction point only when the corresponding time window is determined to be a thermally dominant window. By simultaneously introducing a spatial perturbation weighting mechanism and a time thermally dominant constraint, misjudgments caused solely by local geometric anomalies or instantaneous perturbations are avoided.

[0139] Compared with existing methods for identifying hot spots based on a single temperature threshold or local load peak, this invention achieves stable and interpretable localization of the hot spot-induced region by integrating geometric evolution, disturbance propagation relationship, and alternating thermal and grinding states, providing reliable spatial prior information for stress propagation constraints and stress distribution prediction.

[0140] In this embodiment, step five specifically includes:

[0141] A discrete model of the liner is constructed using the measurement point regions corresponding to the updated geometric parameters of the liner as the basic units. The surface of the liner is discretized in plane according to the measurement point regions, and the volume elements corresponding to each measurement point region are divided into layers along the thickness direction of the liner to form a set of volume elements consisting of surface volume elements, middle volume elements and bottom volume elements.

[0142] In this invention, when dividing the volume elements corresponding to each measuring point area into layers along the thickness direction of the liner, the current total thickness value of the liner corresponding to the measuring point area is first obtained, and the total thickness value is used as the layering reference. Specifically, the liner surface is used as the starting reference plane, and the total thickness is divided into three equal segments along the normal direction. The thickness interval near the liner surface is divided into surface volume elements, the middle thickness interval is divided into middle volume elements, and the thickness interval near the back of the liner is divided into bottom volume elements. The same layering method is used for each measuring point area to ensure that the number of volume elements and the corresponding level in the thickness direction of each measuring point area are consistent. After the layering is completed, a unique level identifier is assigned to each volume element, and the correspondence between the surface volume elements, middle volume elements and bottom volume elements is established to determine the transmission path of disturbance and stress in the thickness direction during the layer-by-layer propagation calculation.

[0143] A multi-field coupling model is constructed, which includes a perturbation field sub-model, an evolution field sub-model, and a stress field sub-model. The perturbation field sub-model, the evolution field sub-model, and the stress field sub-model are established on the same set of volume elements and share the spatial adjacency relationship between volume elements.

[0144] In the disturbance field sub-model, the node attributes corresponding to each measurement point region in the disturbance behavior map are used as the disturbance input vector of the corresponding surface volume unit, and the edge weights between measurement point regions in the disturbance behavior map are used as the disturbance transmission coefficients between adjacent surface volume units to describe the lateral propagation relationship of disturbance between volume units.

[0145] In the evolution field sub-model, for the surface volume element corresponding to each measurement point region, the updated thickness parameter, curvature parameter, and surface roughness parameter of the corresponding measurement point region are read and weighted summation is performed to obtain the evolution modulation coefficient of the surface volume element. In specific implementation, the normalized thickness parameter, curvature parameter, and surface roughness parameter are assigned weight coefficients of 0.5, 0.3, and 0.2, respectively, and weighted summation is performed to obtain the evolution modulation coefficient of the corresponding surface volume element. The evolution modulation coefficient is a dimensionless scalar in nature representing the modulation intensity of the current geometric structure state of the measurement point region on the subsequent stress evolution process, and is used in the multi-field coupling model to participate in the layer-by-layer stress propagation calculation.

[0146] The set of hot spot induction points is used as a stress propagation constraint. In the multi-field coupling model, propagation constraint markers are set for the volume elements corresponding to the measurement point regions belonging to the set of hot spot induction points. The perturbation transmission coefficient and evolution modulation coefficient corresponding to the volume elements with propagation constraint markers are amplified and updated, while the perturbation transmission coefficient and evolution modulation coefficient corresponding to the volume elements without propagation constraint markers are attenuated and updated.

[0147] In this invention, the set of hot spot induction points serves as a stress propagation constraint. By setting propagation constraint markers on the corresponding volume elements of the measurement point region, the perturbation transmission coefficient and evolution modulation coefficient are updated in a differentiated manner. In specific implementation, propagation constraint markers are first set for the volume elements corresponding to the measurement points in the hot spot induction point set. For volume elements with propagation constraint markers, the perturbation transmission coefficient is updated by amplification. The original perturbation transmission coefficient is multiplied by the amplification factor of 1.3 to obtain the updated perturbation transmission coefficient. At the same time, the evolution modulation coefficient of the volume element is multiplied by the amplification factor of 1.3 to obtain the updated evolution modulation coefficient. For volume elements without propagation constraint markers, the perturbation transmission coefficient is updated by attenuation. The original perturbation transmission coefficient is multiplied by the attenuation factor of 0.7 to obtain the updated perturbation transmission coefficient. At the same time, the evolution modulation coefficient of the volume element is multiplied by the attenuation factor of 0.7 to obtain the updated evolution modulation coefficient. The above factors are set as fixed constants during the system initialization phase and remain unchanged within the same prediction period. This allows the hot spot induction point area to obtain higher propagation weights in the lateral perturbation propagation and longitudinal layer-by-layer propagation process, while the propagation weights of non-hot spot areas are relatively suppressed. This results in differentiated stress propagation paths and propagation intensity distributions in multi-field coupling calculations.

[0148] In the stress field sub-model, layer-by-layer propagation calculations are performed in the order of surface volume elements to middle volume elements and then to bottom volume elements. The perturbation input vector of the middle volume element is obtained by multiplying the perturbation input vector of the corresponding surface volume element with the updated perturbation transfer coefficient component by component.

[0149] The perturbation input vector of the bottom volume element is obtained by multiplying the perturbation input vector of the corresponding middle volume element with the updated perturbation transfer coefficient component by component.

[0150] In each layer of volume element, the perturbation input vector of the corresponding layer of volume element is multiplied component by component with the updated evolution modulation coefficient, and the multiplication results are summed to obtain the equivalent stress value of the corresponding layer of volume element.

[0151] After completing the layer-by-layer propagation calculation of all volume elements, according to the spatial positional relationship of the measuring point area on the liner surface, the equivalent stress values ​​of the surface volume element, middle volume element and bottom volume element corresponding to the same measuring point area are summed and averaged, and mapped back to the corresponding position on the liner surface to form an initial predicted stress distribution covering the entire area of ​​the liner. The initial predicted stress distribution refers to the regional representative stress distribution formed by collecting the equivalent stress values ​​obtained by the multi-layer volume elements corresponding to the measuring point area in the layer-by-layer propagation calculation with the measuring point area as the smallest output unit.

[0152] In this invention, the surface of the liner is discretized according to the measurement point regions, and the corresponding volume elements of each measurement point region are layered along the thickness direction to construct a set of volume elements consisting of surface volume elements, middle volume elements, and bottom volume elements. This allows the geometry, disturbance input, and stress response of the liner to be modeled within a unified spatial discretization framework. Based on this, the disturbance field sub-model, evolution field sub-model, and stress field sub-model are established on the same set of volume elements and share the spatial adjacency relationships between volume elements. This achieves a consistent expression of multi-field information in the spatial structure and provides a stable computational foundation for layer-by-layer propagation calculations.

[0153] In the multi-field coupling model, the perturbation field sub-model introduces the perturbation information generated by the material-liner interaction through the perturbation behavior spectrum, so that the external impact and vibration effects can act on the surface volume element in the form of a vector. The evolution field sub-model forms the evolution modulation coefficient of the surface volume element by weighted summing of the updated thickness parameter, curvature parameter and surface roughness parameter, so that the geometric state of the liner has a direct impact on the perturbation propagation process. By coupling the perturbation field and the evolution field at the volume element level, the stress calculation is not only affected by the instantaneous perturbation, but also constrained by the geometric evolution state, avoiding the problem of the separation of geometric factors and load factors in traditional stress prediction.

[0154] Furthermore, by introducing a set of hot spot induction points as a stress propagation constraint, propagation constraint markers are set for the volume elements corresponding to the measurement point regions belonging to the set of hot spot induction points, and the perturbation transmission coefficient and evolution modulation coefficient are updated differentially. This allows the hot spot region to obtain a higher propagation weight in the process of lateral perturbation propagation and longitudinal layer-by-layer propagation, while the propagation weight of non-hot spot regions is suppressed. This results in the formation of a spatially directional stress propagation path in the multi-field coupling model. This constraint mechanism can effectively highlight the dominant role of high thermal risk regions in the stress distribution formation process and avoid the stress from being too uniformly propagated in space, thus masking local risks.

[0155] Through the synergistic effect of the above multi-field coupling model and hot spot propagation constraints, the obtained initial predicted stress distribution can simultaneously reflect the disturbance propagation characteristics, geometric evolution state, and hot spot concentration effect, making the prediction results more closely resemble the actual operating conditions in terms of spatial distribution, and providing a reliable basis for geometric backpropagation feedback and final stress distribution correction.

[0156] In this embodiment, the geometric backtracking feedback mechanism is specifically as follows:

[0157] The absolute value of the stress residual is taken to obtain the stress deviation amplitude, and the sign of the stress residual is determined. A positive stress residual indicates that the initial predicted stress value is greater than the actual monitored stress value, and a negative stress residual indicates that the initial predicted stress value is less than the actual monitored stress value.

[0158] The stress deviation amplitude is multiplied by the updated thickness parameter, curvature parameter and surface roughness parameter of the corresponding measuring point area, and the product is divided by the absolute value of the initial predicted stress value of the corresponding measuring point area in the initial predicted stress distribution to obtain the corresponding thickness correction amount, curvature correction amount and roughness correction amount.

[0159] When the stress residual is positive, the thickness correction, curvature correction, and roughness correction are subtracted from the updated thickness parameters, curvature parameters, and surface roughness parameters of the corresponding measurement point area, respectively.

[0160] When the stress residual is negative, the thickness correction amount, curvature correction amount, and roughness correction amount are added to the updated thickness parameter, curvature parameter, and surface roughness parameter of the corresponding measuring point area, respectively.

[0161] After correcting the geometric parameters of all measuring point areas, the corrected geometric structural parameters of the liner are obtained.

[0162] In this embodiment, step seven specifically includes:

[0163] Based on the back-corrected liner geometry parameters, the evolution modulation coefficients in the multi-field coupling model are recalculated to obtain an updated multi-field coupling model.

[0164] In the updated multi-field coupling model, following the layer-by-layer propagation calculation process from the disturbance field and evolution field to the stress field, the corrected equivalent stress values ​​of the surface, middle and bottom layers of the body element corresponding to each measuring point area are obtained. The three corrected equivalent stress values ​​are summed again and averaged, and then mapped back to the corresponding position on the liner surface to obtain the final predicted stress distribution of the liner.

[0165] In this invention, a geometric backtracking feedback mechanism is introduced to correct the geometric parameters of the liner, so that the geometric state in the multi-field coupling model can be consistent with the actual stress monitoring results. After the geometric backtracking correction is completed, the backtracked thickness parameters, curvature parameters and surface roughness parameters are reused to calculate the evolution modulation coefficient, so that the evolution field in the multi-field coupling model reflects the geometric state after feedback correction, and avoids the initial geometric modeling error being continuously amplified in stress prediction.

[0166] In the updated multi-field coupling model, the calculation process of perturbation field and evolution field propagating to stress field layer by layer remains unchanged. Only the evolution modulation coefficient is updated to readjust the perturbation propagation process at the geometric level. By calculating the corrected equivalent stress values ​​for the surface volume element, middle volume element and bottom volume element respectively, the stress results of the three layers are summed and averaged at the measurement point area scale, which can comprehensively reflect the overall stress state in the thickness direction of the liner plate, so that the final predicted stress distribution includes both surface impact effect and internal structural response. Example

[0167] To verify the feasibility of this invention in practice, it was applied to the stress prediction scenario of the liner of a medium-to-large crushing equipment that is in a long-term continuous operation. In actual operation, this type of crushing equipment has a wide particle size distribution and obvious feed fluctuations. The liner is subjected to frequent impact loads, sliding friction and local heat accumulation for a long time. Stress concentration and abnormal wear are prone to occur in local areas of the liner. However, traditional methods that rely on experience or static finite element models cannot accurately reflect the actual stress state of the liner. As a result, the problem of delayed liner failure prediction and maintenance decisions relying on human experience has existed for a long time.

[0168] In this embodiment, dynamic information generated by the interaction between the material and the liner is first collected synchronously during the operation of the crusher. Multiple accelerometers and impact sensors are deployed at different measuring points on the liner, and material motion image data acquired by an industrial camera is used to continuously monitor the material's trajectory within the crushing chamber and its contact behavior with the liner. Based on the collected dynamic information, a disturbance behavior map is constructed. Each measuring point on the liner is used as a map node, with average acceleration, impact frequency, and impact intensity as node attributes. The correlation frequency of the material's trajectory passing through different measuring point areas in consecutive frames is used as the edge weight, thus forming a disturbance behavior map that reflects the disturbance propagation path and intensity.

[0169] After the disturbance behavior map is constructed, it is used as the driving input to synchronously collect information on temperature rise, impact frequency, and friction work during the operation of the liner. A thermal-wear alternation analysis model is constructed within a unified time window. By using adjacent measuring point regions with edge connections as combined units, combined characteristic values ​​of temperature rise, impact frequency, and friction work are formed respectively. The change amplitude of each combined characteristic value is compared within a continuous time window to determine the thermal-dominant window and the wear-dominant window. The number of state switching is counted to obtain the thermal-wear alternation factor corresponding to each measuring point region. This thermal-wear alternation factor can quantitatively reflect the frequency of alternation between thermal and wear effects in local areas of the liner during operation.

[0170] Subsequently, based on the aforementioned thermal grinding interleaving factor, a deformation tendency generation network is used to generate a geometric latent variable field describing the inherent deformation tendency of the liner under the current operating state. In this process, each measuring point area of ​​the liner is used as a network processing unit. The current geometric structure parameters of the liner are input to the geometric state input layer. The thickness parameter, curvature parameter, and surface roughness parameter are modulated component by component through the thermal grinding interleaving factor modulation layer. Vector difference calculation is performed within adjacent time windows to form a geometric change vector, thereby obtaining the deformation change direction component vector and the deformation change amplitude component vector. After splicing and mapping, geometric latent variables are generated. The geometric latent variables of all measuring point areas together constitute the geometric latent variable field.

[0171] Based on this, historical wear monitoring data is introduced to construct a wear evolution trend vector. The geometric potential field is then fused with this vector to update the liner's geometric parameters. This ensures that the geometric parameters reflect both the deformation potential under current operating conditions and are constrained by historical wear patterns. Based on the updated liner geometric parameters, combined with the disturbance behavior map and thermal wear crossover factor, a set of hot spot induction points on the liner is further obtained. By calculating the modulus of the updated modulated geometric feature vector, accumulating the disturbance-weighted eigenvalues, and combining this with thermal dominance window determination, measurement point areas with significant heat accumulation risk are identified.

[0172] The measurement point regions corresponding to the updated liner geometry parameters are discretized and modeled, and divided into surface, middle, and bottom volume elements along the liner thickness direction. In the multi-field coupling model, the perturbation behavior map is mapped to a perturbation field sub-model, and the geometry parameters are mapped to an evolution field sub-model. The set of hot spot induction points is used as a stress propagation constraint to differentially update the perturbation transfer coefficient and evolution modulation coefficient. In the stress field sub-model, layer-by-layer propagation calculations are performed in the order of surface volume elements to middle volume elements, and then to bottom volume elements to obtain the initial predicted stress distribution of the liner.

[0173] After obtaining the initial predicted stress distribution, it is compared with the actual monitored stress data to calculate the stress residual, and a geometric backtracking feedback mechanism is introduced. By performing numerical backtracking corrections on the thickness, curvature, and surface roughness parameters based on the magnitude and sign of the stress residual, the corrected liner geometry parameters are formed. Finally, based on the corrected liner geometry parameters, the multi-field coupling model is updated, and the layer-by-layer propagation calculation is re-executed to obtain the final predicted stress distribution.

[0174] To verify the beneficial effects of the present invention, the stress prediction results of the traditional static finite element method and the method of the present invention were compared in multiple measurement point areas. The relevant statistical results are shown in Table 1.

[0175] Table 1. Comparison of stress prediction errors in different measuring point areas

[0176]

[0177] As can be seen from the data in Table 1 above, when the actual monitored stress levels are similar across different measuring point areas, there is a significant difference in prediction accuracy between the traditional method and the method of this invention. Taking measuring point area A1 as an example, the actual monitored stress is 182.4 MPa, while the traditional method predicts a value of 213.7 MPa with a relative error of 17.2%, whereas the method of this invention predicts a stress of 187.9 MPa with a relative error of only 3.0%. The prediction result is significantly closer to the actual monitored value. A similar trend exists in other measuring point areas, indicating that this difference is not an isolated phenomenon.

[0178] In measurement areas A2 and A5, the relative errors of the traditional method were 15.3% and 15.0%, respectively, while the relative errors of the method of this invention were reduced to 2.9% and 2.5%, respectively, with an error reduction of over 80%. This indicates that in the medium stress level region, the present invention can effectively suppress the prediction deviation caused by operating condition fluctuations and local disturbances. For measurement areas A3 and A4 with higher stress levels, the relative errors of the traditional method reached 17.3% and 17.4%, respectively, while the method of this invention still controlled the relative errors within 3.1% and 2.8%, demonstrating good prediction stability even in high stress and strong disturbance regions.

[0179] As can be seen from this embodiment, the present invention has a significantly improved ability to predict the stress distribution of the liner under complex crushing conditions compared with the prior art. It no longer relies on static geometric assumptions or single load models, but instead constructs a disturbance behavior map to uniformly model multi-source dynamic information such as material movement, impact, and vibration, fundamentally improving the ability to characterize the disturbance propagation relationship. On this basis, a thermal-wearing interleaved analysis model is introduced to distinguish and quantify thermal and wear effects in time and space dimensions, effectively avoiding the problem of the superposition of thermal and wear effects in traditional methods, which are difficult to identify.

[0180] Meanwhile, this invention transforms the passive result of changes in the liner's geometric structure into a perceptible and moduloable intermediate state expression by constructing a deformation trend generation network and a geometric latent field. Furthermore, it combines historical wear evolution trends to achieve stable updates of geometric parameters, ensuring continuity and physical consistency in the geometric evolution process. Based on the updated geometric parameters and the set of hot spot induction points, stress propagation constraints are introduced into the multi-field coupling model, enhancing the stress transfer characteristics in high-risk areas and improving the spatial resolution of stress prediction. In addition, a geometric backtracking feedback mechanism establishes a closed-loop correction relationship between the stress prediction results and actual monitoring data, effectively suppressing the amplified impact of instantaneous disturbances and accumulated model errors on the prediction results.

[0181] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for predicting stress distribution in crusher liners based on multi-field coupling, characterized in that, Includes the following steps: Step 1: Collect dynamic information on the interaction between material and liner during crusher operation, and construct a disturbance behavior map based on the dynamic information; Step 2: Using the disturbance behavior map as the driving input, collect information on temperature rise, impact frequency, and friction work during the operation of the liner, construct a thermal-grind cross-analysis model, and extract the thermal-grind cross-factor; Step 3: Based on the aforementioned thermal wear crossover factor, a deformation trend generation network is used to generate a geometric latent field describing the inherent deformation trend of the liner under the current operating state. The geometric latent field is then fused with the wear evolution trend vector to update the geometric structure parameters of the liner. Step 4: Based on the updated liner geometry parameters, and combined with the disturbance behavior map and thermal wear crossover factor, obtain the set of hot spot induction points on the liner; Step 5: Construct a multi-field coupling model and use the set of hot spot induction points as stress propagation constraints. In the multi-field coupling model, realize the layer-by-layer propagation of the perturbation field and evolution field to the stress field to obtain the initial predicted stress distribution. Step 6: Calculate the stress residual between the initial predicted stress distribution and the actual monitored stress data, and introduce a geometric backtracking feedback mechanism to backtrack and correct the updated liner geometric parameters. Step 7: Update the multi-field coupling model based on the back-corrected geometric parameters of the liner to obtain the final predicted stress distribution.

2. The method for predicting stress distribution in crusher liners based on multi-field coupling according to claim 1, characterized in that, The dynamic information specifically includes multi-point acceleration signals, impact signals, and material motion image data.

3. The method for predicting stress distribution in crusher liners based on multi-field coupling according to claim 1, characterized in that, The construction of the disturbance behavior map based on dynamic information specifically includes: Based on the multi-point acceleration and impact signals in the dynamic information, the average acceleration, impact frequency and impact intensity of each measuring point per unit time are extracted. The motion image data of the material is processed frame by frame to identify the measurement point area where the material contacts the liner and obtain the material's motion trajectory. Using the measurement points of the liner as the nodes of the graph, the average acceleration, impact frequency and impact intensity of each measurement point per unit time are used as the node attributes. The correlation frequency between nodes is determined by calculating the number of times the material movement trajectory passes through two measurement point areas in a continuous frame. The correlation frequency is then normalized to establish edge connections and obtain the corresponding edge weights. A perturbation behavior graph is constructed based on the node attributes, edge connections, and corresponding edge weights.

4. The method for predicting stress distribution in crusher liners based on multi-field coupling according to claim 1, characterized in that, Step two specifically involves: Taking the liner measuring point area corresponding to each node in the disturbance behavior map as the analysis object, the temperature rise change information, impact frequency information and friction work information of the corresponding measuring point area are collected in the same time window, and the thermal grinding characteristic time series of the corresponding measuring point area is formed in time order. Based on the edge connection relationships in the disturbance behavior map, adjacent measurement point regions with edge connections are selected as combined units; Within the combined unit, the temperature rise change information of adjacent measuring point areas is multiplied by the corresponding edge weights in the disturbance behavior map, and the results of the multiplication are accumulated within the combined unit to obtain the combined temperature rise characteristic value of the corresponding time window. Within the combined unit, the impact frequency information of adjacent measurement point areas is multiplied by the corresponding edge weights in the disturbance behavior spectrum, and the results of the multiplication are accumulated within the combined unit to obtain the combined impact frequency characteristic value of the corresponding time window. Within the combined unit, the friction work information of adjacent measuring point areas is multiplied by the corresponding edge weights in the disturbance behavior spectrum, and the results of the multiplication are accumulated within the combined unit to obtain the combined characteristic value of friction work for the corresponding time window. Using the time series formed by the combined characteristic values ​​of temperature rise, impact frequency, and friction work within a continuous time window as input, a thermal grinding cross-analysis model constrained by the edge connection relationship of the perturbation behavior graph is constructed. In the thermal grinding cross-analysis model, the variation range of the temperature rise combination characteristic value and the variation range of the impact frequency combination characteristic value and the friction work combination characteristic value within a continuous time window are compared window by window. When the variation range of the temperature rise combination characteristic value is greater than the variation range of the impact frequency combination characteristic value and greater than the variation range of the friction work combination characteristic value, the time window is determined to be a thermally dominated window. Otherwise, the time window is determined to be the grinding-dominated window; The number of state transitions between the heat-dominated window and the grinding-dominated window is counted sequentially over time and used as the heat-grind interleaving factor.

5. The method for predicting stress distribution in crusher liners based on multi-field coupling according to claim 1, characterized in that, The process of generating a geometric latent field describing the inherent deformation tendency of the liner under the current operating state using a deformation tendency generation network based on the aforementioned thermal grinding interleaving factor is as follows: Using the measurement point area of ​​the liner plate as a network processing unit, the deformation tendency generation network includes a geometric state input layer, a thermal grinding interleaving factor modulation layer, a deformation tendency calculation layer, and a geometric latent output layer. In the geometric state input layer, the current liner geometric parameters of each measuring point area are received and organized so that the liner geometric parameters are aligned in the time window and measuring point area dimensions. The liner geometric parameters include the thickness parameters, curvature parameters and surface roughness parameters of the measuring point area. In the thermally ground interlacing factor modulation layer, the thermally ground interlacing factor of the corresponding measurement point area is used as the modulation factor, and multiplied by the thickness parameter, curvature parameter and surface roughness parameter respectively to obtain the modulation geometric feature vector of the corresponding measurement point area. The modulation geometric feature vector is input to the deformation tendency calculation layer, and the modulation geometric feature vector is subjected to vector difference operation within adjacent time windows to obtain the geometric change vector. The sign of each component of the geometric change vector is determined to obtain the deformation change direction component vector of the corresponding measurement point area; The absolute values ​​of each component of the geometric change vector are taken and normalized to obtain the deformation change amplitude component vector of the corresponding measurement point area. The deformation change direction component vector and the deformation change amplitude component vector are concatenated along the feature dimension to form the deformation trend feature vector of the corresponding measurement point area. The deformation tendency feature vector is input to the geometric latent variable output layer. Each component of the deformation tendency feature vector is multiplied by the corresponding set geometric mapping coefficient, and the results of the multiplication are summed to obtain the geometric latent variable corresponding to the measurement point area in the current operating state. The geometric latent variables of all measurement point areas together constitute the geometric latent variable field.

6. The method for predicting stress distribution in crusher liners based on multi-field coupling according to claim 1, characterized in that, The process of fusing the geometric latent field with the wear evolution trend vector to update the liner's geometric parameters specifically involves: Taking each measuring point area of ​​the liner as the processing object, historical wear monitoring data of the corresponding measuring point area is collected within multiple consecutive historical time windows. The historical wear monitoring data includes liner thickness data, curvature data and surface roughness data. Within adjacent historical time windows, the thickness data, curvature data, and surface roughness data are respectively subjected to difference calculation to obtain the thickness wear change, curvature wear change, and surface roughness wear change of the corresponding measuring point area under each historical time window; The thickness wear variation obtained within multiple consecutive historical time windows is arranged into a thickness wear variation sequence in chronological order; the curvature variation obtained within multiple consecutive historical time windows is arranged into a curvature variation sequence in chronological order; and the roughness wear variation obtained within multiple consecutive historical time windows is arranged into a roughness wear variation sequence in chronological order. The thickness wear change sequence, curvature wear change sequence, and roughness wear change sequence are respectively subjected to least squares linear fitting operation to obtain the thickness wear change slope, curvature wear change slope, and roughness wear change slope of the corresponding measuring point area, and the thickness wear change slope, curvature wear change slope, and roughness wear change slope are combined to form a wear evolution trend vector; The geometric latent variables of the corresponding measurement point area are multiplied by each component of the wear evolution trend vector to obtain the fused modulation vector; Each component in the fused modulation vector is used as a correction increment for the thickness parameter, curvature parameter, and surface roughness parameter of the corresponding measurement point area. These components are then added one by one to the corresponding current thickness parameter, curvature parameter, and surface roughness parameter to obtain the updated geometric structure parameters.

7. The method for predicting stress distribution in crusher liners based on multi-field coupling according to claim 1, characterized in that, Step four specifically involves: Using the measurement point regions corresponding to the updated liner geometric parameters as the analysis objects, in the disturbance behavior graph, the graph nodes corresponding to each measurement point region are determined, and the edge connection relationships and corresponding edge weights connected to the nodes are obtained. The thermal grinding interleaving factor of the corresponding measuring point area is multiplied with the updated liner geometric structure parameters one by one to obtain the updated modulation geometric feature vector of the corresponding measuring point area. Based on the edge connection relationship of the perturbation behavior map, the vector magnitude of the updated modulation geometric feature vector of all adjacent measurement point regions that are connected to the measurement point region is calculated respectively. The obtained vector magnitude is multiplied by the corresponding edge weight, and the multiplication results are accumulated to obtain the perturbation weighted feature value of the target measurement point region. The perturbation weighted feature value is compared with the set hot spot determination threshold, and combined with the dominant state of the thermal wear interleaving factor of the corresponding measurement point area in the current time window, when the perturbation weighted feature value is greater than the set hot spot determination threshold and the corresponding current time window is a thermal dominant window, the corresponding measurement point area is determined as a hot spot induction point. By collecting all the measurement points that have been identified as hot spot induction points, a set of hot spot induction points on the liner is obtained.

8. The method for predicting stress distribution in crusher liners based on multi-field coupling according to claim 1, characterized in that, Step five specifically involves: A discrete model of the liner is constructed using the measurement point regions corresponding to the updated geometric parameters of the liner as the basic units. The surface of the liner is discretized in plane according to the measurement point regions, and the volume elements corresponding to each measurement point region are divided into layers along the thickness direction of the liner to form a set of volume elements consisting of surface volume elements, middle volume elements and bottom volume elements. A multi-field coupling model is constructed, which includes a perturbation field sub-model, an evolution field sub-model, and a stress field sub-model. The perturbation field sub-model, the evolution field sub-model, and the stress field sub-model are established on the same set of volume elements and share the spatial adjacency relationship between volume elements. In the disturbance field sub-model, the node attributes corresponding to each measurement point region in the disturbance behavior map are used as the disturbance input vector of the corresponding surface volume unit, and the edge weights between measurement point regions in the disturbance behavior map are used as the disturbance transmission coefficients between adjacent surface volume units to describe the lateral propagation relationship of disturbance between volume units. In the evolution field sub-model, for the surface volume unit corresponding to each measurement point region, the updated thickness parameters, curvature parameters and surface roughness parameters of the corresponding measurement point region are read respectively, and a weighted summation operation is performed to obtain the evolution modulation coefficient of the surface volume unit. The set of hot spot induction points is used as a stress propagation constraint. In the multi-field coupling model, propagation constraint markers are set for the volume elements corresponding to the measurement point regions belonging to the set of hot spot induction points. The perturbation transmission coefficient and evolution modulation coefficient corresponding to the volume elements with propagation constraint markers are amplified and updated, while the perturbation transmission coefficient and evolution modulation coefficient corresponding to the volume elements without propagation constraint markers are attenuated and updated. In the stress field sub-model, layer-by-layer propagation calculations are performed in the order of surface volume elements to middle volume elements and then to bottom volume elements. The perturbation input vector of the middle volume element is obtained by multiplying the perturbation input vector of the corresponding surface volume element with the updated perturbation transfer coefficient component by component. The perturbation input vector of the bottom volume element is obtained by multiplying the perturbation input vector of the corresponding middle volume element with the updated perturbation transfer coefficient component by component. In each layer of volume element, the perturbation input vector of the corresponding layer of volume element is multiplied component by component with the updated evolution modulation coefficient, and the multiplication results are summed to obtain the equivalent stress value of the corresponding layer of volume element. After completing the layer-by-layer propagation calculations for all body elements, the equivalent stress values ​​of the surface body elements, middle body elements, and bottom body elements corresponding to the same body element area are summed and averaged according to the spatial positional relationship of the measuring point area on the liner surface. This average value is then mapped back to the corresponding position on the liner surface to form the initial predicted stress distribution covering the entire area of ​​the liner.

9. The method for predicting stress distribution in crusher liners based on multi-field coupling according to claim 1, characterized in that, The geometric back-pull feedback mechanism is specifically as follows: The absolute value of the stress residual is taken to obtain the stress deviation amplitude, and the sign of the stress residual is determined. The stress deviation amplitude is multiplied by the updated thickness parameter, curvature parameter and surface roughness parameter of the corresponding measuring point area, and the product is divided by the absolute value of the initial predicted stress value of the corresponding measuring point area in the initial predicted stress distribution to obtain the corresponding thickness correction amount, curvature correction amount and roughness correction amount. When the stress residual is positive, the thickness correction, curvature correction, and roughness correction are subtracted from the updated thickness parameters, curvature parameters, and surface roughness parameters of the corresponding measurement point area, respectively. When the stress residual is negative, the thickness correction amount, curvature correction amount, and roughness correction amount are added to the updated thickness parameter, curvature parameter, and surface roughness parameter of the corresponding measuring point area, respectively. After correcting the geometric parameters of all measuring point areas, the corrected geometric structural parameters of the liner are obtained.

10. The method for predicting stress distribution in crusher liners based on multi-field coupling according to claim 1, characterized in that, Step seven specifically involves: Based on the back-corrected liner geometry parameters, the evolution modulation coefficients in the multi-field coupling model are recalculated to obtain an updated multi-field coupling model. In the updated multi-field coupling model, following the layer-by-layer propagation calculation process from the disturbance field and evolution field to the stress field, the corrected equivalent stress values ​​of the surface, middle and bottom layers of the corresponding volume elements in each measuring point area are obtained. The three corrected equivalent stress values ​​are summed again and averaged, and then mapped back to the corresponding position on the liner surface to obtain the final predicted stress distribution of the liner.

Images