A solar silicon wafer bearing basket real-time monitoring system based on an internet of things

Abstract

The application discloses a solar silicon wafer bearing basket real-time monitoring system based on an internet of things, and relates to the technical field of solar silicon wafer production equipment monitoring. The system comprises an identification filing module, which is used for configuring a unique identity for a bearing basket and establishing a full-life-cycle monitoring file corresponding to the bearing basket; and a process collection module, which is used for collecting process event data of the bearing basket in the processes of loading, carrying, buffering, cleaning and unloading, and writing the process event data into a corresponding process memory chain in chronological order. The application relates to the technical field of solar silicon wafer production equipment monitoring. The solar silicon wafer bearing basket real-time monitoring system based on the internet of things can realize fine identification of abnormal areas and slot-level position output by locally inverting parameters of each container groove area and comparing real-time bearing state parameters with dynamic risk judgment thresholds of corresponding areas point by point, and can discover local stiffness attenuation, partial load deformation and local instability hazards earlier compared with an integral basket-level coarse-grained judgment mode.

Description

Technical Field

[0001] This invention relates to the field of monitoring solar silicon wafer production equipment, and more specifically, to a real-time monitoring system for solar silicon wafer-carrying flower baskets based on the Internet of Things. Background Technology

[0002] In the production of solar silicon wafers, the load-bearing basket serves as a key turnover and load-bearing component, widely used in multiple processes such as loading, handling, cleaning, buffering, and unloading. Due to its long-term high-frequency use, it must withstand repeated mechanical impacts, vibration loads, and chemical corrosion during the cleaning process, which can easily lead to structural performance degradation, such as localized stiffness reduction, structural instability, or uneven load distribution.

[0003] In existing technologies, the management of flower baskets mainly relies on manual inspections or simple replacement strategies based on usage frequency. These methods have significant shortcomings: firstly, manual inspections are inherently delayed and subjective, making it difficult to detect early, hidden damage; secondly, experience-based or fixed-cycle replacement methods fail to reflect actual service conditions, easily leading to over-maintenance or safety hazards. Furthermore, traditional monitoring methods typically focus only on the overall structural condition, lacking detailed analysis of localized areas at the basket level, making it difficult to accurately locate anomalies and assess risk levels.

[0004] With the development of Industrial Internet of Things (IIoT) technology, although some devices have achieved data acquisition and remote monitoring, these are mostly focused on single-parameter monitoring, lacking the ability to fuse and analyze multi-source data, and failing to combine structural response characteristics with historical process data for comprehensive evaluation. This results in the accuracy of monitoring results and the ability to provide early warnings still need improvement. Therefore, there is an urgent need for an intelligent monitoring system that can integrate multi-process data, achieve structural state inversion analysis, and possess dynamic risk assessment capabilities. Summary of the Invention

[0005] The purpose of this invention is to provide a real-time monitoring system for solar silicon wafers carrying flower baskets based on the Internet of Things. This system solves the problem that existing real-time monitoring systems for solar silicon wafers carrying flower baskets are mostly focused on monitoring a single parameter, lack the ability to integrate and analyze multi-source data, and fail to combine structural response characteristics with historical process data for comprehensive evaluation. As a result, the accuracy of monitoring results and early warning capabilities still need to be improved, and they cannot meet the usage requirements.

[0006] This invention achieves the above objective through the following technical solution: a real-time monitoring system for flower baskets supported by solar silicon wafers based on the Internet of Things, the system comprising: The module includes: identification and filing module, process acquisition module, excitation application module, response acquisition module, parameter inversion module, damage calculation module, threshold generation module, early warning output module, and data update module. The identification and filing module is used to configure a unique identifier for the flower basket and to establish a full life cycle monitoring file corresponding to the flower basket. The process acquisition module is used to collect process event data of the flower basket during the loading, handling, buffering, cleaning and unloading stages, and write them into the corresponding process memory chain in chronological order. The excitation application module is used to apply a standardized micro-excitation signal to the target structural part of the flower basket within a preset detection window; The response acquisition module is used to acquire the structural dynamic response data of the flower basket under the action of the standardized micro-excitation signal; The parameter inversion module is used to perform inversion analysis on the dynamic response data of the structure based on a preset structural constraint model, and to obtain the local stiffness parameters, eccentric load parameters and instability trend parameters corresponding to each trough region. The damage calculation module is used to calculate the damage accumulation index and remaining safety margin parameter of the corresponding flower basket based on the historical process event data in the process memory chain. The threshold generation module is used to couple and analyze the local stiffness parameter, the off-center load parameter, the instability trend parameter, the damage accumulation index, and the remaining safety margin parameter to generate a dynamic risk judgment threshold corresponding to the current service status. The early warning output module is used to compare the real-time bearing status parameters of each tank area with the corresponding dynamic risk judgment threshold, and output the tank-level abnormal location, risk level and early warning result. The data update module is used to write back the monitoring results, early warning results, and handling results to the process memory chain to update the subsequent monitoring baseline of the flower basket.

[0007] Furthermore, the identification and documentation module includes: Identity encoding unit, parameter binding unit, baseline writing unit, and lifetime tracking unit; The identification coding unit is used to generate a unique identification code based on the batch information, specification information, and production information of the flower baskets. The parameter binding unit is used to establish a mapping relationship between the unique identification code and the material parameters, trough layout parameters, design cycle number, initial stiffness parameters, and design safety margin parameters of the flower basket. The baseline writing unit is used to write the baseline response data and baseline threshold data obtained from the factory inspection of the flower basket. The lifespan tracking unit is used to associate and store the monitoring records, early warning records, disposal records, and re-inspection records of the flower basket during its subsequent service life, using the unique identification code as an index.

[0008] Furthermore, the process acquisition module includes: Event recognition unit, timing arrangement unit, intensity quantization unit, and memory chain generation unit; The event recognition unit is used to identify process events generated during the loading impact, handling vibration, buffering and dwelling, cleaning corrosion and unloading and transfer of the flower basket; The timing arrangement unit is used to arrange the process events sequentially according to the event occurrence time, duration, and working position. The intensity quantification unit is used to extract the impact amplitude, vibration level, corrosion duration, and load change amplitude of each process event; The memory chain generation unit is used to write the occurrence frequency, location, level of action, and quantification results of different process events into the corresponding process memory chain carrying the flower basket.

[0009] Furthermore, the excitation application module includes a window triggering unit, an action positioning unit, and an excitation control unit, and the response acquisition module includes a vibration acquisition unit, a strain acquisition unit, a displacement acquisition unit, and a time synchronization unit; The window triggering unit is used to activate the detection window when the flower basket is in a preset static and stable state. The positioning unit is used to determine the target structural part corresponding to the trough area; The excitation control unit is used to apply a standardized micro-excitation signal with controlled amplitude and frequency band to the target structural part; The time synchronization unit is used to synchronize the time of the multi-source acquisition channels; The vibration acquisition unit, the strain acquisition unit, and the displacement acquisition unit are used to synchronously acquire the frequency response, strain response, and displacement response of the flower basket under micro-excitation.

[0010] Furthermore, the parameter inversion module includes: Constraint modeling unit, modality recognition unit, region decoupling unit, and parameter solving unit; The constraint modeling unit is used to construct a preset structural constraint model that matches the boundary conditions and trough distribution conditions of the load-bearing basket structure. The modal recognition unit is used to extract modal features characterizing the structural state of the trough region based on the frequency response, the strain response, and the displacement response; The region decoupling unit is used to separate the coupling response effects between adjacent reservoir regions; The parameter solving unit is used to combine the preset structural constraint model, the modal characteristics, and the decoupling results to solve the local stiffness parameters, eccentric load parameters, and instability trend parameters of each trough region.

[0011] Furthermore, the damage calculation module includes: Event weighting unit, cumulative calculation unit, margin assessment unit, and limit correction unit; The event weighting unit is used to assign damage weights to various process events based on the differences in the contribution of different process events to the damage of the load-bearing basket structure. The cumulative calculation unit is used to calculate the damage accumulation index based on the cumulative occurrence, impact level, quantification results, and damage weight of various process events; The margin assessment unit is used to calculate the remaining safety margin parameters based on the preset damage limit threshold and the damage accumulation index. The limit correction unit is used to correct the preset damage limit threshold based on the material decay information of the flower basket and the number of cycles.

[0012] Furthermore, the threshold generation module includes: State coupling unit, threshold correction unit, region mapping unit, and baseline correction unit; The state coupling unit is used to jointly characterize local stiffness parameters, off-center load parameters, instability trend parameters, damage accumulation index, and remaining safety margin parameters. The threshold correction unit is used to generate a dynamic risk judgment threshold that is adapted to the current service status based on the joint characterization results and benchmark response data, on the basis of the initial static risk threshold at the factory. The region mapping unit is used to allocate the dynamic risk determination threshold to the corresponding trough region; The baseline correction unit is used to call up the monitoring baseline formed in the previous monitoring cycle and continuously correct the dynamic risk judgment threshold of the current monitoring cycle.

[0013] Furthermore, the early warning output module includes: Slot-by-slot comparison unit, location positioning unit, level determination unit, and processing output unit; The slot-by-slot comparison unit is used to compare the real-time bearing status parameters of each slot area with the corresponding dynamic risk judgment threshold point by point. The location unit is used to determine the location of the abnormal area by combining the trough number and the target structural part number; The level determination unit is used to determine three risk levels—low risk, medium risk, and high risk—based on the extent and duration of the real-time bearing status parameters exceeding the dynamic risk determination threshold. The disposal output unit is used to output prompts, warnings, or emergency shutdown warnings according to the risk level, and send re-inspection instructions or isolation instructions to the corresponding workstations.

[0014] Furthermore, the preset structural constraint model is a structural inversion prediction model, which includes a feature extraction subunit, a temporal correlation subunit, a multi-task output subunit, and a training management subunit. The feature extraction subunit is used to extract local features from the structural dynamic response data; The time-series correlation subunit is used to establish response correlations between different sampling times; The multi-task output subunit is used to output local stiffness parameters, eccentric load parameters, and instability tendency parameters respectively; The training management subunit is used to train the structure inversion prediction model based on the multi-task joint loss function and output model parameters that meet the deployment conditions.

[0015] Furthermore, the parameter inversion module also includes: Online update unit, weight calibration unit, and deployment switching unit; The online update unit is used to incrementally update the structure inversion prediction model by calling newly labeled samples when the preset monitoring number condition or the preset early warning confirmation number condition is reached. The weight calibration unit is used to calibrate the task weights in the multi-task joint loss function based on the differences in the magnitude of loss of each task and the recall results of on-site fault samples. The deployment switching unit is used to switch the updated model to an online monitoring model when the updated model meets the accuracy threshold and false alarm threshold conditions, so as to maintain long-term inversion accuracy and early warning stability.

[0016] The beneficial effects of this invention are as follows: 1. By establishing a unique identifier, a full lifecycle monitoring file, and a process memory chain for each flower basket, different usage stages, different workstation processes, and different disposal results can be uniformly linked to the same monitoring object, overcoming the problems of scattered status data and discontinuous historical trajectories in existing technologies.

[0017] 2. By incorporating historical process event data such as material feeding impact, handling vibration, cleaning corrosion, cyclic load, and temperature alternation into damage accumulation analysis, and using them together with local stiffness parameters, off-center load parameters, and instability trend parameters to participate in risk assessment, we can avoid making one-sided judgments based solely on single real-time detection results, thereby improving the consistency between early warning results and actual service conditions.

[0018] 3. By performing local parameter inversion on each container area and comparing the real-time load-bearing state parameters with the dynamic risk judgment threshold of the corresponding area point by point, it is possible to achieve refined identification of abnormal areas and container-level position output. Compared with the whole basket level coarse-grained judgment method, it can detect local stiffness attenuation, off-center load deformation and local instability risks earlier.

[0019] 4. By coupling structural state parameters with damage accumulation index and remaining safety margin parameters, a dynamic risk judgment threshold adapted to the current service status is formed based on the initial threshold at the factory. This can avoid the judgment bias caused by applying fixed thresholds uniformly to new and old baskets or baskets with different damage levels, and improve the pertinence and stability of anomaly judgment.

[0020] 5. By writing back the monitoring results, early warning results, and handling results to the process memory chain and updating the monitoring baseline, the system can inherit historical experience and correct judgment criteria in subsequent monitoring processes, thereby improving the monitoring continuity, adaptability, and early warning reliability under long-term operating conditions. Attached Figure Description

[0021] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a system block diagram of the present invention; Figure 2 This is a flowchart of the parameter inversion module of the present invention; Figure 3 This is a flowchart of the damage calculation module of the present invention. Detailed Implementation

[0022] The present application will now be described in further detail with reference to the accompanying drawings. It should be noted that the following specific embodiments are only used to further illustrate the present application and should not be construed as limiting the scope of protection of the present application. Those skilled in the art can make some non-essential improvements and adjustments to the present application based on the above application content.

[0023] Example 1: Please see Figure 1-3 This invention provides a technical solution: a real-time monitoring system for flower baskets supported by solar silicon wafers based on the Internet of Things, the system comprising: The identification and documentation module is used to assign a unique identifier to the flower basket supporting the solar silicon wafer and to establish a corresponding full life cycle monitoring file. Among them, the unique identification code is a unique identification code assigned to each solar silicon wafer carrier basket to accurately distinguish different carrier baskets, similar to a person's ID number; the full life cycle monitoring file records all relevant information of the carrier basket from its use to its disposal, including data from all aspects such as production, use, maintenance, and testing. The process acquisition module is used to collect process event data during the loading, handling, buffering, cleaning and unloading of the flower basket, and write it into the corresponding process memory chain in chronological order; Among them, the process event data carries various data generated during the production processes of the flower basket, such as loading, handling, buffering, cleaning and unloading, including information such as process start time, end time, operators, and equipment status; the process memory chain records the chain structure of the flower basket in each process event data in chronological order, which can clearly present the experience and status changes of the flower basket in the entire production process. The excitation application module is used to apply standardized micro-excitation signals to the target structural parts carrying the flower basket within a preset detection window; The system includes: a preset detection window, which is a pre-defined time period for detecting the flower basket; a target structural part, which is the key structural part of the flower basket that needs to be detected, as these parts may have a significant impact on the performance and safety of the flower basket; and a standardized micro-excitation signal, which is a micro-excitation signal formulated according to a unified standard to stimulate the structure of the flower basket to produce a dynamic response, so as to collect relevant data for analysis. The strength and form of this signal are precisely designed and standardized. The response acquisition module is used to acquire dynamic response data of the structure carrying the flower basket under the action of micro-excitation signals; Among them, the structural dynamic response data is the dynamic change data of the structure of the flower basket after being subjected to micro-excitation signals, such as vibration frequency, amplitude, displacement, etc. These data reflect the current state and performance of the flower basket structure. The parameter inversion module is used to perform inversion analysis on the dynamic response data of the structure based on the preset structural constraint model, and to obtain the local stiffness parameters, eccentric load parameters and instability trend parameters of each trough region. Among them, the preset structural constraint model is a pre-established mathematical model of the load-bearing basket structure. This model considers factors such as the structural characteristics and material properties of the load-bearing basket and sets certain constraints to perform inversion analysis on the collected structural dynamic response data; the local stiffness parameter reflects the structural resistance to deformation of each compartment area of ​​the load-bearing basket. The greater the stiffness, the less likely the structure is to deform under stress; the eccentric load parameter describes the degree of uneven load distribution during the load-bearing process. Eccentric loading may cause excessive local stress on the load-bearing basket, affecting its service life and safety; the instability trend parameter is used to determine whether there is an instability trend in each compartment area of ​​the load-bearing basket, such as structural buckling. By analyzing this parameter, potential safety hazards can be detected in advance. The damage calculation module is used to calculate the cumulative damage index and remaining safety margin parameters of the corresponding carrying basket based on the historical process event data in the process memory chain. Among them, the damage accumulation index, calculated based on historical data of the load-bearing basket in each process, taking into account various factors such as load size, duration of action, and environmental conditions, is an indicator reflecting the degree of accumulated damage to the load-bearing basket; the higher the value, the more severe the damage. The remaining safety margin parameter... This represents the difference between the safe load that the flower basket can withstand in its current state and the actual load it may withstand. It is used to assess the safety of the flower basket for continued use. The larger the remaining safety margin, the safer the flower basket is. The threshold generation module is used to couple and analyze local stiffness parameters, off-center load parameters, and instability trend parameters with damage accumulation index and remaining safety margin parameters to generate dynamic risk judgment thresholds corresponding to the real-time service status of the current load-bearing basket. Among them, the dynamic risk judgment threshold is a threshold generated by coupling analysis based on local stiffness parameters, off-center load parameters, instability trend parameters, damage accumulation index and remaining safety margin parameters to judge the risk level of the current real-time service status of the load-bearing basket. When the real-time monitoring data exceeds the threshold, the load-bearing basket is considered to be at risk. The early warning output module is used to compare the real-time bearing status parameters of each tank area with the corresponding dynamic risk judgment threshold, and output the tank-level abnormal location, risk level and early warning result. The system includes: real-time load-bearing status parameters, which are the current status parameters of each compartment of the load-bearing basket collected in real time, such as actual load and deformation; compartment-level anomaly location, which is the specific compartment location with an anomaly determined by comparing it with the dynamic risk judgment threshold; risk level, which is the classification of the risk degree of the abnormal situation of the load-bearing basket based on the comparison result of the real-time load-bearing status parameters and the dynamic risk judgment threshold, which can generally be divided into different levels such as low risk, medium risk, and high risk; and early warning result, which is an early warning message about the current status of the load-bearing basket output by combining information such as compartment-level anomaly location and risk level, to remind relevant personnel to take appropriate measures. The data update module is used to write back the monitoring results, early warning results, and handling results to the process memory chain to update the subsequent monitoring baseline for the flower basket. The monitoring results include all data obtained from real-time monitoring of the flower basket, including structural dynamic response data and parameter inversion results; early warning results include early warning information about the current status of the flower basket generated by the early warning output module; handling results include feedback on the handling of corresponding measures after the early warning results, such as whether repairs or replacements were performed and the effects of the operations; and the monitoring baseline is a reference standard used for subsequent monitoring of the flower basket's status. By writing the monitoring results, early warning results, and handling results back to the process memory chain, the monitoring baseline can be updated, making subsequent monitoring more accurate and effective.

[0024] It should be noted that during use, the identification and filing module assigns a unique identifier to the flower basket and files it for easy tracking and management. The process acquisition module records process data over time, clearly presenting the production process of the flower basket. The excitation application and response acquisition module works together to obtain dynamic response data of the structure, providing a basis for analysis. The parameter inversion module can obtain key parameters and understand the status of each compartment of the flower basket. The damage calculation module can assess the damage and safety margin of the flower basket. The threshold generation module generates dynamic risk judgment thresholds to achieve accurate early warning. The early warning output module outputs abnormal information in a timely manner to ensure production safety. The data update module writes back the results and continuously optimizes the monitoring baseline, making subsequent monitoring more accurate and effective. The overall design realizes real-time and accurate monitoring of the entire life cycle of the flower basket, improving production quality and safety.

[0025] In one embodiment, the dynamic response data of the structure is inverted based on a preset structural constraint model to obtain the local stiffness parameters, eccentric load parameters, and instability trend parameters of each trough region, specifically including: Vibration frequency, displacement amplitude, and strain response data of the flower basket under standardized micro-excitation are collected using IoT vibration sensors and strain sensors. This data is then input into a preset structural constraint model, and combined with modal parameter identification and finite element iterative inversion calculations, the local stiffness parameters of each trough region are obtained. The calculation formula is: in, For the first The standardized micro-excitation force experienced by each reservoir region For the first Dynamic displacement response of a trough region under excitation; Based on local stiffness parameters, the eccentric loading parameters characterizing the differences in stiffness uniformity in the trough region are calculated. The calculation formula is: in, This is the arithmetic mean of the local stiffness of all the trough areas that support the flower basket; By combining the initial stiffness of the load-bearing basket at the factory with the current measured stiffness, the instability trend parameter reflecting the overall deterioration trend of the structure is calculated. The calculation formula is: in, The initial local stiffness of the flower basket at the time of manufacture is determined by the factory inspection test calibration. The total number of areas that can hold flower baskets.

[0026] This design, through the collection of dynamic response data by multiple sensors and the input of the data into a preset model combined with specific calculations to obtain parameters, can accurately acquire key parameters of each compartment area. The comprehensive data collection using multiple sensors ensures complete and accurate information. The preset structural constraint model, combined with modal parameter identification and finite element iterative inversion calculation, makes the calculation results scientific and reliable. Through these parameters, the local stiffness, stiffness uniformity differences, and overall degradation trends of each compartment in the basket can be clearly understood, providing a solid foundation for subsequent evaluation of the basket's condition. This helps to identify potential problems in a timely manner, ensures the stability and reliability of the basket during the production process, and improves the production quality of solar silicon wafers.

[0027] In one embodiment, the damage accumulation index and remaining safety margin parameters are calculated based on historical process event data from the process memory chain. The specific calculation process is as follows: Historical process event data, including material loading impact, handling vibration, cleaning corrosion, cyclic load, and temperature alternation, are extracted from the process memory chain. The cumulative occurrence frequency of each type of process event is calculated. Combined with the damage weighting coefficient of different processes on the basket structure, the cumulative damage index of the basket is calculated. The calculation formula is: in, For the first The damage weight coefficients for process-related events are determined by the following rules: based on finite element fatigue simulation, accelerated aging experiments, and long-term on-site operation and maintenance data statistics, the process that causes the greatest damage to the basket structure is assigned the highest weight. For the remaining processes, values ​​are normalized based on the actual damage percentage, with cleaning corrosion and cyclic load weights determined accordingly. Vibration weighting during transport Impact of material feeding on weight ; For the first The cumulative number of events occurring in the same process step This represents the total number of process event types. Based on the damage accumulation index and the preset damage limit threshold, the remaining safety margin parameter reflecting the safety reserve capacity of the basket structure is calculated. The calculation formula is: in, To withstand the maximum pre-set damage limit threshold allowed within the design life of the flower basket, the threshold is determined based on: the fatigue strength of the flower basket material, the design cycle life, and industry safety standards, taking the maximum allowable damage value at the end of the design life. .

[0028] This design extracts data from the process memory chain and calculates the damage accumulation index and remaining safety margin parameters by combining weighting coefficients. The process memory chain records process events throughout the entire life cycle of the flower basket, providing comprehensive data. By statistically analyzing the cumulative number of various process events and combining them with weighting coefficients to calculate the damage accumulation index, the impact of different processes on the flower basket can be comprehensively considered, resulting in more accurate results. The remaining safety margin parameters calculated based on the damage accumulation index can intuitively reflect the safety reserve capacity of the flower basket structure. This helps to predict the lifespan of the flower basket in advance, rationally arrange maintenance and replacement, avoid production disruptions due to flower basket damage, reduce production costs, and improve production efficiency and safety.

[0029] In one embodiment, performing coupling analysis to generate a dynamic risk assessment threshold specifically includes: By weighted coupling of structural state parameters such as local stiffness, eccentric loading, and instability tendency with service state parameters such as damage accumulation index and remaining safety margin, the limitations of single-parameter judgment are eliminated, and a dynamic risk assessment threshold adapted to the current real-time service status of the flower basket is generated. The calculation formula is: in, To support the initial static risk threshold set at the factory for the flower basket, the threshold is determined based on the average measured response of the new flower basket structure, plus an additional value. Safety margin determined; The instability trend weighting coefficient. The cumulative damage weighting coefficients are determined using the following rules: The Analytic Hierarchy Process (AHP) is used in conjunction with field fault samples for calibration, with instability trend weights as the basis. Damage cumulative weight The sum of the weights is no greater than 1.

[0030] This design uses a weighted coupling of structural and service status parameters to generate a dynamic risk assessment threshold, eliminating the limitations of single-parameter assessment. Different parameters reflect the condition of the flower basket from different perspectives, and a single parameter is difficult to comprehensively and accurately assess the risk. The weighted coupling comprehensively considers multiple factors, making the generated dynamic risk assessment threshold more suitable for the real-time service status of the flower basket. Based on this threshold, the risk of the flower basket can be judged more accurately, and timely measures can be taken to prevent accidents. At the same time, the weight determination rules are scientific and reasonable, ensuring the accuracy and reliability of the threshold generation, effectively improving the accuracy of the monitoring system's assessment of the flower basket's risks, and ensuring production safety.

[0031] In one embodiment, the output of the tank-level anomaly location, risk level, and early warning result includes: The load-bearing status parameters collected in real time for each tank area are compared point by point with the corresponding dynamic risk assessment threshold. When the real-time load-bearing status parameter is greater than the dynamic risk assessment threshold, the load-bearing status parameter is considered higher than the dynamic risk assessment threshold. At that time, it was determined that there was a structural anomaly in the container area; Risk levels are determined based on the ratio of real-time parameters to dynamic risk assessment thresholds. The threshold for grading is determined based on: safety specifications for silicon wafer manufacturing equipment, statistics on basket failure cases, and equipment downtime risk level standards. The grading rules are as follows:

[0032] It outputs alerts, warnings, and emergency shutdown warnings according to low, medium, and high risk levels, respectively.

[0033] This design, by comparing real-time parameters with thresholds to determine anomalies, classifies risk levels, and outputs early warning results, can accurately locate abnormal flower baskets, promptly identify the root cause of problems, and classify risk levels based on ratios. The rules are determined by combining multiple standards, making it scientific and reasonable. It can accurately assess the degree of risk and output corresponding early warning results according to different risk levels, allowing staff to quickly understand the severity of the situation and take corresponding measures. Low-risk warnings can remind people to pay attention, medium-risk alarms require preparation for handling, and high-risk emergency shutdown warnings can immediately stop production to avoid losses, effectively ensuring production safety and improving production efficiency and product quality.

[0034] In one embodiment, the preset structural constraint model is a deep learning-based structural inversion prediction model, and the specific steps for model training are as follows: Dataset construction: Collect no less than 5,000 sets of dynamic response data of the load-bearing basket structure as the training set. The data includes typical working condition samples such as new state, normal wear, local crack, eccentric load deformation, and corrosion failure. The data is divided into training set, validation set and test set in a ratio of 7:2:1. The sample labels include local stiffness, eccentric load, instability trend, damage index and risk level true value. The model architecture and initialization employ a hybrid network architecture combining one-dimensional convolutional layers and bidirectional LSTM. The input layer dimension is the number of sensor channels × time sequence length. Three convolutional layers are set with kernel sizes of 3, 5, and 7, respectively, and ReLU is used as the activation function. The LSTM hidden layer dimensions are 128 / 64 / 32, with a dropout rate of 0.2. The output layer corresponds to three parameter branches: local stiffness, off-center loading, and instability tendency, respectively, and uses linear activation. Training hyperparameter settings, AdamW optimizer selected, initial learning rate value The cosine annealing decay is used, with a batch size of 32, a maximum number of training epochs of 100, and an early stopping strategy of patience of 10. The loss function is constructed using a joint loss function for multiple tasks.

[0035] in, , , These are the mean square error losses for stiffness, eccentric loading, and instability trend, respectively. Model training and validation: forward propagation calculates the predicted values ​​and loss, backpropagation updates the network parameters, and the mean absolute error is evaluated on the validation set after each training round. With the coefficient of determination ,when Training should be stopped if the validation set loss does not decrease for 10 consecutive rounds. Model testing and deployment: Verify the model's generalization ability on the test set, requiring a mean absolute percentage error. Once the accuracy requirements are met, the model weights are solidified and deployed to IoT edge computing nodes for real-time inversion calculations.

[0036] This design employs a deep learning structure inversion prediction model with pre-defined structural constraints and detailed training steps. The dataset construction covers various typical working conditions, ensuring the model can learn rich features and improve generalization ability. The reasonable model architecture and initialization, combined with one-dimensional convolution and bidirectional LSTM, effectively extract data features. The scientifically set training hyperparameters help the model converge quickly. The multi-task joint loss function comprehensively considers all parameters, improving model accuracy. The rigorous model training, verification, testing, and deployment processes ensure that the model is put into use after meeting the accuracy standards. It can perform real-time and accurate inversion calculations, providing reliable support for monitoring the status of flower baskets and improving the level of intelligent production.

[0037] In one embodiment, the weights of the multi-task joint loss function , , The rules are defined, specifically including: The priority determination rule takes the degree of impact on structural safety as the core priority. Local stiffness directly determines the load-bearing capacity and silicon wafer safety, so it has the highest priority. Instability trend directly reflects the risk of structural failure, so it has the second highest priority. Off-center loading reflects the difference in uniformity, so it has a general priority. Numerical normalization rules are used, and the importance of each output parameter is scored using the Analytic Hierarchy Process (AHP) combined with the Fault Tree Analysis (FTA). The importance score of local stiffness is given as a percentage. Off-center load ratio The proportion of unstable trends The weights are obtained after normalization. , , And satisfy ; The dynamic calibration rule uses fixed weights in the early stage of model training, and adaptively scales the weights according to the difference in the loss magnitude of each task in the validation set in the later stage of training, so that the loss values ​​of the three tasks are on the same order of magnitude, and avoids the dominant gradient update of one task due to the difference in the magnitude. The engineering verification rules use the recall rate of failure samples such as field cracks, eccentric loading, and instability as the verification indicator. When the failure recall rate... And false alarm rate At that time, the current loss weight is locked as the final deployment weight.

[0038] This design employs clear rules for determining the weights of the multi-task joint loss function. The priority labeling rule focuses on the degree of impact on structural safety, ensuring sufficient attention is paid to key parameters of basket safety. The numerical normalization rule uses scientific methods to score and normalize, ensuring reasonable weight allocation. The dynamic calibration rule avoids problems caused by differences in units during training, guaranteeing collaborative optimization among tasks. The engineering verification rule uses the recall rate and false alarm rate of actual fault samples as indicators, ensuring the model's effectiveness in practical applications. These rules work together to ensure that the loss function weights are scientifically and reasonably determined, improving model training effectiveness and prediction accuracy, and better serving basket status monitoring.

[0039] In one embodiment, the model further includes an online incremental update step, where labeled samples are automatically extracted for incremental training every 1000 monitoring cycles or after 5 or more warning confirmation events, with the learning rate set to [value missing]. Freeze the bottom convolutional layers and only update the parameters of the LSTM and fully connected layers to maintain the long-term prediction accuracy of the model.

[0040] This design incorporates an online incremental update step. As the number of monitoring sessions increases or warning confirmation events occur, the flower basket status data continuously changes. Regular incremental training allows the model to learn new data features, freezes the underlying convolutional layers, and only updates the LSTM and fully connected layer parameters. This ensures the stability of the model's existing features while adapting to new data changes, maintaining long-term prediction accuracy, and preventing performance degradation due to data changes. It ensures that the model can always accurately invert and calculate flower basket parameters, providing a reliable basis for real-time monitoring, effectively improving the monitoring system's adaptability to changes in flower basket status, and ensuring stable production.

[0041] Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0042] The above embodiments provide a detailed description of the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A real-time monitoring system for flower baskets supported by solar silicon wafers based on the Internet of Things, characterized in that, The system includes: The module includes: identification and filing module, process acquisition module, excitation application module, response acquisition module, parameter inversion module, damage calculation module, threshold generation module, early warning output module, and data update module. The identification and filing module is used to configure a unique identifier for the flower basket and to establish a full life cycle monitoring file corresponding to the flower basket. The process acquisition module is used to collect process event data of the flower basket during the loading, handling, buffering, cleaning and unloading stages, and write them into the corresponding process memory chain in chronological order. The excitation application module is used to apply a standardized micro-excitation signal to the target structural part of the flower basket within a preset detection window; The response acquisition module is used to acquire the structural dynamic response data of the flower basket under the action of the standardized micro-excitation signal; The parameter inversion module is used to perform inversion analysis on the dynamic response data of the structure based on a preset structural constraint model, and to obtain the local stiffness parameters, eccentric load parameters and instability trend parameters corresponding to each trough region. The damage calculation module is used to calculate the damage accumulation index and remaining safety margin parameter of the corresponding flower basket based on the historical process event data in the process memory chain. The threshold generation module is used to couple and analyze the local stiffness parameter, the off-center load parameter, the instability trend parameter, the damage accumulation index, and the remaining safety margin parameter to generate a dynamic risk judgment threshold corresponding to the current service status. The early warning output module is used to compare the real-time bearing status parameters of each tank area with the corresponding dynamic risk judgment threshold, and output the tank-level abnormal location, risk level and early warning result. The data update module is used to write back the monitoring results, early warning results, and handling results to the process memory chain to update the subsequent monitoring baseline of the flower basket.

2. The real-time monitoring system for flower baskets supported by solar silicon wafers based on the Internet of Things as described in claim 1, characterized in that, The identification and filing module includes: Identity encoding unit, parameter binding unit, baseline writing unit, and lifetime tracking unit; The identification coding unit is used to generate a unique identification code based on the batch information, specification information, and production information of the flower baskets. The parameter binding unit is used to establish a mapping relationship between the unique identification code and the material parameters, trough layout parameters, design cycle number, initial stiffness parameters, and design safety margin parameters of the flower basket. The baseline writing unit is used to write the baseline response data and baseline threshold data obtained from the factory inspection of the flower basket. The lifespan tracking unit is used to associate and store the monitoring records, early warning records, handling records, and re-inspection records of the flower basket during its subsequent service life, using the unique identification code as an index.

3. The real-time monitoring system for flower baskets supported by solar silicon wafers based on the Internet of Things as described in claim 1, characterized in that, The process acquisition module includes: Event recognition unit, timing arrangement unit, intensity quantization unit, and memory chain generation unit; The event recognition unit is used to identify process events generated during the loading impact, handling vibration, buffering and dwelling, cleaning corrosion and unloading and transfer of the flower basket; The timing arrangement unit is used to arrange the process events sequentially according to the event occurrence time, duration, and working position. The intensity quantification unit is used to extract the impact amplitude, vibration level, corrosion duration, and load change amplitude of each process event; The memory chain generation unit is used to write the occurrence frequency, location, level of action, and quantification results of different process events into the corresponding process memory chain carrying the flower basket.

4. The real-time monitoring system for flower baskets supported by solar silicon wafers based on the Internet of Things as described in claim 1, characterized in that: The excitation application module includes a window triggering unit, an action positioning unit, and an excitation control unit; the response acquisition module includes a vibration acquisition unit, a strain acquisition unit, a displacement acquisition unit, and a time synchronization unit. The window triggering unit is used to activate the detection window when the flower basket is in a preset static and stable state. The positioning unit is used to determine the target structural part corresponding to the trough area; The excitation control unit is used to apply a standardized micro-excitation signal with controlled amplitude and frequency band to the target structural part; The time synchronization unit is used to synchronize the time of the multi-source acquisition channels; The vibration acquisition unit, the strain acquisition unit, and the displacement acquisition unit are used to synchronously acquire the frequency response, strain response, and displacement response of the flower basket under micro-excitation.

5. The real-time monitoring system for flower baskets supported by solar silicon wafers based on the Internet of Things as described in claim 4, characterized in that, The parameter inversion module includes: Constraint modeling unit, modality recognition unit, region decoupling unit, and parameter solving unit; The constraint modeling unit is used to construct a preset structural constraint model that matches the boundary conditions and trough distribution conditions of the load-bearing basket structure. The modal recognition unit is used to extract modal features characterizing the structural state of the trough region based on the frequency response, the strain response, and the displacement response; The region decoupling unit is used to separate the coupling response effects between adjacent reservoir regions; The parameter solving unit is used to combine the preset structural constraint model, the modal characteristics, and the decoupling results to solve the local stiffness parameters, eccentric load parameters, and instability trend parameters of each trough region.

6. The real-time monitoring system for flower baskets supported by solar silicon wafers based on the Internet of Things as described in claim 3, characterized in that, The damage calculation module includes: Event weighting unit, cumulative calculation unit, margin assessment unit, and limit correction unit; The event weighting unit is used to assign damage weights to various process events based on the differences in the contribution of different process events to the damage of the load-bearing basket structure. The cumulative calculation unit is used to calculate the damage accumulation index based on the cumulative occurrence, impact level, quantification results, and damage weight of various process events; The margin assessment unit is used to calculate the remaining safety margin parameters based on the preset damage limit threshold and the damage accumulation index. The limit correction unit is used to correct the preset damage limit threshold based on the material decay information of the flower basket and the number of cycles.

7. The real-time monitoring system for flower baskets supported by solar silicon wafers based on the Internet of Things as described in claim 6, characterized in that, The threshold generation module includes: State coupling unit, threshold correction unit, region mapping unit, and baseline correction unit; The state coupling unit is used to jointly characterize local stiffness parameters, off-center load parameters, instability trend parameters, damage accumulation index, and remaining safety margin parameters. The threshold correction unit is used to generate a dynamic risk judgment threshold that is adapted to the current service status based on the joint characterization results and benchmark response data, on the basis of the initial static risk threshold at the factory. The region mapping unit is used to allocate the dynamic risk determination threshold to the corresponding trough region; The baseline correction unit is used to call up the monitoring baseline formed in the previous monitoring cycle and continuously correct the dynamic risk judgment threshold of the current monitoring cycle.

8. The real-time monitoring system for flower baskets supported by solar silicon wafers based on the Internet of Things as described in claim 7, characterized in that, The early warning output module includes: Slot-by-slot comparison unit, location positioning unit, level determination unit, and processing output unit; The slot-by-slot comparison unit is used to compare the real-time bearing status parameters of each slot area with the corresponding dynamic risk judgment threshold point by point. The location unit is used to determine the location of the abnormal area by combining the trough number and the target structural part number; The level determination unit is used to determine three risk levels—low risk, medium risk, and high risk—based on the extent and duration of the real-time bearing status parameters exceeding the dynamic risk determination threshold. The disposal output unit is used to output prompts, warnings, or emergency shutdown warnings according to the risk level, and send re-inspection instructions or isolation instructions to the corresponding workstations.

9. The real-time monitoring system for flower baskets supported by solar silicon wafers based on the Internet of Things as described in claim 5, characterized in that: The preset structural constraint model is a structural inversion prediction model, which includes a feature extraction subunit, a temporal correlation subunit, a multi-task output subunit, and a training management subunit. The feature extraction subunit is used to extract local features from the structural dynamic response data; The time-series correlation subunit is used to establish response correlations between different sampling times; The multi-task output subunit is used to output local stiffness parameters, eccentric load parameters, and instability tendency parameters respectively; The training management subunit is used to train the structure inversion prediction model based on the multi-task joint loss function and output model parameters that meet the deployment conditions.

10. The real-time monitoring system for flower baskets supported by solar silicon wafers based on the Internet of Things as described in claim 9, characterized in that, The parameter inversion module also includes: Online update unit, weight calibration unit, and deployment switching unit; The online update unit is used to incrementally update the structure inversion prediction model by calling newly labeled samples when the preset monitoring number condition or the preset early warning confirmation number condition is reached. The weight calibration unit is used to calibrate the task weights in the multi-task joint loss function based on the differences in the magnitude of loss of each task and the recall results of on-site fault samples. The deployment switching unit is used to switch the updated model to an online monitoring model when the updated model meets the accuracy threshold and false alarm threshold conditions, so as to maintain long-term inversion accuracy and early warning stability.

Images