Warehouse logistics visual monitoring system based on digital twinning

Abstract

The application discloses a warehouse logistics visual monitoring system based on digital twinning, belongs to the technical field of warehouse logistics fault monitoring, and specifically comprises the following steps: collecting various data in a warehouse environment, including warehouse architecture, article position, inventory information, equipment state data and personnel activity, and performing preprocessing; based on the various data in the preprocessed warehouse environment, a hierarchical digital twinning model is constructed; a virtual-real space real-time mapping system is constructed; the operation of warehouse equipment is simulated and predicted by using the digital twinning model; warehouse logistics is monitored in multiple dimensions; and a fault prediction model is constructed to predict and analyze the faults of warehouse equipment; the application uses Kalman filtering and an optimization algorithm to perform virtual-real compensation optimization, improves the accuracy of the digital twinning model, can identify potential fault risks in advance, issues early warnings in time, greatly reduces the equipment fault rate, reduces maintenance costs, and improves the normal operation time of equipment.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of warehouse logistics fault monitoring, and in particular relates to a warehouse logistics visual monitoring system based on digital twinning. BACKGROUND

[0002] With the rapid development of the logistics industry, modern warehouse management has become increasingly complex, especially in large-scale and intelligent warehouses. How to efficiently and accurately manage goods and equipment has become a problem that needs to be solved. Traditional warehouse logistics monitoring methods mostly rely on manual management, experience judgment, etc., and have certain limitations. These methods can usually only achieve real-time monitoring of a single device or a small amount of data in the warehouse, and cannot comprehensively and accurately monitor and optimize the entire warehouse logistics system.

[0003] Digital twinning, as a technology based on the mutual mapping of the physical world and virtual models, can map the physical space, equipment operation, goods status, environmental conditions, and other factors in the warehouse to the virtual world in real time in the field of warehouse logistics. Digital twinning technology can form a dynamic, comprehensive, and real-time virtual model. This virtual model not only reflects the real-time status of the actual warehouse environment, but also can be simulated and predicted to further optimize warehouse operations. However, there are still some defects: in various equipment and warehouse logistics scenarios, how to ensure the accuracy of the digital twinning model and the accuracy of virtual-real data transmission is still a technical problem.

[0004] Patent application No. CN119005858A discloses a material warehouse management system based on digital twinning technology, which includes a data acquisition and monitoring module, a digital twinning warehouse model module, an inventory management module, a logistics and transportation management module, an order and demand forecasting module, a warehouse operation management module, a safety and risk management module, and a data analysis and decision support module. The data acquisition and monitoring module collects and monitors warehouse environment and material state data in real time, the digital twinning warehouse model module dynamically simulates the actual situation of the warehouse, the inventory management module manages the records, tracking, classification, and automatic replenishment of inventory, the logistics and transportation management module manages the in-out warehouse process of materials, the order and demand forecasting module processes orders and performs demand analysis, the warehouse operation management module manages personnel and equipment, the safety and risk management module monitors warehouse safety, and the data analysis and decision support module provides optimization decision support. This technical solution improves the work efficiency, safety, and accuracy of material warehouse.

[0005] The patent application with the publication number CN118229182A discloses a warehouse logistics workshop monitoring system and method based on digital twin technology, which comprises a user login module, a 3D interaction module, a warning module, a roaming module, a video monitoring module, a digital workshop module, a device management module, a production efficiency module, a inventory management module and a data analysis dashboard module; a warehouse logistics workshop digital twin is constructed by means of digital twin technology, the real-time synchronization of the device state and the operation state is realized in the digital twin, and the device state and the operation state are one-to-one mapped, by means of data acquisition and data analysis, the visualization of the workshop operation data is realized, the production is transparent, the data fusion and convergence across systems are realized, the interconnection of each business link is realized, the full-factor expression of the workshop and the global visualization are realized, so that the real-time state of the workshop is better monitored, and the production efficiency is improved.

[0006] The above technical solution has the following defects: the data transmission between the virtual and real spaces may have certain errors. SUMMARY

[0007] In view of the deficiencies of the prior art, the present application provides a warehouse logistics visual monitoring system based on digital twin, which realizes the all-around, real-time and dynamic visual monitoring of the warehouse logistics environment by combining digital twin, Internet of Things, data analysis and artificial intelligence and other advanced technologies to compensate for the data transmission between the virtual and real spaces, and improves the warehouse equipment fault prediction accuracy.

[0008] To achieve the above object, the present application provides the following technical solution:

[0009] The warehouse logistics visual monitoring system based on digital twin comprises a data acquisition module, a digital twin construction module, a virtual-real mapping module and a prediction analysis module.

[0010] The data acquisition module is used to acquire various data in the warehouse environment, including warehouse architecture, article position, inventory information, device state data and personnel activity, and perform preprocessing.

[0011] The digital twin construction module constructs a hierarchical digital twin model based on the preprocessed various data in the warehouse environment.

[0012] The virtual-real mapping module is used to construct a virtual-real space real-time mapping system, and simulate and predict the operation of the warehouse equipment by using the digital twin model.

[0013] The prediction analysis module is used to monitor the warehouse logistics in multiple dimensions, and construct a fault prediction model to predict and analyze the fault of the warehouse equipment.

[0014] Specifically, the digital twin construction module comprises a first-layer digital twin model construction unit, a second-layer digital twin model construction unit and a third-layer digital twin model construction unit.

[0015] The first-layer digital twin model construction unit is configured to construct a device-level digital twin body, model parameters of the device in different operating states, and adjust based on data driving to feed back the operating state of the device in real time.

[0016] The second-layer digital twin model construction unit is configured to construct a storage location-level digital twin body, and assign different color heat values to different storage locations according to the residence time and turnover frequency of the goods in the storage locations.

[0017] The third-layer digital twin model construction unit is configured to construct a system-level digital twin body to optimize the overall logistics path planning in the warehouse.

[0018] Specifically, the virtual-real mapping module comprises a virtual-real mapping unit and a state prediction unit.

[0019] The virtual-real mapping unit is configured to synchronize data between the warehouse equipment and the digital twin model through Internet of Things technology, and perform virtual-real deviation compensation.

[0020] The state prediction unit is configured to predict the operating state of the warehouse equipment.

[0021] Specifically, the virtual-real deviation compensation comprises:

[0022] The Kalman filter is adopted to dynamically filter the virtual-real space data to obtain preliminary correction data for correcting the deviation of the virtual-real space data.

[0023] The preliminary correction data is taken as an initial population, the fitness function is defined as the weighted sum of the state deviation of the warehouse equipment and the path planning efficiency, the crossover individuals are selected according to the fitness, and the elite individuals are dynamically adjusted, when the warehouse equipment is in the starting stage and the state fluctuation rate is greater than 15%, the top n% individuals are reserved as the crossover individuals, when the warehouse equipment is in the stable stage and the state fluctuation rate is less than 5%, m% of the crossover individuals are reserved, and m>n is satisfied.

[0024] The adaptive crossover iteration is adopted, when the warehouse equipment is in the starting stage and the state fluctuation rate is greater than 15%, the crossover probability is adjusted to 0.8, when the warehouse equipment is in the stable stage and the state fluctuation rate is less than 5%, the crossover probability is adjusted to 0.3, and if the optimal solution is not updated for 20 iterations or the fitness variance is less than a set threshold for 50 consecutive iterations, part of the population is reset.

[0025] Randomly select 30% of the non-crossed individuals, and randomly initialize or generate new individuals based on historical data for the selected non-crossed individuals, and perform crossover and mutation operations on the randomly initialized individuals or generated new individuals;

[0026] When convergence or the maximum evolution number is reached, output the optimal bias correction scheme.

[0027] Specifically, the virtual-real space data is data of interaction between the warehouse and the goods and the digital twin model.

[0028] Specifically, the running state of the warehouse equipment is predicted, including:

[0029] According to the pre-processed equipment state data and the first-layer digital twin model, a warehouse equipment real-time state equation is established;

[0030] According to the state change rate of the warehouse equipment within △t, the real-time state of the warehouse equipment is updated;

[0031] The running state of the warehouse equipment is simulated in real time on the cloud server by using the digital twin model, and the future state of the warehouse equipment is predicted by inputting the current state and input signals.

[0032] Specifically, the prediction analysis module includes a fault prediction model construction and training unit and a prediction analysis unit.

[0033] The fault prediction model construction and training unit is configured to construct a fault prediction model and perform training.

[0034] The prediction analysis unit uses the trained fault prediction model to predict and analyze the fault of the warehouse equipment.

[0035] Specifically, the trained fault prediction model is used to predict and analyze the fault of the warehouse equipment, including:

[0036] A machine learning algorithm is selected to construct a fault prediction model, and a data set D train The fault prediction model is trained, and the training target of the model is set to minimize the loss function.

[0037] The fault prediction model is trained until the minimum loss function of the training target of the model converges and remains unchanged, and the training is stopped to obtain the trained fault prediction model.

[0038] Specifically, the trained fault prediction model is used to predict and analyze the fault of the warehouse equipment, including:

[0039] Key features are extracted from the historical data set D of the warehouse equipment, including statistical features and frequency domain features, and a feature set F new, F new = {Ft1, Ft2,..., Ft m}, Ft m represents the mth feature with the most discriminative ability for fault judgment;

[0040] The extracted feature set F new is input into the trained warehouse equipment fault identification model to obtain a warehouse equipment fault prediction probability.

[0041] According to actual requirements, a fault warning threshold is set, if the warehouse equipment fault prediction probability is greater than the threshold, an alarm is triggered, otherwise, no alarm is triggered.

[0042] Compared with the prior art, the beneficial effects of the present application are:

[0043] 1. The present application proposes a warehouse logistics visual monitoring system based on digital twinning, by constructing a digital twinning model of the warehouse logistics system, using Kalman filtering and optimization algorithm for virtual-real compensation optimization, which can monitor each link in the warehouse in real time, including goods storage, equipment operation, logistics path and other information, improving the accuracy of the digital twinning model.

[0044] 2. The present application proposes a warehouse logistics visual monitoring system based on digital twinning, based on real-time collected data, combining digital twinning model and artificial intelligence algorithm, which can intelligently optimize each operation in the warehouse logistics system, including: scheduling and path planning of equipment, real-time tracking of inventory situation and dynamic adjustment of goods storage position, etc., so as to reduce the redundancy of transportation path, optimize resource allocation and improve logistics efficiency.

[0045] 3. The present application proposes a warehouse logistics visual monitoring system based on digital twinning, through continuous monitoring and analysis of equipment operation data, the digital twinning model can realize real-time evaluation and prediction of equipment state, combined with big data analysis and machine learning algorithm, which can identify potential fault risk in advance and issue warning in time, guide warehouse managers to carry out maintenance and replacement operation, greatly reduce equipment failure rate, reduce maintenance cost and improve equipment uptime. BRIEF DESCRIPTION OF DRAWINGS

[0046] Figure 1 The present application provides a warehouse logistics visual monitoring system architecture based on digital twinning;

[0047] Figure 2 The present application provides a flowchart for bias correction of virtual-real space data. DETAILED DESCRIPTION

[0048] The application will be described in detail below with specific examples. The following examples will help those skilled in the art to further understand the application, but do not limit the application in any form. It should be noted that, for those skilled in the art, without departing from the concept of the application, a number of modifications and improvements can be made. These are within the scope of protection of the application.

[0049] In order to make the purpose, technical scheme and advantages of the application more clear, the application will be further described in detail below in combination with the drawings and examples. It should be understood that the specific examples described here are only used to explain the application and do not limit the application.

[0050] It should be noted that, if there is no conflict, each feature in the embodiments of the application can be combined with each other, and all within the scope of protection of the application. In addition, although the functional modules are divided in the device schematic diagram, and the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than the module division in the device or the order in the flowchart. In addition, the "first", "second", "third" and the like used in the application do not limit the data and execution order, but only distinguish the same items or similar items with basically the same function and effect.

[0051] Unless otherwise defined, all technical and scientific terms used in the specification have the same meaning as understood by those skilled in the art to which the application belongs. The terms used in the specification of the application are only for the purpose of describing the specific embodiments and are not used to limit the application. The term "and / or" used in the specification includes any and all combinations of one or more related listed items.

[0052] Example 1

[0053] Please refer to Figure 1 、 Figure 2 , the application provides an embodiment: a warehouse logistics visual monitoring system based on digital twinning, comprising: a data acquisition module, a digital twinning construction module, a virtual-real mapping module and a prediction analysis module;

[0054] The data acquisition module is used to acquire various data in the warehouse environment, including warehouse architecture, article location, inventory information, device state data and personnel activity, etc., and to preprocess;

[0055] Various data in the warehouse environment are collected by means of high-precision surveying and mapping unmanned aerial vehicles, laser radars, sensors, RFID, cameras and other equipment;

[0056] The preprocessing includes denoising, cleaning the data, removing abnormal data and redundant data, denoising the pictures, and removing noise points in the images.

[0057] The digital twin construction module constructs a hierarchical digital twin model based on the pre-processed data in the warehouse environment.

[0058] The digital twin construction module includes a first-level digital twin model construction unit, a second-level digital twin model construction unit, and a third-level digital twin model construction unit.

[0059] The first-level digital twin model construction unit is used to construct a device-level digital twin, model parameters of the device under different operating states, and adjust based on data-driven, real-time feedback on the operating state of the device.

[0060] The construction of the device-level digital twin models the parameters of the device under different operating states and adjusts based on data-driven, real-time feedback on the operating state of the device, including:

[0061] Sensors such as accelerometers, displacement sensors, force sensors, etc. are arranged on AGV, stacker, etc. to collect real-time data of the device's motion state, speed, load information, working environment data (e.g. temperature, vibration, speed, etc.) and state data (e.g. running, standby, fault, etc.).

[0062] Based on the mechanical properties and kinematics of the device, a dynamic model of AGV, stacker, etc. is constructed to model the parameters of the device under different operating states (such as current, speed, acceleration, etc.), ensuring that the device twin can reflect the actual operating conditions of the device in real time.

[0063] For example, the Newtonian mechanics is used to establish the dynamic equation of AGV to simulate the motion trajectory of the device and the mechanical properties of the load.

[0064] Using historical data and sensor data, machine learning algorithms such as regression analysis and deep neural networks are used to model the motion process of the device, based on data-driven model adjustment to real-time feedback on the operating state of the device, automatically correcting the differences between the physical model and the actual situation.

[0065] The second-level digital twin model construction unit is used to construct a storage location-level digital twin, and according to the residence time and turnover frequency of goods in the storage location, different storage locations are assigned different color heat values.

[0066] The construction of the storage location-level digital twin assigns different storage locations different color heat values according to the residence time and turnover frequency of goods in the storage location, including:

[0067] Through RFID, bar code scanning and other technologies, real-time acquisition of the location information of each cargo in the warehouse, storage state (for example: taken away, to be replenished, etc.), collection of cargo location environment data by environmental sensors (such as temperature and humidity sensors) for judging the influence of cargo storage conditions on the logistics process;

[0068] Using cargo location information and environmental data, combined with regional division and cargo storage density, a heat map generation algorithm is used to generate a heat map of different regions inside the warehouse in real time.

[0069] In this embodiment, the heat map reflects the storage density and activity state of the goods in different areas of the warehouse, such as a region may cause a logistics bottleneck due to storing too many goods, or cause abnormal storage environment due to high temperature / humidity and other factors;

[0070] Based on real-time data, the heat map is updated in real time, and the storage location of the goods is dynamically adjusted to avoid inefficiency or congestion in the logistics operation caused by improper space layout.

[0071] The third-level digital twin model construction unit is configured to construct a system-level digital twin and optimize the overall logistics path planning in the warehouse.

[0072] The system-level digital twin is constructed to optimize the overall logistics path planning in the warehouse, including:

[0073] Based on the first-level digital twin model and the second-level digital twin model, a global model of warehouse logistics is constructed to reflect the mutual relationship and dynamic change between each warehouse device, goods, and warehouse area.

[0074] A path planning algorithm (such as A* algorithm, Dijkstra algorithm, etc.) is used to optimize the logistics path of AGV, stacker and other warehouse devices, and the shortest distance, minimum waiting time and other indicators of the logistics path are dynamically calculated to automatically select the optimal path.

[0075] In this embodiment, considering the real-time changes in the warehouse (such as changes in the storage location of goods and changes in the operating state of devices), the optimization algorithm can update the logistics path in real time under different conditions to avoid congestion or inefficient transportation routes.

[0076] A multi-objective scheduling algorithm is used to balance multiple objectives such as device load, path length, transportation time, etc., and according to the working state, priority and real-time demand of the device, the working order of AGV, stacker and other devices is intelligently scheduled.

[0077] In the embodiment, the construction of the hierarchical twin model architecture can provide real-time and dynamic monitoring and optimization support in warehouse logistics management, model and optimize the warehouse logistics environment from the equipment layer, the goods storage layer and the global logistics system layer respectively, ensure the efficiency, real-time and intelligence of digital twin in the warehouse logistics process, not only improve the operation efficiency of warehouse logistics, but also reduce the manual intervention, predict and avoid potential problems in advance.

[0078] The virtual-real mapping module is configured to construct a virtual-real space real-time mapping system and simulate and predict the operation of the warehouse equipment by using the digital twin model.

[0079] The virtual-real mapping module comprises a virtual-real mapping unit and a state prediction unit.

[0080] The virtual-real mapping unit synchronizes data between the warehouse equipment and the digital twin model by using the Internet of Things technology and performs virtual-real deviation compensation.

[0081] The virtual-real deviation compensation comprises:

[0082] The Kalman filter is used to dynamically filter the virtual-real space data to obtain preliminary correction data for correcting the deviation of the virtual-real space data, and the virtual-real space data is the data exchanged between the warehouse and the goods and the digital twin model.

[0083] The preliminary correction data is used as the initial population, the fitness function is defined as the weighted sum of the warehouse equipment state deviation and the path planning efficiency, the crossover individuals are selected according to the fitness, and the elite individuals are dynamically adjusted, when the warehouse equipment is in the starting stage and the state fluctuation rate is greater than 15%, the top n% individuals are reserved as the crossover individuals, when the warehouse equipment is in the stable stage and the state fluctuation rate is less than 5%, m% of the crossover individuals are reserved, and m>n.

[0084] In the embodiment, when the warehouse equipment is in the starting stage and the state fluctuation rate is greater than 15%, n can be between 5 and 10, and when the warehouse equipment is in the stable stage and the state fluctuation rate is less than 5%, m can be between 15 and 20.

[0085] Adaptive crossover iteration is used, when the warehouse equipment is in the starting stage and the state fluctuation rate is greater than 15%, the crossover probability is adjusted to 0.8, when the warehouse equipment is in the stable stage and the state fluctuation rate is less than 5%, the crossover probability is adjusted to 0.3, and if the optimal solution is not updated for 20 iterations or the fitness variance is less than the set threshold for 50 consecutive iterations, the part of the population is reset.

[0086] Randomly select 30% of the non-crossover individuals, completely randomly initialize the selected non-crossover individuals or generate new individuals based on historical data, and perform crossover and mutation operations on the randomly initialized individuals or the generated new individuals.

[0087] In this embodiment, when the device anomaly (such as AGV frequently deviating from the planned path) is detected, part of the population is reset, 30% of the AGV path planning individuals are reset, a new path is generated based on real-time sensor data, and iteration is performed to select the optimal solution. The 30% is not a fixed value and can be 25%-30%;

[0088] When convergence or maximum evolution generation is reached, the optimal deviation correction scheme is output.

[0089] In this embodiment, in the real-time data mapping process between the virtual space and the physical space, due to sensor errors, device delays, network fluctuations and other reasons, deviations may occur between the virtual twin and the actual device. Identifying and analyzing these deviation sources, focusing on device motion trajectories, cargo positions and sensor reading errors in the physical environment, Kalman filtering can optimize the state estimation of devices and goods in the virtual model by comparing real-time sensor data with the expected state of the digital twin, and real-time correction of virtual-real deviations. However, when the data fluctuates greatly, the accuracy is obviously insufficient.

[0090] Genetic algorithm has high requirements for the selection of initial population, and the quality of initial population has a great influence on the result. Kalman filtering is used to solve the problem of initial population quality. In the cross iteration process, it is easy to fall into local optimum or premature convergence. The present application uses adaptive cross iteration. In the early stage of iteration, when the population individuals are greatly different, the cross probability is appropriately increased to promote the wide spread of excellent genes and quickly explore the solution space. With the deepening of iteration, the population tends to be homogenized, and the cross probability is reduced to finely mine local potential better solutions. For example, in the case of warehouse equipment operation parameter deviation compensation, the parameter fluctuation is large in the starting stage of the equipment, and high cross probability (0.8) helps to quickly locate the appropriate parameter range. After stable operation, fine tuning is performed using low cross probability (0.3) to ensure accurate deviation compensation. If the cross probability is a fixed value, it is easy to converge prematurely. The present application realizes adaptive optimization by introducing a device fluctuation rate threshold (15%, 5%), which significantly improves the robustness of the algorithm in dynamic warehouse scenarios.

[0091] The state prediction unit is configured to predict the running state of the warehouse equipment.

[0092] The predicted running state of the warehouse equipment includes:

[0093] According to the preprocessed device state data and the first-layer digital twin model, a real-time state equation of the warehouse equipment is established, and the specific formula is:

[0094]

[0095] wherein, x(t) represents the state vector of the warehouse equipment at time t, f() represents the dynamic relationship function, x(t-△t) represents the state vector of the warehouse equipment at the previous time, u(t) represents the input signal, p represents the warehouse equipment parameters, i.e. parameters related to its physical properties, and △t represents the time interval;

[0096] The parameter p represents the physical properties related to the system characteristics, such as mass, friction coefficient, elastic modulus, etc. These parameters are usually determined through experiments or theoretical derivation during the system modeling stage, and determine the response characteristics of the system to the input signal. Overall, the principle of the state equation is based on the time-varying characteristics of dynamic systems, and the current state of the system depends not only on its past state, but also on external input and system characteristics. Through recursive definition of the state, this equation realizes the continuity and causality in time;

[0097] The dynamic relationship function f() represents the dynamics of the warehouse equipment, describing how to calculate the current state based on the previous state, the current input signal and the system parameters. This function is usually nonlinear and can capture the complex behavior and interaction of the system;

[0098] In practical applications, this state equation can be used for real-time state prediction, fault diagnosis and optimization control strategy. By continuously updating the state and making predictions, the health status of the system can be monitored in real time, potential faults can be detected in a timely manner, and the safe and reliable operation of the warehouse equipment can be ensured;

[0099] The real-time state of the warehouse equipment is updated, and the update formula is:

[0100]

[0101] where, x(t+△t) represents the state vector of the warehouse equipment at time t+△t, represents the state change rate;

[0102] The running state of the warehouse equipment is simulated in real time on the cloud server using the digital twin model, and the future state of the warehouse equipment is predicted by inputting the current state and the input signal u(t).

[0103] The prediction analysis module is used to monitor the warehouse logistics in multiple dimensions and construct a fault prediction model to predict and analyze the faults of the warehouse equipment.

[0104] The prediction analysis module includes a fault prediction model construction training unit and a prediction analysis unit.

[0105] The fault prediction model construction training unit is used to construct and train the fault prediction model.

[0106] The construction of the failure prediction model and the training include:

[0107] The machine learning algorithm is selected to construct the failure prediction model, and the labeled data set D train The failure prediction model is trained, and the training target of the model is set to minimize the loss function, and the specific formula of the loss function is:

[0108]

[0109] Wherein, L represents the minimized loss function of the training target of the failure prediction model, Γ(·) represents the loss function, xl represents the number of data in the labeled data set D train , y a represents the label of the data in the labeled data set D train , that is, whether the data corresponds to the time when the failure occurs, represents the prediction result of the a-th data by the failure prediction model;

[0110] The specific formula of the loss function Γ(·) is:

[0111] Wherein, represents the class weight of the a-th data, which is used to process the imbalance of the class, for example, if a certain class sample is less, it can be given higher weight;

[0112] Cross-entropy loss is a common method to measure the difference between two probability distributions, which evaluates the distance between the predicted distribution and the true distribution of the model output in classification problems, and the weight is introduced to balance the influence of class imbalance. In many practical applications, the samples of some classes may be significantly less than those of other classes, resulting in the model's bias towards the main classes. By giving higher weights to minority class samples, the model can pay more attention to these samples during training, thereby improving its prediction performance. The minimization target of this loss function is to adjust the model parameters through iterative optimization (such as gradient descent) so that the prediction result is closer to the true label. This method not only improves the accuracy of the model, but also enhances the recognition ability of the minority class failure;

[0113] The failure prediction model is trained until the minimized loss function of the training target of the model converges and remains unchanged, and the training is stopped, and the trained failure prediction model is obtained.

[0114] The prediction and analysis unit uses the trained failure prediction model to predict and analyze the failure of the warehouse equipment.

[0115] The use of the trained failure prediction model to predict and analyze the failure of the warehouse equipment includes:

[0116] Extract key features from the warehouse equipment historical data set D, including statistical features and frequency domain features, and select the feature set F with the most discriminant ability for fault judgment new , new F m} where Ft m represents the mth feature with the most discriminant ability for fault judgment

[0117] Specifically, the statistical features include mean, standard deviation, maximum value, minimum value, etc., and the frequency domain features are extracted by fast Fourier transform (FFT) to extract frequency components

[0118] The extracted feature set F new is input into the trained warehouse equipment fault recognition model to calculate the warehouse equipment fault prediction probability, and the specific formula is:

[0119]

[0120] where P(y=c|F new ) represents the probability of the occurrence of the cth fault given the feature F new , y represents the fault category, represents the model parameters of the cth fault, which is usually a weight vector obtained by training, indicating the relationship between the feature and the fault category, affecting the model's prediction of this category, k represents the index of all possible fault categories, and T represents the total number of fault categories

[0121] In this embodiment, the principle of the formula is that the formula uses a soft-max function to convert the linear combination of model outputs (i.e. ) into a probability distribution, which ensures that the sum of the prediction probabilities of all categories is 1, so that the output can be directly interpreted as a probability

[0122] The model parameters of each category, i.e., the weights, are learned through training data and reflect the importance of each feature in different fault categories. The greater the weight, the greater the impact of the feature on identifying the fault of this category

[0123] By calculating the probability of each category, the model can determine the fault type that the input feature F new is most likely to correspond to, and select the category with the maximum probability as the final prediction result. This formula and its parameters are crucial in fault prediction, and through the analysis of real-time features and the application of corresponding model parameters, potential fault types can be effectively identified to provide real-time monitoring and maintenance recommendations

[0124] According to actual needs, set a fault warning threshold Ψ, if P(y=c|F newIf yes, an alarm is triggered, otherwise, no alarm is triggered.

[0125] In addition, parts of the above technical solutions provided in the embodiments of the present application that are consistent with the implementation principles of corresponding technical solutions in the prior art are not described in detail to avoid excessive repetition.

[0126] The specific embodiments described above are further explained in further detail by the purpose, technical solutions and beneficial effects of the present application. It should be understood that the above description is only a specific embodiment of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. A warehousing and logistics visualization monitoring system based on digital twins, characterized in that, include: Data acquisition module, digital twin construction module, virtual-real mapping module, and predictive analysis module; The data acquisition module is used to collect various data in the warehousing environment, including: warehousing structure, item location, inventory information, equipment status data and personnel activities, and to preprocess them; The digital twin construction module constructs a hierarchical digital twin model based on various data in the preprocessed warehousing environment. The virtual-real mapping module is used to construct a real-time virtual-real space mapping system and use a digital twin model to simulate and predict the operation of warehousing equipment. The predictive analysis module is used to monitor warehousing logistics in multiple dimensions and build a fault prediction model to predict and analyze the faults of warehousing equipment. The virtual-real mapping module includes: a virtual-real mapping unit and a state prediction unit; The virtual-real mapping unit is used to synchronize data between warehousing equipment and digital twin models through Internet of Things (IoT) technology, and to compensate for virtual-real discrepancies. The status prediction unit is used to predict the operating status of the storage equipment; The virtual-to-real deviation compensation includes: Kalman filtering is used to dynamically filter the virtual and real space data to obtain preliminary correction data for the deviation correction of the virtual and real space data; Using the initial corrected data as the initial population, the fitness function is defined as the weighted sum of the state deviation of the storage equipment and the path planning efficiency. Crossover individuals are selected based on fitness, and elite individuals are dynamically adjusted. When the storage equipment is in the startup phase and the state volatility is >15%, the top n% of individuals are retained as crossover individuals. When the storage equipment is in the stable phase and the state volatility is <5%, m% of crossover individuals are retained, satisfying m>n. An adaptive crossover iteration is adopted. When the storage equipment is in the startup phase and the state volatility is >15%, the crossover probability is adjusted to 0.

8. When the storage equipment is in the stable phase and the state volatility is <5%, the crossover probability is adjusted to 0.

3. If the optimal solution is not updated in 20 iterations or the fitness variance is less than the set threshold for 50 consecutive iterations, a part of the population is reset. Randomly select 30% of non-crossover individuals, perform completely random initialization on the selected non-crossover individuals or generate new individuals based on historical data, and perform crossover mutation operation on the randomly initialized individuals or the generated new individuals; When convergence or the maximum number of generations is reached, the optimal deviation correction scheme is output. The virtual and physical space data refers to the data generated by the interaction between the warehouse and the goods and the digital twin model.

2. The warehousing and logistics visualization monitoring system based on digital twins as described in claim 1, characterized in that, The digital twin construction module includes: a first-layer digital twin model construction unit, a second-layer digital twin model construction unit, and a third-layer digital twin model construction unit; The first-layer digital twin model building unit is used to build a device-level digital twin, model the parameters of the device under different operating conditions, and make adjustments based on data-driven approaches, providing real-time feedback on the device's operating status. The second-layer digital twin model construction unit is used to construct a storage location-level digital twin. Based on the storage time and turnover frequency of goods in the storage location, different colors are assigned heat values ​​to different storage locations. The third-layer digital twin model building unit is used to construct a system-level digital twin and optimize the overall logistics path planning within the warehouse.

3. The warehousing and logistics visualization monitoring system based on digital twins as described in claim 2, characterized in that, The predicted operating status of the warehousing equipment includes: Based on the preprocessed equipment status data and the first-layer digital twin model, establish the real-time state equation of the warehousing equipment; The real-time status of the storage equipment is updated based on the rate of change of the equipment's status within Δt. The operation status of warehousing equipment is simulated in real time using a digital twin model on a cloud server. By inputting the current status and input signals, the future status of the warehousing equipment can be predicted.

4. The warehousing and logistics visualization monitoring system based on digital twins as described in claim 1, characterized in that, The predictive analysis module includes: a fault prediction model construction and training unit and a predictive analysis unit; The fault prediction model construction training unit is used to construct and train the fault prediction model. The predictive analysis unit uses a trained fault prediction model to predict and analyze faults in warehousing equipment.

5. The warehousing and logistics visualization monitoring system based on digital twins as described in claim 4, characterized in that, The method of using a trained fault prediction model to predict and analyze faults in warehousing equipment includes: We selected a machine learning algorithm to build a fault prediction model, using a labeled dataset D. train Train the fault prediction model, setting the training objective of the model as minimizing the loss function; Train the fault prediction model until the loss function that minimizes the training objective of the model converges and remains constant, then stop training to obtain the trained fault prediction model.

6. The warehousing and logistics visualization monitoring system based on digital twins as described in claim 5, characterized in that, The method of using a trained fault prediction model to predict and analyze faults in warehousing equipment includes: Extract key features from the historical data set D of the storage equipment. These key features include statistical features and frequency domain features. Then, select the feature set F that has the strongest discriminative power for fault diagnosis. new F new ={Ft1,Ft2,...,Ft m }, Ft m This represents the m-th feature that is most discriminative in fault diagnosis. The extracted feature set Fnew is input into the trained warehousing equipment fault identification model to obtain the fault prediction probability of warehousing equipment. Set fault warning thresholds according to actual needs. If the probability of warehouse equipment failure is greater than If the condition is met, an alarm will be triggered; otherwise, no alarm will be triggered.

Images