Warehouse operation data visual analysis method and device

Abstract

This application provides a method and apparatus for visual analysis of warehousing operation data. The method includes: collecting multi-dimensional data during the warehousing operation process in real time; performing preliminary data processing at edge nodes and then sending the data to a hierarchical database on a central data platform for storage; performing initial screening of the multi-dimensional data using a statistical detection algorithm; performing secondary screening of the multi-dimensional data using a pre-trained isolated forest algorithm; performing final screening of the multi-dimensional data using a time series model; finally determining the anomaly probability value of abnormal data in each dimension; constructing a visualization model of the real warehousing space and mapping the multi-dimensional data into the visualization model; determining the warning level based on the detected anomaly probability value; and providing warning prompts in the visualization model based on the warning level. This application can improve the efficiency and accuracy of warehousing management.

Description

Technical Field

[0001] This application relates to the field of data processing, specifically to a method and apparatus for visualizing and analyzing warehouse operation data. Background Technology

[0002] In modern logistics and supply chain management, the efficiency and accuracy of warehousing operations play a crucial role in overall operational efficiency. With the rapid development of information technology, the application of modern warehouse management systems (WMS) and automated equipment has become increasingly widespread, bringing unprecedented convenience and efficiency improvements to warehousing operations. However, in terms of data-driven, refined operations, current warehouse management systems still face many challenges and shortcomings, specifically in the following aspects:

[0003] 1. Isolated and one-sided data dimensions. Most current warehouse management systems focus primarily on isolated business outcome data, such as inbound and outbound volumes and inventory turnover rates. While these data reflect the basic situation of warehouse operations, they lack comprehensive collection and correlation analysis of multi-dimensional data, including operational processes, equipment status, environmental parameters, and personnel behavior. This "data silo" phenomenon makes it difficult for managers to fully and accurately grasp the actual situation of warehouse operations, thus affecting the scientific and accurate nature of decision-making.

[0004] 2. Severe lag in analysis and decision-making. Existing warehouse management systems mostly rely on offline, batch processing for data analysis, and the generated reports often reflect the situation at a specific point in time, failing to provide real-time decision support for on-site operations.

[0005] 3. Lack of spatial and contextual awareness. While traditional two-dimensional reports and dashboards can display certain data information, they cannot intuitively reflect the physical spatial layout and dynamic processes of the work site.

[0006] 4. Passive and Delayed Response. Existing systems rely heavily on fixed threshold rules for anomaly detection. While these rules are simple and easy to implement, they lack the ability to intelligently identify complex and interconnected anomalies. Therefore, when anomalies occur in warehousing operations, the system often fails to issue timely warnings, leading to significant losses by the time the problem is discovered.

[0007] Given the numerous shortcomings of existing warehouse management systems in terms of data-driven, refined operations, there is an urgent need for a data visualization and analysis method for warehouse operations that can improve the efficiency and accuracy of warehouse management. Summary of the Invention

[0008] To address the problems in the existing technology, this application provides a method and apparatus for visualizing and analyzing warehouse operation data, which can improve the efficiency and accuracy of warehouse management.

[0009] To solve at least one of the above problems, this application provides the following technical solution:

[0010] Firstly, this application provides a method for visualizing and analyzing warehouse operation data, including:

[0011] Multi-dimensional data during the warehousing operation is collected in real time according to a preset sampling frequency. The multi-dimensional data includes operation status data, equipment operation data, environmental parameter data, personnel behavior data, and inventory dynamic data. The multi-dimensional data is transmitted to edge nodes for preliminary data processing through a preset transmission protocol and then sent to the central data platform for storage in a preset hierarchical storage database.

[0012] Outlier filtering is performed on the multi-dimensional data in the hierarchical storage database according to the preset statistical detection algorithm to determine the corresponding initial screening data. The initial screening data is then input into the pre-trained isolated forest algorithm for association anomaly identification to determine the corresponding secondary screening data. The secondary screening data is then input into the preset time series model for continuous anomaly identification to determine the corresponding final screening data. Based on the initial screening data, secondary screening data, and final screening data, the anomaly probability value of each dimension of the data is determined.

[0013] The multi-dimensional data is mapped to a visualization model of the real warehouse space using a preset 3D digital twin engine. The abnormal probability values ​​are classified into warning levels, and the corresponding warning levels are determined. Based on the warning levels, the data of each dimension is visualized and warned in the visualization model to realize the visualization analysis of warehouse operation data. The 3D digital twin engine is communicatively connected to the central data platform.

[0014] Furthermore, the step of collecting multi-dimensional data in real time during the warehousing operation according to a preset sampling frequency includes:

[0015] Personnel behavior data is collected every second via ultra-wideband (UWB) or Bluetooth AOA positioning systems.

[0016] Equipment operation data is collected every 3 seconds from programmable logic controllers (PLCs), IoT gateways, or the equipment's built-in sensors.

[0017] Operation status data is collected every 5 seconds via the warehouse management system (WMS) interface or a handheld PDA terminal.

[0018] Environmental parameter data is collected from distributed environmental sensing nodes deployed in the warehouse every 10 seconds.

[0019] Inventory dynamic data is collected from the warehouse management system (WMS) database or machine vision inventory system every 60 seconds.

[0020] Furthermore, the step of transmitting the multi-dimensional data to edge nodes via a preset transmission protocol for preliminary data processing before sending it to the central data platform and storing it in a preset hierarchical storage database includes:

[0021] The multi-dimensional data is transmitted to the edge node through a preset transmission protocol for preliminary data processing to determine the corresponding pre-processed data. The preliminary data processing includes at least one of data filtering and data cleaning.

[0022] The preprocessed data is sent to the central data platform and stored in a preset hierarchical storage database. The hierarchical storage database includes a hot data storage layer and a cold data storage layer. The hot data storage layer is used to store data generated within a preset time threshold and accessed frequently, while the cold data storage layer is used to store historical data accessed infrequently beyond the preset time threshold.

[0023] Further, before inputting the initial screening data into the pre-trained Isolation Forest algorithm for association anomaly identification and determining the corresponding rescreening data, the following steps are included:

[0024] Data marked as normal operating cycles are extracted from historical multi-dimensional warehouse data to construct an initial training set. An initial isolated forest model is trained based on the initial training set. The initial isolated forest model is used to predict the initial training set, and the training data samples with the highest initial outlier scores are removed. The corresponding purified training dataset is determined and the corresponding training data features are extracted.

[0025] A final isolated forest model composed of multiple isolation trees is constructed. A dedicated data subset is randomly assigned to each isolation tree from the purified training dataset. At the same time, a dedicated feature subset is randomly extracted from the training data features for each isolation tree. The final isolated forest model is trained based on the dedicated data subset and the dedicated feature subset to determine the corresponding pre-trained isolated forest algorithm. The dedicated data subset is obtained from the training dataset through sliding time window sampling to learn data patterns in different time segments. The dedicated feature subset contains features from at least two different data dimensions to isolate anomalies in a cross-dimensional feature space.

[0026] Further, the step of inputting the initial screening data into a pre-trained isolated forest algorithm for association anomaly identification to determine the corresponding rescreening data includes:

[0027] The initial screening data is standardized and vectorized to determine the corresponding unified feature vector, wherein the standardization adopts the Z-score standardization method;

[0028] The unified feature vector is input into the pre-trained isolated forest algorithm to determine the anomaly score output by the isolated forest algorithm. Based on the anomaly score, abnormal sample points with anomaly scores higher than the threshold are selected to determine the corresponding re-screening data.

[0029] Furthermore, the step of mapping the multi-dimensional data to a visualization model of the real warehouse space based on a preset 3D digital twin engine includes:

[0030] A static 3D scene model built using the Unity engine, which corresponds 1:1 to the physical space layout, shelves, and equipment locations of a real warehouse.

[0031] Establish a dynamic binding relationship between the multi-dimensional data and the corresponding entity objects in the static three-dimensional scene model, and determine the corresponding visualization model. Among them, the equipment operation data is bound to the corresponding equipment model to present the equipment operation status, location coordinates and fault status in the three-dimensional scene; the inventory dynamic data is bound to the corresponding shelf model to present the shelf occupancy status in the three-dimensional scene; and the personnel behavior data is bound to the corresponding personnel avatar model to present the personnel movement trajectory and location in the three-dimensional scene.

[0032] Furthermore, the step of classifying the abnormal probability values ​​into early warning levels, determining the corresponding early warning levels, and providing visual early warning prompts for each dimension of data in the visualization model based on the early warning levels includes:

[0033] The numerical range of the abnormal probability value is dynamically divided to determine the corresponding warning level, which includes Level I, Level II, Level III and Level IV;

[0034] If the warning level is Level I, then the warning-related objects will be highlighted in the visualization model;

[0035] If the warning level is Level II, then sound alerts will be given to the relevant objects in the visualization model, and the relevant objects will be controlled to flash in the model.

[0036] If the warning level is Level III, then a moderate alarm will be issued to the relevant objects in the visualization model, and the relevant work processes will be suspended.

[0037] If the warning level is Level IV, a severe alarm will be issued to the relevant objects in the visualization model, and the associated equipment will be automatically shut down, and the emergency plan will be activated.

[0038] Secondly, this application provides a warehouse operation data visualization and analysis device, comprising:

[0039] The multi-dimensional data acquisition module is used to collect multi-dimensional data in real time during the warehousing operation according to a preset sampling frequency. The multi-dimensional data includes operation status data, equipment operation data, environmental parameter data, personnel behavior data, and inventory dynamic data. The multi-dimensional data is transmitted to the edge node for preliminary data processing through a preset transmission protocol and then sent to the central data platform for storage in a preset hierarchical storage database.

[0040] The abnormal data prediction module is used to filter outliers from the multi-dimensional data in the hierarchical storage database according to a preset statistical detection algorithm, determine the corresponding initial screening data, input the initial screening data into a pre-trained isolated forest algorithm for association anomaly identification, determine the corresponding secondary screening data, input the secondary screening data into a preset time series model for continuous anomaly identification, determine the corresponding final screening data, and determine the anomaly probability value of each dimension of data based on the initial screening data, secondary screening data, and final screening data.

[0041] The abnormal status visualization module is used to map the multi-dimensional data to a visualization model of the real warehouse space based on a preset three-dimensional digital twin engine, classify the abnormal probability value into warning levels, determine the corresponding warning levels, and provide visual warning prompts for each dimension of data in the visualization model according to the warning levels, thereby realizing the visualization analysis of warehouse operation data. The three-dimensional digital twin engine is communicatively connected to the central data platform.

[0042] Thirdly, this application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the warehouse operation data visualization and analysis method.

[0043] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the described warehouse operation data visualization and analysis method.

[0044] Fifthly, this application provides a computer program product, including a computer program / instructions, which, when executed by a processor, implement the steps of the described warehouse operation data visualization and analysis method.

[0045] As can be seen from the above technical solution, this application provides a method and device for visual analysis of warehousing operation data. It collects multi-dimensional data during the warehousing operation process in real time, performs preliminary data processing at edge nodes, and then sends the data to a hierarchical database on a central data platform for storage. The multi-dimensional data is initially screened using statistical detection algorithms, further screened using a pre-trained isolated forest algorithm, and finally screened using a time-series model. This process determines the anomaly probability value of abnormal data in each dimension, constructs a visualization model of the real warehousing space, maps the multi-dimensional data into the visualization model, determines the warning level based on the detected anomaly probability value, and provides warning prompts in the visualization model according to the warning level. This improves the efficiency and accuracy of warehousing management. Attached Figure Description

[0046] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0047] Figure 1 This is one of the flowcharts illustrating the warehouse operation data visualization and analysis method in the embodiments of this application;

[0048] Figure 2 This is a structural diagram of the warehouse operation data visualization and analysis device in the embodiments of this application;

[0049] Figure 3 This is a schematic diagram of the structure of the electronic device in the embodiments of this application.

[0050] Figure label:

[0051] Electronic device 9600, central processing unit 9100, memory 9140, communication module 9110, input unit 9120, audio processor 9130, display 9160, power supply 9170, buffer memory 9141, application / function storage unit 9142, data storage unit 9143, driver storage unit 9144, antenna 9111, speaker 9131, microphone 9132. Detailed Implementation

[0052] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0053] The acquisition, storage, use, and processing of data in this application comply with relevant laws and regulations.

[0054] Considering the current shortcomings of warehouse management systems, such as isolated and fragmented data dimensions, lack of data correlation analysis and context awareness, and low efficiency, this application provides a method and apparatus for visual analysis of warehouse operation data. It collects multi-dimensional data during real-time warehouse operations, performs preliminary data processing at edge nodes, and then sends the data to a hierarchical database on a central data platform. The multi-dimensional data is initially screened using statistical detection algorithms, further screened using a pre-trained isolated forest algorithm, and finally screened using a time-series model. This process determines the anomaly probability value of abnormal data in each dimension, constructs a visualization model of the real warehouse space, and maps the multi-dimensional data into the visualization model. Based on the detected anomaly probability values, a warning level is determined, and warning prompts are displayed in the visualization model according to the warning level. This improves the efficiency and accuracy of warehouse management.

[0055] To improve the efficiency and accuracy of warehouse management, this application provides an embodiment of a warehouse operation data visualization and analysis method, see [link to embodiment]. Figure 1 The aforementioned method for visualizing and analyzing warehouse operation data specifically includes the following:

[0056] Step S101: Collect multi-dimensional data in real time during the warehousing operation according to the preset sampling frequency. The multi-dimensional data includes operation status data, equipment operation data, environmental parameter data, personnel behavior data, and inventory dynamic data. The multi-dimensional data is transmitted to the edge node for preliminary data processing through the preset transmission protocol and then sent to the central data platform for storage in the preset hierarchical storage database.

[0057] Optionally, in this embodiment, the task of this step is to achieve real-time collection, data perception, and aggregation of data throughout the entire warehousing operation process.

[0058] Optional multi-dimensional data acquisition aims to achieve comprehensive situational awareness and break down the "data silos" in traditional systems. This is achieved by constructing an interconnected data acquisition network covering five core dimensions. The real-time acquisition scheme for multi-source heterogeneous data is shown in the table below:

[0059]

[0060] Specifically, the system collects operational status data: through an API interface, it deeply integrates with the Warehouse Management System (WMS) or directly from operational devices such as handheld terminals (PDAs) to capture key node information such as the creation, allocation, start, completion, and time elapsed of picking lists at a frequency of approximately once every 5 seconds. This enables micro-level tracking of the task flow.

[0061] Specifically, for equipment operation data acquisition: For automated warehouses, conveyor lines, AGVs, stacker cranes, and other equipment, the system communicates with the programmable logic controller (PLC) via an IoT gateway or directly, collecting their core operating parameters at a high frequency of approximately once every 3 seconds. These parameters include motor speed, real-time power and load of the AGV, three-dimensional coordinate position of the stacker crane, and various fault codes and status words reported by the equipment itself. This is equivalent to installing a real-time operating black box for each critical piece of equipment.

[0062] Specifically, environmental parameter data collection: By deploying a distributed sensor network at key nodes within the warehouse (such as temperature-controlled warehouses, high-value goods areas, and high-intensity work areas), the system collects data on environmental factors such as temperature, humidity, light intensity, noise, and even dust concentration approximately every 10 seconds. This brings environmental factors that have a potential impact on operational quality and safety into the monitoring scope.

[0063] Specifically, personnel behavior data collection: Utilizing high-precision indoor positioning technologies such as UWB (Ultra-Wideband) or Bluetooth AOA (Angle of Arrival), the system can track the precise location of pickers and operators within the warehouse in real time (accurate to the centimeter level) at an extremely high frequency of once per second, thereby generating their movement trajectories, dwell time at specific storage locations, and operation frequency. This provides unprecedentedly detailed data for analyzing personnel efficiency and optimizing workflow design.

[0064] Specifically, the system collects dynamic inventory data by monitoring change logs in the WMS database and combining this with machine vision inventory results from cameras fixed to shelves or inspection robots. It updates the occupancy status of all storage locations and the inventory quantity changes for specific SKUs approximately every minute, and calculates inventory accuracy. This achieves near real-time monitoring of inventory consistency between records and actual stock.

[0065] By cross-covering these five dimensions, the system no longer views warehouse operations in a one-dimensional way, but constructs a comprehensive digital image encompassing "people, machines, materials (inventory), methods (operations), and environment." The collected multi-dimensional data is naturally aligned on the timeline, enabling in-depth analysis of complex correlation issues.

[0066] If massive amounts of real-time data were to flow directly into the central platform, it would put enormous pressure on network and computing resources. Therefore, all of the above data is aggregated to edge computing nodes via the lightweight MQTT protocol for preliminary processing, and then transmitted to the central data platform efficiently and reliably.

[0067] Edge computing nodes perform critical preprocessing tasks, including:

[0068] Data cleaning and formatting: Filter out obviously invalid data and convert data from different sources and with different formats into a standardized JSON or Avro format.

[0069] Preliminary aggregation and calculation: Perform preliminary aggregation on high-frequency data. For example, calculate the average speed and approximate location per second for personnel location data once per second at the edge before uploading, thereby reducing the amount of data transmission.

[0070] Edge rule triggering: Executes simple rule judgments with extremely high real-time requirements. For example, when UWB data detects that a person has entered a dangerous area, the edge node can respond in milliseconds and send instructions directly to the on-site audible and visual alarms without waiting for cloud-based decisions.

[0071] By introducing edge computing nodes, some computing power is pushed to the edge, enabling rapid local response to emergencies. In case of network outages or other abnormal situations, edge nodes can also temporarily cache data to ensure that data is not lost.

[0072] The processed data is eventually persisted, but the access value and frequency of data vary at different times, so an intelligent tiered storage strategy is needed.

[0073] Specifically, the database adopts a cold and hot storage architecture.

[0074] Hot data layer: Data generated recently (e.g., within the last 30 days) that requires real-time monitoring, frequent querying, and analysis is stored in high-performance time-series databases (such as TimescaleDB) or in-memory databases (such as Redis). These databases are deeply optimized for time-series data, supporting fast writing of massive amounts of data and millisecond-level query response, perfectly meeting the needs of real-time monitoring dashboards and instant analysis.

[0075] Cold data layer: Historical data exceeding a specific time threshold (e.g., 30 days) has seen a significant decrease in access frequency, but still needs to be retained for compliance, long-term trend analysis, and model training purposes. This data is automatically and seamlessly archived to lower-cost object storage services (such as Amazon S3 and MinIO). Object storage offers near-unlimited capacity and extremely low storage costs. Cold data is efficiently compressed using a columnar storage format (Parquet). By partitioning and indexing by time, task type, and device ID, the query efficiency of historical data is greatly improved.

[0076] Meanwhile, the system is configured with daily full snapshot backup and incremental log backup strategies, supporting data rollback at any point in time and disaster recovery across nodes.

[0077] This step establishes a data supply chain through the acquisition network, transmission preprocessing pipeline, and hierarchical storage strategy, laying a solid data foundation for subsequent visualization and data anomaly analysis.

[0078] Step S102: Filter outliers from the multi-dimensional data in the hierarchical storage database according to a preset statistical detection algorithm to determine the corresponding initial screening data. Input the initial screening data into a pre-trained isolated forest algorithm for association anomaly identification to determine the corresponding secondary screening data. Input the secondary screening data into a preset time series model for continuous anomaly identification to determine the corresponding final screening data. Determine the anomaly probability value of each dimension of the data based on the initial screening data, secondary screening data, and final screening data.

[0079] Optionally, in this embodiment, this step is to perform intelligent anomaly detection on multi-dimensional data, which is key to improving the efficiency of anomaly detection and warehouse management.

[0080] Optionally, in this embodiment, a hybrid multi-stage anomaly detection model is used for anomaly detection. This model employs a "layer-by-layer, complementary" precision filtering and identification mechanism, aiming to accurately and efficiently identify various anomalies, from simple to complex, from massive and complex warehouse data. Its execution process can be divided into three interconnected stages: first, computationally efficient statistical methods are used to quickly filter obvious anomalies; second, a machine learning model is used to identify complex correlation anomalies in static data; and finally, a deep learning model is used to capture dynamic time-series pattern anomalies that are difficult to detect using traditional methods.

[0081] Specifically, the first stage involves statistically-based rapid outlier filtering.

[0082] The first phase, serving as the first line of defense for anomaly detection, primarily involves high-speed, low-computational-cost preliminary screening. The system reads or batches key performance indicator data streams from a hierarchical storage database as needed. Subsequently, for each individual indicator to be monitored (e.g., the real-time power of a stacker crane, the instantaneous temperature of a specific area, or the number of operations per unit time for a picker), the system dynamically calculates its statistical characteristics, such as mean and standard deviation, within a sliding time window. Next, using methods such as Z-score normalization or box plotting, the deviation of the current data point from the data distribution within that window (i.e., the Z-value) is calculated. The system presets a strict threshold boundary (e.g., |Z-score|>3); any data point exceeding this boundary is identified as a statistically significant "outlier" and thus quickly identified.

[0083] The core function of this stage is its ability to identify very obvious and sudden anomalies with extremely low latency, such as sudden equipment shutdown, rapid temperature spikes or drops, and momentary interruptions in personnel operation. This acts as a highly efficient "coarse screen," quickly filtering out approximately 5% of the most extreme anomalies, greatly reducing the burden on subsequent complex algorithm models and ensuring the system's rapid response capability to major and sudden failures.

[0084] Specifically, the second stage involves identifying association anomalies based on isolated forests.

[0085] The data that passes the first stage of screening is not necessarily completely normal; it is then fed into the second stage of deep analysis using the Isolation Forest algorithm. The goal of this stage is to discover more subtle anomalies hidden within the cross-relationships of multidimensional data. The system aligns and combines the initially screened data from different dimensions (such as equipment, environment, and personnel) at specific points in time, forming a unified multidimensional feature vector. The pre-trained Isolation Forest model then analyzes these vectors. The algorithm's mechanism lies in constructing multiple isolation trees and randomly selecting features and split points to isolate each data point. Its core idea is that outliers, due to their significantly different feature patterns from normal points, can be isolated more quickly. The model calculates an "anomaly score" by averaging the path lengths required for isolation.

[0086] The core capability at this stage lies in identifying correlational or contextual anomalies. These anomalies are characterized by the fact that any single indicator may fall within the normal statistical range, but the combination of several indicators constitutes an abnormal state. For example, a conveyor (equipment dimension) may operate at a normal rate, but the power consumption of its driving motor (another equipment dimension) may be abnormally low, indicating potential faults such as belt slippage. Or, inventory checks at a certain location (operation dimension) may occur at a normal frequency, but the operator's movement within that area (personnel dimension) may show an abnormal wandering pattern, potentially indicating difficulty in locating items or process inefficiencies. The Isolation Forest algorithm can keenly capture such cross-dimensional inconsistencies, identifying seemingly normal but actually abnormal data points as secondary screening data, achieving in-depth mining of complex anomaly patterns.

[0087] Specifically, the third stage involves dynamic anomaly identification based on time series models.

[0088] The data that was not identified as abnormal in the second stage, along with its corresponding time series context, is fed into the pre-set time series model (an LSTM-based autoencoder) in the third stage. This model has been trained on a large amount of historical normal data and has learned the patterns of normal behavior of core entities (such as equipment and operational processes) in warehousing operations over time. The model's working mechanism is as follows: it receives a segment of time series data, attempts to encode it internally, and then decodes and reconstructs the data. The system continuously compares the error (reconstruction error) between the model's "reconstructed data" and the "actual input data."

[0089] The ultimate goal of this stage is to capture trend anomalies and gradual pattern changes. These anomalies are almost imperceptible at a single point in time, but their development can lead to serious problems. For example, the battery life of an AGV may be slowly but steadily declining, even though its voltage is within the normal range each time it is detected; this decline in the trend itself is a significant anomaly. The order processing speed of a picking station may be decreasing at an imperceptible rate each day. A sensor reading may begin to show small, periodic fluctuations. LSTM autoencoders are extremely sensitive to these slowly changing dynamic patterns. When the actual data begins to deviate from the normal pattern it has learned, the reconstruction error increases significantly.

[0090] Finally, by combining the anomalies from the initial screening, secondary screening, and final screening, and using deep learning reconstruction error as the main framework, with the initial screening and secondary screening serving as confidence weighting factors and explanatory sources, the entire decision-making framework is integrated.

[0091] Specifically, the first stage (statistical detection) acts as a hard veto. If a data point is determined to be an extreme outlier by the first stage, it is directly assigned an extremely high probability value for anomalies, or the probability value for anomalies is strongly revised upwards to ensure that it immediately triggers the highest level of alarm.

[0092] Specifically, the second stage (Isolation Forest) acts as a confidence amplifier. If the anomaly probability value in the third stage shows a moderate anomaly (e.g., 0.6), and the anomaly probability value in the second stage also shows a clear correlation anomaly (e.g., a score of 0.8), then these two mutually reinforcing signals will greatly increase the confidence of the system's judgment. A weighted up-adjustment is then performed to obtain the final anomaly probability value. This probability value is a continuous value between 0 and 1, quantitatively describing the degree of confidence that the current state deviates from its historical normal pattern.

[0093] Step S103: Based on the preset 3D digital twin engine, the multi-dimensional data is mapped to a visualization model of the real warehouse space. The abnormal probability value is divided into warning levels, the corresponding warning level is determined, and the data of each dimension is visualized and warned in the visualization model according to the warning level, so as to realize the visualization analysis of warehouse operation data. The 3D digital twin engine is communicatively connected to the central data platform.

[0094] Optionally, in this embodiment, this step is the core of the visualization presentation.

[0095] Optionally, in this embodiment, an innovative three-layer visualization architecture is adopted to meet different management needs from macro-level situation to micro-level details. The framework includes:

[0096] 3D Digital Twin Scene: A 1:1 digital twin model of the warehouse is built based on the Unity / Three.js engine. This model not only statically reproduces the warehouse structure, but also dynamically presents the real-time status of shelves (inventory levels), automated equipment (running / stopping / faulting), and personnel (location and trajectory) through real-time data-driven visualization, using color coding (e.g., red - alarm, green - normal) and information labels.

[0097] Multidimensional analysis dashboard: It integrates key performance indicators (KPIs) such as task completion rate, space efficiency analysis, energy consumption curve, and overall equipment efficiency (OEE), and uses chart libraries such as ECharts to realize dynamic and interactive data drill-down and exploration analysis.

[0098] Intelligent Situation Overview: Provides real-time summaries of the core status, highest alert level, and critical events for each work area in a cockpit mode. Supports quick switching of perspectives or retrieval of detailed information for specific areas via voice or touch interaction.

[0099] Specifically, a three-dimensional digital twin scene.

[0100] First, based on CAD drawings, point cloud scan data, or manual modeling of the actual warehouse, a 1:1 static 3D scene is constructed using a 3D engine such as Unity or Three.js, perfectly matching the layout, dimensions, and equipment models of the physical warehouse. This scene includes all key entities such as shelves, conveyor lines, AGVs, stacker cranes, workstations, and personnel avatars.

[0101] Subsequently, a dynamic data binding mechanism is established, defining which attribute of which entity object in the 3D scene is associated with the real-time data stream flowing in from the central data platform. For example:

[0102] Equipment operation data is bound to the corresponding equipment model, driving its state changes. The "stacker's current position coordinates" data transmitted from the PLC will drive the movement of the stacker crane in the 3D model; the "conveyor line running speed" data will drive the animation speed of the conveyor belt in the 3D model; the "fault code" will trigger the equipment model to turn red and stop the animation.

[0103] The dynamic inventory data is linked to the corresponding shelf locations. The "inventory quantity" or "occupancy rate" of each location is mapped to the visual attributes of the location model, such as dynamically presenting the inventory level through color gradients (from blue / empty to red / full) or height stretching, forming an intuitive "inventory heat map".

[0104] Personnel behavior data is used to bind the real-time coordinate data stream of personnel to the corresponding personnel avatar model through the UWB positioning system, so that the employee's movement trajectory and current position can be reproduced in real time in the three-dimensional scene.

[0105] Specifically, a dynamic hierarchical early warning mechanism. This mechanism intelligently assesses and prioritizes abnormal situations.

[0106] The system receives anomaly probability values ​​output from the anomaly detection model, classifies these probabilities into levels based on a pre-set dynamic grading strategy, and takes corresponding early warning response measures. The level classification and early warning response measures are shown in the table below:

[0107]

[0108] This example demonstrates how this embodiment constructs a digital twin model of the warehouse, maps real-time collected multidimensional data into a virtual space, and utilizes an original hybrid anomaly detection algorithm to achieve real-time monitoring of the operation process, intelligent early warning of potential risks, and in-depth traceability of historical data, thereby improving the transparency, efficiency, and security of warehouse operations.

[0109] As described above, the warehousing operation data visualization and analysis method provided in this application can collect multi-dimensional data during the warehousing operation process in real time, perform preliminary data processing at edge nodes, and then send the data to the hierarchical database of the central data platform for storage. It performs initial screening of the multi-dimensional data using statistical detection algorithms, secondary screening using a pre-trained isolated forest algorithm, and final screening using a time-series model. Ultimately, it determines the anomaly probability value of abnormal data in each dimension, constructs a visualization model of the real warehousing space, maps the multi-dimensional data to the visualization model, determines the warning level based on the detected anomaly probability value, and provides warning prompts in the visualization model according to the warning level. This improves the efficiency and accuracy of warehousing management.

[0110] In one embodiment of the warehouse operation data visualization and analysis method of this application, it may further include the following:

[0111] Step S201: Collect personnel behavior data at a frequency of once per second using an ultra-wideband (UWB) or Bluetooth AOA positioning system.

[0112] Step S202: Collect equipment operation data from the programmable logic controller (PLC), IoT gateway, or the equipment's built-in sensors at a frequency of once every 3 seconds.

[0113] Step S203: Collect the operation status data every 5 seconds through the warehouse management system (WMS) interface or handheld PDA terminal.

[0114] Step S204: Collect environmental parameter data from the distributed environmental sensing nodes deployed in the warehouse at a frequency of once every 10 seconds.

[0115] Step S205: Collect inventory dynamic data from the warehouse management system (WMS) database or machine vision inventory system every 60 seconds.

[0116] Optionally, in this embodiment, multi-dimensional data acquisition is used to achieve full-domain situational awareness, break the "data silos" in traditional systems, and build an interconnected data acquisition network by covering five core dimensions.

[0117] Specifically, the system collects operational status data: through an API interface, it deeply integrates with the Warehouse Management System (WMS) or directly from operational devices such as handheld terminals (PDAs) to capture key node information such as the creation, allocation, start, completion, and time elapsed of picking lists at a frequency of approximately once every 5 seconds. This enables micro-level tracking of the task flow.

[0118] Specifically, for equipment operation data acquisition: For automated warehouses, conveyor lines, AGVs, stacker cranes, and other equipment, the system communicates with the programmable logic controller (PLC) via an IoT gateway or directly, collecting their core operating parameters at a high frequency of approximately once every 3 seconds. These parameters include motor speed, real-time power and load of the AGV, three-dimensional coordinate position of the stacker crane, and various fault codes and status words reported by the equipment itself. This is equivalent to installing a real-time operating black box for each critical piece of equipment.

[0119] Specifically, environmental parameter data collection: By deploying a distributed sensor network at key nodes within the warehouse (such as temperature-controlled warehouses, high-value goods areas, and high-intensity work areas), the system collects data on environmental factors such as temperature, humidity, light intensity, noise, and even dust concentration approximately every 10 seconds. This brings environmental factors that have a potential impact on operational quality and safety into the monitoring scope.

[0120] Specifically, personnel behavior data collection: Utilizing high-precision indoor positioning technologies such as UWB (Ultra-Wideband) or Bluetooth AOA (Angle of Arrival), the system can track the precise location of pickers and operators within the warehouse in real time (accurate to the centimeter level) at an extremely high frequency of once per second, thereby generating their movement trajectories, dwell time at specific storage locations, and operation frequency. This provides unprecedentedly detailed data for analyzing personnel efficiency and optimizing workflow design.

[0121] Specifically, the system collects dynamic inventory data by monitoring change logs in the WMS database and combining this with machine vision inventory results from cameras fixed to shelves or inspection robots. It updates the occupancy status of all storage locations and the inventory quantity changes for specific SKUs approximately every minute, and calculates inventory accuracy. This achieves near real-time monitoring of inventory consistency between records and actual stock.

[0122] Through step S205, this embodiment constructs a full-element digital image covering "people, machines, materials, methods, and environment", enabling in-depth analysis of complex correlation issues.

[0123] In one embodiment of the warehouse operation data visualization and analysis method of this application, it may further include the following:

[0124] Step S301: Transmit the multi-dimensional data to the edge node for preliminary data processing through a preset transmission protocol to determine the corresponding pre-processed data. The preliminary data processing includes at least one of data filtering and data cleaning.

[0125] Step S302: Send the preprocessed data to the central data platform and store it in a preset hierarchical storage database. The hierarchical storage database includes a hot data storage layer and a cold data storage layer. The hot data storage layer is used to store data generated within a preset time threshold and accessed frequently. The cold data storage layer is used to store historical data accessed infrequently beyond the preset time threshold.

[0126] Optionally, in this embodiment, preliminary data processing is performed.

[0127] If massive amounts of real-time data were to flow directly into the central platform, it would put enormous pressure on network and computing resources. Therefore, all of the above data is aggregated to edge computing nodes via the lightweight MQTT protocol for preliminary processing, and then transmitted to the central data platform efficiently and reliably.

[0128] Edge computing nodes perform critical preprocessing tasks, including:

[0129] Data cleaning and formatting: Filter out obviously invalid data and convert data from different sources and with different formats into a standardized JSON or Avro format.

[0130] Preliminary aggregation and calculation: Perform preliminary aggregation on high-frequency data. For example, calculate the average speed and approximate location per second for personnel location data once per second at the edge before uploading, thereby reducing the amount of data transmission.

[0131] Edge rule triggering: Executes simple rule judgments with extremely high real-time requirements. For example, when UWB data detects that a person has entered a dangerous area, the edge node can respond in milliseconds and send instructions directly to the on-site audible and visual alarms without waiting for cloud-based decisions.

[0132] By introducing edge computing nodes in step S302, this embodiment reduces system pressure by shifting some computing power to the edge.

[0133] In one embodiment of the warehouse operation data visualization and analysis method of this application, it may further include the following:

[0134] Step S401: Extract data marked as normal operating cycle from historical multi-dimensional warehouse data to construct an initial training set. Train an initial isolated forest model based on the initial training set. Make predictions on the initial training set based on the initial isolated forest model and remove the training data sample with the highest initial outlier score. Determine the corresponding purified training dataset and extract the corresponding training data features.

[0135] Step S402: Construct a final isolated forest model composed of multiple isolated trees. Randomly assign a dedicated data subset to each isolated tree from the purified training dataset. Simultaneously, randomly extract a dedicated feature subset from the training data features for each isolated tree. Train the final isolated forest model based on the dedicated data subset and the dedicated feature subset to determine the corresponding pre-trained isolated forest algorithm. The dedicated data subset is obtained from the training dataset through sliding time window sampling and is used to learn data patterns in different time segments. The dedicated feature subset contains features in at least two different data dimensions and is used for anomaly isolation in a cross-dimensional feature space.

[0136] Optionally, in this embodiment, this step is the construction and pre-training process of the Isolation Forest algorithm.

[0137] Specifically, firstly, there's the self-purification process of the model training data. The core lies in extracting normal behavioral patterns from historical data through an iterative bootstrapping method.

[0138] First, the system retrieves historical data from the warehouse information platform. This data comes from a period that has been explicitly marked as normal operation by domain experts or business rules. Based on this data, the system constructs an initial training set, which theoretically represents the baseline of the health status of warehouse operations.

[0139] Subsequently, using this initial training set, the system trains and generates an initial Isolation Forest model. This initial model is not directly used for production environment detection, but rather serves as a data filter. After training, the system immediately applies this initial model inversely to its training data itself—the initial training set—for a comprehensive predictive analysis. The model calculates an initial anomaly score for each data sample in the initial training set, quantifying the degree of deviation of each sample point from the overall data pattern.

[0140] Next, the system automatically identifies and removes the most anomalous data points from the initial training set according to a preset elimination strategy (e.g., eliminating the top 5% of samples with the highest outlier scores). These eliminated samples may be minor anomalies that went unnoticed during historical normal cycles, transient noise during data collection, or data generated under special but unrecorded temporary operations. After eliminating these samples, the remaining data constitutes a purified training dataset. The features contained in this new dataset are the training data features extracted and confirmed by the system to represent the core normal patterns of warehouse operations.

[0141] This step effectively solves a common problem in machine learning in the industrial field: the contamination of training data. In real-world warehousing environments, absolutely pure, normal data is almost nonexistent; any historical data segment may contain unrecorded flaws. If noisy data is used directly to train the model, the model will mistakenly identify these abnormal patterns as part of the normal process, resulting in decreased sensitivity in production environments and an inability to effectively identify genuine anomalies.

[0142] Optionally, in this embodiment, the model is trained in a structured manner.

[0143] After obtaining high-quality training data, a final isolated forest model that integrates the characteristics of warehouse operation data is constructed. The main body of the model consists of a large set of isolated trees. For each isolated tree in the set, the system performs a targeted data and feature allocation:

[0144] In terms of data, the system introduces a sliding time window sampling mechanism. Data within a specific continuous time period is extracted from the timeline of the training set and assigned as a subset to the current isolation tree. This means that each tree "sees" a complete but partial fragment of business activity.

[0145] In terms of features, when the system randomly selects a subset of features for each tree, it imposes a mandatory constraint: the specific subset of features must contain features from at least two different data dimensions. For example, a subset may contain both "transport line rate" from the "equipment operation data" dimension and "ambient temperature" from the "environmental parameter data" dimension, rather than just "transport line rate" and its historical average, which are features of the same dimension.

[0146] Based on the above rules, the system defines a unique training perspective for each isolated tree. This perspective is applied not only within a specific time context but also across different data dimensions. Using these carefully configured subsets of data and features, each tree in the entire isolated forest model is constructed in parallel, thereby completing the training of the final model and obtaining a pre-trained isolated forest algorithm that can be used online.

[0147] This step of building an isolated forest targets two inherent characteristics of warehouse data: temporal correlation and multi-dimensional correlation.

[0148] Specifically, by sampling through a sliding time window, the trees in the model can learn normal behavioral patterns over a specific time span. This allows the model to not only focus on the numerical values ​​at a single point in time but also to perceive how various parameters should reasonably coordinate and change within an operational cycle. For example, it can identify normal patterns with temporal logic, such as "after the picking task starts, the movement trajectory of personnel in the area and the AGV scheduling frequency should show a matching increase within a certain period of time," thus making it easier to detect anomalies that violate this temporal logic.

[0149] Forced assignment of cross-dimensional feature subsets is key to enabling models to discover complex, hidden anomalies. It compels each tree to consider the interactions between features from different business dimensions during its construction. For example, a model node might simultaneously judge based on "stall crane energy consumption" and "warehouse ambient temperature," thereby learning to identify correlated anomalies such as "abnormally high energy consumption, but the ambient temperature has not risen synchronously," which cannot be revealed by any single dimension and indicates a potential equipment failure.

[0150] Through step S402, this embodiment successfully obtained a pre-trained isolated forest through the purification of the training dataset and the structural design of the isolated forest, laying the foundation for subsequent data anomaly detection.

[0151] In one embodiment of the warehouse operation data visualization and analysis method of this application, it may further include the following:

[0152] Step S501: Standardize and vectorize the initial screening data to determine the corresponding unified feature vector, wherein the standardization adopts the Z-score standardization method;

[0153] Step S502: Input the unified feature vector into the pre-trained isolated forest algorithm, determine the anomaly score output by the isolated forest algorithm, filter out the abnormal sample points with anomaly scores higher than the threshold based on the anomaly scores, and determine the corresponding re-screening data.

[0154] Optionally, in this embodiment, since the initial screening data comes from multiple dimensions such as equipment, environment, and operation, and their numerical units and magnitudes vary greatly, the system adopts the Z-score standardization method. For each data feature, the mean and standard deviation are calculated from its historical data. Then, the current value is subtracted from the mean and divided by the standard deviation, thereby transforming the values ​​of all features to the same standard scale with a mean of 0 and a standard deviation of 1.

[0155] Subsequently, vectorization processing is performed. The system will collect multiple standardized heterogeneous data (such as standardized temperature values, rate values, and power values) from the same time slice or a very short time window, and splice and combine them in a preset order to form a numerical array with a unified structure and fixed dimensions, namely a unified feature vector.

[0156] Z-score standardization eliminates dimensional differences, preventing features with large numerical ranges from having excessive weight when calculating distances or partitioning data spaces, ensuring that the model treats all features equally. Vectorization, on the other hand, integrates the originally loose and heterogeneous multi-source data streams into standardized mathematical inputs that machine learning models can directly recognize and process. This forms the data foundation for anomaly learning.

[0157] Optionally, in this embodiment, a unified feature vector is input into the trained Isolation Forest model. For each input feature vector, the algorithm traverses all decision trees and calculates the average path length required to reach a leaf node from the root. The shorter the path, the more likely the data point is to be isolated, and the greater the probability that it is considered an outlier. Finally, the algorithm outputs a continuous outlier score.

[0158] This anomaly score is compared with a preset threshold. All original data samples corresponding to feature vectors with anomaly scores higher than this threshold will be filtered out and marked as anomaly candidates for the second stage.

[0159] Through step S502, this embodiment successfully and effectively controls the amount and quality of data entering the next stage of the deep learning model.

[0160] In one embodiment of the warehouse operation data visualization and analysis method of this application, it may further include the following:

[0161] Step S601: Construct a static 3D scene model based on the Unity engine that corresponds 1:1 to the physical space layout, shelves, and equipment locations of the real warehouse;

[0162] Step S602: Establish dynamic binding relationships between the multi-dimensional data and the corresponding entity objects in the static three-dimensional scene model, and determine the corresponding visualization model. Among them, the equipment operation data is bound to the corresponding equipment model to present the equipment operation status, location coordinates and fault status in the three-dimensional scene; the inventory dynamic data is bound to the corresponding shelf model to present the shelf occupancy status in the three-dimensional scene; and the personnel behavior data is bound to the corresponding personnel avatar model to present the personnel movement trajectory and location in the three-dimensional scene.

[0163] Optionally, in this embodiment, this step constructs a three-dimensional digital twin scene.

[0164] A 1:1 digital twin model of the warehouse is built using the Unity / Three.js engine. This model not only statically reproduces the warehouse structure, but also dynamically presents the real-time status of shelves (inventory levels), automated equipment (running / stopping / failure), and personnel (location and trajectory) through real-time data. It is visualized using color coding (e.g., red - alarm, green - normal) and information labels.

[0165] Specifically, based on CAD drawings, point cloud scan data, or manual modeling of the actual warehouse, a 1:1 static 3D scene is constructed using 3D engines such as Unity or Three.js, perfectly matching the layout, dimensions, and equipment models of the physical warehouse. This scene includes all key entities such as shelves, conveyor lines, AGVs, stacker cranes, workstations, and personnel avatars.

[0166] Subsequently, a dynamic data binding mechanism is established, defining which attribute of which entity object in the 3D scene is associated with the real-time data stream flowing in from the central data platform. For example:

[0167] Equipment operation data is bound to the corresponding equipment model, driving its state changes. The "stacker's current position coordinates" data transmitted from the PLC will drive the movement of the stacker crane in the 3D model; the "conveyor line running speed" data will drive the animation speed of the conveyor belt in the 3D model; the "fault code" will trigger the equipment model to turn red and stop the animation.

[0168] The dynamic inventory data is linked to the corresponding shelf locations. The "inventory quantity" or "occupancy rate" of each location is mapped to the visual attributes of the location model, such as dynamically presenting the inventory level through color gradients (from blue / empty to red / full) or height stretching, forming an intuitive "inventory heat map".

[0169] Personnel behavior data is used to bind the real-time coordinate data stream of personnel to the corresponding personnel avatar model through the UWB positioning system, so that the employee's movement trajectory and current position can be reproduced in real time in the three-dimensional scene.

[0170] Through step S602, this embodiment successfully constructed a digital twin simulation framework, which enables the monitoring and understanding of real-time data by mapping real-time collected multidimensional data to this virtual space.

[0171] In one embodiment of the warehouse operation data visualization and analysis method of this application, it may further include the following:

[0172] Step S701: Dynamically divide the numerical range of the abnormal probability value to determine the corresponding warning level, which includes Level I, Level II, Level III and Level IV;

[0173] Step S702: If the warning level is Level I, then the warning-related objects are highlighted in the visualization model;

[0174] Step S703: If the warning level is Level II, then sound an alarm for the relevant objects in the visualization model and control the relevant objects to flash in the model;

[0175] Step S704: If the warning level is Level III, then issue a moderate alarm to the relevant objects in the visualization model, and suspend the relevant work processes at the same time;

[0176] Step S705: If the warning level is Level IV, a serious alarm will be issued to the relevant objects in the visualization model, and the associated equipment will be automatically shut down and the emergency plan will be activated.

[0177] Optionally, in this embodiment, this step defines a dynamic hierarchical early warning mechanism, which intelligently judges and prioritizes abnormal situations.

[0178] The system receives anomaly probability values ​​from the anomaly detection model, classifies the anomaly probability values ​​into levels according to a preset dynamic grading strategy, and takes corresponding early warning response measures.

[0179] Specifically, there are four levels in total.

[0180] When the alert level is determined to be Level I (information alert), the system simply highlights the relevant object (such as a piece of equipment or a storage location) with a color or adds additional information in the digital twin model. The effect is to provide managers with potential anomaly clues without interfering with normal operations, achieving silent situational awareness.

[0181] When the alert level is upgraded to Level II (minor warning), the system maintains visual brightness while adding a gentle audible alert and controls the object model to flash periodically on the interface. This multimodal alert is designed to proactively attract the operator's attention, ensuring that minor anomalies are detected and addressed promptly to prevent escalation.

[0182] Upon reaching Level III (moderate alarm), the response strategy shifts from reactive alerts to in-process intervention. In addition to providing strong visual warnings on the interface using more prominent symbols (such as red borders and warning icons), the core action of the system is to automatically send a command to the control system to suspend the work processes associated with the abnormal object. This directly prevents the abnormal situation from worsening, buying time for manual intervention and creating safer conditions.

[0183] Level IV (Critical Alarm), the highest level, indicates that the system determines a significant risk. At this point, the system will trigger the highest-level emergency response mechanism: firstly, it will send an emergency shutdown command to the relevant production line equipment, forcing it into a safe mode to physically block the risk; secondly, it will simultaneously activate pre-set emergency plans, such as automatically notifying the emergency response team and pushing out handling flowcharts. The ultimate effect of this series of automated operations is to achieve a closed loop from prediction to control, maximizing the safety of personnel, equipment, and assets, and minimizing potential losses.

[0184] Through step S705, this embodiment successfully improves the management efficiency of warehouse operation data by classifying anomalies into levels and corresponding anomaly handling schemes.

[0185] To improve the efficiency and accuracy of warehouse management, this application provides an embodiment of a warehouse operation data visualization and analysis device for implementing all or part of the aforementioned warehouse operation data visualization and analysis method. See [link to embodiment]. Figure 2 The warehouse operation data visualization and analysis device specifically includes the following components:

[0186] The multi-dimensional data acquisition module 10 is used to collect multi-dimensional data in real time during the warehousing operation according to a preset sampling frequency. The multi-dimensional data includes operation status data, equipment operation data, environmental parameter data, personnel behavior data, and inventory dynamic data. The multi-dimensional data is transmitted to the edge node for preliminary data processing through a preset transmission protocol and then sent to the central data platform for storage in a preset hierarchical storage database.

[0187] The abnormal data prediction module 20 is used to filter outliers in the multi-dimensional data in the hierarchical storage database according to a preset statistical detection algorithm, determine the corresponding initial screening data, input the initial screening data into a pre-trained isolated forest algorithm for association anomaly identification, determine the corresponding secondary screening data, input the secondary screening data into a preset time series model for continuous anomaly identification, determine the corresponding final screening data, and determine the anomaly probability value of each dimension of data according to the initial screening data, secondary screening data and final screening data.

[0188] The abnormal status visualization module 30 is used to map the multi-dimensional data to a visualization model of the real warehouse space according to the preset three-dimensional digital twin engine, classify the abnormal probability value into warning levels, determine the corresponding warning levels, and provide visual warning prompts for each dimension of data in the visualization model according to the warning levels, so as to realize the visualization analysis of warehouse operation data. The three-dimensional digital twin engine is communicatively connected to the central data platform.

[0189] As described above, the warehousing operation data visualization and analysis device provided in this application embodiment can collect multi-dimensional data during the warehousing operation process in real time, perform preliminary data processing at edge nodes, and then send the data to the hierarchical database of the central data platform for storage. It performs initial screening of the multi-dimensional data using statistical detection algorithms, secondary screening using a pre-trained isolated forest algorithm, and final screening using a time-series model. Ultimately, it determines the anomaly probability value of abnormal data in each dimension, constructs a visualization model of the real warehousing space, maps the multi-dimensional data to the visualization model, determines the warning level based on the detected anomaly probability value, and provides warning prompts in the visualization model according to the warning level. This improves the efficiency and accuracy of warehousing management.

[0190] From a hardware perspective, in order to improve the efficiency and accuracy of warehouse management, this application provides an embodiment of an electronic device for implementing all or part of the aforementioned warehouse operation data visualization and analysis method. The electronic device specifically includes the following components:

[0191] The system comprises a processor, memory, a communications interface, and a bus; wherein the processor, memory, and communications interface communicate with each other via the bus; the communications interface is used to realize information transmission between the warehouse operation data visualization and analysis method and core business systems, user terminals, and related databases and other related devices; the logic controller can be a desktop computer, tablet computer, or mobile terminal, etc., and this embodiment is not limited to these. In this embodiment, the logic controller can be implemented with reference to the embodiments of the warehouse operation data visualization and analysis method in the present embodiment, and the content of the embodiments of the warehouse operation data visualization and analysis method is incorporated herein, and repeated parts will not be described again.

[0192] It is understood that the user terminal may include smartphones, tablet computers, network set-top boxes, portable computers, desktop computers, personal digital assistants (PDAs), in-vehicle devices, smart wearable devices, etc. Among these, the smart wearable devices may include smart glasses, smartwatches, smart bracelets, etc.

[0193] In practical applications, some parts of the warehouse operation data visualization and analysis method can be executed on the electronic device side as described above, or all operations can be completed on the client device. The choice can be made based on the processing power of the client device and the limitations of the user's usage scenario. This application does not impose any limitations on this. If all operations are completed on the client device, the client device may further include a processor.

[0194] The aforementioned client device may have a communication module (i.e., a communication unit) that can communicate with a remote server to achieve data transmission. The server may include a server on the task scheduling center side; in other implementation scenarios, it may also include a server on an intermediate platform, such as a server on a third-party server platform that has a communication link with the task scheduling center server. The server may include a single computer device, a server cluster consisting of multiple servers, or a distributed server structure.

[0195] Figure 3 This is a schematic block diagram illustrating the system configuration of the electronic device 9600 according to an embodiment of this application. Figure 3 As shown, the electronic device 9600 may include a central processing unit 9100 and a memory 9140; the memory 9140 is coupled to the central processing unit 9100. It is worth noting that... Figure 3 This is an example; other types of structures can also be used to supplement or replace this structure to achieve telecommunications functions or other functions.

[0196] In one embodiment, the warehouse operation data visualization and analysis method can be integrated into the central processing unit 9100. The central processing unit 9100 can be configured to perform the following controls:

[0197] Step S101: Collect multi-dimensional data in real time during the warehousing operation according to the preset sampling frequency. The multi-dimensional data includes operation status data, equipment operation data, environmental parameter data, personnel behavior data, and inventory dynamic data. The multi-dimensional data is transmitted to the edge node for preliminary data processing through the preset transmission protocol and then sent to the central data platform for storage in the preset hierarchical storage database.

[0198] Step S102: Filter outliers from the multi-dimensional data in the hierarchical storage database according to a preset statistical detection algorithm to determine the corresponding initial screening data. Input the initial screening data into a pre-trained isolated forest algorithm for association anomaly identification to determine the corresponding secondary screening data. Input the secondary screening data into a preset time series model for continuous anomaly identification to determine the corresponding final screening data. Determine the anomaly probability value of each dimension of the data based on the initial screening data, secondary screening data, and final screening data.

[0199] Step S103: Based on the preset 3D digital twin engine, the multi-dimensional data is mapped to a visualization model of the real warehouse space. The abnormal probability value is divided into warning levels, the corresponding warning level is determined, and the data of each dimension is visualized and warned in the visualization model according to the warning level, so as to realize the visualization analysis of warehouse operation data. The 3D digital twin engine is communicatively connected to the central data platform.

[0200] As described above, the electronic device provided in this application collects multi-dimensional data during the warehousing operation in real time. After preliminary data processing at the edge nodes, the data is sent to the hierarchical database of the central data platform for storage. The multi-dimensional data is initially screened according to a statistical detection algorithm, then further screened according to a pre-trained isolated forest algorithm, and finally screened according to a time series model. The abnormal probability value of abnormal data in each dimension is determined, a visualization model of the real warehousing space is constructed, and the multi-dimensional data is mapped to the visualization model. The warning level is determined according to the detected abnormal probability value, and a warning prompt is given in the visualization model according to the warning level. This can improve the efficiency and accuracy of warehousing management.

[0201] In another embodiment, the warehouse operation data visualization and analysis method can be configured separately from the central processing unit 9100. For example, the warehouse operation data visualization and analysis method can be configured as a chip connected to the central processing unit 9100, and the function of the warehouse operation data visualization and analysis method can be realized through the control of the central processing unit.

[0202] like Figure 3 As shown, the electronic device 9600 may further include: a communication module 9110, an input unit 9120, an audio processor 9130, a display 9160, and a power supply 9170. It is worth noting that the electronic device 9600 does not necessarily need to include these components. Figure 3 All components shown; in addition, the electronic device 9600 may also include Figure 3 For components not shown, please refer to existing technologies.

[0203] like Figure 3 As shown, the central processing unit 9100, sometimes also referred to as a controller or operating control, may include a microprocessor or other processor device and / or logic device, which receives inputs and controls the operation of various components of the electronic device 9600.

[0204] The memory 9140 may be, for example, one or more of a cache, flash memory, hard drive, removable media, volatile memory, non-volatile memory, or other suitable devices. It may store the aforementioned failure-related information, and also store a program for executing that information. The central processing unit 9100 may execute the program stored in the memory 9140 to perform information storage or processing, etc.

[0205] Input unit 9120 provides input to central processing unit 9100. Input unit 9120 may be, for example, a keypad or touch input device. Power supply 9170 provides power to electronic device 9600. Display 9160 displays images and text. Display may be, for example, an LCD display, but is not limited thereto.

[0206] The memory 9140 can be a solid-state memory, such as a read-only memory (ROM), random access memory (RAM), a SIM card, etc. It can also be a memory that retains information even when power is off, can be selectively erased, and contains more data; examples of this type of memory are sometimes referred to as EPROMs. The memory 9140 can also be some other type of device. The memory 9140 includes a buffer memory 9141 (sometimes referred to as a buffer). The memory 9140 may include an application / function storage unit 9142 for storing application programs and function programs or processes for executing the operation of the electronic device 9600 via the central processing unit 9100.

[0207] The memory 9140 may also include a data storage unit 9143 for storing data, such as contacts, digital data, pictures, sounds, and / or any other data used by the electronic device. The driver storage unit 9144 of the memory 9140 may include various drivers for the electronic device's communication functions and / or for performing other functions of the electronic device (such as messaging applications, address book applications, etc.).

[0208] The communication module 9110 is a transmitter / receiver that sends and receives signals via the antenna 9111. The communication module 9110 is coupled to the central processing unit 9100 to provide input signals and receive output signals, which is the same as in a conventional mobile communication terminal.

[0209] Based on different communication technologies, multiple communication modules 9110 can be configured in the same electronic device, such as cellular network modules, Bluetooth modules, and / or wireless LAN modules. The communication module 9110 is also coupled to a speaker 9131 and a microphone 9132 via an audio processor 9130 to provide audio output via the speaker 9131 and receive audio input from the microphone 9132, thereby realizing typical telecommunications functions. The audio processor 9130 may include any suitable buffer, decoder, amplifier, etc. Furthermore, the audio processor 9130 is also coupled to a central processing unit 9100, enabling on-device recording via the microphone 9132 and on-device playback of stored sound via the speaker 9131.

[0210] Embodiments of this application also provide a computer-readable storage medium capable of implementing all steps of the warehouse operation data visualization and analysis method with a server or client as the execution subject in the above embodiments. The computer-readable storage medium stores a computer program that, when executed by a processor, implements all steps of the warehouse operation data visualization and analysis method with a server or client as the execution subject in the above embodiments. For example, when the processor executes the computer program, it implements the following steps:

[0211] Step S101: Collect multi-dimensional data in real time during the warehousing operation according to the preset sampling frequency. The multi-dimensional data includes operation status data, equipment operation data, environmental parameter data, personnel behavior data, and inventory dynamic data. The multi-dimensional data is transmitted to the edge node for preliminary data processing through the preset transmission protocol and then sent to the central data platform for storage in the preset hierarchical storage database.

[0212] Step S102: Filter outliers from the multi-dimensional data in the hierarchical storage database according to a preset statistical detection algorithm to determine the corresponding initial screening data. Input the initial screening data into a pre-trained isolated forest algorithm for association anomaly identification to determine the corresponding secondary screening data. Input the secondary screening data into a preset time series model for continuous anomaly identification to determine the corresponding final screening data. Determine the anomaly probability value of each dimension of the data based on the initial screening data, secondary screening data, and final screening data.

[0213] Step S103: Based on the preset 3D digital twin engine, the multi-dimensional data is mapped to a visualization model of the real warehouse space. The abnormal probability value is divided into warning levels, the corresponding warning level is determined, and the data of each dimension is visualized and warned in the visualization model according to the warning level, so as to realize the visualization analysis of warehouse operation data. The 3D digital twin engine is communicatively connected to the central data platform.

[0214] As described above, the computer-readable storage medium provided in this application collects multi-dimensional data during real-time warehousing operations, performs preliminary data processing at edge nodes, and then sends the data to a hierarchical database on a central data platform for storage. It performs initial screening of the multi-dimensional data using statistical detection algorithms, secondary screening using a pre-trained isolated forest algorithm, and final screening using a time-series model. This process ultimately determines the anomaly probability value of abnormal data in each dimension, constructs a visualization model of the real warehousing space, maps the multi-dimensional data to the visualization model, determines the warning level based on the detected anomaly probability value, and provides warning prompts in the visualization model according to the warning level. This improves the efficiency and accuracy of warehousing management.

[0215] Embodiments of this application also provide a computer program product capable of implementing all steps in the warehouse operation data visualization and analysis method with the execution subject being a server or client in the above embodiments. When this computer program / instruction is executed by a processor, it implements the steps of the warehouse operation data visualization and analysis method. For example, the computer program / instruction implements the following steps:

[0216] Step S101: Collect multi-dimensional data in real time during the warehousing operation according to the preset sampling frequency. The multi-dimensional data includes operation status data, equipment operation data, environmental parameter data, personnel behavior data, and inventory dynamic data. The multi-dimensional data is transmitted to the edge node for preliminary data processing through the preset transmission protocol and then sent to the central data platform for storage in the preset hierarchical storage database.

[0217] Step S102: Filter outliers from the multi-dimensional data in the hierarchical storage database according to a preset statistical detection algorithm to determine the corresponding initial screening data. Input the initial screening data into a pre-trained isolated forest algorithm for association anomaly identification to determine the corresponding secondary screening data. Input the secondary screening data into a preset time series model for continuous anomaly identification to determine the corresponding final screening data. Determine the anomaly probability value of each dimension of the data based on the initial screening data, secondary screening data, and final screening data.

[0218] Step S103: Based on the preset 3D digital twin engine, the multi-dimensional data is mapped to a visualization model of the real warehouse space. The abnormal probability value is divided into warning levels, the corresponding warning level is determined, and the data of each dimension is visualized and warned in the visualization model according to the warning level, so as to realize the visualization analysis of warehouse operation data. The 3D digital twin engine is communicatively connected to the central data platform.

[0219] As described above, the computer program product provided in this application collects multi-dimensional data during the warehousing operation in real time, performs preliminary data processing at edge nodes, and then sends the data to the hierarchical database of the central data platform for storage. It performs initial screening of the multi-dimensional data using a statistical detection algorithm, secondary screening using a pre-trained isolated forest algorithm, and final screening using a time-series model. This process ultimately determines the anomaly probability value of abnormal data in each dimension, constructs a visualization model of the real warehousing space, maps the multi-dimensional data to the visualization model, determines the warning level based on the detected anomaly probability value, and provides warning prompts in the visualization model according to the warning level. This improves the efficiency and accuracy of warehousing management.

[0220] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, apparatus, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied 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.

[0221] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (devices), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in one or more blocks of the flowchart illustrations and / or one or more blocks of the block diagrams.

[0222] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means that implement the functions specified in one or more flowcharts and / or one or more block diagrams.

[0223] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, such that the instructions, which execute on the computer or other programmable apparatus, provide steps for implementing the functions specified in one or more flowcharts and / or one or more block diagrams.

[0224] Specific embodiments have been used to illustrate the principles and implementation methods of this invention. The descriptions of the embodiments above are only for the purpose of helping to understand the method and core ideas of this 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 this invention. Therefore, the content of this specification should not be construed as a limitation of this invention.

Claims

1. A method for visualizing and analyzing warehouse operation data, characterized in that, The method includes: Multi-dimensional data during the warehousing operation is collected in real time according to a preset sampling frequency. The multi-dimensional data includes operation status data, equipment operation data, environmental parameter data, personnel behavior data, and inventory dynamic data. The multi-dimensional data is transmitted to edge nodes for preliminary data processing through a preset transmission protocol and then sent to the central data platform for storage in a preset hierarchical storage database. Outlier filtering is performed on the multi-dimensional data in the hierarchical storage database according to the preset statistical detection algorithm to determine the corresponding initial screening data. The initial screening data is then input into the pre-trained isolated forest algorithm for association anomaly identification to determine the corresponding secondary screening data. The secondary screening data is then input into the preset time series model for continuous anomaly identification to determine the corresponding final screening data. Based on the initial screening data, secondary screening data, and final screening data, the anomaly probability value of each dimension of the data is determined. The multi-dimensional data is mapped to a visualization model of the real warehouse space using a preset 3D digital twin engine. The abnormal probability values ​​are classified into warning levels, and the corresponding warning levels are determined. Based on the warning levels, the data of each dimension is visualized and warned in the visualization model to realize the visualization analysis of warehouse operation data. The 3D digital twin engine is communicatively connected to the central data platform.

2. The method for visualizing and analyzing warehouse operation data according to claim 1, characterized in that, The process of collecting multi-dimensional data in real time during warehousing operations according to a preset sampling frequency includes: Personnel behavior data is collected every second via ultra-wideband (UWB) or Bluetooth AOA positioning systems. Equipment operation data is collected every 3 seconds from programmable logic controllers (PLCs), IoT gateways, or the equipment's built-in sensors. Operation status data is collected every 5 seconds via the warehouse management system (WMS) interface or a handheld PDA terminal. Environmental parameter data is collected from distributed environmental sensing nodes deployed in the warehouse every 10 seconds. Inventory dynamic data is collected from the warehouse management system (WMS) database or machine vision inventory system every 60 seconds.

3. The method for visualizing and analyzing warehouse operation data according to claim 1, characterized in that, The process of transmitting the multi-dimensional data to edge nodes via a preset transmission protocol for preliminary data processing, followed by transmission to the central data platform and storage in a preset hierarchical storage database, includes: The multi-dimensional data is transmitted to the edge node through a preset transmission protocol for preliminary data processing to determine the corresponding pre-processed data. The preliminary data processing includes at least one of data filtering and data cleaning. The preprocessed data is sent to the central data platform and stored in a preset hierarchical storage database. The hierarchical storage database includes a hot data storage layer and a cold data storage layer. The hot data storage layer is used to store data generated within a preset time threshold and accessed frequently, while the cold data storage layer is used to store historical data accessed infrequently beyond the preset time threshold.

4. The method for visualizing and analyzing warehouse operation data according to claim 1, characterized in that, Before inputting the initial screening data into the pre-trained Isolation Forest algorithm for association anomaly identification and determining the corresponding rescreening data, the process includes: Data marked as normal operating cycles are extracted from historical multi-dimensional warehouse data to construct an initial training set. An initial isolated forest model is trained based on the initial training set. The initial isolated forest model is used to predict the initial training set, and the training data samples with the highest initial outlier scores are removed. The corresponding purified training dataset is determined and the corresponding training data features are extracted. A final isolated forest model composed of multiple isolation trees is constructed. A dedicated data subset is randomly assigned to each isolation tree from the purified training dataset. At the same time, a dedicated feature subset is randomly extracted from the training data features for each isolation tree. The final isolated forest model is trained based on the dedicated data subset and the dedicated feature subset to determine the corresponding pre-trained isolated forest algorithm. The dedicated data subset is obtained from the training dataset through sliding time window sampling to learn data patterns in different time segments. The dedicated feature subset contains features from at least two different data dimensions to isolate anomalies in a cross-dimensional feature space.

5. The method for visualizing and analyzing warehouse operation data according to claim 4, characterized in that, The step of inputting the initial screening data into a pre-trained isolated forest algorithm for association anomaly identification to determine the corresponding rescreening data includes: The initial screening data is standardized and vectorized to determine the corresponding unified feature vector, wherein the standardization adopts the Z-score standardization method; The unified feature vector is input into the pre-trained isolated forest algorithm to determine the anomaly score output by the isolated forest algorithm. Based on the anomaly score, abnormal sample points with anomaly scores higher than the threshold are selected to determine the corresponding re-screening data.

6. The method for visualizing and analyzing warehouse operation data according to claim 1, characterized in that, The visualization model that maps the multi-dimensional data to the real warehouse space based on a preset 3D digital twin engine includes: A static 3D scene model built using the Unity engine, which corresponds 1:1 to the physical space layout, shelves, and equipment locations of a real warehouse. Establish dynamic binding relationships between the multi-dimensional data and corresponding entity objects in the static 3D scene model, and determine the corresponding visualization model. Specifically, equipment operation data is bound to the corresponding equipment model to present the equipment's operating status, location coordinates, and fault status in the 3D scene; inventory dynamic data is bound to the corresponding shelf model to present the shelf occupancy status in the 3D scene; and personnel behavior data is bound to the corresponding personnel avatar model. Binding is used to represent the movement trajectory and position of people in the three-dimensional scene.

7. The method for visualizing and analyzing warehouse operation data according to claim 1, characterized in that, The process of classifying the abnormal probability values ​​into warning levels, determining the corresponding warning levels, and providing visual warning prompts for each dimension of data in the visualization model based on the warning levels includes: The numerical range of the abnormal probability value is dynamically divided to determine the corresponding warning level, which includes Level I, Level II, Level III and Level IV; If the warning level is Level I, then the warning-related objects will be highlighted in the visualization model; If the warning level is Level II, then sound alerts will be given to the relevant objects in the visualization model, and the relevant objects will be controlled to flash in the model; If the warning level is Level III, then a moderate alarm will be issued to the relevant objects in the visualization model, and the relevant work processes will be suspended. If the warning level is Level IV, a severe alarm will be issued to the relevant objects in the visualization model, and the associated equipment will be automatically shut down, and the emergency plan will be activated.

8. A data visualization and analysis device for warehouse operations, characterized in that, The device includes: The multi-dimensional data acquisition module is used to collect multi-dimensional data in real time during the warehousing operation according to a preset sampling frequency. The multi-dimensional data includes operation status data, equipment operation data, environmental parameter data, personnel behavior data, and inventory dynamic data. The multi-dimensional data is transmitted to the edge node for preliminary data processing through a preset transmission protocol and then sent to the central data platform for storage in a preset hierarchical storage database. The abnormal data prediction module is used to filter outliers from the multi-dimensional data in the hierarchical storage database according to a preset statistical detection algorithm, determine the corresponding initial screening data, input the initial screening data into a pre-trained isolated forest algorithm for association anomaly identification, determine the corresponding secondary screening data, input the secondary screening data into a preset time series model for continuous anomaly identification, determine the corresponding final screening data, and determine the anomaly probability value of each dimension of data based on the initial screening data, secondary screening data, and final screening data. The abnormal status visualization module is used to map the multi-dimensional data to a visualization model of the real warehouse space based on a preset three-dimensional digital twin engine, classify the abnormal probability value into warning levels, determine the corresponding warning levels, and provide visual warning prompts for each dimension of data in the visualization model according to the warning levels, thereby realizing the visualization analysis of warehouse operation data. The three-dimensional digital twin engine is communicatively connected to the central data platform.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the warehouse operation data visualization and analysis method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the warehouse operation data visualization and analysis method according to any one of claims 1 to 7.

Images