Intelligent server assembly method and device, electronic equipment and storage medium

Abstract

The embodiment of the invention provides an intelligent server assembling method and device, electronic equipment and a storage medium, and is used for the technical field of Internet of Things. The intelligent server assembly method comprises the following steps: updating an offline simulation model and an online simulation model according to a model structure data set and synchronously acquired real-time state data flow, and performing multi-objective optimization on a mixed production task through the offline simulation model according to received service parameters to obtain a candidate resource allocation data set, and evaluating and screening the candidate resource allocation data set through the online simulation model to obtain optimal resource allocation data, and performing conflict detection and routing planning on the optimal resource allocation data according to the space-time resource constraint set to generate a server assembly adjustment instruction sequence. According to the method, through multi-source real-time perception, candidate generation through offline simulation, online proxy rapid screening and space-time resolution and closed-loop execution, optimal resource allocation and high-precision traceability in the mixed line assembly production process are achieved.

Description

Technical Field

[0001] This application relates to the field of Internet of Things (IoT) technology, and in particular to a server intelligent assembly method, apparatus, electronic device, and storage medium. Background Technology

[0002] The increasing demand for customized servers for enterprise and cloud data centers, coupled with a wide variety of product models and rapid updates, coupled with customers' frequent requests for short lead times and small-batch, multi-variety supply, makes traditional dedicated line mass production unable to meet the market's elastic demand. Mixed-line production can process servers with different configurations in parallel on the same production line, shortening delivery cycles, reducing inventory and switchover costs, and responding to order fluctuations by adjusting capacity allocation in real time. This achieves a better balance between cost, lead time, and service levels, which is the core driving force behind the market's push for mixed-line transformation.

[0003] Currently, most server assembly lines still rely on static rules and part-time manual scheduling for mixed-line resource allocation. Common practices include dividing the workshop into several functional areas and putting them into production in batches, using schedules exported from ERP / MES for pre-production scheduling, or using a combination of simple heuristic scheduling (such as first-come-first-served, priority scheduling) and manual intervention to allocate AGVs and workstation tasks. In terms of path and vehicle scheduling, they mostly rely on traditional path following and local obstacle avoidance strategies, and most scheduling decisions are updated offline or semi-real-time, lacking automated support for complex multi-objective trade-offs.

[0004] The existing solutions mentioned above rely on static rules, offline scheduling, and human experience. Their main drawback is that they cannot quickly calculate and issue verified alternative scheduling solutions that are acceptable in terms of multiple objectives (completion time, energy consumption, and idle rate) when short-term disturbances (such as equipment failure, order insertion, or material delays) occur, resulting in delayed response, low resource utilization, and incomplete quality traceability.

[0005] It should be noted that the above content is not necessarily prior art, nor is it intended to limit the scope of patent protection of this application. Summary of the Invention

[0006] This application provides a server intelligent assembly method, apparatus, electronic device, and storage medium to solve or alleviate one or more of the technical problems mentioned above.

[0007] One aspect of this application provides a server intelligent assembly method, the method comprising: A real-time status data stream is generated based on synchronously collected production line sensor data, and the preset offline simulation model and the preset online simulation model are updated based on the preset model structure dataset and the real-time status data stream. The updated offline simulation model is used to perform multi-objective optimization on the preset mixed-line production tasks based on the received business parameters, and a candidate resource allocation dataset is obtained. The candidate resource allocation dataset is evaluated and screened using the updated online simulation model to obtain the optimal resource allocation data. Based on a preset set of spatiotemporal resource constraints, conflict detection and dynamic routing planning are performed on the optimal resource allocation data to generate a sequence of server assembly and adjustment instructions.

[0008] Optionally, generating a real-time status data stream based on synchronously collected production line sensor data includes: The production line sensor data is collected synchronously at preset time intervals through a pre-set multi-source heterogeneous sensor network. Based on the preset data format and collection timestamp, the production line sensor data is classified and aggregated and packaged in a time series to obtain a real-time status dataset. The real-time status dataset is quality verified and sorted according to time sequence to generate a real-time status data stream.

[0009] Optionally, updating the preset offline simulation model and the preset online simulation model based on the preset model structure dataset and the real-time state data stream includes: Based on the preset model structure dataset, the real-time status data stream is classified and processed to generate production status parameter groups corresponding to each model structure in each time period. The production state parameter group is used to perform periodic parallel reset processing on the corresponding model structure state parameters to update the preset offline simulation model and the preset online simulation model.

[0010] Optionally, the step of performing multi-objective optimization processing on the preset mixed-line production task based on the received business parameters using the updated offline simulation model to obtain a candidate resource allocation dataset includes: Using a pre-defined non-dominated sorting genetic algorithm, the pre-defined mixed-line production tasks are sorted by multi-objective optimization based on the received business parameters to generate an initial set of production task execution sequences. The updated offline simulation model is used to perform numerical simulation and execution evaluation on the initial production task execution sequence set to obtain the evaluation score corresponding to each production task execution sequence. Based on the evaluation scores, the initial production task execution sequence set is sorted non-dominated to generate a non-dominated level sequence, and the congestion degree corresponding to each production task execution sequence in the initial production task execution sequence set is calculated. By using a preset iterative filtering method, production task execution sequences are selected from the initial production task execution sequence set based on the non-dominant level sequence, the evaluation score, and the crowding degree, generating a candidate resource allocation dataset.

[0011] Optionally, the step of evaluating and filtering the candidate resource allocation dataset using the updated online simulation model to obtain the optimal resource allocation data includes: The updated online simulation model is used to perform numerical simulation processing on the candidate resource allocation dataset to generate instantaneous predicted state data corresponding to each candidate resource allocation dataset. Based on the preset set of business preference weight coefficients and the instantaneous predicted state data, calculate the weighted comprehensive score corresponding to each candidate resource allocation data; Based on the weighted composite score, the candidate resource allocation data with the highest score is selected from the candidate resource allocation dataset as the optimal allocation data to be verified; The updated online simulation model performs short-term simulation verification and fine-tuning on the optimal allocation data to be verified according to a preset time period to obtain the optimal resource allocation data.

[0012] Optionally, the step of performing conflict detection and dynamic routing planning on the optimal resource allocation data based on a preset set of spatiotemporal resource constraints to generate a server assembly adjustment instruction sequence includes: Step S41: Classify and parse the optimal resource allocation data according to the preset resource allocation device tag set to obtain the resource allocation time sequence data corresponding to each resource allocation device; Step S42: Based on the preset spatiotemporal resource constraint set, perform spatiotemporal conflict detection on the resource allocation time series data for path nodes; Step S43: According to the preset business priority conflict resolution rules, recursively adjust the identified conflict path nodes to update the corresponding resource allocation time sequence data. Repeat steps S42 to S43 until there are no conflicting path nodes between all resource allocation timing data, so as to synthesize the server assembly adjustment instruction sequence corresponding to each resource allocation device according to the updated resource allocation timing data through a preset drive instruction synthesis method.

[0013] Optionally, the method further includes: Obtain the latest real-time status data stream, and extract the target dataset from the real-time status data stream and the current server assembly adjustment instruction sequence according to the preset target data type; Based on the order data in the target dataset and the business parameters, the task status of the preset mixed-line production task and the preset business priority conflict resolution rules are modified in real time to update the mixed-line production task and the business priority conflict resolution rules. Based on the resource consumption data in the target dataset, the inventory resource data of the preset spatiotemporal resource constraint set is corrected in real time to update the spatiotemporal resource constraint set.

[0014] Another aspect of this application provides a server intelligent assembly apparatus, the apparatus comprising: The model update module is used to generate a real-time status data stream based on synchronously collected production line sensor data, so as to update the preset offline simulation model and the preset online simulation model according to the preset model structure dataset and the real-time status data stream. The task optimization module is used to perform multi-objective optimization processing on the preset mixed-line production tasks based on the received business parameters through the updated offline simulation model, and obtain a candidate resource allocation dataset. The evaluation and screening module is used to evaluate and screen the candidate resource allocation dataset through the updated online simulation model to obtain the optimal resource allocation data. The instruction planning module is used to perform conflict detection and dynamic routing planning on the optimal resource allocation data based on a preset set of spatiotemporal resource constraints, and generate a server assembly adjustment instruction sequence.

[0015] Another aspect of this application provides a computer device, including: At least one processor; and A memory that is communicatively connected to the at least one processor; Wherein: the memory stores instructions that can be executed by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the method as described above.

[0016] Another aspect of this application provides a computer-readable storage medium storing computer instructions that, when executed by a processor, implement the method described above.

[0017] Another aspect of this application provides a computer program product including a computer program that, when executed by a processor, implements the method described above.

[0018] The embodiments of this application employing the above-described technical solution may have the following advantages: 1. Through multi-source real-time sensing and unified time-series data flow, achieve transparency of the entire production process and product / process level traceability.

[0019] 2. Pareto candidates are generated through offline high-precision simulation, and then quickly screened and verified by an online agent, which significantly improves the efficiency and resource utilization of production scheduling and path decision-making.

[0020] 3. By adopting a combination of spatiotemporal conflict resolution, dynamic routing and MPC closed-loop execution, it can quickly respond to equipment failures or order interruptions, reduce the risk of line stoppages and collisions, and ensure execution accuracy and quality control. Attached Figure Description

[0021] The accompanying drawings exemplify embodiments and form part of the specification, serving together with the textual description to explain exemplary implementations of the embodiments. The illustrated embodiments are for illustrative purposes only and do not limit the scope of the claims. Throughout the drawings, the same reference numerals refer to similar but not necessarily identical elements.

[0022] Figure 1 The schematic diagram illustrates a process flow of a server intelligent assembly method according to Embodiment 1 of this application; Figure 2 This schematic diagram illustrates the functional block diagram of a server intelligent assembly device according to Embodiment 2 of this application; Figure 3 A schematic diagram of the hardware architecture of a computer device according to Embodiment 3 of this application is shown. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application. All other embodiments obtained by those skilled in the art based on the embodiments in this application without inventive effort are within the scope of protection of this application.

[0024] It should be noted that the descriptions involving "first," "second," etc., in the embodiments of this application are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, features defined with "first" or "second" may explicitly or implicitly include at least one feature. Furthermore, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed in this application.

[0025] In the description of this application, it should be understood that the numerical labels before the steps do not indicate the order of the steps, but are only used to facilitate the description of this application and to distinguish each step, and therefore should not be construed as a limitation of this application.

[0026] For ease of understanding, an exemplary runtime environment is provided below: The server intelligent assembly method provided in this application operates within a mixed-line production control system for server assembly. This mixed-line production control system includes, but is not limited to, a multi-source heterogeneous sensor network, a resource allocation equipment cluster, an edge computing gateway, and a cloud computing platform. The multi-source heterogeneous sensor network is deployed on various entities on the production line (such as trolleys, workbenches, warehouses, and work-in-process), and includes UWB positioning modules, RFID readers, vision sensors, torque sensors, and IMUs. The multi-source heterogeneous sensor network is used to collect data from each execution node in the mixed-line production process, converting the physical world's state (position, identity, image, torque) into the raw data signals of this application. The resource allocation equipment cluster includes programmable execution units such as intelligent trolleys, unmanned forklifts, and robotic arms. In an optional embodiment, the resource allocation equipment cluster can be an independent intelligent warehousing system. As the executor of resource allocation instructions, the resource allocation equipment cluster accurately completes physical operations such as movement, handling, and assembly according to the received instructions, and feeds back its own state (such as position and power) to the mixed-line production control system. The edge computing gateway is deployed near the production line. Its core function is to perform preliminary filtering, timestamp synchronization (following the PTP protocol), and format unification on the raw data collected by the multi-source heterogeneous sensor network to generate a regular real-time status data stream. In a preferred embodiment, to improve system response speed and robustness, the edge computing gateway can be further configured as a distributed computing node to execute local offline simulation tasks in parallel within its area. The cloud computing platform serves as the computing hub of the mixed-line production control system of this application, carrying and running the digital twin engine, intelligent decision engine, and adaptive updater to execute the core algorithm processes of simulation model updating, multi-objective optimization, online screening, dynamic routing planning, and adaptive adjustment of system parameters in the technical solution of this application.

[0027] For example, the mixed-line production control system deeply integrates a dynamic path planning mechanism and a multi-source data sensing network on the basis of a circular production line structure. The system's hardware architecture includes a circular conveyor line, intelligent trolleys, and modular workbenches. The modular workbenches include conveyor belts that integrate multi-dimensional sensing systems. The intelligent trolleys travel on the circular conveyor line, serving as a component transport tool between the warehouse and the production line. The circular conveyor line is a pre-deployed transport route between the warehouse and several modular workbenches.

[0028] The technical solution of this application will be described below through multiple embodiments, using a mixed-line production control system as the main implementer. It should be understood that these embodiments can be implemented in many different forms and should not be construed as being limited to the embodiments described herein.

[0029] Example 1 like Figure 1 The diagram shown is a flowchart of a server intelligent assembly method provided in an embodiment of this application. The server intelligent assembly method provided in this embodiment includes the following steps.

[0030] Step S1: Generate a real-time status data stream based on the synchronously collected production line sensor data, and update the preset offline simulation model and the preset online simulation model according to the preset model structure dataset and the real-time status data stream.

[0031] To achieve intelligent decision-making for dynamic path planning and real-time status monitoring in a mixed-line server production system, a complete data processing pipeline from physical sensing to digital model synchronization is first required. This application uses a multi-source heterogeneous sensor network architecture to collect real-time sensor data from each node of the production line. The obtained production line sensor data refers to the raw signal set generated by sensing devices distributed in intelligent sliding carts, modular workbenches, and material buffer areas, such as UWB positioning modules, optical landmark recognition units, inertial measurement units, anti-metal RFID systems, structured light scanners, and torque sensors, synchronously collecting the states of various physical entities during server assembly. The production line sensor data acquisition process in this application follows a strict hierarchical collaborative mechanism: At the intelligent trolley level, the UWB module continuously transmits signals at a frequency of 3.5-6.5 GHz. It calculates the absolute position coordinates by measuring the signal flight time with a fixed base station. The accuracy of this signal is periodically calibrated using zero-drift correction triggered by optical QR code landmarks at preset points. Meanwhile, the inertial measurement unit (IMU) captures the vehicle's instantaneous acceleration and angular velocity changes at a sampling frequency of 200 Hz. These three types of data are input in real-time to an onboard embedded fusion processor. This processor executes a dedicated algorithm based on error-state Kalman filtering. This algorithm uses the absolute coordinates of the UWB as observations to suppress integral drift in inertial navigation, uses the millimeter-level precision data from the optical landmarks as hard constraints to reset accumulated errors, and utilizes the high-frequency characteristics of the inertial data to perform motion interpolation within the filtering framework. Finally, it outputs an optimal pose estimation sequence containing three-dimensional spatial coordinates, attitude angles, and velocity vectors.

[0032] In the material management stage, the RFID system deployed in metallic environments employs anti-metal tags with electromagnetic isolation design. The readers maintain a stable electromagnetic field in complex multipath environments through adaptive power control and frequency hopping communication algorithms. Its access strategy uses a time-division multiplexing-based polling mechanism, dividing the 100-millisecond communication window into multiple orthogonal time slots. A central scheduler activates readers in different physical areas in a time-division manner, enabling stable batch reading of nearly a hundred tag identification information and material status even in environments with strong interference. For critical assembly processes, the structured light scanner does not continuously upload point cloud data. Instead, a lightweight convolutional neural network runs on an FPGA chip integrated in the field, performing real-time analysis of the acquired depth image stream. Only when specific events such as component installation being in place, screw hole alignment deviation exceeding 0.2 mm, or cable loosening are detected, is a single frame of a JPEG-XS compressed key image, along with its corresponding bounding box coordinates and pose estimation results, triggered for uploading. All sensing units achieve microsecond-level time synchronization through a Precision Time Protocol (PTP) network based on the IEEE 1588 standard. The master clock is a rubidium atomic clock deployed in the plant area. Synchronization signals are distributed to each sensing node through a time-sensitive network switch to ensure that each data packet is marked with a precise PTP timestamp at the source.

[0033] The core role of production line sensor data lies in constructing a complete digital mirror of the physical assembly line in the information space, providing a unique and reliable data foundation for subsequent simulation model updates and optimization decisions. For example, when a server chassis of model "SRV-2840" enters the assembly station, the active RFID tag on its tray is instantly identified by the reader, triggering the system to load the complete process route for this model from the database. At the same time, the structured light scanner confirms that the CPU heatsink installation angle deviation is 0.3 degrees, the torque sensor uploads the final torque value of the motherboard fixing screws in real time as 2.4 N·m with the curve conforming to the standard paradigm, and the intelligent trolley reports its relative distance to the target station as 1.2 meters through the multimodal positioning unit and is expected to complete docking in 3 seconds. All these heterogeneous data streams spanning different devices and sampling rates, after time alignment, jointly define the precise production status of "SRV-2840 is in the motherboard fastening process at station A3".

[0034] After data synchronization and acquisition are completed, due to differences in sampling frequencies among different sensing units (e.g., 200Hz for inertial measurement units and burst-triggered RFID events), the edge computing gateway immediately performs standardized preprocessing on the raw data stream. The gateway's built-in stream processing engine first aligns all input data according to the acquisition timestamps, resampling it to a unified 100Hz time series. Subsequently, data quality verification rules are executed, including outlier removal based on the Laida criterion and filling in instantaneous data gaps caused by network jitter using a time series linear interpolation algorithm. The cleaned data is then encapsulated into a unified Avro serialization format, which contains standardized data packets (i.e., real-time state datasets) containing metadata such as the data body, timestamps accurate to microseconds, data source device IDs, and quality identifiers. The final collection of these data packets arranged chronologically constitutes the real-time state data stream. This standardized intermediate product ensures the continuity of data in time and the reliability of data quality. For example, when a set of inertial data from the trolley is lost due to communication interference, the verification module will mark the data in that interval as invalid, and the interpolation algorithm will reconstruct the motion trajectory based on the normal data points before and after, thereby ensuring the integrity of the dataset.

[0035] Subsequently, the real-time status data stream is pushed to the message queue, driving the update process of both offline and online simulation models. The model structure dataset used in this application defines the parameter mapping relationships required by the simulation engine. For example, it maps the "sliding car position" field to the state variables of the kinematic model and the "torque reading" to the input conditions of the quality inspection logic. The classification processing module extracts the parameter subsets corresponding to each entity from the dataset based on this mapping relationship and the data's temporal order. For example, for the simulation model of "Workstation 5," the system filters all visual recognition results, torque upload events, and work-in-process RFID status change records associated with that workstation from the dataset, aggregates them into a unified view centered on the workstation, and generates the production status parameter group for that workstation within the current time window.

[0036] Based on the production state parameter set obtained above, differentiated update operations are performed on the two types of simulation models. This application employs a periodic parallel reset strategy for the high-precision offline simulation model, where the offline simulation model runs on distributed computing nodes composed of edge computing gateways, with each edge node responsible for simulating a production line segment within its assigned physical area. Every complete production cycle or when the system accumulates significant state changes, the scheduler pauses the simulation process, forcibly resetting the state of each virtual entity to the latest physical value extracted from the real-time state dataset. For example, when the RFID system reports that the inventory of "Memory Bar Box B-12" has fallen below the safety threshold, the edge node responsible for that area will immediately update the inventory variable in the simulation model during the next reset cycle, ensuring that subsequent optimization sorting reflects real material constraints. In conjunction with this, this application employs a near real-time synchronous update mechanism for the online simulation model. This simplified model runs on a cloud computing platform, dynamically adjusting its internal state mirror by continuously subscribing to key state change events in the real-time state dataset. When a smart trolley in the physical world stops due to a sudden malfunction, the event of its status changing from "running" to "shutdown" is preprocessed by the edge gateway and quickly uploaded to the cloud simulation model. The model then dynamically reconstructs all affected material flow paths, providing an instant response capability for upcoming dynamic routing planning requests.

[0037] Through the aforementioned layered update mechanism that combines edge and cloud collaboration, the system in this application ensures both the accuracy of offline optimization and the agility of online decision-making, laying a solid state foundation for the intelligent control of the entire server mixed-line production system.

[0038] Step S2: Using the updated offline simulation model, perform multi-objective optimization processing on the preset mixed-line production tasks based on the received business parameters to obtain a candidate resource allocation dataset.

[0039] After completing the real-time state synchronization and model update in step S1, the system enters the production scheme decision-making stage, which focuses on multi-objective optimization. In this application, the production scheme decision-making stage utilizes a pre-defined non-dominated sorting genetic algorithm to deeply optimize mixed-line production tasks. The candidate resource allocation dataset specifically refers to the set of alternative production schemes formed through multiple rounds of genetic iteration and simulation evaluation, achieving a balance among multiple optimization objectives. Each scheme in this dataset includes a complete product launch sequence, equipment scheduling strategy, and resource allocation plan.

[0040] First, an initial population (i.e., an initial set of production task execution sequences) is constructed based on the received business parameters. These business parameters include the newly added "SRV-5280" high-priority server order, the capacity increase requirement for production line A, and the production progress data of the five currently executing server models. The non-dominated sorting genetic algorithm used in this application integrates the dynamically input business parameters with the preset mixed-line production tasks, generating an initial population containing 200 chromosomes. Each chromosome employs a double-stranded encoding structure: the product sorting chain uses a process-based encoding method, discretizing the assembly processes of different server models into gene units, such as sequences like [SRV-2840-motherboard installation, SRV-3640-power supply testing, SRV-5280-CPU assembly] representing the mixed production order; the resource allocation chain uses real-number encoding, defining continuous parameters such as the speed coefficient of each intelligent trolley and the safety stock threshold of the buffer area. This dual-chain design can simultaneously optimize discrete production sequencing and continuous resource allocation. For example, when the system needs to prioritize the urgent order of "SRV-5280", the non-dominated sorting genetic algorithm will generate more chromosome sequences containing early processes of this model during the initialization phase.

[0041] After obtaining the initial production task execution sequence set, the updated offline simulation model from step S1 above is invoked for batch numerical simulation. The high-precision offline simulation model used in this application runs on distributed edge computing nodes. Each node carries a detailed digital image of the corresponding physical area, including the precise kinematic model of the intelligent trolley, the random fault simulation of the workbench, and the dynamic prediction of material consumption. The offline simulation model decodes the chromosome into specific control instructions, simulates an eight-hour production cycle at ultra-real-time speed, and records three key performance indicators: total completion time measures the duration from the start of the first process to the last product coming off the line; total transportation energy consumption accumulates the power consumption of all intelligent trolleys during the simulation cycle; and total resource idle time counts the idle time of equipment waiting for materials or instructions. For example, when simulating a specific chromosome, the model may show that the total completion time is extended by 47 minutes because the power testing process of "SRV-3640" and the firmware burning of "SRV-2840" compete for the same test station. At the same time, the third intelligent trolley generates an additional 12% energy consumption due to unreasonable path planning. These precisely quantified outputs constitute the evaluation score vector corresponding to each chromosome.

[0042] Based on the evaluation score set obtained from the above simulation operations, a hierarchical structure of solutions is established through non-dominated sorting. The sorting adopted in this application follows the Pareto dominance principle: if solution A is not inferior to solution B on all objectives and is strictly superior on at least one objective, then A is said to dominate B. The first non-dominated layer contains all elite solutions that are not dominated by any other solution, the second layer contains solutions dominated only by solutions in the first layer, and so on. For example, when dealing with a population containing 200 chromosomes, the algorithm may identify 28 non-dominated solutions forming the first layer, which embody different trade-off strategies. For example, one solution reduces the completion time by 20% at the cost of an additional 15% energy consumption, while another solution reduces both energy consumption and completion time by increasing equipment idle time. To maintain the diversity of solution distribution in the objective space, the crowding degree of each solution within each non-dominated layer is calculated. This index measures the distance between a solution and its neighbors in the objective space. The formula is that the crowding degree of solution i within its non-dominated layer is equal to the normalized sum of the differences in function values ​​between adjacent solutions on each objective function. Specifically, this is expressed as calculating the difference in function values ​​between solutions i-1 and i+1 along each objective dimension, dividing it by the range of that objective in the entire population, and finally summing over all objectives. A high crowding degree indicates a large blank area around the solution, giving it priority in the selection process. This mechanism ensures that the optimization process explores all potential trade-offs rather than converging to a single local optimum.

[0043] Finally, candidate data are screened through evolutionary optimization of the population. This application employs a multi-round genetic operation combining non-dominated rank sequences, evaluation scores, and crowding levels. The selection operation uses a binary tournament strategy, randomly selecting two chromosomes each time, prioritizing individuals with higher non-dominated ranks, and selecting individuals with higher crowding levels when ranks are the same. This mechanism ensures elite solutions are preserved while promoting solution set diversity. The crossover operation employs differentiated strategies for the bi-chain structure: for the product sorting chain, a method similar to sequential crossover is used, randomly selecting process sub-sequences from the parent chromosome and passing them on to the offspring while maintaining relative order; for the resource allocation chain, simulated binary crossover is used, interpolating between parent parameter vectors to generate offspring parameters. The mutation operation randomly alters the chromosome structure with low probability, including reverse mutation of product sequences and Gaussian perturbation of resource parameters. The offspring population generated in each iteration is merged with the parent population, and a new generation is re-selected through rapid non-dominated sorting and crowding calculation, repeating this cycle until a preset termination condition of 200 generations is reached. The final output candidate resource allocation dataset contains dozens of non-dominated solutions distributed on the Pareto front. Each solution corresponds to a complete mixed-line production scheme. For example, the schemes in the dataset may include specific decisions such as "prioritizing the completion of the SRV-5280 emergency order, adjusting the No. 3 sliding car to high-speed mode, and increasing the safety stock of the No. 5 buffer area to 25 units". These schemes provide a wealth of optimization options for subsequent online real-time decision-making.

[0044] Step S3: Evaluate and filter the candidate resource allocation dataset using the updated online simulation model to obtain the optimal resource allocation data.

[0045] After generating a candidate resource allocation dataset containing multiple Pareto optimal solutions through step S2, the process proceeds to the online rapid decision-making stage. The core task of this application's decision-making stage is to select the unique execution plan that best suits the current production scenario from multiple alternatives.

[0046] First, the candidate resource allocation dataset is numerically simulated using an updated online simulation model. The online simulation model, as a simplified digital image running on a cloud computing platform, retains the core business logic but significantly simplifies physical details compared to the offline model. For example, the motion model of the intelligent sliding cart is simplified to uniform motion, and the failure modes of the workbench are simplified to fixed-probability events. This simplification allows the complete simulation of a single scheme to be completed within seconds. The numerical simulation processing used in this application uses real-time status data synchronized by the front-end edge gateway as initial conditions, including the work-in-process status of each workstation, the immediate inventory level of the material buffer, and the equipment availability status. This embodiment of the application extrapolates the production process for each candidate scheme over the next 15 minutes, outputting the instantaneous predicted status data for each scheme. This instantaneous predicted status data includes estimated values ​​for three key performance indicators: the expected number of orders to be completed in the next stage, the predicted total system energy consumption, and the average equipment utilization rate. For example, when the system detects that the battery level of the No. 3 intelligent sliding cart is below 30%, the online simulation accurately reflects the impact of the additional charging time required for the vehicle in subsequent tasks, providing a longer completion time prediction for schemes involving tasks with that vehicle.

[0047] After obtaining the instantaneous predicted state data of each scheme, a weighted comprehensive score is calculated based on a preset set of business preference weight coefficients. The business preference weight coefficient set used in this application is a dynamically configurable three-dimensional vector, corresponding to the preference levels for completion time, energy consumption level, and equipment utilization rate, respectively. During normal production periods, the business preference weight coefficient set may be set to [0.4, 0.3, 0.3] to indicate a balanced consideration of various objectives; when urgent orders require expedited completion, the production administrator can adjust the time weight to 0.7, and reduce other weights accordingly. This dynamic adjustment mechanism ensures that the system's decisions are always consistent with real-time business objectives. The weighted comprehensive score calculation in this application uses a normalized weighted sum method. First, each objective value is subjected to minimum-maximum normalization, converting indicators of different dimensions into dimensionless values ​​between 0 and 1. Then, the normalized objective values ​​are multiplied by their corresponding weights and summed. A higher score indicates that the scheme better aligns with current business preferences. For example, a solution predicts that a critical order can be completed within 12 minutes in an emergency order scenario. Although the energy consumption is high, its time dimension normalization score can reach 0.9, and it can still obtain a high overall score with a time weight of 0.7.

[0048] Based on the weighted composite score set obtained from the above calculations, a sorting operation is then performed, and the highest-scoring scheme from the candidate resource allocation dataset is selected as the optimal allocation data to be verified. This selection process considers not only the absolute value of the score but also verifies the feasibility constraints of the scheme, such as checking whether the scheme requires the use of equipment currently under maintenance or whether it involves materials with insufficient inventory. When the selected scheme to be verified has such minor constraint violations, the system records these anomalies as key areas of focus for subsequent fine-tuning. For example, when the highest-scoring scheme has a potential conflict because its planned path passes through area 5 which is under maintenance, the scheme is still selected as the scheme to be verified, but the path conflict is marked as a critical point that needs optimization.

[0049] Finally, the updated online simulation model was used to perform short-term simulation verification and fine-tuning of the optimal allocation data to be verified. The verification process in this application adopted a higher-precision simulation mode, with the time period set to the future 30 minutes and the simulation step size refined from the conventional 1 second to 0.1 seconds, focusing on monitoring various conflicts and bottlenecks that may occur in the early stages of implementation. The fine-tuning process dynamically adjusted the scheme parameters based on simulation feedback. When a resource conflict was detected, a local search algorithm was launched to generate multiple variant schemes in the neighborhood of the scheme to be verified and to perform rapid evaluation. For example, when it was found that two intelligent sliding cars would have a time conflict at an intersection, the fine-tuning algorithm would generate multiple adjustment schemes: delaying the departure time of car 1 in scheme A, advancing the running speed of car 2 in scheme B, or replanning an alternative path for one of the cars in scheme C. The effect of each adjustment scheme was verified through rapid simulation, and finally the adjustment strategy with the greatest overall improvement was selected. Through multiple rounds of simulation-based exploratory adjustments, potential conflicts in the solution were eliminated one by one, performance indicators were further optimized, and the final output was optimal resource allocation data that fully adapted to the current production environment. This data not only contained theoretical superiority but also demonstrated practical feasibility, providing a fully validated input basis for subsequent dynamic routing planning. The entire screening and validation process is typically completed within 2-3 minutes, ensuring that the system can respond promptly to changes in the production environment while guaranteeing decision quality that meets the complex needs of mixed-line production.

[0050] Step S4: Based on the preset spatiotemporal resource constraint set, perform conflict detection and dynamic routing planning on the optimal resource allocation data to generate a server assembly adjustment instruction sequence.

[0051] After obtaining verified optimal resource allocation data, the process moves to the detailed planning phase of production instructions. The core task of this application in the planning phase is to transform the macro-level resource allocation scheme into executable, conflict-free micro-level operational instructions.

[0052] First, the optimal resource allocation data is categorized and parsed based on a pre-defined set of resource allocation equipment tags. This resource allocation equipment tag set serves as a unique identifier for each device, containing equipment codes such as "AGV-001" to "AGV-020" for intelligent sliding carts and "FORK-01" to "FORK-05" for unmanned forklifts. Each tag is associated with a database of the device's performance parameters, including key attributes such as maximum operating speed, load capacity, and battery capacity. Based on these tags, the parsing engine breaks down the unified resource allocation scheme into device-specific task sequences, generating resource allocation time-series data for each resource allocation device. This data structure precisely defines the action blueprint of each device in the time dimension, including elements such as task start time, target location coordinates, and expected operation duration. For example, when the resource allocation plan requires the “SRV-2840” server chassis to be transferred from the assembly area to the testing area, the parser will select “AGV-007” with sufficient power to perform the task based on the equipment status, and insert timing instruction points accurate to the second into its timing data, such as “10:15:30 arrives at assembly station A3, 10:15:45 completes loading, 10:16:20 arrives at the entrance of the testing area”.

[0053] After obtaining the complete resource allocation time-series dataset, conflict detection is then performed based on a preset set of spatiotemporal resource constraints. The spatiotemporal resource constraint set used in this application is a dynamically maintained digital map of the production environment, containing the network topology of the ring transport path, physical size limitations of each path node, maximum parallel capacity of the shared area, and information on currently occupied spatiotemporal blocks. The conflict detection algorithm is based on the principle of time window intersection. For each pair of potentially interactive device time-series data, it checks whether their planned time windows on each shared path node overlap. The specific detection formula is: for any two devices planning to use the same node's time windows [StartTime1, EndTime1] and [StartTime2, EndTime2], a spatiotemporal conflict is determined to exist when max(StartTime1, StartTime2) < min(EndTime1, EndTime2). For example, the system detected that “AGV-007” planned to use intersection N5 from 10:16:00 to 10:16:15, while “AGV-012” also planned to use the same node from 10:16:10 to 10:16:20. The two time windows overlapped by 5 seconds, and the system immediately marked this conflict as a high-risk event.

[0054] Upon detecting a conflict, the system initiates a recursive adjustment process based on preset business priority conflict resolution rules. The business priority conflict resolution rules adopted in this application include a multi-level processing strategy: the primary rule is order priority-oriented, granting path priority to devices carrying urgent orders; the secondary rule is optimal system efficiency, selecting the solution with the lowest overall latency after adjustment; supplementary rules include considerations such as prioritizing device power warnings and minimizing path modification costs. In this embodiment, the recursive adjustment employs a depth-first search strategy, first performing local replanning on the device time-series data related to the conflict node to generate multiple adjustment candidate solutions, including different strategies such as delay-first solutions, advance-later solutions, and path detour solutions. For example, when "AGV-007" and "AGV-012" collide at intersection N5, the resolver first checks that the "SRV-5280" order carried by "AGV-007" has a higher priority. Therefore, it generates a solution to delay "AGV-012" by 30 seconds. However, it then discovers that this adjustment will cause "AGV-012" to have a new secondary conflict with a third device. So the system backtracks to the original conflict point and adopts an alternative solution of having "AGV-007" depart 5 minutes earlier. Through this recursive trial and error, a conflict-free adjustment path is finally found.

[0055] This application continuously executes an iterative process of conflict detection and recursive adjustment until all spatiotemporal conflicts between devices are completely eliminated. Each adjustment generates a time-series data update that is synchronized in real-time to the global constraint set, ensuring that subsequent conflict detection is based on the latest state. When no new conflicts are detected in three consecutive rounds of detection, the system determines that the current resource allocation time-series data has reached a stable state and then initiates the drive instruction synthesis program. This synthesizer transforms abstract time-series data into specific control instructions that can be parsed by the devices, encapsulating them in standard JSON format, including machine-readable information such as target coordinate sequences, velocity curve configuration files, and action trigger conditions. For example, the instruction package generated for the intelligent sliding car precisely specifies executable commands such as "move along path points P15-P18 at a speed of 0.8 m / s, trigger position detection at beacon B27, and send an acknowledgment signal to the central controller after successful docking." The final output server assembly adjustment instruction sequence is distributed to the corresponding resource allocation devices via a time-sensitive network, while a timeline diagram of the entire scheduling scheme is visualized on the workstation terminal, providing production managers with a complete execution preview. This sophisticated instruction generation mechanism ensures accurate conversion from optimization schemes to physical execution, providing reliable scheduling guarantees for mixed-line server production.

[0056] After completing a full round of production scheduling instruction generation, the system enters a continuous optimization adaptive learning phase. In this application, this phase uses a real-time feedback mechanism to ensure that the execution system of this application can dynamically respond to environmental changes.

[0057] First, the system acquires the latest real-time status data stream, which continuously flows into the system from a multi-source heterogeneous sensor network. This stream includes rapidly changing production information such as the real-time location coordinates of the intelligent trolley, inventory changes in the material buffer area, and the work completion status of the workstations. The system then performs precise data extraction from this massive data stream based on a preset target data type. The target data type defines the key decision-making indicators that the system needs to focus on, including structured data categories such as order completion status change events, equipment downtime reports, and material consumption threshold alarms. The extraction process also cross-validates with the currently executing server assembly adjustment command sequence. For example, if the command sequence indicates that "AGV-007" should arrive at the test area at 10:25:00, but the real-time location data shows that the vehicle is still lingering in the assembly area, the system will immediately mark this discrepancy as critical anomaly data. The final extracted target dataset becomes an authoritative snapshot reflecting the system's true operating status, providing factual basis for subsequent updates to decision-making rules.

[0058] Based on order data and dynamic business parameters in the target dataset, the system initiates real-time correction of the preset mixed-line production task and business priority conflict resolution rules. The order data analysis module continuously tracks the progress status of each order. When it detects that the urgent order "SRV-5280" faces delivery risk due to delays in preceding processes, the task correction engine immediately initiates a replanning process, removing completed processes from the current mixed-line production task, recalculating the time constraints of the remaining processes, and integrating the newly added standard order "SRV-4720" to form an updated task set. Business parameters synchronously provide external decision context, such as customer-adjusted delivery deadlines and energy efficiency control indicators issued by the production management department. These parameters directly affect the weight configuration of priority rules. The update of business priority conflict resolution rules adopts a case-based reasoning mechanism. The system matches the current production scenario with historical decision cases, and automatically adjusts the priority judgment logic in the rule base when similar scenarios are identified. For example, when multiple cases occur where devices become congested in narrow areas due to following the "shortest distance" rule, the system will increase the weight of the "path smoothness" indicator in the rule base and correspondingly reduce the priority of "travel distance". This rule evolution based on actual operating results ensures that conflict resolution strategies are continuously optimized.

[0059] Simultaneously, real-time corrections are made to the preset spatiotemporal resource constraint set based on resource consumption data in the target dataset. The resource consumption monitor continuously tracks the requisition records and inventory changes of each material. When the RFID system reports that the inventory of "Memory Module MC-284" has suddenly dropped from 125 units to 15 units, the inventory resource data correction module responds immediately. It first verifies whether the consumption matches the issued production instructions. After confirming that there is no error, it updates the available quantity of the material to the constraint set and triggers the inventory warning flag. The update of spatiotemporal constraints is more complex, requiring comprehensive processing of multi-dimensional information such as changes in equipment status and changes in path availability. When the visual sensor detects that the effective passage width of the No. 3 main road has been reduced by 40% due to the placement of temporary equipment, the spatial constraint corrector recalculates the passage capacity of the path segment, adjusts the maximum number of parallel devices from 3 to 1, and updates the topology attributes of the path network. In terms of time, the system dynamically adjusts the standard time estimates for each process by analyzing the deviation between the actual operating time and the planned time of the equipment. For example, if it is found that the average time of the "firmware burning" process increases by 25% due to version upgrades, the time budget parameters of this process in the constraint set will be modified accordingly.

[0060] Through the aforementioned adaptive update process, a complete control loop is formed. The updated mixed-line production tasks, business priority conflict resolution rules, and spatiotemporal resource constraint sets take effect immediately, providing accurate input conditions for the next round of scheduling decisions. This continuous self-optimization mechanism enables the production system to cope with various uncertainties such as order changes, equipment anomalies, and resource fluctuations, always maintaining a high degree of consistency between the scheduling scheme and the actual environment. For example, when the system simultaneously receives two events: "insufficient material inventory" and "new order insertion," the updated business priority rules will prioritize ensuring the continuity of existing orders, automatically scheduling new orders until material replenishment is complete. At the same time, the updated constraint set ensures that the scheduling algorithm generates feasible production sequences under resource-constrained conditions. This dynamic coordination capability significantly improves the robustness and adaptability of the server-side mixed-line production system.

[0061] The server intelligent assembly method provided in this embodiment updates the offline and online simulation models based on the model structure dataset and synchronously collected real-time status data streams. The offline simulation model performs multi-objective optimization of mixed-line production tasks based on received business parameters to obtain a candidate resource allocation dataset. The online simulation model evaluates and filters the candidate resource allocation dataset to obtain the optimal resource allocation data. Based on a set of spatiotemporal resource constraints, conflict detection and routing planning are performed on the optimal resource allocation data to generate a server assembly adjustment instruction sequence. This application achieves optimal resource allocation and high-precision traceability through multi-source real-time perception, offline simulation to generate candidates, rapid online agent filtering, spatiotemporal resolution, and MPC closed-loop execution.

[0062] Example 2 like Figure 2 The diagram shown illustrates the functional block diagram of a server intelligent assembly device provided in an embodiment of this application. This device can be divided into one or more program modules. One or more program modules are stored in a storage medium and executed by one or more processors to complete the embodiment of this application. The program module referred to in this embodiment is a series of computer program instruction segments capable of performing a specific function. The following description will specifically introduce the function of each program module in this embodiment. Figure 2 As shown, the server intelligent assembly device 1000 may include: a model update module 1100, a task optimization module 1200, an evaluation and screening module 1300, an instruction planning module 1400, and a real-time correction module 1500, wherein: The model update module 1100 is used to generate a real-time status data stream based on synchronously collected production line sensor data, so as to update the preset offline simulation model and the preset online simulation model according to the preset model structure dataset and the real-time status data stream. The task optimization module 1200 is used to perform multi-objective optimization processing on the preset mixed-line production tasks based on the received business parameters through the updated offline simulation model, and obtain a candidate resource allocation dataset. The evaluation and screening module 1300 is used to evaluate and screen the candidate resource allocation dataset through the updated online simulation model to obtain the optimal resource allocation data. The instruction planning module 1400 is used to perform conflict detection and dynamic routing planning on the optimal resource allocation data according to a preset set of spatiotemporal resource constraints, and generate a server assembly adjustment instruction sequence.

[0063] As an optional embodiment, the model update module 1100 is specifically used for: The production line sensor data is collected synchronously at preset time intervals through a pre-set multi-source heterogeneous sensor network. Based on the preset data format and collection timestamp, the production line sensor data is classified and aggregated and packaged in a time series to obtain a real-time status dataset. The real-time status dataset is quality-verified and sorted according to time order to generate a real-time status data stream. Based on the preset model structure dataset, the real-time status data stream is classified and processed to generate production status parameter groups corresponding to each model structure in each time period. The production state parameter group is used to perform periodic parallel reset processing on the corresponding model structure state parameters to update the preset offline simulation model and the preset online simulation model.

[0064] As an optional embodiment, the task optimization module 1200 is specifically used for: Using a pre-defined non-dominated sorting genetic algorithm, the pre-defined mixed-line production tasks are sorted by multi-objective optimization based on the received business parameters to generate an initial set of production task execution sequences. The updated offline simulation model is used to perform numerical simulation and execution evaluation on the initial production task execution sequence set to obtain the evaluation score corresponding to each production task execution sequence. Based on the evaluation scores, the initial production task execution sequence set is sorted non-dominated to generate a non-dominated level sequence, and the congestion degree corresponding to each production task execution sequence in the initial production task execution sequence set is calculated. By using a preset iterative filtering method, production task execution sequences are selected from the initial production task execution sequence set based on the non-dominant level sequence, the evaluation score, and the crowding degree, generating a candidate resource allocation dataset.

[0065] As an optional embodiment, the evaluation and screening module 1300 is specifically used for: The updated online simulation model is used to perform numerical simulation processing on the candidate resource allocation dataset to generate instantaneous predicted state data corresponding to each candidate resource allocation dataset. Based on the preset set of business preference weight coefficients and the instantaneous predicted state data, calculate the weighted comprehensive score corresponding to each candidate resource allocation data; Based on the weighted composite score, the candidate resource allocation data with the highest score is selected from the candidate resource allocation dataset as the optimal allocation data to be verified; The updated online simulation model performs short-term simulation verification and fine-tuning on the optimal allocation data to be verified according to a preset time period to obtain the optimal resource allocation data.

[0066] As an optional embodiment, the instruction planning module 1400 is specifically used for: Step S41: Classify and parse the optimal resource allocation data according to the preset resource allocation device tag set to obtain the resource allocation time sequence data corresponding to each resource allocation device; Step S42: Based on the preset spatiotemporal resource constraint set, perform spatiotemporal conflict detection on the resource allocation time series data for path nodes; Step S43: According to the preset business priority conflict resolution rules, recursively adjust the identified conflict path nodes to update the corresponding resource allocation time sequence data. Repeat steps S42 to S43 until there are no conflicting path nodes between all resource allocation timing data, so as to synthesize the server assembly adjustment instruction sequence corresponding to each resource allocation device according to the updated resource allocation timing data through a preset drive instruction synthesis method.

[0067] As an optional embodiment, the server intelligent assembly device 1000 further includes a real-time correction module 1500, which is specifically used for: Obtain the latest real-time status data stream, and extract the target dataset from the real-time status data stream and the current server assembly adjustment instruction sequence according to the preset target data type; Based on the order data in the target dataset and the business parameters, the task status of the preset mixed-line production task and the preset business priority conflict resolution rules are modified in real time to update the mixed-line production task and the business priority conflict resolution rules. Based on the resource consumption data in the target dataset, the inventory resource data of the preset spatiotemporal resource constraint set is corrected in real time to update the spatiotemporal resource constraint set.

[0068] Example 3 Figure 3 This illustration schematically depicts the hardware architecture of a computer device 10000 suitable for implementing a server intelligent assembly method according to Embodiment 3 of this application. In some embodiments, the computer device 10000 may be a terminal device such as a smartphone, wearable device, tablet computer, personal computer, in-vehicle terminal, game console, virtual device, workbench, digital assistant, set-top box, or robot. In other embodiments, the computer device 10000 may be a rack server, blade server, tower server, or cabinet server (including independent servers or server clusters composed of multiple servers), etc. Figure 3 As shown, the computer device 10000 includes, but is not limited to: a memory 10010, a processor 10020, and a network interface 10030 that can communicate and be linked with each other via a system bus. Wherein: The memory 10010 includes at least one type of computer-readable storage medium, including flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 10010 may be an internal storage module of a computer device 10000, such as the hard disk or memory of the computer device 10000. In other embodiments, the memory 10010 may also be an external storage device of the computer device 10000, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the computer device 10000. Of course, the memory 10010 may also include both the internal storage module and the external storage device of the computer device 10000. In this embodiment, the memory 10010 is typically used to store the operating system and various application software installed on the computer device 10000, such as the program code for the server intelligent assembly method. In addition, the memory 10010 can also be used to temporarily store various types of data that have been output or will be output.

[0069] In some embodiments, processor 10020 may be a central processing unit (CPU), controller, microcontroller, microprocessor, or other chip. Processor 10020 is typically used to control the overall operation of computer device 10000, such as performing control and processing related to data interaction or communication with computer device 10000. In this embodiment, processor 10020 is used to run program code stored in memory 10010 or process data.

[0070] Network interface 10030 may include a wireless network interface or a wired network interface, which is typically used to establish a communication link between computer device 10000 and other computer devices. For example, network interface 10030 is used to connect computer device 10000 to an external terminal via a network, establishing a data transmission channel and communication link between computer device 10000 and the external terminal. The network may be an intranet, the Internet, Global System for Mobile Communication (GSM), Wideband Code Division Multiple Access (WCDMA), 4G network, 5G network, Bluetooth, Wi-Fi, or other wireless or wired networks.

[0071] It should be pointed out that, Figure 3 Only computer devices with components 10010-10030 are shown; however, it should be understood that it is not required to implement all of the shown components, and more or fewer components may be implemented instead.

[0072] In this embodiment, the server intelligent assembly method stored in memory 10010 can also be divided into one or more program modules and executed by one or more processors (such as processor 10020) to complete the embodiment of this application.

[0073] Obviously, those skilled in the art should understand that the modules or steps of the embodiments of this application described above can be implemented using general-purpose computer devices. They can be centralized on a single computer device or distributed across a network of multiple computer devices. Optionally, they can be implemented using computer-executable program code, thereby storing them in a storage device for execution by a computer device. In some cases, the steps shown or described can be performed in a different order than those presented here, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the embodiments of this application are not limited to any particular combination of hardware and software.

[0074] It should be noted that the above are merely preferred embodiments of this application and do not limit the scope of patent protection of this application. Any equivalent structural or procedural changes made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of this application.

Claims

1. A method of intelligent assembly of servers, characterized by, The method comprises: According to the synchronous acquisition of production line sensor data, real-time state data stream is generated, and preset offline simulation model and preset online simulation model are updated according to preset model structure data set and the real-time state data stream; Through the updated offline simulation model, multi-objective optimization processing is performed on the preset mixed line production task according to the received business parameters, and candidate resource allocation data set is obtained; Through the updated online simulation model, the candidate resource allocation data set is evaluated and screened, and the optimal resource allocation data is obtained; According to the preset space-time resource constraint set, the optimal resource allocation data is conflict detected and dynamically routed, and a server assembly adjustment instruction sequence is generated.

2. The method of claim 1, wherein, The real-time state data stream is generated according to the synchronous acquisition of production line sensor data, which comprises: According to the preset time interval, the production line sensor data is synchronously acquired through the preset multi-source heterogeneous sensor network; According to the preset data format and the acquisition time stamp, the production line sensor data is time series classified and collected and packaged, and the real-time state data set is obtained; According to the time sequence, the real-time state data set is quality verified and sorted, and the real-time state data stream is generated.

3. The method of claim 1, wherein, The preset offline simulation model and the preset online simulation model are updated according to the preset model structure data set and the real-time state data stream, which comprises: According to the preset model structure data set, the real-time state data stream is classified and processed, and the production state parameter group corresponding to each model structure in each time period is generated; According to the state parameters of the production state parameter group corresponding to the model structure, periodic parallel reset processing is performed to update the preset offline simulation model and the preset online simulation model.

4. The method of claim 1, wherein, The candidate resource allocation data set is obtained by performing multi-objective optimization processing on the preset mixed line production task according to the received business parameters through the updated offline simulation model, which comprises: According to the received business parameters, the initial production task execution sequence set is generated by performing multi-objective optimization sorting on the preset mixed line production task through the preset non-dominated sorting genetic algorithm; Through the updated offline simulation model, the initial production task execution sequence set is numerically simulated and evaluated, and the evaluation score corresponding to each production task execution sequence is obtained; According to the evaluation score, the initial production task execution sequence set is non-dominantly sorted to generate a non-dominant level sequence, and the crowding degree corresponding to each production task execution sequence in the initial production task execution sequence set is calculated; Through the preset iterative screening method, the production task execution sequence is screened from the initial production task execution sequence set according to the non-dominant level sequence, the evaluation score and the crowding degree, and the candidate resource allocation data set is generated.

5. The method of claim 1, wherein, The optimal resource allocation data is obtained by evaluating and screening the candidate resource allocation data set through the updated online simulation model, which comprises: Through the updated online simulation model, the candidate resource allocation data set is numerically simulated, and the instantaneous prediction state data corresponding to each candidate resource allocation data is generated; According to the preset service preference weight coefficient set and the instantaneous prediction state data, a weighted comprehensive score corresponding to each candidate resource allocation data is calculated; According to the weighted comprehensive score, the candidate resource allocation data with the highest score is selected from the candidate resource allocation data set as the to-be-verified optimal allocation data; Through the updated online simulation model, short-term simulation verification and fine-tuning of the to-be-verified optimal allocation data are performed according to a preset time period, and optimal resource allocation data is obtained.

6. The method of claim 1, wherein, The conflict detection and dynamic routing planning of the optimal resource allocation data according to the preset space-time resource constraint set include: Step S41, classifying and analyzing the optimal resource allocation data according to a preset resource allocation device tag set to obtain resource allocation time sequence data corresponding to each resource allocation device; Step S42, performing space-time conflict detection on the resource allocation time sequence data according to the preset space-time resource constraint set; Step S43, according to the preset service priority conflict resolution rule, the conflict path node is identified and adjusted recursively to update the corresponding resource allocation time sequence data; Repeat the above step S42 to the above step S43 until there is no conflict path node between all resource allocation time sequence data, and synthesize the server assembly adjustment instruction sequence corresponding to each resource allocation device according to the updated resource allocation time sequence data through the preset driving instruction synthesis manner.

7. The method of claim 6, wherein, The method further comprises: Obtaining the current latest real-time state data stream to extract a target data set from the real-time state data stream and the current server assembly adjustment instruction sequence according to a preset target data type; According to the order data in the target data set and the service parameters, the task state of the preset mixed production task and the preset service priority conflict resolution rule is real-time corrected to update the mixed production task and the service priority conflict resolution rule; According to the resource consumption data in the target data set, the inventory resource data of the preset space-time resource constraint set is real-time corrected to update the space-time resource constraint set.

8. A server intelligent assembly device applied to the server intelligent assembly method of claim 1, characterized in that, The device comprises: A model updating module is configured to generate a real-time state data stream according to synchronously collected production line sensor data, and update a preset offline simulation model and a preset online simulation model according to a preset model structure data set and the real-time state data stream; A task optimization module is configured to perform multi-objective optimization processing on a preset mixed production task according to received service parameters through the updated offline simulation model to obtain a candidate resource allocation data set; An evaluation and screening module is configured to evaluate and screen the candidate resource allocation data set through the updated online simulation model to obtain optimal resource allocation data; An instruction planning module is configured to perform conflict detection and dynamic routing planning on the optimal resource allocation data according to a preset space-time resource constraint set to generate a server assembly adjustment instruction sequence.

9. A computer device, comprising: Comprise: At least one processor; And The memory is in communication connection with the at least one processor; wherein: The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer instructions, the computer instructions being executed by a processor to implement the method of any one of claims 1 to 7.

Images