Complex heavy equipment collaborative design task and personnel rematching method under disturbance condition

Abstract

The invention belongs to the technical field of collaborative design of complex heavy equipment, and particularly relates to a task and personnel rematching method for collaborative design of complex heavy equipment under a disturbance condition, which comprises the following steps: S1, monitoring a collaborative design process of the complex heavy equipment, judging whether disturbance occurs or not, and if so, turning to S2; s2, the disturbance type is judged, if personnel change disturbance exists, the step S3 is executed, if emergency task increment disturbance exists, the step S4 is executed, and if task technology change disturbance exists, the step S5 is executed; s3, constructing a personnel disturbance multi-objective optimization model, solving the model, obtaining a re-matching scheme, and turning to S6; s4, constructing an order insertion disturbance multi-objective optimization model, solving the model, obtaining a re-matching scheme, and turning to S6; s5, a technical disturbance multi-objective optimization model is constructed and solved, a re-matching scheme is obtained, and then the step S6 is executed; and S6, personnel transfer and distribution are carried out by using a re-matching scheme. According to the method, in the collaborative design of complex heavy equipment, the robustness and the overall efficiency of the design process to disturbance can be improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of collaborative design of complex heavy equipment, and in particular relates to a method for re-matching collaborative design tasks and personnel of complex heavy equipment under disturbance conditions. Background Art

[0002] Under the dual pressures of innovation-driven development and increasingly fierce market competition, the complex heavy equipment industry is undergoing profound transformation and upgrading. This type of equipment is generally characterized by highly complex structures, intensive technological integration, lengthy design cycles, and significant capital investment. Its successful development relies heavily on the deep integration and efficient collaboration of multidisciplinary expertise (such as mechanics, electrical engineering, hydraulics, control, and materials). However, traditional, relatively isolated design models struggle to meet the market's stringent demands for cost-effective, efficient delivery, and high reliability.

[0003] It is against this backdrop that the collaborative design of complex heavy equipment has emerged. By integrating cross-disciplinary talent resources and enabling multi-agent parallel collaboration, it aims to significantly improve overall design efficiency and reduce R&D cycles and costs. Focusing on specific collaborative design practices, the process exhibits distinct characteristics of high customization, high variety, and low-volume production. It also features multidisciplinary cross-fertilization, multi-agent participation, and dynamic task flow evolution. Therefore, achieving efficient, precise, and dynamic optimization of cross-disciplinary design tasks and diverse human resources has become a core challenge for improving collaborative design efficiency and ensuring on-time and high-quality project delivery. The allocation of design tasks directly determines project progress and costs.

[0004] However, the actual design process does not always run smoothly according to the established matching plan. In today's manufacturing environment, where personnel turnover is frequent, changes in customer demand lead to an increase in urgent order tasks, and accelerated technology iterations lead to frequent updates of design task parameters, process disturbances have become the norm. These disturbances have a serious impact on the established execution plan, typical of which include: Personnel change disturbances: The original executor is unable to complete the task as planned due to various reasons (such as job transfer, resignation, health problems, etc.), and their execution time window changes from "available" to "unavailable". Urgent task increment (insertion) disturbances: It is necessary to temporarily insert urgent design tasks that are not pre-planned and have a higher priority. Task technology change disturbances: Existing design tasks have undergone substantial changes in technical specifications, parameters, plans, etc., resulting in changes in task content, technical requirements or required skill combinations.

[0005] Once such a disturbance occurs, how to quickly and effectively re-match personnel to restore task execution to an efficient state becomes a key step in minimizing project delays and cost overruns. Existing re-matching methods have the following major limitations: 1. Response to a single type of disturbance: Current methods primarily focus on "personnel changes," primarily designing strategies for personnel replacement. However, for "urgent task increments (insert orders)" and "technical changes to tasks," two equally common and significant disturbances, there is a lack of systematic and targeted response mechanisms. 2. Lag in processing disturbance information: Especially for "technical changes to tasks," existing technologies rely heavily on manual intervention to understand and parse the changed technical text (such as revised drawings, specifications, etc.), and then convert the manual interpretation results into the input required by the re-matching model. This manual processing approach is inefficient, resulting in significant lags in technical analysis, making it impossible to support rapid re-matching decisions, further prolonging the disturbance response time. 3. Insufficient dynamic optimization capabilities: Existing methods lack the ability to comprehensively optimize the situation under multiple disturbance interweaving or complex disturbance scenarios, which affects the optimality of the rematching solution in terms of overall time and cost-effectiveness.

[0006] Therefore, how to improve the robustness of the design process to disturbances and the overall efficiency in the collaborative design of complex heavy equipment has become an urgent problem to be solved. Summary of the Invention

[0007] In response to the above-mentioned deficiencies in the existing technology, the present invention provides a method for re-matching tasks and personnel in the collaborative design of complex heavy equipment under disturbance conditions, which can improve the robustness of the design process to disturbances and the overall efficiency in the collaborative design of complex heavy equipment.

[0008] In order to solve the above technical problems, the present invention adopts the following technical solutions:

[0009] A method for re-matching complex heavy equipment collaborative design tasks and personnel under disturbance conditions includes the following steps:

[0010] S1. Obtain the original allocation plan of collaborative design personnel, which includes the corresponding responsible personnel for each task; monitor the collaborative design process of complex heavy equipment based on the original allocation plan, and determine whether a disturbance occurs. If so, proceed to S2; if not, continue monitoring;

[0011] S2. Determine the type of disturbance. If it is a personnel change disturbance, go to S3; if it is an urgent task increment disturbance, go to S4; if it is a task technology change disturbance, go to S5. The personnel change disturbance is when the time period for a person to perform a task changes from available to unavailable.

[0012] S3. Taking the minimization of the total time and total cost of the remaining tasks as the optimization goal, a multi-objective optimization model of personnel disturbance is constructed and solved to obtain a rematching solution; then go to S6;

[0013] S4: Add the urgent task increment as an insertion task to the remaining tasks, recording it as the total task after the insertion. Taking the minimization of the total time and total cost of the total task after the insertion as the optimization goal, construct a multi-objective optimization model for the insertion disturbance. Solve it and obtain a rematching solution. Then go to S6.

[0014] S5. Analyze the technical content of the change to obtain the remaining tasks after the change; take the minimization of the total time and total cost of the remaining tasks after the change as the optimization goal, build a technical disturbance multi-objective optimization model; solve it and obtain the re-matching solution; then go to S6;

[0015] S6. Use the obtained rematching plan to carry out corresponding personnel transfer and allocation.

[0016] Compared with the prior art, the present invention has the following beneficial effects:

[0017] 1. Accurate response and processing of multi-source disturbances. This method can intelligently identify and distinguish three common and significant types of process disturbances (personnel changes, urgent task increments / insertion orders, and task technology changes). Based on the unique impact mechanism of each disturbance type (such as personnel failure freeing up tasks, insertion orders introducing new tasks, and technological changes changing task requirements), a dedicated rematching strategy and optimization model are developed. This is significantly superior to existing technologies (such as a single response mechanism for personnel changes only), achieving a more comprehensive and tailored automated response to complex disturbance scenarios.

[0018] 2. Efficiently handle technical change disturbances. The solution specifically designs a process for handling task technical change disturbances (step S5), introducing a "change technical content analysis" step. This design effectively overcomes the inefficiency and lag of existing technologies that rely on manual interpretation of technical change texts. Through (partial) automated analysis, it quickly understands the change requirements and converts them into computable "remaining tasks after the change." This provides accurate input for subsequent re-matching optimization, significantly shortening the response decision time under technical disturbances.

[0019] 3. Comprehensively optimize the design cycle and cost. For all three types of disturbances (steps S3, S4, and S5), the rematching model is constructed with the multi-objective optimization goal of minimizing the total time and total cost of the remaining tasks (or the total tasks after the insertion order / the remaining tasks after the change) as the objective. This modeling approach systematically drives the rematching solution to converge toward shortening the overall remaining time and controlling additional costs, helping to more effectively control project delays and cost overruns after a disturbance occurs.

[0020] 4. Improved response speed and automation. The solution establishes a complete closed-loop process from disturbance monitoring, type identification, model building, and solution (steps S2-S5). This process automates monitoring, decision-making, and reallocation solution generation, significantly reducing the time required for traditional manual intervention and coordination. This is particularly true for analyzing technology changes and solving optimization models. Compared to manual response methods, it can generate feasible reallocation solutions more quickly.

[0021] In summary, this method can significantly improve the robustness of the design process to disturbances and the overall efficiency in the collaborative design of complex heavy equipment by identifying and distinguishing various disturbance types, automatically analyzing technical changes, and quickly generating efficient rematching solutions based on a multi-objective optimization model.

[0022] Preferably, in S3, the objective function of the personnel disturbance multi-objective optimization model is:

[0023]

[0024] Where, T a2 Represents the total time required for the remaining tasks to be matched, T max 、T min They represent the longest and shortest time required to complete the remaining tasks by different personnel; C2 represents the total cost required after the remaining tasks are matched, and C max 、C min They represent the maximum cost and minimum cost of the remaining tasks required to be completed by different personnel; ω1 and ω2 are weight coefficients respectively.

[0025] Such a setting 1. Dual optimization goal fusion. This model creatively integrates the two key goals of "time optimization" and "cost control" into a unified objective function (Z0). Through normalization processing Indicators with different dimensions (time units, cost units) and huge range differences (such as the longest / shortest time, maximum / minimum cost that different personnel may incur) are converted into comparable, dimensionless values ​​in the [0,1] interval, eliminating the impact of dimensional differences on optimization.

[0026] 2. Precise balance and trade-offs. Using adjustable weight coefficients (ω1, ω2), the model achieves a dynamic balance and precise trade-off between the two objectives of "shortening total time" and "reducing total cost." Decision-makers can adjust the weights based on the project's current priorities (e.g., prioritizing schedule improvement or cost control), guiding the model to generate the rematching solution that best meets project goals (maximizing Z0).

[0027] 3. Focus on the optimal allocation of resources after disturbance. The objective function value (Z0) calculated by the model is directly related to the actual effect of the personnel re-matching plan (T a2By seeking Max(Z0), the model automatically guides the optimization process toward the direction that can minimize the overall execution time of the remaining tasks, minimize the cost, or achieve the best combination of the two under a given personnel resource pool, significantly improving the scientific nature and effectiveness of resource allocation.

[0028] In summary, the objective function, through the normalized dual-objective fusion and weight adjustment mechanism, provides a personnel rematching decision-making basis that can scientifically quantify, dynamically weigh, and comprehensively optimize the remaining task execution time and cost for coping with personnel change disturbances in the collaborative design of complex heavy equipment, thereby effectively improving the ability to quickly respond to personnel disturbances and the overall benefits of resource allocation.

[0029] Preferably, in S4, the objective function of the single-interference multi-objective optimization model is:

[0030]

[0031] Where, T a3 , C3, L ave3 They represent the total time, total cost, and average load of designers after redistributing the total task after the insertion order as a whole; T a4 , C4, L ave4 They represent the total time, total cost, and average load of designers for the total tasks after the insertion of the task, with the remaining tasks remaining unchanged and only the personnel redistributed for the insertion task; T max 、T min They represent the longest and shortest time required for different personnel to complete the total task after inserting the order; C max 、C min They represent the maximum cost and minimum cost of the total task after inserting the order due to the use of different personnel; L max , L min They represent the maximum average load and minimum average load of the total task after inserting the order due to the use of different personnel; ω3, ω4, and ω5 are weight coefficients.

[0032] This setup innovatively integrates three key performance indicators (KPIs)—time (T), cost (C), and workload balance (L)—into the objective function (Z0). This transcends the limitations of traditional optimization models, which often focus on a single or dual objective (e.g., time and cost alone), and more comprehensively addresses the multidimensional objectives of actual project management. The three indicators, T, C, and L, are normalized. This design cleverly eliminates differences in dimensionality and value ranges between the different indicators, making them comparable within a unified [0, 1] scale. This addresses the optimization bias caused by imbalanced indicators in multi-objective optimization.

[0033] 2. The function design is compatible with the evaluation of two task allocation strategies. When i = 3, the global redistribution strategy (redistributes all tasks after inserting an order). When i = 4, the incremental allocation strategy (allocates only the inserted order tasks, leaving the remaining tasks unchanged). This design allows the model to flexibly compare the overall benefits (Z0) under different strategies and identify the optimal solution (Max(Z0)), supporting dynamic decision-making that is more tailored to project realities.

[0034] 3. Compared to many existing scheduling models that primarily focus on time and cost, this model explicitly incorporates "average workload of designers" into its optimization objectives. This helps avoid significant workload imbalances within teams (e.g., overloading some staff, leading to decreased efficiency / quality or delays, while others are idle and waste resources) when responding to interruptions, thereby improving the efficiency and stability of team collaboration, a critical dimension often overlooked by existing technologies.

[0035] In summary, this method integrates the triple normalization objectives of time, cost, and personnel load, embeds a flexible weight adjustment mechanism, and supports the evaluation of multiple allocation strategies. It provides a rematching decision-making tool for the collaborative design of complex heavy equipment to cope with insertion disturbances, which can scientifically quantify and comprehensively optimize task execution efficiency, resource input economy, and human resource utilization balance, thereby effectively improving the robustness of the design process to disturbances and the overall efficiency.

[0036] Preferably, in S5, the objective function of the technical perturbation multi-objective optimization model is:

[0037]

[0038] Where, T max 、T min They represent the longest and shortest time required to complete the total task after the change due to the use of different personnel; C max 、C min They represent the maximum cost and minimum cost of the total task after the change due to the use of different personnel; T a5 , C5 are the total time and total cost after the total task is redistributed after the change; ω6 and ω7 are the weight coefficients respectively.

[0039] This setup, 1. The model focuses on two core performance indicators, time (T) and cost (C), to construct the objective function (Z0). This is highly consistent with the core demands of task technology changes—changed tasks inevitably lead to changes in execution time and resource input costs due to technical requirements, directly determining the risk of project delays and cost overruns.

[0040] 2. Compared to existing technologies that rely on inefficient manual interpretation of technical change text and adjustment of rematching solutions, this model, as a key step in the automated decision-making process (S5), works in conjunction with the pre-process "analysis of technical change content" to rapidly transform change semantics into calculable remaining tasks after the change. Automatically optimizing based on a normalized dual-objective model, it rapidly generates rematching solutions. This significantly shortens the time from sensing a technical change to producing a response solution, resolving the pain point of "lagging technical analysis" associated with existing technologies.

[0041] Preferably, in S3, S4, and S5, when calculating the total time after redistribution, the actual time T for the redistributed person i to complete the j-th task is aij The calculation formula is:

[0042] T aij =Min{T ej (1-p max ),T ej (1-p ij )};

[0043] Where p max is the preset maximum proficiency; p ij T represents the proficiency of person i in task j; ej (1-p max ) represents the maximum proficiency p based on the preset max Calculate the shortest time, T ej (1-p ij ) indicates that the ij Calculate the theoretical time it takes for person i to perform task j.

[0044] This setting can 1. Accurately quantify the impact of personnel ability differences on efficiency. ij Personalized calculation of the theoretical time required to perform task j ej (1-p ij ), directly mapping the differences in skill levels of different personnel into comparable time efficiency indicators, avoiding the errors caused by averaging estimation in existing technologies and significantly improving the accuracy of man-hour prediction. max Limit, forcing the actual time consumption to be no less than the theoretical limit T ej (1-p max This mechanism effectively avoids infeasible scheduling plans caused by over-reliance on individual "theoretical optimal personnel" (such as super-capable personnel who are assumed to not exist), and enhances the engineering feasibility of the re-matching plan.

[0045] 2. Double constraints balance efficiency and fairness. The design of the minimum function Min{·} dynamically selects a value that is closer to reality: when the personnel capacity is close to the upper limit (p ij ≈p max), use time estimates close to the optimal value to maximize efficiency potential; when personnel capabilities are weak (p ij <<p max ) when using its own theoretical time consumption to objectively reflect resource bottlenecks. This mechanism not only encourages efficiency improvement, but also prevents the over-exploitation of highly skilled personnel and improves team stability.

[0046] 3. Accurate output T aij This design provides high-fidelity underlying time data for the perturbation optimization model (minimizing total time and cost) for S3-S5. Compared to existing methods that use fixed hour quotas or simple weighted averages, this design more accurately reflects the actual efficiency of post-perturbation redistribution, ensuring the effectiveness of the upper-level optimization results.

[0047] Preferably, in S3, S4, and S5, when calculating the total cost after redistribution, the actual cost C of the redistributed person i completing the j-th task is ij The calculation formula is:

[0048] T ij =T aij *[1+ξ(k-1)];

[0049] In the formula, k is the personnel level, and ξ represents the salary coefficient that changes with the promotion of job level.

[0050] Such a setting, 1. Converts traditional empirical pricing into a structured mathematical model, truly reflecting the business logic of "high skills = high cost".

[0051] 2. Deep coupling with dynamic working hours Taij (including the influence of proficiency): working hours T aij Compression (efficient personnel) can partially offset high-level costs (such as experts completing tasks quickly); avoid the "double high costs" (time-consuming and high unit price) generated by inefficient high-level personnel; provide a micro-computing basis for the goal of minimizing total costs and help optimize global resources.

[0052] 3. Parameter ξ acts as a configurable lever. Companies can adjust ξ values ​​based on project priorities (e.g., cost-sensitive vs. quality-sensitive projects). Industry differences (e.g., high-end equipment vs. general-purpose equipment) can be customized through ξ to achieve cost models. This breaks the rigidity of one-size-fits-all cost accounting and adapts to complex management scenarios.

[0053] Preferably, the personnel proficiency p ij The calculation process includes:

[0054] Constructing a task information table and a personnel information table; the task information table includes information of various dimensions of tasks; the personnel information table includes information of various dimensions of personnel;

[0055] Calculate the similarity between tasks and construct a task similarity matrix;

[0056] Construct a proficiency prediction model to calculate the proficiency p of person i in task j based on the information in the person information table and the task similarity matrix ij ;

[0057]

[0058] Where m represents the historical completion frequency, which is obtained by counting the number of historical executions of similar tasks; o represents the task similarity, which is determined by calculating the mean similarity between historical tasks and current tasks through the task similarity matrix; k represents the personnel title level; α, β, γ, and b are the parameters of the fitting curve to be determined.

[0059] The compression ratio of the actual working time of historical tasks to the standard working time is used to quantify the corresponding proficiency p ij , combined with the corresponding m, k, o, the parameters α, β, γ, and b of the proficiency prediction model are iteratively optimized by the least squares method;

[0060] Use the quantitative prediction model to calculate the proficiency p of each person ij .

[0061] This setup 1. breaks through reliance on subjective experience and achieves objective quantification of capabilities. By constructing a task similarity matrix and integrating a multi-dimensional indicator model, employee proficiency is deconstructed into three quantifiable factors: historical completion frequency (m), professional title level (k), and task similarity (o). This completely replaces the traditional, crude method of relying on manager experience, significantly improving the scientific nature and consistency of employee capability assessment.

[0062] 2. Fusion of multi-source heterogeneous features accurately captures the essence of capabilities. The model innovatively integrates three types of differentiated features: behavioral features (m) reflect accumulated practical experience in similar tasks; qualification features (k) reflect the theoretical capability benchmark corresponding to the job title; and semantic features (o) quantify the relevance of skill transfer between new and historical tasks. Using a logarithmic function and a normalized denominator (m+k+o), the model balances the magnitude differences of multiple features, avoiding bias in a single indicator and more comprehensively capturing the essence of capabilities.

[0063] 3. Data-driven parameter self-optimization mechanism. Using the historical task work-hour compression rate as a supervisory signal (i.e., a proxy variable for true proficiency), the least squares method is used to iteratively optimize the parameters (α, β, γ, b) in reverse. This design establishes a closed-loop feedback loop between business data (actual work hours) and model parameters, ensuring that the model continuously aligns with actual production practices. It dynamically adapts to evolving trends in enterprise personnel capabilities (such as skill improvement and technological iteration), significantly improving the timeliness and adaptability of prediction results compared to fixed-parameter models.

[0064] 4. Support refined personnel-task matching decision-making. As the basic capability assessment unit of the personnel re-matching method (steps S3-S5), the output p ij Directly serve: Task time model (T aij Calculation) → Optimize time target; cost model (C ij Calculation) → Optimize cost targets; personnel load balance analysis provides high-precision personnel capability input for the upper-level multi-objective optimization model, improving the scientific nature of the re-matching plan from the root.

[0065] Preferably, the process of constructing the task similarity matrix includes:

[0066] Construct the skill similarity evaluation matrix q required for the task:

[0067]

[0068] Where q ij Characterizes the matching degree between task i and benchmark task in terms of the skill j dimension required by the task; n is the total number of tasks, u is the total number of required skill dimensions;

[0069] Construct the task type similarity evaluation matrix l:

[0070]

[0071] Where, l ij represents the matching degree between task i and benchmark task in terms of task type j dimension; v is the total number of task type dimensions;

[0072] Construct a comprehensive task similarity matrix:

[0073]

[0074] Where s i =(q i1 ,…,q iu ,l i1 ,…,l iv ) represents the comprehensive similarity code of task i, and n is the total number of design tasks;

[0075] The generalized Jaccard similarity is used to calculate the intersection-over-union ratio of the skill sets of the two tasks as the similarity between the two tasks:

[0076]

[0077] Get the task similarity matrix I:

[0078]

[0079] This setup 1. Decoupling multi-dimensional features enhances the depth of semantic parsing. A unique dual-matrix structure (skill matrix q + type matrix l) deconstructs task attributes: q focuses on technical capability requirements (e.g., hydraulic system design requires "fluid mechanics / sealing technology" skills), while l captures business process characteristics (e.g., differences between conceptual design and detailed design stages). This breaks through the limitations of traditional methods that rely solely on single features (e.g., skill labels), and more accurately captures the multifaceted nature of heavy equipment design tasks.

[0080] 2. Comprehensive coding realizes heterogeneous feature fusion. Through vector splicing s i =(q i1 ,…,q iu ,l i1 ,…,l iv ) unifies discrete dimensions into a computable semantic space, eliminates the dimensional barriers of skill and type features, provides standardized input for subsequent similarity calculations, and solves the matching bias caused by feature heterogeneity in existing technologies.

[0081] 3. The generalized Jaccard algorithm enhances the robustness of similarity measurement. It improves the limitation of traditional Jaccard that only considers binary features and introduces vector modulus operation (||S i || 2 ) to adapt to continuous matching data, through S i ∩S j The intersection reflects the strength superposition of the common skills / types, making the similarity calculation compatible with partial matching scenarios (such as task A requiring 80% hydraulic skills + 20% control skills vs. task B requiring 60% hydraulic skills). It can also avoid the defect of cosine similarity being insensitive to vector length.

[0082] 4. Support the precision of upper-level personnel-task matching. The output similarity matrix I is directly applied to the personnel proficiency model (p ij Calculation) provides an objective correlation basis for "task-personnel" capability matching: high-similarity tasks can reuse historical personnel performance data → improve the reliability of proficiency prediction, and low-similarity tasks trigger cross-domain personnel reorganization → optimize the adaptability of the rematching plan.

[0083] Preferably, in S5, the BertTextCNN model is used to automatically process technical text information and parse the changed technical content.

[0084] This setup, through the introduction of the BertTextCNN model, enables automated and precise parsing of technical change text. This solution not only significantly improves the response efficiency and accuracy of the S5 process but also becomes the core enabling technology for the rapid closure of the entire technical disturbance rematching process. It effectively enhances the real-time parsing capability and responsiveness of complex heavy equipment collaborative design to task technical change disturbances, significantly improving the robustness of the design process to disturbances and the overall efficiency.

[0085] Preferably, in S3, an elite-preserving genetic algorithm is used to solve the multi-objective optimization model of personnel changes.

[0086] This setup, through elite-preserving guided evolution, dual-objective collaborative ranking, and adaptive operational control, provides a high-quality, diverse, and timely Pareto-optimal solution set for rematching decisions under personnel change disturbances. This significantly enhances the design process's dynamic responsiveness to personnel disturbances and the scientific nature of its decision-making, thereby improving overall robustness and efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0087] In order to make the purpose, technical solutions and advantages of the invention more clear, the present invention will be further described in detail below with reference to the accompanying drawings, in which:

[0088] Figure 1 Flowchart of this method;

[0089] Figure 2 The Gantt chart of the original matching plan before the personnel change disturbance occurs in Example 2;

[0090] Figure 3 Execute the Gantt chart for the personnel change disturbance rematching task in Example 2;

[0091] Figure 4 The Gantt chart of the original matching plan before the incremental disturbance of the urgent task in Example 2 occurs;

[0092] Figure 5 Insert the time for the urgent task in Example 2;

[0093] Figure 6 Gantt chart for the overall rematch execution of the incremental disturbance of the emergency task in Example 2;

[0094] Figure 7 Re-matching and executing the Gantt chart for the incremental disturbance part of the urgent task in Example 2;

[0095] Figure 8 The Gantt chart for the incremental perturbation and rematching execution of the merged tasks in Example 2;

[0096] Figure 9 Execute the Gantt chart for the original matching plan before the task technology change in Example 2;

[0097] Figure 10 A Gantt chart is shown for the technical change rematching task in Example 2. DETAILED DESCRIPTION

[0098] The following is a further detailed description through specific implementation methods:

[0099] Example 1

[0100] like Figure 1 As shown, this embodiment discloses a method for re-matching complex heavy equipment collaborative design tasks and personnel under disturbance conditions, including the following steps:

[0101] S1. Obtain the original allocation plan of collaborative design personnel, which includes the corresponding responsible personnel for each task; monitor the collaborative design process of complex heavy equipment based on the original allocation plan, and determine whether a disturbance occurs. If so, proceed to S2; if not, continue monitoring;

[0102] S2. Determine the type of disturbance. If it is a personnel change disturbance, go to S3; if it is an urgent task increment disturbance, go to S4; if it is a task technology change disturbance, go to S5. The personnel change disturbance is when the time period for a person to perform a task changes from available to unavailable.

[0103] S3. Taking the minimization of the total time and total cost of the remaining tasks as the optimization goal, a multi-objective optimization model of personnel disturbance is constructed; and the model is solved to obtain a rematching solution; then go to S6.

[0104] Among them, the objective function of the multi-objective optimization model of personnel disturbance is:

[0105]

[0106] Where, T a2 Represents the total time required for the remaining tasks to be matched, T max 、T min They represent the longest and shortest time required to complete the remaining tasks by different personnel; C2 represents the total cost required after the remaining tasks are matched, and C max 、C min They represent the maximum cost and minimum cost of the remaining tasks required to be completed by different personnel; ω1 and ω2 are weight coefficients respectively.

[0107] In this way, the model creatively integrates the two key goals of "time optimization" and "cost control" into a unified objective function (Z0). Indicators with different dimensions (time units, cost units) and huge range differences (such as the longest / shortest time, maximum / minimum cost that different personnel may incur) are converted into comparable, dimensionless values ​​in the interval [0,1], eliminating the impact of dimensional differences on optimization. In addition, using adjustable weight coefficients (ω1, ω2), the model achieves a dynamic balance and precise trade-off between the two goals of "shortening total time" and "reducing total cost". Decision makers can adjust the weights according to the current focus of the project (such as prioritizing progress or controlling costs) to guide the model to generate a rematching plan that best meets the project goals (maximizing Z0). In addition, the objective function value (Z0) calculated by the model is directly related to the actual effect of the personnel rematching plan (T a2 By seeking Max(Z0), the model automatically guides the optimization process toward the direction that can minimize the overall execution time of the remaining tasks, minimize the cost, or achieve the best combination of the two under a given personnel resource pool, significantly improving the scientific nature and effectiveness of resource allocation.

[0108] In specific implementation, an elite-preserving genetic algorithm is used to solve a multi-objective optimization model for personnel changes. This approach, through elite-preserving guided evolution, dual-objective collaborative sorting, and adaptive operational control, provides a high-quality, highly diverse, and timely Pareto-optimal solution set for rematching decisions under personnel change disturbances. This significantly enhances the design process's dynamic responsiveness to personnel disturbances and the scientific nature of its decision-making, thereby improving overall robustness and efficiency.

[0109] S4: Add the urgent task increment as an insertion task to the remaining tasks, recording it as the total task after the insertion. Taking the minimization of the total time and total cost of the total task after the insertion as the optimization goal, construct a multi-objective optimization model for the insertion disturbance. Solve it and obtain a rematching solution. Then go to S6.

[0110] The objective function of the single-interference multi-objective optimization model is:

[0111]

[0112] Where, T a3 , C3, L ave3 They represent the total time, total cost, and average load of designers after redistributing the total task after the insertion order as a whole; T a4 , C4, L ave4 They represent the total time, total cost, and average load of designers for the total tasks after the insertion of the task, with the remaining tasks remaining unchanged and only the personnel redistributed for the insertion task; T max 、T min They represent the longest and shortest time required for different personnel to complete the total task after inserting the order; C max 、C minThey represent the maximum cost and minimum cost of the total task after inserting the order due to the use of different personnel; L max , L min They represent the maximum average load and minimum average load of the total task after inserting the order due to the use of different personnel; ω3, ω4, and ω5 are weight coefficients.

[0113] In this way, the three key performance indicators (KPIs) of time (T), cost (C), and staff load balance (L) are innovatively integrated into the objective function (Z0). This overcomes the limitations of traditional optimization models, which often focus on a single or dual objective (e.g., time-cost alone), and more comprehensively meets the multi-dimensional objectives required by actual project management. The three indicators T, C, and L are normalized. This design cleverly eliminates differences in dimensionality and value ranges among the different indicators, making them comparable within a unified [0, 1] scale, thus addressing the optimization bias caused by imbalanced indicators in multi-objective optimization. Furthermore, the function design is compatible with the evaluation of two task allocation strategies: i = 3, a global reallocation strategy (which redistributes the total number of tasks after inserting an order); i = 4, an incremental allocation strategy (which allocates only the "inserted order tasks," leaving the remaining tasks unchanged). This design enables the model to flexibly compare the comprehensive benefits (Z0) of different strategies and identify the optimal solution (Max(Z0)), supporting dynamic decision-making that is more tailored to project realities. Furthermore, unlike many existing scheduling models that primarily focus on time and cost, this model explicitly incorporates the "average workload of designers" into its optimization objectives. This helps avoid significant workload imbalances within teams (e.g., overloading some staff, leading to decreased efficiency / quality or delays, while others remain idle and waste resources) when responding to interruptions. This improves the efficiency and stability of team collaboration, a crucial dimension often overlooked by existing technologies.

[0114] S5. Analyze the changed technical content to obtain the remaining tasks after the change; take the minimization of the total time and total cost of the remaining tasks after the change as the optimization goal, build a technical disturbance multi-objective optimization model; solve it and obtain the re-matching solution; then go to S6.

[0115] During implementation, the BertTextCNN model was used to automatically process technical text information and parse the technical content of the changes. By introducing the BertTextCNN model to enable automated and precise parsing of technical change text, this solution not only significantly improved the response efficiency and accuracy of the S5 process but also became the core enabling technology for the rapid closure of the entire technical perturbation rematching process. This effectively enhanced the real-time parsing capability and responsiveness of complex heavy equipment collaborative design to mission-related technical change disturbances, significantly improving the robustness of the design process to disturbances and the overall efficiency.

[0116] In specific implementation, the objective function of the technical perturbation multi-objective optimization model is:

[0117]

[0118] Where, T max 、T min They represent the longest and shortest time required to complete the total task after the change due to the use of different personnel; C max 、C min They represent the maximum cost and minimum cost of the total task after the change due to the use of different personnel; T a5 , C5 are the total time and total cost after the total task is redistributed after the change; ω6 and ω7 are the weight coefficients respectively.

[0119] In this way, the model focuses on the two core performance indicators of time (T) and cost (C) to construct the objective function (Z0). This is highly consistent with the core demand of task technology changes - after the change, the task will inevitably change due to changes in technical requirements, which will inevitably lead to changes in execution time and resource input costs, directly determining the risk of project delays and cost overruns. Moreover, compared with the inefficient way of existing technologies that rely on manual interpretation of technical change texts and manual adjustment of rematching plans, this model serves as a key link in the automated decision-making process (S5): It is linked with the previous "change technical content analysis" to quickly convert the change semantics into calculable remaining tasks after the change. Automatic optimization based on the normalized dual-objective model to quickly generate rematching plans. It significantly shortens the time from perceiving technical changes to outputting response plans, solving the pain point of "technical analysis lag" of existing technologies.

[0120] It should be noted that, during specific implementation, the multi-objective optimization models in S3-S5 must meet the following constraints: the skills mastered by the personnel fully cover the task requirements, and the personnel's workload is less than the preset load threshold;.

[0121] In real time, when calculating the total time after redistribution in S3, S4, and S5, the actual time T for the redistributed person i to complete the jth task is aij The calculation formula is:

[0122] T aij =Min{T ej (1-p max ),T ej (1-p ij )};

[0123] Where p max is the preset maximum proficiency; p ij T represents the proficiency of person i in task j; ej (1-p max ) represents the maximum proficiency p based on the preset max Calculate the shortest time, T ej (1-p ij) indicates that the ij Calculate the theoretical time it takes for person i to perform task j.

[0124] Thus, through the personnel-specific proficiency p ij Personalized calculation of the theoretical time required to perform task j ej (1-p ij ), directly mapping the differences in skill levels of different personnel into comparable time efficiency indicators, avoiding the errors caused by averaging estimation in existing technologies and significantly improving the accuracy of man-hour prediction. max Limit, forcing the actual time consumption to be no less than the theoretical limit T ej (1-p max This mechanism effectively avoids infeasible scheduling solutions caused by over-reliance on individual “theoretical optimal personnel” (such as the assumption that super-capable personnel do not exist), and enhances the engineering feasibility of the re-matching solution. In addition, the design of the minimum function Min{·} dynamically selects a value that is closer to reality: when the personnel capacity is close to the upper limit (p ij ≈p max ), use time estimates close to the optimal value to maximize efficiency potential; when personnel capabilities are weak (p ij <<p max ) when using its own theoretical time consumption to objectively reflect resource bottlenecks. This mechanism not only encourages efficiency improvement, but also prevents the over-exploitation of highly skilled personnel and improves team stability.

[0125] In S3, S4, and S5, when calculating the total cost after redistribution, the actual cost C of the redistributed person i completing the jth task is ij The calculation formula is:

[0126] C ij =T aij *[1+ξ(k-1)];

[0127] In the formula, k is the personnel level, and ξ represents the salary coefficient that changes with the promotion of job level.

[0128] In this way, traditional experience pricing is transformed into a structured mathematical model, which truly reflects the business logic of "high skills = high cost". In addition, it is deeply coupled with the dynamic working hours Taij (including the influence of proficiency). Working hours T aijCompression (of highly efficient personnel) can partially offset high-level costs (e.g., experts completing tasks quickly); avoid the "double high costs" (long time and high unit price) of inefficient, high-level personnel; and provide a micro-calculation basis for minimizing total costs, facilitating global resource optimization. Furthermore, the parameter ξ serves as a configurable lever. Enterprises can adjust the value of ξ based on project priorities (e.g., cost-sensitive vs. quality-sensitive); and customize cost models based on industry differences (e.g., high-end equipment vs. general-purpose equipment). This breaks the rigidity of one-size-fits-all cost accounting and adapts to complex management scenarios.

[0129] During specific implementation, personnel proficiency p ij The calculation process includes:

[0130] Construct a task information table and a personnel information table; the task information table includes information on various dimensions of the task, and the personnel information table includes information on various dimensions of the personnel. In specific implementations, the task information table includes the design task's name, task ID, required skills, task status, and task type; task statuses include unexecuted, in progress, and completed. The personnel information table includes the designer's number, technical level, skill type, maximum working hours, and personnel status; personnel statuses include idle, partially loaded, and fully loaded.

[0131] Calculate the similarity between tasks and construct a task similarity matrix;

[0132] Construct a proficiency prediction model to calculate the proficiency p of person i in task j based on the information in the person information table and the task similarity matrix ij ;

[0133]

[0134] Where m represents the historical completion frequency, which is obtained by counting the number of historical executions of similar tasks; o represents the task similarity, which is determined by calculating the mean similarity between historical tasks and current tasks through the task similarity matrix; k represents the personnel title level; α, β, γ, and b are the parameters of the fitting curve to be determined.

[0135] The compression ratio of the actual working time of historical tasks to the standard working time is used to quantify the corresponding proficiency p ij , combined with the corresponding m, k, o, the parameters α, β, γ, and b of the proficiency prediction model are iteratively optimized by the least squares method;

[0136] Use the quantitative prediction model to calculate the proficiency p of each person ij .

[0137] This breaks free from reliance on subjective experience and achieves objective quantification of competence. By constructing a task similarity matrix and integrating it with a multi-dimensional indicator model, employee proficiency is deconstructed into three quantifiable factors: historical completion frequency (m), professional title level (k), and task similarity (o). This completely replaces the traditional, crude approach of relying on managerial experience, significantly improving the scientific and consistent nature of competence assessment. Furthermore, the model innovatively integrates three types of differentiated features: behavioral features (m) reflect accumulated practical experience with similar tasks; qualification features (k) reflect the theoretical competence benchmark corresponding to the professional title; and semantic features (o) quantify the skill transfer relevance between new tasks and historical tasks. A logarithmic function and a normalized denominator (m+k+o) are used to reconcile the magnitude differences among multiple features, avoiding bias from a single indicator and more comprehensively capturing the essence of competence. Furthermore, using the historical task time compression rate as a supervisory signal (i.e., a proxy for true proficiency), the least squares method is used to iteratively optimize the parameters (α, β, γ, b) in reverse order. This design establishes a closed-loop feedback loop between business data (actual working hours) and model parameters, allowing the model to continuously adapt to actual production rules; it dynamically adapts to the evolution trend of enterprise personnel capabilities (such as skill improvement and technology iteration), and compared with fixed parameter models, it greatly improves the timeliness and scenario adaptability of prediction results.

[0138] As the basic capability assessment unit of the personnel re-matching method (steps S3-S5), the output p ij Directly serve: Task time model (T aij Calculation) → Optimize time target; cost model (C ij Calculation) → Optimize cost targets; personnel load balance analysis provides high-precision personnel capability input for the upper-level multi-objective optimization model, improving the scientific nature of the re-matching plan from the root.

[0139] To help those skilled in the art better understand the task information table and the personnel information table, the following description is given.

[0140] The task (i.e., design task) information table description model can be expressed as follows:

[0141] TaskInfo={ID,TaskTime,TaskSkills,TaskState,TaskType};

[0142] Among them, ID: PName-FTask-STask, the design task number, encodes the design task and includes the product name to which the design task belongs, the names of the first-level subtask and the second-level subtask, and other information. These three types of information together determine the unique number of the subtask.

[0143] PName represents the product name, as shown in Table 1;

[0144] Table 1 Product name code

[0145]

[0146] Ftask represents the name of the first-level subtask, as shown in Table 2:

[0147] Table 2 Level 1 task coding

[0148]

[0149] STask represents the name of the second-level subtask, as shown in Table 3:

[0150] Table 3 Secondary task coding

[0151]

[0152] From the above coding method, it can be seen that the unique task ID of the hydraulic design oil tank and its corresponding component design subtask in the extruder design project is "010201".

[0153] TaskTime: Design task time information, mainly stipulates the standard working time T of the jth task ej and the actual working time T of person i to complete the jth task after considering proficiency aij , in days, as well as the start and end times of the task.

[0154] TaskSkills: Design task skills, specifying the specific skills required to complete a task. These skills include product type and specific skill requirement information. The combination of product type and specific skill requirements uniquely identifies the skill requirements for a task. Skill requirements reflect the specific technical characteristics of the task. Table 4 lists the specific skill attributes corresponding to skill requirement identifiers.

[0155] Table 4 Task requirement skill identification code

[0156]

[0157] From the above coding method, it can be seen that for the hydraulic design of the oil tank and its related component design subtask in the extruder design project, the task required skills are "drawing / geometric modeling" and "hydraulic structure design", and the unique coding of its task required skills is "0101, 0106".

[0158] TaskState: Design task status, describes the current status of the design task, and has three forms: unexecuted, executing, and completed.

[0159] TaskType: Design task type, which describes the type of design task and is determined by the product type and task code name, as shown in Table 5.

[0160] Table 5 Design task type coding

[0161]

[0162] From the above, we can see that for the hydraulic design of the oil tank and its associated component design subtask in the extruder design project, its task code is "010201", and its task type code is "0102".

[0163] The personnel information table can be expressed as follows:

[0164] esignerInfo={ID,Level,Skills,MaxLoad,DeState};

[0165] Among them, ID: designer number, which encodes the designer and includes the designer's name information.

[0166] Level: Designer level, indicating the designer's professional title level. Different professional levels have different salary coefficients.

[0167] Skills: Designer skills, representing all types of skills that designers possess. The skills mastered by the designer and the skills required for the task are considered here as different expressions of the same attribute. Therefore, the specific attributes of the skills mastered by the designer are shown in Table 4.

[0168] MaxLoad: maximum working time T of the designer i , T i It represents the maximum working time that designer i can allocate in a collaborative design project cycle, in days.

[0169] DeState: Designer status, which indicates the different working states of designers due to the number of design tasks they are currently handling. It can be divided into three states: idle state, underloaded state, and fully loaded state. The workload of designers can be defined as:

[0170]

[0171] Where, T aij / T i It represents the ratio of the time that person i spends completing the jth task to the longest working time of designer i, L i is the accumulated load value of designer i, L0 is the initial load value of designer i, and L is the total load value of designer i.

[0172] In specific implementation, the process of constructing the task similarity matrix includes:

[0173] Construct the skill similarity evaluation matrix q required for the task:

[0174]

[0175] Where q ij Characterizes the matching degree between task i and benchmark task in terms of the skill j dimension required by the task; n is the total number of tasks, u is the total number of required skill dimensions;

[0176] Construct the task type similarity evaluation matrix l:

[0177]

[0178] Where, l ij represents the matching degree between task i and benchmark task in terms of task type j dimension; v is the total number of task type dimensions;

[0179] Construct a comprehensive task similarity matrix:

[0180]

[0181] Where s i =(q i1 ,…,q iu ,l i1 ,…,l iv ) represents the comprehensive similarity code of task i, and n is the total number of design tasks;

[0182] The generalized Jaccard similarity is used to calculate the intersection-over-union ratio of the skill sets of the two tasks as the similarity between the two tasks:

[0183]

[0184] Get the task similarity matrix I:

[0185]

[0186] In this way, the original dual matrix structure (skill matrix q + type matrix l) deconstructs the task attributes: q focuses on technical capability requirements (such as hydraulic system design requires "fluid mechanics / sealing technology" skills), l captures business process characteristics (such as the difference between conceptual design and detailed design stages), breaking through the limitations of single characteristics of traditional methods (such as only using skill labels), and more accurately depicting the multifaceted nature of heavy equipment design tasks. In addition, through vector splicing s i =(q i1 ,…,q iu ,l i1 ,…,l iv ) unifies discrete dimensions into a computable semantic space, eliminates the dimensionality barrier of skill and type features, provides standardized input for subsequent similarity calculations, and solves the matching deviation caused by feature heterogeneity in existing technologies. In addition, the traditional Jaccard is improved by only considering binary features and introducing vector modulus operation (||Si || 2 ) to adapt to continuous matching data, through S i ∩S j The intersection reflects the strength superposition of the common skills / types, making the similarity calculation compatible with partial matching scenarios (such as task A requiring 80% hydraulic skills + 20% control skills vs. task B requiring 60% hydraulic skills). It can also avoid the defect of cosine similarity being insensitive to vector length.

[0187] The output similarity matrix I is directly applied to the proficiency model of the person in the historical dialogue (p ij Calculation) provides an objective correlation basis for "task-personnel" capability matching: high-similarity tasks can reuse historical personnel performance data → improve the reliability of proficiency prediction, and low-similarity tasks trigger cross-domain personnel reorganization → optimize the adaptability of the rematching plan.

[0188] S6. Use the obtained rematching plan to carry out corresponding personnel transfer and allocation.

[0189] This method can intelligently identify and distinguish three common and significant types of process disturbances (personnel changes, urgent task increments / insertion orders, and task technology changes). According to the unique impact mechanism of each type of disturbance (such as personnel failure freeing up tasks, insertion orders introducing new tasks, and technical changes changing task requirements), a dedicated rematching strategy and optimization model are formulated. This is significantly better than existing technologies (such as a single response mechanism only for personnel changes), and achieves a more comprehensive and more appropriate automated response to complex disturbance scenarios. In addition, the solution specially designs a processing flow for task technology change disturbances (step S5), and introduces the "change technology content analysis" link. This design effectively overcomes the inefficiency and lag problems of existing technologies that rely on manual interpretation of technical change texts. Through (partial) automated analysis, the change requirements are quickly understood and converted into computable "remaining tasks after the change", providing accurate input for subsequent rematching optimization, thereby significantly shortening the response decision time under technical disturbances. Moreover, for all three types of disturbances (steps S3, S4, and S5), when constructing the rematching model, it is clear that minimizing the "total time and total cost of the remaining tasks (or the total tasks after inserting the order / the remaining tasks after the change)" is used as the multi-objective optimization goal. This modeling approach systematically drives the rematching plan to converge in the direction of shortening the overall remaining time and controlling additional costs, which helps to more effectively control project delays and cost overruns after the disturbance occurs. In addition, the solution constructs a complete closed-loop process from disturbance monitoring, type identification, model building and solution (steps S2-S5). This process automatically connects monitoring, decision-making and redistribution plan generation, greatly reducing the time required for traditional manual intervention and coordination, especially in analyzing technical changes and solving optimization models. Compared with human-led response methods, it can produce feasible rematching solutions faster.

[0190] This method can significantly improve the robustness of the design process to disturbances and the overall efficiency in the collaborative design of complex heavy equipment by identifying and distinguishing various disturbance types, automatically analyzing technical changes, and quickly generating efficient rematching solutions based on a multi-objective optimization model.

[0191] Example 2

[0192] In order for those skilled in the art to better understand the effect of this method, the following example analysis is performed.

[0193] 1. Regarding personnel change disturbances

[0194] Taking the 120MN extruder design project released by Z company on the collaborative manufacturing platform as an example, after formulating the matching plan (i.e. the original matching plan), the "task-personnel" matching plan when designing product parts according to the original matching plan is as follows: Figure 2 shown.

[0195] On the 9th and 15th days of the mission, designers numbered 8 and 9 applied for 14 days of personal leave and 6 days of sick leave respectively.

[0196] The project management system receives the designer's abnormal condition information and analyzes and processes it:

[0197] (1) Regarding designer 8, he applied for personal leave on the 9th day of executing the design task set. After checking the matching plan at that time, this person was not in the matching sequence. Therefore, his abnormal situation did not affect the progress of the collaborative project, and there was no need to re-match the person.

[0198] (2) Regarding designer 9, he applied for sick leave on the 15th day of the design task set execution. Checking the matching plan at this time, this person is in the matching sequence. In addition, the time period in which he was in an abnormal state was from the 15th to the 21st day, and the task started on the 15th and 47th day. Therefore, this person's abnormal condition will affect the development of the collaborative project, and the person needs to be re-matched. Task 010209 is put back into the collaborative design task pool, and the genetic algorithm with the elite retention strategy mentioned above is used to re-match this task. The new matching result is person 6. The load of person 6 at this time is 0.19. After taking on this task, the load becomes 0.58, which does not exceed the rated maximum load of 2, and can be assigned. The estimated completion time for designer 6 to complete this task is 3.10 days.

[0199] After changing the number of designers responsible for task 010209 from 9 to 6, the execution time of the design task set is 18.57 days and the cost is 7263.44 yuan. Figure 3 shown.

[0200] Furthermore, a simulated annealing algorithm was used to solve this problem for rematching. The result was that the number of designers responsible for task 010209 was changed from 9 to 14. The design task execution time was now 18.75 days, and the cost was 7,240.55 yuan. Comparing these results with those of the elite-preserving genetic algorithm after normalization, it can be concluded that the elite-preserving genetic algorithm improved the overall performance of the simulated annealing algorithm by approximately 2.41% when handling the rematching problem caused by personnel changes. The results are shown in Table 6.

[0201] Table 6 Comparison of the effects of different personnel rematching methods

[0202]

[0203] 2. About the Incremental Disturbance of Urgent Tasks

[0204] Taking the 120MN extruder design project published by Z company on the collaborative manufacturing platform as an example, the "task-personnel" matching scheme for product component design according to the original matching scheme is as follows: Figure 4 shown.

[0205] On the 7th day of collaborative task execution, the project management center received two new urgent tasks, such as Figure 5 As shown in Table 7, it is necessary to start the allocation and execution of urgent tasks immediately after the current tasks are completed. The specific task information is shown in Table 7.

[0206] Table 7 Emergency task information

[0207]

[0208] For the two new tasks, we use the overall rematching method and the partial rematching method respectively, and the new matching results are as follows:

[0209] (1) Overall rematching: The matching relationship between completed and ongoing tasks remains unchanged, and the urgent task and subsequent unfinished tasks are merged to obtain a new matching sequence such as Figure 6 As shown in the figure, the predicted time is 13.44 days, the predicted cost is 5285 yuan, and the average load rate of the designers is 0.92.

[0210] 2) Partial rematching: The original task matching relationship remains unchanged, and only urgent tasks are matched. The matching scheme of urgent tasks and unfinished tasks is merged, such as Figure 7 As shown in the figure, the predicted time is 13.55 days, the predicted cost is 5491.26 yuan, and the average load factor of the designers is 0.69.

[0211] Through the objective function, the objective function values ​​of the two re-matching methods are calculated. The overall re-matching effect is better than the partial re-matching effect by about 18.75%, as shown in Table 8.

[0212] Table 8 Comparison of the effects of different task rematching methods

[0213]

[0214] Therefore, we choose the overall rematching approach to solve the rematching problem of task incremental disturbance, and get the overall matching solution as follows: Figure 8 As shown in the figure, at this time, the total time to complete the task set is 21.78 days and the total cost is 8531.38 yuan.

[0215] 3. Regarding Mission Technology Change Disturbance

[0216] Taking the 120MN extruder design project released by Z company on the collaborative manufacturing platform as an example, the "task-personnel" matching scheme for product component design according to the original matching scheme is as follows: Figure 9 shown.

[0217] On the 13th day of collaborative task execution, the project management center received notifications of technical changes for three tasks. After processing by the BertTextCNN model, the task IDs and required skill IDs obtained by mapping the technical change text content are shown in Table 9.

[0218] Table 9 BertTextCNN model processing results

[0219]

[0220] The required skills are fed back to the project dynamic monitoring system, and the corresponding design task information model is updated. Due to the improvement of task technology, the standard time for task completion increases by about 1 / 3 based on the original time. The new task information is shown in Table 10.

[0221] Table 10 Task information after updating skill requirements

[0222]

[0223] Combine the tasks after the technical change with the unfinished tasks, rematch them according to the above method, and check the personnel load to get the following results: Figure 10 As shown, the prediction time is 12.79 days and the prediction cost is 5006.77 yuan.

[0224] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit the technical solutions. Those skilled in the art should understand that modifications or equivalent replacements of the technical solutions of the present invention that do not depart from the purpose and scope of the technical solutions of the present invention should be included in the scope of the claims of the present invention.

Claims

1. A method for re-matching complex heavy equipment collaborative design tasks and personnel under disturbance conditions, characterized by: The following steps are involved: S1. Obtain the original allocation plan of collaborative design personnel, which includes the corresponding responsible personnel for each task; monitor the collaborative design process of complex heavy equipment based on the original allocation plan, and determine whether a disturbance occurs. If so, proceed to S2; if not, continue monitoring; S2. Determine the type of disturbance. If it is a personnel change disturbance, go to S3; if it is an urgent task increment disturbance, go to S4; if it is a task technology change disturbance, go to S5. The personnel change disturbance is when the time period for a person to perform a task changes from available to unavailable. S3. Taking the minimization of the total time and total cost of the remaining tasks as the optimization goal, a multi-objective optimization model of personnel disturbance is constructed and solved to obtain a rematching solution; then go to S6; S4: Add the urgent task increment as an insertion task to the remaining tasks, recording it as the total task after the insertion. Taking the minimization of the total time and total cost of the total task after the insertion as the optimization goal, construct a multi-objective optimization model for the insertion disturbance. Solve it and obtain a rematching solution. Then go to S6. S5. Analyze the changed technical content and obtain the remaining tasks after the change; Taking the minimization of the total time and total cost of the remaining tasks after the change as the optimization goal, a multi-objective optimization model of technical disturbance is constructed and solved to obtain a rematching solution; then go to S6; S6. Use the obtained rematching plan to carry out corresponding personnel transfer and allocation.

2. The method for re-matching complex heavy equipment collaborative design tasks and personnel under disturbance conditions as claimed in claim 1, characterized in that: In S3, the objective function of the multi-objective optimization model of personnel disturbance is: Where, T a2 Represents the total time required for the remaining tasks to be matched, T max 、T min They represent the longest and shortest time required to complete the remaining tasks by different personnel; C2 represents the total cost required after the remaining tasks are matched, and C max 、C min They represent the maximum cost and minimum cost of the remaining tasks required to be completed by different personnel; ω1 and ω2 are weight coefficients respectively.

3. The method for re-matching complex heavy equipment collaborative design tasks and personnel under disturbance conditions as claimed in claim 2, characterized in that: In S4, the objective function of the single-disturbance multi-objective optimization model is: Where, T a3 , C3, L ave They represent the total time, total cost, and average load of designers after redistributing the total task after the insertion order as a whole; T a4 , C4, L ave4 They represent the total time, total cost, and average load of designers for the total tasks after the insertion of the task, with the remaining tasks remaining unchanged and only the personnel redistributed for the insertion task; T max 、T min They represent the longest and shortest time required for different personnel to complete the total task after inserting the order; C max 、C min They represent the maximum cost and minimum cost of the total task after inserting the order due to the use of different personnel; L max , L min They represent the maximum average load and minimum average load of the total task after inserting the order due to the use of different personnel; ω3, ω4, and ω5 are weight coefficients.

4. The method for re-matching complex heavy equipment collaborative design tasks and personnel under disturbance conditions as claimed in claim 3, characterized in that: In S5, the objective function of the technical perturbation multi-objective optimization model is: Where, T max 、T min They represent the longest and shortest time required to complete the total task after the change due to the use of different personnel; C max 、C min They represent the maximum cost and minimum cost of the total task after the change due to the use of different personnel; T a5 , C5 are the total time and total cost after the total task is redistributed after the change; ω6 and ω7 are the weight coefficients respectively.

5. The method for re-matching complex heavy equipment collaborative design tasks and personnel under disturbance conditions as claimed in claim 4, characterized in that: In S3, S4, and S5, when calculating the total time after redistribution, the actual time T for the redistributed person i to complete the jth task is aij The calculation formula is: T aij =Min{T ej (1-p max ),T ej (1-p ij )}; Where p max is the preset maximum proficiency; p ij T represents the proficiency of person i in task j; ej (1-p max ) represents the maximum proficiency p based on the preset max Calculate the shortest time, T ej (1-p ij ) indicates that the ij Calculate the theoretical time it takes for person i to perform task j.

6. The method for re-matching complex heavy equipment collaborative design tasks and personnel under disturbance conditions as claimed in claim 5, characterized in that: In S3, S4, and S5, when calculating the total cost after redistribution, the actual cost C of the redistributed person i completing the jth task is ij The calculation formula is: C ij =T aij *[1+ξ(k-1)]; In the formula, k is the personnel level, and ξ represents the salary coefficient that changes with the promotion of job level.

7. The method for re-matching complex heavy equipment collaborative design tasks and personnel under disturbance conditions as claimed in claim 6, characterized in that: Personnel proficiency p ij The calculation process includes: Constructing a task information table and a personnel information table; the task information table includes information of various dimensions of tasks; the personnel information table includes information of various dimensions of personnel; Calculate the similarity between tasks and construct a task similarity matrix; Construct a proficiency prediction model to calculate the proficiency p of person i in task j based on the information in the person information table and the task similarity matrix ij ; Where m represents the historical completion frequency, which is obtained by counting the number of historical executions of similar tasks; o represents the task similarity, which is determined by calculating the mean similarity between historical tasks and current tasks through the task similarity matrix; k represents the personnel title level; α, β, γ, and b are the parameters of the fitting curve to be determined. The compression ratio of the actual working time of historical tasks to the standard working time is used to quantify the corresponding proficiency p ij , combined with the corresponding m, k, o, the parameters α, β, γ, and b of the proficiency prediction model are iteratively optimized by the least squares method; Use the quantitative prediction model to calculate the proficiency p of each person ij .

8. The method for re-matching complex heavy equipment collaborative design tasks and personnel under disturbance conditions as claimed in claim 7, characterized in that: The process of constructing the task similarity matrix includes: Construct the skill similarity evaluation matrix q required for the task: Where q ij Characterizes the matching degree between task i and benchmark task in terms of the skill j dimension required by the task; n is the total number of tasks, u is the total number of required skill dimensions; Construct the task type similarity evaluation matrix l: Where, l ij represents the matching degree between task i and benchmark task in terms of task type j dimension; v is the total number of task type dimensions; Construct a comprehensive task similarity matrix: Where s i =(q i1 ,…,q iu ,l i1 ,…,l iv ) represents the comprehensive similarity code of task i, and n is the total number of design tasks; The generalized Jaccard similarity is used to calculate the intersection-over-union ratio of the skill sets of the two tasks as the similarity between the two tasks: Get the task similarity matrix I:

9. The method for re-matching complex heavy equipment collaborative design tasks and personnel under disturbance conditions as claimed in claim 1, characterized in that: In S5, the BertTextCNN model is used to automatically process technical text information and parse the changed technical content.

10. The method for re-matching complex heavy equipment collaborative design tasks and personnel under disturbance conditions as claimed in claim 1, characterized in that: In S3, an elite-preserving genetic algorithm is used to solve the multi-objective optimization model of personnel changes.

Images