A method for evaluating operation and maintenance effect of a ship production area

Abstract

The application discloses a kind of ship production area operation effect evaluation method, comprising: obtaining operation area work content generation technical response factor;According to production schedule, production link priority is divided;Dynamic evaluation threshold function is constructed to modify technical response factor;Obtain operating state and generate state feature vector;Based on technical response factor, priority and state feature vector, generate comprehensive evaluation result;Whether maintenance is executed is determined by threshold comparison;When triggering maintenance, locate abnormal source and generate maintenance scheme.The application quantifies the technical features of the operation area through technical response factors, dynamically adjusts the evaluation threshold by combining production priorities, constructs a multi-level comprehensive evaluation model to achieve accurate state assessment, and quickly locates the root cause of the problem using parallel evaluation channels, solving the problem of strong subjectivity and evaluation lag in traditional manual inspection, and achieving intelligent monitoring and accurate maintenance of ship production area operation.

Description

Technical Field

[0001] This invention relates to the field of ship manufacturing and processing technology, and specifically to a method for evaluating the operational effectiveness of ship production areas. Background Technology

[0002] Shipbuilding is a typical complex manufacturing process, involving collaborative operations across multiple work areas, cross-process production, and the dynamic allocation of a large number of personnel, equipment, and materials. The operational status of the production area directly affects product quality, production efficiency, and safety management levels. As shipbuilding develops towards refinement and intelligence, higher demands are placed on real-time monitoring, scientific evaluation, and precise maintenance of the production area.

[0003] Currently, the operation and management of ship production areas mainly rely on manual inspections, experience-based judgment, and paper records. Production management personnel conduct regular inspections to manually check the operating status of equipment, the placement and sequence of materials, the compliance of personnel operations, and environmental parameters in the work area, recording the inspection results in paper forms or electronic documents. This approach has the following technical shortcomings: Data collection methods are outdated, relying on human senses and experience-based judgment, and lacking objective and quantitative detection methods. The existing evaluation system is mainly based on monthly or quarterly centralized inspections, with evaluation indicators mostly qualitative descriptions and lacking scientific quantitative models. This fails to reflect dynamic changes in production areas and makes it difficult to promptly detect and warn of potential problems. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and provide an evaluation method for the operation and maintenance effect of ship production areas, so that the production mode can respond more effectively to the changes in production conditions and requirements.

[0005] To achieve the above objectives, the technical solution provided by the present invention is as follows.

[0006] A method for evaluating the operational effectiveness of a ship production area includes the following steps: S10. Obtain the work content of the ship production area, extract the process characteristic parameters and work attribute information of each work area, and generate technical response factors based on the preset quantitative model. S20. Obtain the ship production schedule, extract the process connection relationship and task time distribution characteristics, and divide each work area into several production links with different priorities according to the duration distribution and task density of the production schedule. S30. Based on the priority of the production process, construct a dynamic evaluation threshold function, correct the threshold of the technical response factor, and generate a comprehensive evaluation scheme that adapts to the current production status. S40. According to the overall evaluation scheme, obtain the operating status of the production area, which are the production status feature vectors of different work areas. S50. Based on the technology response factor, the priority of the production process, the state feature vector, and the comprehensive evaluation results of each work area, generate the following: S60. Compare the comprehensive evaluation result with the evaluation threshold function to determine whether to execute the maintenance mechanism; S70. Repeat the above steps until the maintenance mechanism is triggered, locate the root cause of the anomaly and generate a maintenance plan, and determine the timing of maintenance execution in conjunction with the production schedule.

[0007] As a preferred technical solution, in step S10, the quantification model includes a technical response factor vector R=[α,β, γ, δ]; Where α is the process complexity coefficient, expressed by the formula α = ∑(w i × p i w is obtained by calculating ) / N. i For the first i The weighting coefficient of each process, p i For the first i The precision grade coefficient of each process, where N is the total number of processes; β is the operational difficulty coefficient, expressed by the formula β = ∑(s j × t j ) / M is calculated to obtain s j For the first j Skill level requirements for this type of task, t j For the first j The percentage of working hours for each type of task, where M is the total number of task categories; γ represents the device coupling degree. By constructing a device association graph G=(V,E), the degree centrality and betweenness centrality of each node are calculated and then fused to generate the device coupling degree. δ is the quality risk index, which is generated by a time series prediction model based on historical quality data and process stability parameters.

[0008] As a preferred technical solution, in S20, the production scheduling data includes at least the start and end times of tasks in each work area, the process connection relationship, and the resource utilization plan; Based on the process connection relationship, a directed acyclic graph (DAG) of the production task is constructed, and critical and non-critical paths are extracted. Based on the duration distribution of the production schedule, the production cycle is divided into several time windows T = {t1, t2,..., t...}. n}; For each time window, calculate the task density index ρ for each work area = total task hours / available hours and resource utilization rate η = allocated resources / total resources; Based on the task density index ρ and resource utilization rate η, the priority of each work area within the time window is determined by a clustering algorithm.

[0009] In a preferred technical solution, in step S30, the dynamic evaluation threshold function is expressed as Th = f(R,P, t), where R is the technical response factor vector, P is the production process priority, and t is the time window. The dynamic evaluation threshold function modifies the technology response factor R based on the production process priority P, satisfying the following: When P is a high priority, Th = R × (1 + k1), where k1 is the first adjustment coefficient and k1 > 0; When P is of medium priority, Th = R; When P is a low priority, Th = R × (1 - k2), where k2 is the second adjustment coefficient and k2 > 0.

[0010] As a preferred technical solution, in step S30, the value range of the first adjustment coefficient k1 is 0.2~0.5, and the value range of the second adjustment coefficient k2 is 0.1~0.3.

[0011] As a preferred technical solution, in S40, the operating status data includes at least one type of characterizing data reflecting the operating status of equipment in the work area, production progress status, quality status, personnel status, and material status. The acquired operational status data is preprocessed, and the preprocessing includes at least one of data cleaning, normalization, and time-series alignment. Feature extraction is performed on the preprocessed data to generate a state feature vector S = [s1, s2, ..., s...]. m ].

[0012] As a preferred technical solution, in S50, a multi-level comprehensive evaluation model is constructed, the model including an indicator layer, a criterion layer and a target layer; The indicator layer is composed of the dimensions of the state feature vector S, the criterion layer is composed of the dimensions of the technical response factor R, and the target layer is the comprehensive score of the work area. Determine the baseline weighting coefficients for each level of indicators, including: Construct the judgment matrix A = [a] of the criterion layer ij ], where a ij Indicates the first i The factor relative to the firstj The importance of each factor; Calculate the eigenvectors of the judgment matrix A, and after normalization, obtain the weight vectors w of each factor in the criterion layer; Perform a consistency check on the judgment matrix. If the consistency ratio CR < 0.1, the weight coefficients are considered valid. Determine the weight matrix U of each indicator in the indicator layer relative to the criterion layer; Calculate the comprehensive weight vector W of each indicator in the indicator layer relative to the target layer. base = U × w; The baseline weight coefficient is dynamically adjusted based on the production process priority P to obtain the adjusted weight coefficient W. adj ; The state feature vector S is input into a multi-level comprehensive evaluation model, and a comprehensive evaluation result Score = ∑(W) is generated through weighted aggregation calculation. adji × s i ).

[0013] As a preferred technical solution, in step S50, an identification model for comprehensive evaluation results is constructed. The identification model includes a main evaluation channel and several special evaluation channels set in parallel. The main evaluation channel uses the multi-level comprehensive evaluation model to generate a comprehensive score for the work area. main ; The specific evaluation channels correspond to each dimension of the technical response factor R. Each specific evaluation channel adopts a weight allocation scheme dominated by its corresponding technical response factor to generate specific scores for process complexity (Scoreα), operation difficulty (Scoreβ), equipment coupling (Scoreγ), and quality risk (Scoreδ). The overall score is compared and analyzed with the scores of each specific item to construct a difference vector D = [dα, dβ, dγ, dδ], where dα = |Score main - Scoreα|,dβ = |Score main - Scoreβ|,dγ = |Score main - Scoreγ|,dδ = |Score main - Scoreδ|; When the overall score is lower than the threshold and the maintenance mechanism is triggered, the special evaluation channel corresponding to the maximum value in the difference vector D is identified as the problem-indicating link, which serves as the basis for root cause analysis.

[0014] As a preferred technical solution, the method further includes constructing a coupled correlation model between the technical response factor and the comprehensive evaluation result; The coupled correlation model is used to establish a dynamic mapping relationship between the technical response factor R and the comprehensive evaluation result Score, including: Collect historical operational data to construct a training sample set, with each sample containing a technology response factor vector R. i The corresponding comprehensive evaluation result, Score i And the production process priority P i ; Construct a deeply coupled network, which includes an input layer, a coupling layer, and an output layer; The input layer is used to receive the technical response factor vector R and the current production process priority P; The coupling layer includes several parallel coupling units, each corresponding to a technical response factor dimension, used to extract the single-dimensional coupling features between that dimension and the comprehensive evaluation result; The output layer is used to fuse the outputs of each coupled unit to generate a predicted evaluation score. pred ; The predicted evaluation score is Score pred Compared with the actual comprehensive evaluation result Score generated in step S50 actual Perform a comparison and calculate the coupling deviation δ = |Score pred - Score actual |; When the coupling deviation δ exceeds the preset deviation threshold, it is determined that the coupling relationship between the current technical response factor and the comprehensive evaluation result is abnormal, and the coupling reconstruction mechanism is triggered. The coupling reconstruction mechanism includes re-collecting recent operational data and updating the mapping relationship between the technical response factor and the comprehensive evaluation result.

[0015] As a preferred technical solution, the method further includes constructing an evolutionary model of production process priorities and technical response factors; The evolutionary model is used to describe the two-way influence relationship between production process priority P and technological response factor R, including: Collect historical data from multiple consecutive production cycles to construct a time-series dataset, where each time point contains a production process priority distribution P. t Technical response factor vector R t and the overall evaluation result (Score) t ; Construct a bidirectional influence function, which includes a positive influence branch and a negative influence branch; The positive impact branch is used to describe the influence of the technical response factor on the priority, and generates the priority change ΔP; The reverse influence branch is used to describe the impact of priority on the technical response factor and generate the change in the technical response factor ΔR. Based on the aforementioned bidirectional influence function, a co-evolutionary state equation is constructed to describe the coupling relationship between priority and technological response factor over time: P{t+1} = Pt + λP·ΔP R{t+1} = Rt + λR·ΔR Where λP and λR are the learning rate coefficients; Based on the evolutionary state equation, predict the priority distribution of production links and the evolution trend of technology response factors in future time windows; When the prediction results show that a certain technology response factor dimension will change by ±20%, the evaluation threshold of that dimension should be adjusted in advance.

[0016] The advantages and beneficial effects of this invention are as follows: By introducing a technical response factor, this invention achieves a quantitative characterization of the technical characteristics of each work area in the ship production zone. The process complexity coefficient, operation difficulty coefficient, equipment coupling degree, and quality risk index constitute a technical profile of the work area, providing an objective basis for subsequent evaluation.

[0017] This invention constructs a multi-level comprehensive evaluation model, linking state feature vectors and technical response factors through a hierarchical weighting system. Benchmark weights are determined through judgment matrix construction and consistency checks, and then dynamically adjusted based on production process priorities. This ensures the evaluation results reflect the real-time operational status of the work area. The parallel identification model with main and specialized evaluation channels can quickly pinpoint problems using difference vectors when comprehensive evaluation results are abnormal.

[0018] This invention constructs an evolutionary model of production process priorities and technological response factors, describing the bidirectional influence between the two. The positive influence branch characterizes the impact of technological factors on priority distribution, while the negative influence branch characterizes the role of priorities in the evolution of technological factors. Based on this equation, future trends can be predicted, allowing for the adjustment of evaluation thresholds before significant changes in technological factors occur, thus enabling proactive responses to changes in production status. Detailed Implementation

[0019] To make the above-mentioned objectives, features, and advantages of this application more apparent and understandable, the specific embodiments of this application are described in detail below. It is to be understood that the specific embodiments described herein are merely illustrative of this application and not intended to limit it. 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.

[0020] The terms “comprising” and “having”, and any variations thereof, used in this application are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the steps or units listed, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to such process, method, product, or apparatus.

[0021] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly or implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0022] This invention provides a method for evaluating the operational effectiveness of ship production areas, aiming to address the problems of traditional ship production management, such as reliance on manual inspections, subjective evaluation, and delayed response. This method quantifies the inherent technical characteristics of the work area by introducing technical response factors, dynamically prioritizing production stages based on production scheduling, and constructing a dynamic evaluation threshold function and a multi-level comprehensive evaluation model to achieve accurate and real-time assessment of the operational status of the production area. When the evaluation results trigger the maintenance mechanism, this method can quickly locate the root cause of the anomaly and determine the optimal maintenance timing in conjunction with the production plan, thereby achieving a shift from passive response to proactive and intelligent maintenance, and improving the level of refined and intelligent management in shipbuilding.

[0023] This invention constructs a closed-loop adaptive evaluation and maintenance system. The method presented in this invention does not view the various operational states of a production area in isolation, but rather considers it as a complex system influenced by both inherent technological attributes and dynamic production demands. Therefore, the first step of the method is to establish a technical baseline profile of the work area, that is, to transform abstract characteristics such as process complexity, operational difficulty, equipment coupling, and quality risk into a quantifiable and comparable vector R through technical response factors. This provides a relatively stable technical background independent of production tasks for all subsequent evaluations, enabling the interpretation of operational states to be based on the inherent characteristics of the work area and avoiding one-size-fits-all misjudgments. For example, a highly complex work area should have a lower tolerance for short-term equipment fluctuations than a simple assembly area.

[0024] Next, the method incorporates the time dimension of production. Ship production is typically project-based, with significant differences in task intensity and resource utilization across different work areas within different time windows. By analyzing production schedules and constructing a directed acyclic graph (DAG) of tasks, the entire production cycle can be divided into dynamically changing production link priorities P. The underlying mechanism involves identifying critical paths and bottlenecks in the production process and assigning them higher "attention" weights; simulating the thinking and factual basis that production managers focus more on core work segments during busy periods, but quantifying this into objective priority indicators through algorithms.

[0025] To integrate technical benchmarks with dynamic production demands, the method constructs a dynamic evaluation threshold function Th = f(R, P, t). This is a key innovation of the invention. Its core mechanism no longer uses a fixed evaluation standard, but rather allows the standard to "flow" with changes in production status. Specifically, for high-priority production stages (e.g., operations on the critical path, welding, deck and keel assembly; conversely, painting), the evaluation threshold Th is appropriately increased (Th = R × (1 + k1)). This means that in these stages, the tolerance for operational status is lower, and the requirements are more stringent. Any slight deviation may be amplified and trigger an early warning to ensure that the main production line is not disrupted. Conversely, for low-priority stages, the threshold is appropriately relaxed (Th = R × (1 - k2)), allowing for some operational fluctuations. This avoids unnecessary maintenance resource contention caused by minor problems in secondary stages, thereby optimizing overall maintenance efficiency.

[0026] After establishing dynamic evaluation criteria, the method enters the data acquisition and processing phase. In accordance with the requirements of the overall evaluation scheme, real-time operational status data of the production area is acquired, including but not limited to equipment status data such as equipment vibration, temperature, and current; progress data such as production schedule deviations and material consumption rates; and quality inspection data and personnel attendance. This multi-source heterogeneous data, after preprocessing such as cleaning, normalization, and time-series alignment, is characterized into a multi-dimensional state feature vector S. This vector S records the current working state of the work area, providing a factual basis for the final evaluation.

[0027] Finally, the method integrates the aforementioned technical response factor R, production process priority P, and state feature vector S through a multi-level comprehensive evaluation model to generate an intuitive comprehensive evaluation result, Score. This model constructs a mapping relationship from the indicator layer to the criterion layer and then to the target layer using the analytic hierarchy process (AHP), and dynamically adjusts the weights between layers through the priority P. This ensures that the final score not only reflects the quality of the current operating state but also its importance within the current production context. When the Score falls below the dynamic threshold Th, the maintenance mechanism is triggered. At this point, the invention uses a parallel evaluation channel mechanism to quickly pinpoint the problem to specific technical dimensions, thus providing precise guidance for subsequent root cause analysis and maintenance plan development. Ultimately, by combining the production schedule to determine the maintenance timing, seamless integration of maintenance operations with production progress is achieved, minimizing losses caused by unplanned downtime.

[0028] In some embodiments, to more precisely characterize the inherent technical features of the work area, the quantification model described in step S10 is concretized as a four-dimensional technical response factor vector R=[α, β, γ, δ]. Wherein: α is the process complexity coefficient, calculated based on the decomposition of all processes within the work area. For example, for a hull section welding area, the processes can be subdivided into beveling, assembly spot welding, root pass welding, fill pass welding, cap pass welding, etc. α = ∑(w i × p i ) / N, w i For the first i The weight of each process step (e.g., based on time or impact on final quality), p i This is the accuracy level coefficient for this process (e.g., set to 1.0, 1.2, 1.5, etc. based on the tolerance range), where N is the total number of processes. This coefficient quantifies the process complexity of the work area from two dimensions: the number of processes and the accuracy requirements. The higher the α value, the more complex the process in that area, and the higher the requirement for process control stability.

[0029] β is the operational difficulty coefficient, expressed as β = ∑(s) j × t j ) / M is calculated. Where s j For the first j The skill level required for each type of job (e.g., 1 for entry-level worker, 2 for intermediate-level worker, and 3 for advanced-level worker), t j This refers to the proportion of time spent on this type of work to the total working hours. For example, in a pipe processing area, pipe bending operations might require highly skilled workers (s...). j =3), and its working hours account for a relatively high proportion (t). j =0.4), while the polishing work only requires a junior worker (s j =1), with a relatively low proportion of working hours (t) jIf the β value is less than or equal to 0.1, then the calculated β value can comprehensively reflect the region's dependence on human skills. A high β value means that fluctuations in personnel skills have a greater impact on production stability; therefore, more attention should be paid to personnel qualifications and operational standardization in status monitoring.

[0030] γ represents the equipment coupling degree. To quantify the degree of mutual influence between equipment, this method constructs an equipment association graph G=(V,E), where node V represents critical equipment and edge E represents the material flow or information flow relationship between equipment. By calculating the degree centrality (number of connected nodes) and betweenness centrality (number of shortest paths through that node) of each node in the graph and fusing these two indicators, the equipment coupling degree γ can be obtained. A high γ value indicates that the equipment in that area is closely connected, forming a structure similar to a production unit. Once a piece of equipment fails, the impact can easily propagate to other equipment through the association graph, causing a chain reaction. Therefore, for areas with high γ values, maintenance strategies should focus more on preventative maintenance and rapid repair capabilities.

[0031] δ represents the quality risk index. This index is a predictive indicator based on historical data. By collecting past quality non-conformity records, rework rates, and stability data of key process parameters (such as the variance of welding current and voltage fluctuations) for the work area, a time series prediction model (such as an ARIMA model or an LSTM network) is used to predict the quality risk level over a future period. A high δ value indicates a greater probability that the area will enter a period of high quality fluctuations, and should be given higher attention during the evaluation.

[0032] In some embodiments, detailed production scheduling data first needs to be obtained from the production management system. This data includes the start and end times of tasks in each work area, the connection between preceding and subsequent processes (e.g., "segmented painting" can only begin after "segmented outfitting" is completed), and the utilization plan of critical resources (e.g., the usage arrangement of the "800T hydraulic press"). Based on this data, a directed acyclic graph (DAG) of production tasks is constructed, and the critical path that determines the entire production cycle is identified using the Critical Path Method (CPM). The steps on the critical path are initialized as high priority. Next, the entire production cycle is divided into multiple time windows. Within each time window, the task density index ρ and resource utilization rate η of each work area are calculated. Then, clustering algorithms such as K-Means are used to cluster the (ρ, η) data points of each work area within each time window into three categories, corresponding to high, medium, and low priorities, respectively. The cluster with the highest ρ and η values ​​is marked as high priority, and the lowest as low priority. In this way, a work area may have different priorities in different time windows. For example, a segmented manufacturing area may have a medium priority at the beginning of the operation, but will be promoted to a high priority after entering the overall assembly stage because it is on the critical path.

[0033] In some embodiments, according to the actual experience of ship production, for high-priority links, the evaluation criteria are appropriately tightened, but excessive tightening will lead to an increase in the false alarm rate. Therefore, the value range of k1 is set to 0.2 - 0.5. For example, taking k1 = 0.3 means that the threshold of the high-priority link is 30% higher than the technical benchmark R, that is, it is required that its operating state needs to remain above the 70% percentile of the technical benchmark value to be qualified. For low-priority links, in order to save maintenance resources, the criteria are appropriately relaxed, and the value range of k2 is set to 0.1 - 0.3. Taking k2 = 0.2 means that the threshold of the low-priority link is 20% lower than the technical benchmark R, providing a 20% fluctuation redundancy space for its operating state. The specific values of these two coefficients can be determined by simulating and optimizing historical operation data.

[0034] In some embodiments, in step S40, raw operation data is first obtained through various sensors and information systems. For example, the operating condition of the equipment is obtained through vibration sensors, the production progress is obtained through the MES system, the quality condition is obtained through the QIS system, the personnel condition is obtained through the personnel positioning system, and the material condition (stock shortage warning state) is obtained through the WMS system. Then, preprocessing is performed: data cleaning to eliminate sensor noise and network transmission errors; normalization to map data with different dimensions (such as a temperature value of 200 °C and a vibration value of 0.05 mm) to the [0, 1] interval; time series alignment to ensure that data from different sampling frequency sources can be compared at the same timestamp. Finally, through feature extraction (such as calculating the mean, variance, change rate, etc. within a time window), the final state feature vector S is formed. For example, s1 can represent the "effective value of spindle vibration", and s2 represents the "production progress deviation rate".

[0035] In some embodiments, in order to construct the multi-level comprehensive evaluation model described in step S50, the present invention adopts an improved analytic hierarchy process. This method first establishes a three-layer structure: Goal layer: Comprehensive score of the work area (Score).

[0036] Criterion layer: Composed of each dimension of the technical response factor R, that is, [process complexity α, operation difficulty β, equipment coupling degree γ, quality risk δ].

[0037] Index layer: Composed of each dimension of the state feature vector S. For example, for the criterion layer "equipment coupling degree γ", the lower-level index layer dimensions may include "vibration intensity of key equipment s1", "mean time between equipment failures s2", etc.

[0038] The process of determining the baseline weights is as follows: Constructing the judgment matrix A: Domain experts compare the relative importance of the four factors in the criterion layer pairwise using a 1-9 scale. For example, if an expert believes that "equipment coupling degree γ" is significantly more important than "process complexity α", then the matrix element a(γ,α)=5, and its reciprocal a(α,γ)=1 / 5, thus constructing a 4x4 judgment matrix A.

[0039] Calculate the eigenvector corresponding to the largest eigenvalue of matrix A, and normalize it to obtain the baseline weight vector w for each factor in the criterion layer. .

[0040] Calculate the consistency ratio CR. If CR < 0.1, then the consistency of the judgment matrix is ​​acceptable and the weight vector w is valid.

[0041] For each factor in the criterion layer, a judgment matrix is ​​constructed again using expert judgment, relating it to the corresponding indicator layer dimension. The weights of each indicator in the indicator layer relative to its respective criterion layer factor are then calculated. This ultimately forms a matrix U, where U{ ij} indicates the first i The first indicator for the first j The weights of each criterion factor.

[0042] Multiplying the weight matrix U of the indicator layer to the criterion layer with the weight vector w of the criterion layer to the target layer yields the benchmark comprehensive weight vector W of each indicator in the indicator layer relative to the target layer. base = U × w.

[0043] When the priority P of the production process changes, W is adjusted according to preset rules. base After making corrections, we obtain W. adj For example, when P is a high priority, the weights of criterion factors highly correlated with the "critical path" can be appropriately increased, while the weights of other factors are correspondingly decreased, but the sum remains 1. Finally, the state feature vector S and the adjusted weights W are... adj Weighted aggregation is performed to obtain the final comprehensive evaluation result: Score = ∑ .

[0044] In step S20, in addition to prioritizing based on production scheduling, regional spatial location factors and scheduling operation sequence factors are introduced. By constructing a regional spatial adjacency matrix and a material flow path diagram, the spatial coupling degree and scheduling dependency degree between each operation area are calculated as auxiliary basis for priority division. If an area with high spatial coupling degree is also located on the scheduling critical path, fluctuations in its operation and maintenance status may trigger a chain reaction, requiring a higher evaluation weight. The scheduling operation sequence is determined by analyzing the pre- and post-task relationships and material flow direction to identify scheduling bottleneck areas, serving as one of the correction factors in the dynamic evaluation threshold function, further enhancing the evaluation results' ability to reflect the actual production physical process.

[0045] S20 also includes constructing a spatial layout model of the production area, wherein the spatial layout model includes: Obtain physical layout data of the production area, including at least the spatial coordinates of each work area, area boundaries, material transfer paths, and personnel movement planning; Based on the physical layout data, construct the region spatial adjacency matrix Aspat. i al, where matrix element a ij Indicates the first i Work area and the j The degree of spatial adjacency between work areas is calculated based on the straight-line distance, path distance, and logistics frequency between the two work areas. Based on the aforementioned regional spatial adjacency matrix Aspat i al, calculate the spatial coupling degree C of each work area i = ∑(a ij × w ij ) / n, where w ij The weighting coefficient is set based on the intensity of material flow, where n is the total number of work areas; The spatial coupling degree C i As an auxiliary factor for priority division, the priorities generated in step S20 are corrected, when the spatial coupling degree C i When the preset threshold is exceeded, the priority of the work area will be raised by one level.

[0046] In some embodiments, in order to quickly locate the root cause of the problem after the maintenance mechanism is triggered, a set of identification models including one main channel and four special channels are constructed in parallel in step S50.

[0047] The main channel adopts the above-mentioned multi-level comprehensive evaluation model to obtain a comprehensive score. main The four specialized evaluation channels correspond to the four factors in R: process complexity, operational difficulty, equipment coupling, and quality risk. Each specialized channel employs an extreme weighting scheme when calculating its specialized score. For example, for the "equipment coupling score Scoreγ," when constructing the evaluation model, the weight wγ of "equipment coupling γ" in the criterion layer is set to an extremely high value (e.g., 0.7), while the remaining 0.3 weights of the other three factors are evenly distributed. Thus, the Scoreγ result primarily reflects the quality of the equipment coupling-related status. When the comprehensive score Score... main When the value falls below the dynamic threshold Th, triggering the maintenance mechanism, the system automatically calculates the difference vector D = Assume that in the difference vector D, |Score main- A maximum value for Scoreγ indicates that the current operational issues in the work area are mainly related to equipment coupling (such as deterioration of equipment interconnection status or a chain reaction caused by equipment failure), while process, operation, and quality risks are relatively normal. This provides maintenance personnel with valuable clues for root cause analysis, enabling them to get straight to the point, check the equipment network and critical node equipment, thereby significantly shortening troubleshooting time.

[0048] In some embodiments, this method also introduces a self-learning mechanism, namely, constructing a coupled correlation model between the technology response factor and the comprehensive evaluation result. This model is mainly used to learn and determine how the technology response factor R affects the final score. It is implemented through a deeply coupled network. The input to this network is the technology response factor vector R and the current priority P, and the output is a predicted evaluation score. pred Its learning goal is to improve Score. pred As close as possible to the Score calculated from the actual state feature vector S in step S50. actual Through training with a large amount of historical data, the network learned a complex, non-linear mapping relationship between the intrinsic properties of R and the final score under different production priorities.

[0049] The score is calculated in real time during system operation. pred With Score actual The coupling deviation δ. If δ suddenly exceeds the preset threshold, it means that the current operating state S is different from the state Score that should be obtained according to the technical benchmark R. pred A discrepancy has emerged. This could indicate a new problem that the model hasn't learned; for example, although the device state feature vector S shows everything is normal (all S values ​​are good), the actual overall score... actual However, it is very low, resulting in a score that is lower than the predicted score based on the technical benchmark R. pred The deviation is significant. This could indicate a problem that the sensors cannot directly detect (such as a batch-to-batch decline in material quality). In this case, the coupling reconstruction mechanism is triggered, and the system prompts that recent data needs to be collected again to update the model and incorporate this new "coupling relationship" into the learning process.

[0050] In some embodiments, this method goes a step further, constructing an evolutionary model of production process priority and technological response factors to describe their bidirectional interaction over time. The model constructs a time series by collecting historical data from multiple consecutive production cycles. It learns the influence in two directions: Positive impacts include, for example, a sustained increase in the quality risk index δ of a certain work area over several periods, indicating instability in quality within that area. The model learns that when δ rises to a certain level, the production plan will often be proactively adjusted, moving subsequent tasks in this area off the critical path or reducing their task intensity to avoid quality risks impacting the overall schedule. This impact is quantified as a change in priority ΔP.

[0051] Conversely, when a work area is marked as high priority, the most skilled operators are often deployed, and equipment inspection and maintenance are intensified. This may lead to an improvement in the operational difficulty coefficient β and equipment coupling degree γ of that area in subsequent cycles. This effect is quantified as the change in the technology response factor ΔR.

[0052] Based on the bidirectional influence function, a co-evolutionary state equation is constructed: P{t+1} = Pt + λP·ΔP and R{t+1} = Rt + λR·ΔR. This equation describes how production management and technical attributes mutually shape and co-evolve. Using this equation, the priority distribution of production links and the evolution trend of technical response factors in future time windows can be predicted. When the prediction results show that a certain technical response factor dimension (such as the quality risk index δ) may deteriorate by more than 20% within the next week, the system can issue an early warning and suggest adjusting the evaluation threshold of that dimension in advance (for example, increasing the special evaluation weight corresponding to δ in advance) to achieve proactive perception and prevention of potential problems, transforming passive maintenance into truly predictive maintenance.

[0053] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent should be determined by the appended claims.

Claims

1. A method for evaluating the operational effectiveness of a ship production area, characterized in that, Includes the following steps: S10. Obtain the work content of the ship production area, extract the process characteristic parameters and work attribute information of each work area, and generate technical response factors based on the preset quantitative model. S20. Obtain the ship production schedule, extract the process connection relationship and task time distribution characteristics, and divide each work area into several production links with different priorities according to the duration distribution and task density of the production schedule. S30. Based on the priority of the production process, construct a dynamic evaluation threshold function and adjust the threshold of the technical response factor. S40. Obtain the operating status of the production area, which are the production status feature vectors of different work areas; S50. Based on the technology response factor, the priority of the production process, the state feature vector, and the comprehensive evaluation results of each work area, generate the following: S60. Compare the comprehensive evaluation result with the evaluation threshold function to determine whether to execute the maintenance mechanism; S70. Repeat the above steps until the maintenance mechanism is triggered, locate the root cause of the anomaly and generate a maintenance plan, and determine the timing of maintenance execution in conjunction with the production schedule.

2. The evaluation and maintenance method according to claim 1, characterized in that, In S10, the quantification model includes a technology response factor vector R=[α, β, γ, δ]; Where α is the process complexity coefficient, expressed by the formula α = ∑(w i × p i w is obtained by calculating ) / N. i For the first i The weighting coefficient of each process, p i For the first i The precision grade coefficient of each process, where N is the total number of processes; β is the operational difficulty coefficient, expressed by the formula β = ∑(s) j × t j ) / M is calculated to obtain s j For the first j Skill level requirements for this type of task, t j For the first j The percentage of working hours for each type of task, where M is the total number of task categories; γ represents the device coupling degree. By constructing a device association graph G=(V,E), the degree centrality and betweenness centrality of each node are calculated and then fused to generate the device coupling degree. δ is the quality risk index, which is generated by a time series prediction model based on historical quality data and process stability parameters.

3. The evaluation and maintenance method according to claim 1, characterized in that, In S20, the production scheduling data includes at least the start and end times of tasks in each work area, the process connection relationship, and the resource utilization plan; Based on the process connection relationship, a directed acyclic graph (DAG) of the production task is constructed, and critical and non-critical paths are extracted. Based on the duration distribution of the production schedule, the production cycle is divided into several time windows T = {t1, t2, ...,t...} n }; For each time window, calculate the task density index ρ for each work area = total task hours / available hours and resource utilization rate η = allocated resources / total resources; Based on the task density index ρ and resource utilization rate η, the priority of each work area within the time window is determined by a clustering algorithm.

4. The evaluation and maintenance method according to claim 1, characterized in that, In S30, the dynamic evaluation threshold function is expressed as Th = f(R, P, t), where R is the technology response factor vector, P is the production process priority, and t is the time window. The dynamic evaluation threshold function modifies the technology response factor R based on the production process priority P, satisfying the following: When P is of high priority, Th = R × (1 + k1), where k1 is the first adjustment coefficient and k1 > 0; When P is of medium priority, Th = R; When P is a low priority, Th = R × (1 - k2), where k2 is the second adjustment coefficient and k2 > 0.

5. The evaluation and maintenance method according to claim 4, characterized in that, The first adjustment coefficient k1 has a value range of 0.2 to 0.5, and the second adjustment coefficient k2 has a value range of 0.1 to 0.

3.

6. The evaluation and maintenance method according to claim 5, characterized in that, In S40, the operating status data includes at least one type of characterization data reflecting the operating status of equipment in the work area, production progress status, quality status, personnel status, and material status. The acquired operational status data is preprocessed, and the preprocessing includes at least one of data cleaning, normalization, and time-series alignment. Feature extraction is performed on the preprocessed data to generate a state feature vector S = [s1, s2, ..., s...]. m ].

7. The evaluation and maintenance method according to claim 6, characterized in that, As a preferred technical solution, in S50, a multi-level comprehensive evaluation model is constructed, the model including an indicator layer, a criterion layer and a target layer; The indicator layer is composed of the dimensions of the state feature vector S, the criterion layer is composed of the dimensions of the technical response factor R, and the target layer is the comprehensive score of the work area. Determine the baseline weighting coefficients for each level of indicators, including: Construct the judgment matrix A = [a] of the criterion layer ij ], where a ij Indicates the first i The factor relative to the first j The importance of each factor; Calculate the eigenvectors of the judgment matrix A, and after normalization, obtain the weight vectors w of each factor in the criterion layer; Perform a consistency check on the judgment matrix. If the consistency ratio CR < 0.1, the weight coefficients are considered valid. Determine the weight matrix U of each indicator in the indicator layer relative to the criterion layer; Calculate the comprehensive weight vector W of each indicator in the indicator layer relative to the target layer. base = U × w; The baseline weight coefficient is dynamically adjusted based on the production process priority P to obtain the adjusted weight coefficient W. adj ; The state feature vector S is input into a multi-level comprehensive evaluation model, and a comprehensive evaluation result Score = ∑(W) is generated through weighted aggregation calculation. adji × s i ).

8. The evaluation and maintenance method according to claim 6, characterized in that, In S50, a comprehensive evaluation result identification model is constructed, which includes a main evaluation channel and several special evaluation channels set in parallel. The main evaluation channel uses the multi-level comprehensive evaluation model to generate a comprehensive score for the work area. main ; The specific evaluation channels correspond to each dimension of the technical response factor R. Each specific evaluation channel adopts a weight allocation scheme dominated by its corresponding technical response factor to generate specific scores for process complexity (Scoreα), operation difficulty (Scoreβ), equipment coupling (Scoreγ), and quality risk (Scoreδ). The overall score is compared and analyzed with the scores of each specific item to construct a difference vector D = [dα, dβ, dγ, dδ], where dα = |Score main - Scoreα|,dβ = |Score main - Scoreβ|,dγ = |Score main -Scoreγ|,dδ = |Score main - Scoreδ|; When the overall score is lower than the threshold and the maintenance mechanism is triggered, the special evaluation channel corresponding to the maximum value in the difference vector D is identified as the problem-indicating link, which serves as the basis for root cause analysis.

9. The evaluation and maintenance method according to claim 6, characterized in that, The method also includes constructing a coupled correlation model between the technology response factor and the comprehensive evaluation result; The coupled correlation model is used to establish a dynamic mapping relationship between the technical response factor R and the comprehensive evaluation result Score, including: Collect historical operational data to construct a training sample set, with each sample containing a technology response factor vector R. i The corresponding comprehensive evaluation result, Score i And the production process priority P i ; Construct a deeply coupled network, which includes an input layer, a coupling layer, and an output layer; The input layer is used to receive the technical response factor vector R and the current production process priority P; The coupling layer includes several parallel coupling units, each corresponding to a technical response factor dimension, used to extract the single-dimensional coupling features between that dimension and the comprehensive evaluation result; The output layer is used to fuse the outputs of each coupled unit to generate a predicted evaluation score. pred ; The predicted evaluation score is Score pred Compared with the actual comprehensive evaluation result Score generated in step S50 actual Perform a comparison and calculate the coupling deviation δ = |Score pred - Score actual |; When the coupling deviation δ exceeds the preset deviation threshold, it is determined that the coupling relationship between the current technical response factor and the comprehensive evaluation result is abnormal, and the coupling reconstruction mechanism is triggered. The coupling reconstruction mechanism includes re-collecting recent operational data and updating the mapping relationship between the technical response factor and the comprehensive evaluation result.

10. The evaluation and maintenance method according to claim 1, characterized in that, The method also includes constructing an evolutionary model of production process priorities and technological response factors; The evolutionary model is used to describe the two-way influence relationship between production process priority P and technological response factor R, including: Collect historical data from multiple consecutive production cycles to construct a time-series dataset, where each time point contains a production process priority distribution P. t Technical response factor vector R t and the overall evaluation result (Score) t ; Construct a bidirectional influence function, which includes a positive influence branch and a negative influence branch; The positive impact branch is used to describe the influence of the technical response factor on the priority, and generates the priority change ΔP; The reverse influence branch is used to describe the impact of priority on the technical response factor and generate the change in the technical response factor ΔR. Based on the aforementioned bidirectional influence function, a co-evolutionary state equation is constructed to describe the coupling relationship between priority and technological response factor over time: P{t+1} = Pt + λP·ΔP R{t+1} = Rt + λR·ΔR Where λP and λR are the learning rate coefficients; Based on the evolutionary state equation, predict the priority distribution of production links and the evolution trend of technology response factors in future time windows; When the prediction results show that a certain technology response factor dimension will change by ±20%, the evaluation threshold of that dimension should be adjusted in advance.