Integrated management platform for intelligent factory in pipeline equipment industry

Abstract

The invention discloses a pipeline equipment industry intelligent factory integrated management platform, and relates to the technical field of pipeline production management. Comprising a data acquisition module, a data preprocessing module, a production waste point identification module and a waste point processing and feedback module, through multi-source data acquisition and preprocessing, in combination with a clear logic rule and an AI identification model, waste points such as equipment shutdown, material accumulation and process abnormality are accurately identified, and the identification efficiency and accuracy are improved. By means of a stage treatment process, various wastes can be disposed in time, resource loss is reduced, and production efficiency is improved. The model adaptation unit can be automatically adjusted according to pipeline specifications and process parameters, and the adaptability to different scenes is enhanced. The feedback mechanism optimizes the model and the production strategy through data iteration, forms an intelligent closed loop, reduces manual dependence, promotes management precision and automation, assists enterprises in continuously improving the production process, and improves the overall competitiveness.

Description

Technical Field

[0001] The present invention relates to the technical field of pipeline production management, and specifically to an integrated management platform for intelligent factories in the pipeline equipment industry. Background Art

[0002] In the pipeline equipment manufacturing industry, the complexity and high precision of the production process pose stringent requirements on management efficiency. Under the traditional production mode, the monitoring of equipment operating status, the scheduling of material flow, and the regulation of process parameters rely mainly on manual inspections and empirical judgments. There are not only problems such as lagging responses and large data errors, but it is also difficult to accurately identify and real-time process all process waste points.

[0003] Currently, the industry generally faces three typical production waste problems: First, frequent equipment downtime waste occurs, including unplanned downtime, overtime for changeover adjustments, etc. The traditional manual recording method is difficult to trace the reasons for downtime and the responsibility nodes, resulting in a long-term low effective operation rate; Second, material accumulation waste is widespread. Due to the lack of a dynamic monitoring mechanism, the phenomenon of mismatch between material storage and production rhythm occurs from time to time, which not only occupies storage space but may also cause quality losses due to material backlogs; Third, process anomaly waste is highly concealed. If fluctuations in key parameters such as welding temperature and pressure cannot be intervened in a timely manner, it is extremely easy to cause a decrease in product qualification rate, and the traditional post-event quality inspection mode is difficult to achieve process control.

[0004] In the prior art, some enterprises have tried to introduce single-function monitoring systems, such as equipment sensor acquisition modules or material management software, but there is a phenomenon of data silos: Equipment data, material data, and process parameters lack linkage analysis, resulting in fragmented identification of waste points. For example, in a pipeline welding process, abnormal temperature may trigger a chain reaction of equipment downtime and material backlogs, but an isolated system can only identify problems in a single link and cannot achieve full-chain traceability.

[0005] At the same time, the subjectivity of manual judgment further exacerbates management weaknesses. Statistics show that under the traditional mode, approximately 60% of equipment downtime waste is not processed in a timely manner due to omissions in manual inspections, and the response delay rate of material accumulation warnings is as high as 45%. In addition, there are significant differences in process parameters (welding temperature, pressure) corresponding to different pipeline specifications (diameter, length). The monitoring logic with fixed thresholds is difficult to adapt to the flexible production requirements of multiple varieties, resulting in high false alarm rates and missed alarm rates.

[0006] Therefore, building an integrated management platform that can achieve data collection, intelligent identification, dynamic processing, and continuous optimization has become the key breakthrough point for solving the production pain points in the pipeline equipment industry and enhancing core competitiveness. Summary of the Invention

[0007] (I) Technical Problems to be Solved To address the shortcomings of existing technologies, this invention provides an integrated management platform for intelligent factories in the pipeline equipment industry, solving the problems mentioned in the background section.

[0008] (II) Technical Solution To achieve the above objectives, the present invention provides the following technical solution: an integrated management platform for intelligent factories in the pipeline equipment industry, comprising: Data acquisition module: Used to acquire equipment operating parameters, material data and video image data during the production process through multi-source data acquisition terminals, and upload them to the database according to a pre-set fixed cycle; Data preprocessing module: used to remove outliers and reduce image noise in the collected data; Production waste point identification module: used to identify equipment downtime waste points and material accumulation waste points, and at the same time, it uses a pre-trained AI recognition model for auxiliary identification; Waste Point Handling and Feedback Module: This module is used to trigger waste point handling processes for identified waste points and feed back the waste point handling results to the AI ​​recognition model and database to optimize subsequent recognition and management.

[0009] As a further aspect of the present invention: the multi-source data acquisition terminal includes: Sensors installed on production equipment are used to collect data on equipment operating status, operating time, welding temperature T, and pressure P. Material sensors are placed in the material storage area to collect the material stacking height H and stacking volume V; Cameras installed in the workshop are used to collect video image data; The data acquisition period of the source data acquisition terminal is t, and t is a preset value.

[0010] As a further aspect of the present invention: the outlier removal method in the data preprocessing module is as follows: For welding temperature T, extract the pre-set normal range [T]. min ,T max ], and T values ​​that exceed the normal range are identified as outliers and removed; For the material accumulation height H, if the difference between the currently collected H value and the H value collected in the previous data collection cycle exceeds the preset normal fluctuation threshold ΔH... max If the value is not found, it is considered an outlier and removed. For video image data, image noise reduction processing is performed to improve clarity.

[0011] As a further aspect of the present invention, in image noise reduction processing: To address the salt-and-pepper noise in images caused by equipment vibration and dust in pipeline equipment production scenarios, an adaptive median filtering algorithm is used for image denoising. To address the high-frequency noise in images caused by uneven lighting in the workshop and welding arc light, Gaussian low-pass filtering is used for image noise reduction. As a further aspect of the present invention, the logic for identifying and judging equipment downtime waste points is as follows: The equipment downtime is denoted as t. stop The standard duration for model replacement and adjustment is marked as t. change (Based on historical replacement data statistics), the replacement operation trigger signal is S; When S=1, it indicates that a type-change operation has been triggered; when S=0, it indicates that no type-change operation has been triggered. When the equipment stops: If S=1, and t stop >t change If it exceeds the duration t waste =t stop -t change This was identified as an abnormal downtime and waste point. If t stop ≤ tchange The shutdown was deemed a reasonable changeover shutdown. If S=0, and the fluctuation value ΔX of the equipment operating parameters is within the preset normal range ΔX normal Inside, determine t stop The corresponding downtime is an abnormal downtime waste point; Among them, the equipment operating parameter fluctuation value ΔX is the equipment operating parameter extracted within a specified time window length L, and calculated using the standard deviation formula. It can be concluded that; In the formula, T j TP represents the average value of the equipment operating parameters collected within a time window of length L. When the fluctuation values ​​of the equipment operating parameters corresponding to the running time, welding temperature T, and pressure P are all within the normal range ΔX normal If S=0, then t stop The corresponding downtime is an abnormal downtime waste point.

[0012] As a further aspect of the present invention: the identification of material accumulation waste points is assessed through a risk coefficient R, calculated using the following formula: ; Among them, H current H represents the current material stacking height. safe For safety height threshold; V current V represents the current material accumulation volume. safe For safe volume threshold; vs currentFor production rate; when R exceeds the risk threshold R threshold When this occurs, it is determined to be a point of material accumulation and waste.

[0013] As a further aspect of the present invention, the construction of the AI ​​recognition model includes: Training data preparation: Collect labeled production data, including: pipe specifications with diameter D and length L, process parameters with welding temperature T and pressure P, equipment status, material data, and marked waste points and normal status; Model structure: The input layer receives equipment operating parameters, material data and video image data after pre-passing outlier removal and image noise reduction; the hidden layer extracts features; and the output layer outputs the waste point identification results. Loss function: Cross-entropy loss function is used; The formula is: ; Where N is the number of samples, y i For the actual label, where y i =1 indicates wasted points, y i =0 indicates a non-wasteful point), y0 i Predict probabilities for the model.

[0014] As a further aspect of the present invention: the AI ​​recognition model includes a model adaptation unit, which automatically adjusts the weights of internal parameters according to the pipe specifications (D, L) and learns the normal and waste patterns under different combinations of process parameters (T, P); When the welding temperature exceeds the normal range and the duration exceeds the threshold (tT, in seconds), it is identified as a process abnormality and waste point.

[0015] As a further aspect of the present invention: the waste point processing flow is as follows: Abnormal equipment shutdown: Issue an alarm and notify maintenance personnel, and record the handling time, measures taken, and equipment status; Material accumulation: Classified processing based on R value: When R < 0.1, adjust the production schedule; When 0.1 ≤ R < 0.3, then part of the material is transferred; When R≥0.3, an emergency warning is issued and the situation is handled with priority.

[0016] As a further aspect of the present invention, the feedback and optimization mechanism includes: Add waste-related data (data at the time of occurrence, data during processing, and data after processing) to the training set, retrain the AI ​​recognition model to optimize parameters; at the same time, adjust production management strategies based on feedback data, such as optimizing material supply plans.

[0017] (III) Beneficial Effects This invention provides an integrated management platform for intelligent factories in the pipeline equipment industry. Compared with existing technologies, it has the following advantages: First, it significantly improves production efficiency and reduces resource waste. The platform utilizes a data acquisition module to comprehensively and periodically collect equipment operating parameters, material data, and video image data. Combined with outlier removal and image noise reduction by the data preprocessing module, it lays a solid foundation for the accurate identification of production waste points. The production waste point identification module, through clear judgment logic (such as the correlation between equipment downtime waste point duration and changeover signal, and the risk coefficient R assessment of material accumulation waste points) and the assistance of AI recognition models, can promptly and accurately identify various waste points such as equipment downtime, material accumulation, and process anomalies. The waste point handling and feedback module triggers corresponding processing procedures for different waste points. For example, it promptly alerts maintenance personnel when equipment stops abnormally, and handles material accumulation according to the R value, effectively preventing the expansion of waste and reducing resource losses caused by equipment idleness and material backlog, thereby improving overall production efficiency.

[0018] Secondly, the platform enhances the accuracy and timeliness of management. Data collection and processing adhere to strict periodicity and standards. Outlier removal methods employ clear criteria for different parameters (such as welding temperature T and material stacking height H), ensuring data reliability. The identification logic for equipment downtime waste points and material accumulation waste points is clear and quantifiable, enabling managers to accurately grasp the specific circumstances of these waste points. Simultaneously, the AI ​​recognition model, trained on tagged production data, can quickly output recognition results. Furthermore, its model adaptation unit automatically adjusts parameter weights based on pipe specifications, adapting to different combinations of process parameters, further improving recognition accuracy. When waste points occur, the platform rapidly triggers processing flows and provides feedback, allowing managers to take timely countermeasures, achieving precise and real-time production management.

[0019] Secondly, it promotes the intelligent and automated upgrading of the production process. The construction and application of AI recognition models are the core manifestation of the platform's intelligence. These models input processed multi-source data, extract features through hidden layers, and output recognition results, employing a cross-entropy loss function to ensure effective model training. A feedback and optimization mechanism adds waste-related data to the training set to retrain the model, continuously optimizing it and improving its recognition capabilities, forming an intelligent closed loop of "recognition-processing-feedback-optimization." This closed-loop mechanism reduces reliance on human experience, achieving automated upgrades in the identification and handling of production waste points, and making factory management more intelligent.

[0020] Furthermore, the platform's adaptability to different production scenarios has been enhanced. The model adaptation unit of the AI ​​recognition model can automatically adjust the weights of internal parameters according to pipe specifications (D, L) and learn patterns under different combinations of process parameters (T, P), enabling the platform to adapt to diverse pipe production needs. Regardless of different pipe product specifications or different process parameter settings, the platform maintains high recognition and management efficiency, improving its versatility and flexibility.

[0021] Finally, optimizing production management strategies promotes continuous improvement. Feedback and optimization mechanisms not only optimize AI model parameters but also adjust production management strategies based on feedback data, such as optimizing material supply plans. Through continuous analysis and summarization of waste point data, enterprises can continuously identify weaknesses in the production process, thereby improving production processes, refining management systems, achieving continuous optimization of production management, and enhancing the overall competitiveness of the enterprise. Attached Figure Description

[0022] Figure 1 This is a system block diagram of the intelligent factory integrated management platform for the pipeline equipment industry of the present invention. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] Please see Figure 1 As shown, the embodiments of the present invention provide the following technical solutions: As an embodiment of the present invention: This invention is an integrated management platform for intelligent factories in the pipeline equipment industry. This integrated management platform is used for the identification, handling, and optimization of production waste points, including: Data acquisition module: Multiple data acquisition terminals are set up for each stage of pipeline equipment production; Sensors are installed on the equipment in the production workshop to collect equipment operating parameters, including equipment operating status, running time, welding temperature T, and pressure P; operating status refers to power on, power off, and standby. In the material storage area, material sensors are deployed to collect the material's stacking height H and stacking volume V; Meanwhile, video image data is collected through cameras in the workshop for subsequent AI recognition; The data collected by the data acquisition terminal is continuously uploaded to the database of the management platform according to a fixed acquisition cycle t.

[0025] Production waste point identification module: Identifies equipment downtime waste points and material accumulation waste points, and uses a pre-trained AI recognition model for auxiliary identification; The logic for identifying and judging equipment downtime waste points is as follows: Let the downtime of the equipment be t. stop The standard time required for the model changeover adjustment is t. change The standard time for changeover adjustments is determined through statistical analysis of historical data on different changeover operations in pipeline equipment production, such as changing production molds for pipes of different diameters. A corresponding time interval (t) is specified for each type of changeover operation. change value; At the same time, it collects trigger signals for equipment changeover operations, such as the sending signal of mold changeover instructions; Let the type change operation trigger signal be S, where S=1 indicates that a type change operation has been triggered, and S=0 indicates that no type change operation has been triggered; When the equipment stops, that is, when the equipment operating status changes from power-on to power-off, the first step is to determine the changeover operation trigger signal S; If S=1, then further compare the actual downtime t. stop The standard duration t corresponding to this type change operation change .

[0026] If t stop ≤t change The shutdown was deemed a reasonable adjustment and replacement, and was not considered a waste of time. If t stop >t change The downtime exceeding the allowed limit is considered an abnormal downtime waste point, and the excess duration is t. waste =t stop -t change .

[0027] If S=0, then the operating parameters of the equipment are considered, such as whether the equipment was stable before shutdown. This is determined by the fluctuation of the equipment operating parameters. Let the fluctuation value of the equipment operating parameters be ΔX. If ΔX is within the normal fluctuation range ΔXnormal, then the operating state is considered stable. If a shutdown occurs while the equipment is operating stably and there is no changeover operation trigger signal, it is considered an abnormal shutdown waste point, and the shutdown duration is t. stop All of the corresponding shutdowns were wasteful shutdowns.

[0028] For example: Assuming that when producing a pipe of a certain diameter, the standard changeover operation time t is [not specified] for changing the production mold for that pipe. change =300s; At a certain moment, the equipment receives a changeover trigger signal S=1, and then the equipment stops. The actual stoppage time is t.stop =350s; Based on the above method, because t stop =350s>>t change =300s, so the excess t waste The stoppage corresponding to 350−300=50s is an abnormal stoppage waste point; In another scenario, the equipment does not receive the changeover trigger signal S=0, and the equipment operating parameter fluctuation ΔX is within the normal range ΔX. normal A shutdown occurred within a short period of time, with a shutdown duration of t. stop If the downtime is 200 seconds, then this 200-second downtime is considered an abnormal downtime waste point.

[0029] The logic for identifying and judging material accumulation waste points is as follows: To identify material accumulation waste points, the material accumulation height H, accumulation volume V, and production rhythm are considered comprehensively. The production rhythm is represented by the production rate v, with the unit being pieces / second, which is the number of pipeline equipment produced per unit time.

[0030] First, based on the material demand patterns in pipeline equipment production, a safety threshold model for material accumulation is established. Let H be the safe height threshold for material stacking. safe The safe height threshold is determined based on factors such as the space capacity of the workshop material storage area, material characteristics, and the rate of material consumption in the production process. The safe volume threshold is V. safe Safe volume threshold and H safe Correspondingly, this is determined based on the bottom area of ​​the storage region.

[0031] At the same time, the impact of production rhythm on material accumulation should be considered, since the production rate v determines the rate of material consumption. Let the current material accumulation height be H. current The accumulated volume is V current The production rate is v current .

[0032] pass: ; Calculate the risk factor R for material accumulation; In the formula, when the material accumulation height H current Exceeding the safe height threshold H safe When the risk is measured, the ratio of the excess height is multiplied by the reciprocal of the production rate. The lower the production rate, that is, the slower the material consumption, the greater the risk coefficient. When the height does not exceed the limit but the volume exceeds the safe volume threshold V safe Similarly, the volume exceeding the proportion is multiplied by the reciprocal of the production rate to measure this. If none of them exceed the limit, then the risk factor is 0; If R exceeds the pre-set risk threshold R threshold If so, it is determined to be a point of material accumulation and waste, and needs to be dealt with in a timely manner.

[0033] Risk threshold R threshold Based on historical experience, the critical risk situation before material accumulation leads to actual production loss is determined; For example: Suppose a pipeline equipment manufacturing workshop has a safe material stacking height threshold H. safe =1.5m, safe volume threshold V safe =3m 3 Current production rate vs current =0.01 pieces / s; The material accumulation height H was collected at a certain moment. current =1.8m, accumulated volume V current =2.8m 3 ; Because H current =1.8m>H safe =1.5m; Subsequently, according to the formula Calculate the risk coefficient R; If the risk threshold Rthreshold=0.2 is set, then obviously R=20>Rthreshold=0.2, which is determined to be a point of material accumulation and waste. Measures need to be taken, such as accelerating material consumption, adjusting the production rhythm, or carrying out material handling.

[0034] The AI ​​recognition model is constructed as follows: Step H1, Model training data preparation: A large amount of production data from the pipeline equipment industry was collected as a training set, including production data of pipeline products of different specifications, covering multiple diameters D and different lengths L; as well as production process parameters corresponding to welding temperature T and pressure P, equipment operation status data corresponding to start-up, shutdown, and changeover status and changeover time, and material accumulation data corresponding to height and volume.

[0035] These data are then labeled, identifying abnormal downtime waste points, material accumulation waste points, and normal production status, forming a labeled dataset used to train a specialized AI recognition model.

[0036] Step H2, Model Structure Design: Build an AI recognition model structure suitable for the pipeline equipment industry, taking full account of the industry's specific characteristics; The model input layer receives preprocessed multi-source data, including equipment operating parameters, material data, and video image feature data; The hidden layer is configured with multiple neurons, and features related to the identification of production waste points are extracted by learning from data of the pipeline equipment industry. The output layer outputs the identification results of production waste points, including whether it is an abnormal downtime waste point or a material accumulation waste point; During model training, a suitable loss function L is used to measure the difference between the model's prediction results and the actual labeled results; For classification tasks, i.e., identifying whether a point is wasteful, the cross-entropy loss function is used, and the formula is: ; Where N is the number of samples in the training dataset, y i Let y be the actual label of the i-th sample, where y i =1 indicates that points are wasted, y i =0 indicates that the point is not wasted, y0 i This represents the probability that the model predicts the i-th sample, which is the probability that the prediction is a waste point.

[0037] By continuously adjusting the model's parameters to minimize the loss function L, the model's ability to identify production waste points in the pipeline equipment industry is optimized.

[0038] Step H3, Model Fit Optimization: In view of the diverse product specifications and complex production process parameters of pipeline equipment, model adaptation units for product specifications and process parameters are added to the model.

[0039] For pipe products with different diameters D and lengths L, a corresponding weight adjustment mechanism is set in the AI ​​recognition model; When production data for pipes of different specifications are input, the AI ​​recognition model automatically adjusts its internal parameters based on the product specification information to better adapt to the identification of production waste points for that pipe specification. For the production process parameters corresponding to welding temperature T and pressure P, the AI ​​recognition model learns the normal production mode and waste point mode under different parameter combinations, thereby improving the accuracy of identifying waste points caused by abnormal process parameters.

[0040] Taking welding temperature as an example, when producing a pipe of a certain specification, the normal range of welding temperature is T. min To T max The AI ​​recognition model learned during training that temperature fluctuations within this range are normal production conditions. However, when the welding temperature exceeds this range and its duration exceeds a preset duration threshold t, T If the temperature is abnormal, it is considered an abnormal situation that may lead to production waste, such as welding quality problems caused by abnormal temperature, which may result in material waste, equipment damage and other waste points.

[0041] Waste Point Handling and Feedback Module: This module is used to trigger waste point handling processes for identified waste points and feed back the waste point handling results to the AI ​​recognition model and database to optimize subsequent recognition and management.

[0042] The waste point handling process is as follows: Once the above identification methods determine that there are production waste points, the management platform will automatically trigger the processing flow. For any abnormal equipment downtime and wasted resources, an alarm signal should be issued first to notify on-site maintenance personnel. Based on the detailed downtime information provided by the platform, including downtime duration, equipment operating parameters at the time of downtime, and whether it was due to a change in equipment type, maintenance personnel can quickly determine the cause of the anomaly and carry out repairs. Meanwhile, the platform records the handling of this abnormal shutdown, including the handling time, handling measures, and the operating status of the equipment after the handling, for subsequent analysis and optimization.

[0043] For material accumulation and waste points, the platform automatically generates handling suggestions based on the identified risk level (determined by the risk coefficient R).

[0044] If the risk coefficient R is low, such as R < 0.1, it is recommended to adjust the production rhythm and appropriately accelerate the material consumption rate, for example, by increasing the output target of the current production shift (within the range of equipment and personnel capacity). If the risk coefficient R is moderate, such as 0.1≤R<0.3, in addition to adjusting the production rhythm, it is recommended to arrange material handling personnel to transfer some materials to reduce the stacking height and volume; If the risk factor R is high, such as R≥0.3, an emergency warning should be issued immediately, and a dedicated person should be assigned to handle the situation quickly. This could involve suspending some non-critical production processes and prioritizing the handling of material accumulation issues to avoid greater losses.

[0045] The feedback and model optimization methods are as follows: After addressing production waste points, the results are fed back to the AI ​​recognition model and the management platform's database.

[0046] On the one hand, new production data is added to the training dataset, including data on when waste occurs, data on the processing process, and data after processing. The AI ​​recognition model is then retrained, and the model parameters are continuously optimized to improve the model's accuracy and adaptability in recognizing production waste.

[0047] For example, if when dealing with a certain abnormal downtime waste point, it is found that the previous model's judgment on the replacement operation was biased, the model can be retrained by adding the accurate data after this processing, so that the model can more accurately distinguish between reasonable downtime and abnormal downtime when encountering similar replacement downtime situations in the future. On the other hand, the management platform analyzes potential problems in the production process based on feedback data and further optimizes production management strategies.

[0048] For example, by analyzing the feedback from multiple instances of material accumulation and waste, it can be discovered that the material supply speed in a certain production process is not matching the production rhythm. The material supply plan for that process can be adjusted to reduce the generation of material accumulation and waste from the source.

[0049] The integrated intelligent factory management platform for the pipeline equipment industry provided in Example 1 comprehensively acquires equipment operating parameters, material accumulation data, and video image data through a multi-dimensional data acquisition module. Combined with clearly defined identification logic for equipment downtime and material accumulation waste points, it achieves accurate identification of production waste points. Simultaneously, it utilizes a pre-trained AI recognition model to assist in identification, and forms a closed-loop management system through a waste point processing and feedback module. This system not only promptly addresses identified waste points but also continuously optimizes the AI ​​model and production management strategies through data feedback, effectively reducing downtime and material accumulation waste during production, and improving the intelligent management level and production efficiency of pipeline equipment manufacturing.

[0050] As a second embodiment of the present invention: In specific implementation, compared with Embodiment 1, the technical solution of this embodiment differs from that of Embodiment 1 only in that this embodiment also includes a data preprocessing module: used to preprocess the collected data to improve data quality; For equipment operating parameters and material sensor data, outliers are first eliminated. Based on historical experience and process requirements in pipeline equipment manufacturing, the normal range of welding temperature T is extracted [T]. min ,T max If the collected T value exceeds this normal range, it is judged as an outlier and removed. For the material accumulation height H, if the difference between the collected H value and the H value collected in the previous collection cycle exceeds the preset normal fluctuation threshold ΔH... max If so, it is considered an outlier and removed; Example 2 adds a data preprocessing module to Example 1 to remove outliers from equipment operating parameters and material sensor data. This improvement effectively filters out abnormal data caused by data acquisition errors or sudden interference, significantly improving the quality of input data and preventing abnormal data from interfering with subsequent waste point identification logic and AI models, making waste point identification results more accurate and reliable. By ensuring the validity of the data, the stability and scientific nature of the entire management platform are further enhanced, laying a more solid data foundation for the efficient identification and handling of production waste points.

[0051] As an embodiment of the present invention: In specific implementation, compared with Embodiment 1 and Embodiment 2, the technical solution of this embodiment is to combine the solutions of Embodiment 1 and Embodiment 2. The difference between the technical solution of this embodiment and Embodiment 1 and Embodiment 2 is only in this embodiment; the data preprocessing module performs image noise reduction processing on the video image data to improve the image clarity and usability, which is convenient for subsequent AI recognition.

[0052] To address the salt-and-pepper noise in pipeline equipment manufacturing scenarios caused by equipment vibration and dust, such as weld spatter reflections and discrete white / black dots formed by material particle interference, an adaptive median filtering algorithm is employed. The image noise reduction steps are as follows: A set of images from the video image data is labeled TX, and the pixel coordinates are labeled (x, y). The sliding window size w is initialized. In this embodiment, w = 3×3, which can be dynamically adjusted according to the noise density. The more severe the noise, the larger the upper limit of the window. The set of pixels within the window is S. xy .

[0053] Iterate through each pixel (x, y) of the image using:

[0054] Calculate the median m of the pixels within the window. xy The median value is the median value of the pixels within the window after sorting by grayscale. Simultaneously extract the minimum value (min) of the pixels within the window. xy Maximum value (max) xy ; If min xy <m xy <max xy : Where: if the current pixel TX(x,y) equals min xy or max xy If the value is not found, it is considered noise, and TX(x,y) is replaced with m. xy ; If TX(x,y) is within min xy and max xy Pixels between these ranges are considered normal and their original values ​​are retained. If min xy =m xy =max xy If the window size w is increased, in this embodiment, w = 3×3 is expanded to w = 5×5. The above steps are repeated until there is a difference in the grayscale of the pixels within the window, or the maximum window limit is reached. In this embodiment, the maximum window limit is w = 9×9.

[0055] Example 3 integrates the solutions from Examples 1 and 2, and adds noise reduction processing for video image data in the data preprocessing module. It employs an adaptive median filtering algorithm to specifically address the salt-and-pepper noise problem caused by equipment vibration and dust in pipeline production scenarios. This optimization improves the clarity and usability of video images, enabling the AI ​​recognition model to more accurately extract production scene features from images and reducing noise interference in waste point identification. It performs particularly well in visual recognition of material accumulation states and image analysis of equipment operating status, further enhancing the accuracy of production waste point identification and the platform's adaptability to complex production environments.

[0056] As an embodiment of the present invention: In specific implementation, compared with Embodiments 1, 2, and 3, the only difference between this embodiment and Embodiments 1, 2, and 3 is that in this embodiment, Gaussian low-pass filtering is used to address high-frequency noise in images caused by uneven lighting in the workshop and welding arc light, such as background brightness fluctuations and reflective stripes on the pipe surface. The image noise reduction processing steps are as follows: First, perform a two-dimensional Fourier transform on the original image I(x,y) to obtain the frequency domain image F(u,v), as shown in the formula: ; Where M and M are the image dimensions, i.e., the width and height in pixels, u and v are the frequency domain coordinates, and j is the imaginary unit; After the transformation, the center of the frequency domain image represents the low-frequency components, which correspond to the overall image outline, while the edges represent the high-frequency components, which correspond to noise and details.

[0057] Then, construct a Gaussian low-pass filter H(u,v), with the following formula: ; In the formula, σ is the distance from the pixel to the center in the frequency domain, used to measure the frequency level. The farther the distance, the higher the frequency. σ is the standard deviation of the Gaussian function, used to control the filter cutoff frequency. The larger σ is, the more high-frequency components are retained.

[0058] Next, the frequency domain image F(u,v) is multiplied by the filter H(u,v) to obtain the filtered frequency domain image. ; Next, perform an inverse Fourier transform on G(u,v) to reconstruct the spatial image g(x,y), as shown in the formula: .

[0059] Example 4 addresses the issue of high-frequency noise in images caused by uneven lighting in the workshop and welding arc light by employing Gaussian low-pass filtering. This method effectively suppresses high-frequency noise in the image through Fourier transform and filtering operations, optimizing the overall image quality and reducing the impact of interference such as background brightness fluctuations and reflective stripes on pipe surfaces on the AI ​​recognition model. This enables the model to accurately identify equipment operating status and material accumulation under complex lighting and welding arc light environments, significantly improving the accuracy of waste point identification in special production conditions and enhancing the platform's practicality and robustness.

[0060] As an embodiment of the present invention: In specific implementation, compared with Embodiment 1, Embodiment 2, Embodiment 3 and Embodiment 4, the technical solution of this embodiment is to combine the solutions of Embodiment 1, Embodiment 2, Embodiment 3 and Embodiment 4.

[0061] Example 5 integrates all the technical solutions from Examples 1 to 4, including comprehensive data acquisition, precise waste point identification logic, and closed-loop processing feedback mechanisms, as well as a complete data preprocessing module (including outlier removal for equipment and material data, adaptive median filtering for salt-and-pepper noise, and Gaussian low-pass filtering for high-frequency noise). This comprehensive solution achieves all-round improvements in data quality, identification accuracy, and processing efficiency. It can adapt to diverse equipment states, complex material accumulation scenarios, and changing image interference environments in pipeline equipment production. It can identify and process production waste points with high accuracy in all scenarios, minimize production waste, comprehensively optimize production management strategies, and provide an integrated and efficient end-to-end management solution for smart factories.

[0062] It should be stated that all user data collected in this application was collected with the user's consent and authorization, and the use of user data is legal and compliant, and the use and processing of user data comply with the relevant laws, regulations and standards of the relevant regions.

[0063] Furthermore, any content not described in detail in this specification is existing technology known to those skilled in the art.

[0064] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0065] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

[0066] The various embodiments of this application have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or improvement of the technology in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. An integrated management platform for intelligent factories in the pipeline equipment industry, characterized by: include: Data acquisition module: Used to acquire equipment operating parameters, material data and video image data during the production process through multi-source data acquisition terminals, and upload them to the database according to a pre-set fixed cycle; Data preprocessing module: used to remove outliers and reduce image noise in the collected data; Production waste point identification module: used to identify equipment downtime waste points and material accumulation waste points, and at the same time, it uses a pre-trained AI recognition model for auxiliary identification; Waste Point Handling and Feedback Module: This module is used to trigger waste point handling processes for identified waste points and feed back the waste point handling results to the pre-trained AI recognition model and the pre-established database.

2. The integrated management platform for intelligent factories in the pipeline equipment industry according to claim 1, characterized in that: The multi-source data acquisition terminal includes: Sensors installed on production equipment are used to collect equipment operating parameters such as equipment operating status, operating time, welding temperature T, and pressure P; material sensors arranged in the material storage area are used to collect material data corresponding to material stacking height H and stacking volume V; cameras installed in the workshop are used to collect video image data. The data acquisition period of the source data acquisition terminal is t, and t is a preset value.

3. The integrated management platform for intelligent factories in the pipeline equipment industry according to claim 2, characterized in that: The specific methods for outlier removal in the data preprocessing module are as follows: For welding temperature T, extract the pre-set normal range [T]. min ,T max ], and T values ​​that exceed the normal range are identified as outliers and removed; For the material accumulation height H, if the difference between the currently collected H value and the H value collected in the previous data collection cycle exceeds the preset normal fluctuation threshold ΔH... max If the value is not found, it is considered an outlier and removed. For video image data, perform image noise reduction processing.

4. The integrated management platform for intelligent factories in the pipeline equipment industry according to claim 3, characterized in that: In image noise reduction processing: For salt-and-pepper noise in images caused by equipment vibration and dust in pipeline equipment production scenarios, an adaptive median filtering algorithm is used for image noise reduction. To address the high-frequency noise in images caused by uneven lighting in the workshop and welding arc light, Gaussian low-pass filtering is used for image noise reduction.

5. The integrated management platform for intelligent factories in the pipeline equipment industry according to claim 2, characterized in that: The logic for identifying and judging equipment downtime waste points is as follows: Extract and mark the equipment downtime as t stop And extract the pre-set standard duration for model change adjustment, and mark it as t. change The changeover operation trigger signal is marked as S; When S=1, it indicates that a type-change operation has been triggered; when S=0, it indicates that no type-change operation has been triggered. When the equipment stops: If S=1, and t stop >t change If it exceeds the duration t waste =t stop -t change This is determined to be an abnormal downtime waste point; if t stop ≤ tchange The shutdown was deemed a reasonable changeover shutdown. If S=0, and the fluctuation value ΔX of the equipment operating parameters is within the normal range ΔX normal Inside, determine t stop The corresponding downtime is an abnormal downtime waste point; Among them, the fluctuation value ΔX of the equipment operating parameters is obtained by extracting the equipment operating parameters within a specified time window and calculating the standard deviation.

6. The integrated management platform for intelligent factories in the pipeline equipment industry according to claim 5, characterized in that: The identification of material accumulation and waste points is assessed using a risk coefficient R, calculated using the following formula: ； Among them, H current H represents the current material stacking height. safe For safety height threshold; V current V represents the current material accumulation volume. safe For safe volume threshold; vs current For production rate; when R exceeds the risk threshold R threshold When this occurs, it is determined to be a point of material accumulation and waste.

7. The integrated management platform for intelligent factories in the pipeline equipment industry according to claim 6, characterized in that: The construction of AI recognition models includes: Training data preparation: Collect labeled production data, including: pipe specifications with diameter D and length L, process parameters with welding temperature T and pressure P, equipment status, material data, and marked waste points and normal status; Model structure: The input layer receives equipment operating parameters, material data and video image data after pre-passing outlier removal and image noise reduction; the hidden layer extracts features; and the output layer outputs the waste point identification results. Loss function: Cross-entropy loss function is used; The formula is: ; Where N is the number of samples, y i For the actual label, where y i =1 indicates wasted points, y i =0 indicates a non-wasteful point, y0 i Predict probabilities for the model.

8. The integrated management platform for intelligent factories in the pipeline equipment industry according to claim 7, characterized in that: The AI ​​recognition model includes a model adaptation unit, which automatically adjusts the weights of internal parameters according to the pipe specifications and learns the normal and waste patterns under different combinations of process parameters. When the welding temperature exceeds the normal range and its duration exceeds the preset duration threshold t T If so, it is identified as a point of waste in the process.

9. The integrated management platform for intelligent factories in the pipeline equipment industry according to claim 8, characterized in that: The waste point handling process is as follows: Abnormal equipment shutdown: Issue an alarm and notify maintenance personnel, and record the handling time, measures taken, and equipment status; Material accumulation: Classified processing based on R value: When R < 0.1, adjust the production schedule; When 0.1 ≤ R < 0.3, then part of the material is transferred; When R≥0.3, an emergency warning is issued and the situation is handled with priority.

10. The integrated management platform for intelligent factories in the pipeline equipment industry according to claim 9, characterized in that: Feedback and optimization mechanisms include: The relevant data when waste occurs, the relevant data during the processing, and the relevant data after processing are added to the training set to retrain the AI ​​recognition model, and at the same time, the database of the integrated management platform is used.

Images