Injection product overflow trimming quality detection method based on multi-source data fusion

Abstract

The application discloses a method for detecting the quality of injection molding product overflow trimming based on multi-source data fusion, acquires the shape data of the injection molding product overflow through a data acquisition terminal, obtains product quality data through a feature extraction terminal, calculates a state score based on the product quality data through a state evaluation terminal, outputs a trimming scheme based on the product quality data and a trimming strategy knowledge base, acquires power time series data in real time during the trimming process, reacquires the product quality data and calculates the state score after the trimming is completed, calculates a trimming ability score based on the power time series data and the state score through an ability score terminal, iteratively executes until the state score reaches the standard, finally calculates a comprehensive energy efficiency index through a comprehensive evaluation terminal, and feeds back the trimming process parameters to the knowledge base. The application fuses multi-source data, realizes the intelligentization and dynamic optimization of the overflow trimming process, improves the trimming quality and energy efficiency, and is suitable for the quality control scene of automatic trimming of injection molding product overflow.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of industrial product quality control, and specifically discloses a method for detecting the quality of overflow trimming of injection molded products based on multi-source data fusion. BACKGROUND

[0002] As a widely used high-efficiency manufacturing process in modern industry, injection molding has a significant advantage in producing complex structure plastic products; however, overflow defects may occur in the products due to reasons such as mold wear, insufficient clamping force or process parameter fluctuation during the injection molding process, which not only affects the appearance and assembly accuracy of the products, but also may reduce the mechanical properties and service life of the products, and therefore subsequent trimming treatment must be performed.

[0003] Traditional overflow trimming relies on manual visual judgment and manual operation, and has the disadvantages of low efficiency, poor consistency and high labor intensity, in recent years, with the development of machine vision and automation technology, some enterprises have begun to use automatic detection and trimming systems based on image processing, however, such methods usually only rely on single visual data, and have limited adaptability to light changes, background interference and overflow shape diversity, and it is difficult to comprehensively and stably evaluate the trimming quality, in addition, the comprehensive consideration of key production indicators such as energy consumption and time efficiency in the trimming process is lacking, which leads to the lack of optimization basis for trimming strategies and low overall energy efficiency.

[0004] Therefore, it is necessary to invent a method for detecting the quality of overflow trimming of injection molded products based on multi-source data fusion to solve the above problems. SUMMARY

[0005] In order to overcome the defects of the prior art, the application provides a method for detecting the quality of overflow trimming of injection molded products based on multi-source data fusion, which acquires shape data of the overflow of the injection molded products through a data acquisition terminal, obtains product quality data through a feature extraction terminal, calculates a state score based on the product quality data through a state evaluation terminal, outputs a trimming scheme based on the product quality data and a trimming strategy knowledge base, acquires power time series data in real time during the trimming process, reacquires product quality data and calculates a state score after the trimming is completed, calculates a trimming ability score based on the power time series data and the state score through an ability score terminal, iteratively executes until the state score meets the standard, finally calculates a comprehensive energy efficiency index through a comprehensive evaluation terminal, and feeds back trimming process parameters to the knowledge base, thereby effectively solving the problems mentioned in the background art.

[0006] To achieve the above purpose, the application provides the following technical scheme: a method for detecting the quality of overflow trimming of injection molded products based on multi-source data fusion, comprising a data acquisition terminal, a feature extraction terminal, a state evaluation terminal, an ability score terminal, a comprehensive evaluation terminal and a trimming strategy knowledge base, specifically comprising the following steps:

[0007] S1, a data acquisition terminal acquires form data of an overflow edge of an injection molded product;

[0008] S2, a feature extraction terminal extracts features from the form data and outputs product quality data;

[0009] S3, a state evaluation terminal calculates a state score based on the product quality data;

[0010] S4, the product quality data is input into a trimming strategy knowledge base, and a trimming scheme is output;

[0011] S5, the trimming scheme is executed, and the data acquisition terminal synchronously acquires power time series data during the trimming process. After the trimming is completed, steps S1-S3 are executed again to obtain product quality data and a state score after trimming;

[0012] S6, a capability scoring terminal calculates a trimming capability score based on the power time series data and the state score;

[0013] S7, steps S4-S6 are iteratively executed until the state score reaches or exceeds a preset target score, and an injection molded product is output;

[0014] S8, a comprehensive evaluation terminal calculates a comprehensive energy efficiency index based on the state score and the power time series data, and stores trimming process parameters in the strategy knowledge base;

[0015] The state score is calculated in the following manner: S=f(V,D,M), where V is the overflow edge volume, D is the average thickness, M is the distribution density, and f is a preset weighting function;

[0016] The determination process of the trimming scheme further includes a dynamic feedback mechanism: comparing the trimming capability score with a preset threshold value. If the score is lower than the threshold value, a conservative trimming scheme is matched for the next stage. If the score is higher than or equal to the threshold value, an aggressive trimming scheme is matched;

[0017] The comprehensive energy efficiency index is calculated in the following manner: η=S final / (E total +λ˙T total ), where S final is the final state score, E total is the total energy consumption, T total is the total time, and λ is the time weight coefficient.

[0018] Preferably, the form data is point cloud data of the overflow edge region of the injection molded product; the product quality data includes the volume, average thickness, and distribution density of the overflow edge; the power time series data is real-time power during the trimming process; and the trimming process parameters include the number of trimmings, the trimming scheme for each trimming, the comprehensive energy efficiency index, and the state score after the last trimming.

[0019] Preferably, the determination process of the trimming scheme is that the product quality data is taken as an index to traverse the trimming strategy knowledge base, and a trimming scheme with a higher comprehensive energy efficiency index and a higher final state score in history is preferentially recommended.

[0020] Preferably, the calculation method of the trimming capability score is C= (S after -S before ) / E, wherein S after is the post-trimming state score, S before is the pre-trimming state score, and E is the trimming energy consumption.

[0021] Preferably, the trimming energy consumption is calculated based on power time sequence data, and the calculation formula is E=∫P(t)dt, wherein P(t) is the power time sequence data, and t is the trimming duration.

[0022] Preferably, the calculation formula of the total energy consumption is: , wherein n is the number of trimmings, E i is the trimming energy consumption of the i th trimming.

[0023] Technical effects and advantages of the present application:

[0024] 1. The point cloud shape data of the overflow edge of the injection molded product is obtained through the data acquisition terminal, and the product quality data containing the overflow edge volume, average thickness and distribution density is obtained through the feature extraction terminal, thereby providing comprehensive and accurate basic information for state evaluation; the state evaluation terminal calculates the state score by using a preset weighting function or a neural network mapping function S=f(V,D,M), which can comprehensively consider the key features of the overflow edge, accurately reflect the quality state of the product overflow edge before and after trimming, avoid the one-sidedness of single data evaluation, and make the quality judgment more scientific;

[0025] 2. When determining the trimming scheme, the product quality data is taken as an index to traverse the trimming strategy knowledge base, and a scheme with a high comprehensive energy efficiency index and a high final state score in history is preferentially recommended, thereby fully utilizing the historical effective data, improving the rationality of the initial trimming scheme, introducing a dynamic feedback mechanism, comparing the trimming capability score with a preset threshold, matching a conservative trimming scheme when the trimming capability score is lower than the threshold, and matching an aggressive scheme when the trimming capability score is higher than or equal to the threshold. This dynamic adjustment method can flexibly adjust the strategy according to the actual trimming capability, avoid the problems of insufficient trimming or excessive trimming caused by a fixed scheme, and improve the adaptability and effectiveness of the trimming process;

[0026] 3. The capability score terminal calculates the trimming capability score by using the formula C= (S after -S before ) / E, wherein the trimming energy consumption E is calculated based on power time sequence data E=∫P(t)dt, which can intuitively reflect the energy utilization efficiency of each trimming, and provide a basis for evaluating the economy of the trimming process; the comprehensive evaluation terminal adopts the formula η=Sfinal / (E total +λ˙T total The comprehensive energy efficiency index is calculated, taking into account the final state score, total energy consumption, and total time. The time weighting coefficient λ balances energy consumption and time factors, guiding the trimming process to ensure quality while taking into account energy consumption and time costs, thereby optimizing overall energy efficiency and productivity. The trimming process is iteratively executed until the state score meets the standard, ensuring that the final product quality meets the preset requirements, reducing rework caused by substandard quality, indirectly reducing additional energy consumption and time consumption, and further improving overall production efficiency.

[0027] 4. After each trimming is completed, the trimming process parameters are stored in the trimming strategy knowledge base. As data accumulates, the knowledge base can be continuously enriched, providing more high-quality solutions for trimming overflow of similar products in the future, forming a virtuous cycle of "data accumulation - solution optimization - effect improvement". Attached Figure Description

[0028] The present invention will be further described with reference to the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present invention. For those skilled in the art, other drawings can be obtained based on the following drawings without creative effort.

[0029] Figure 1 This is a schematic diagram of the overall structure of the present invention.

[0030] Figure 2 This is a flowchart illustrating the overall steps of the present invention. Detailed Implementation

[0031] 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.

[0032] This invention provides, for example Figure 1 The method for quality inspection of overflow trimming of injection molded products based on multi-source data fusion includes a data acquisition terminal, a feature extraction terminal, a status assessment terminal, a capability scoring terminal, a comprehensive assessment terminal, and a trimming strategy knowledge base.

[0033] like Figure 2 As shown, the specific steps include the following:

[0034] S1. The data acquisition terminal collects the shape data of the overflow of the injection molded product;

[0035] Further, in the above technical solution, the shape data is point cloud data of the overflow edge region of the injection molded product.

[0036] Further, the data acquisition terminal is composed of a three-dimensional scanning module, a data transmission module and an auxiliary positioning module, and in the preferred technical solution of the application, the specific hardware selection and function adaptation are as follows:

[0037] The three-dimensional scanning module: a high-precision structured light three-dimensional scanner is selected, such as model EinScan-Pro 2X, the scanning accuracy of the equipment reaches 0.05mm, the point cloud density supports 50-200 points / mm² adjustable, the scanning range covers 100mm*100mm-1000mm*1000mm, which can adapt to the scanning of the overflow edge region of injection molded products of different sizes, and ensure that the obtained overflow edge shape data can accurately reflect the three-dimensional geometric characteristics of the overflow edge.

[0038] The data transmission module: equipped with an industrial Ethernet interface and a wireless WiFi module, supporting IEEE 802.11ac protocol, redundant design of two transmission modes, of which the Ethernet transmission rate is greater than or equal to 100Mbps, used for high-speed uploading of point cloud data after scanning is completed; the WiFi module is used for real-time transmission of scanning state signals to ensure real-time communication with the feature extraction terminal.

[0039] The auxiliary positioning module: equipped with 2 high-definition industrial cameras and a laser positioner, the camera is used to capture the reference profile of the injection molded product, and the laser positioner emits a cross positioning laser to assist the operator to aim the scanner lens at the overflow edge region, avoiding invalid data acquisition caused by scanning range deviation.

[0040] Further, in the above technical solution, the specific way of collecting the shape data of the overflow edge of the injection molded product is: using a three-dimensional scanning module to scan the overflow edge region of the injection molded product to obtain high-precision point cloud data, i.e. shape data, ensuring that the scanning covers the entire overflow edge region during scanning, including its contour, height, width and other geometric characteristics, preprocessing the collected shape data, including denoising, filtering and registration operation, to improve the data quality, storing the processed shape data in a standardized format such as PLY format, and transmitting the shape data to the feature extraction terminal

[0041] S2, the feature extraction terminal extracts features from the shape data and outputs product quality data;

[0042] Further, in the above technical solution, the product quality data includes the volume, average thickness and distribution density of the overflow edge.

[0043] Further, in the above technical solution, the feature extraction process is:

[0044] The feature extraction terminal receives the shape data from the data acquisition terminal, loads the shape data into a processing system such as the PointCloud Library (PCL), and accurately calculates product quality data from the shape data by using a three-dimensional point cloud processing algorithm. The specific method is as follows:

[0045] Overflow volume: based on the shape data, a triangular mesh model is generated using a Poisson surface reconstruction algorithm. In the preferred technical solution of the present application, the reconstruction depth is set to 8 to ensure that the mesh accurately fits the three-dimensional shape of the overflow edge, and the number of mesh surfaces is controlled to be less than or equal to 10 4 ~ 10 5 Magnitude, balancing accuracy and calculation efficiency; the triangular mesh model is imported into the integration module of SciPy, and Simpson's integration method is used to calculate the volume of the space enclosed by the mesh. During the integration process, the product body surface is used as the reference surface, and only the overflow edge volume outside the reference surface is calculated;

[0046] Average thickness: on the triangular mesh model of the overflow edge region, 100-200 sampling points are selected by uniform mesh sampling method, the sampling density is adaptively adjusted according to the overflow edge area, and the minimum sampling interval is less than or equal to 0.5mm. For each sampling point, the sampling point is numbered, and the normal vector of the point is calculated by the normal estimation algorithm of PCL. The nearest point is searched in the direction of the normal vector to the outside of the overflow edge and the product body respectively, and the distance between the two points is the overflow edge thickness at the sampling point, denoted as d i ; the average thickness is calculated by weighted average method, and the formula is: , where n is the number of sampling points, i is the sampling point number, S i is the area of the triangular mesh where the sampling point is located;

[0047] Distribution density: the region segmentation is performed based on the curvature-based region growing algorithm to accurately separate the overflow edge region and the product body region, and to ensure that the subsequent calculation is only effective for the overflow edge region. The product parting surface of the injection molded product is used as the projection surface, the segmented overflow edge region point cloud model is projected onto the projection surface by the projection module of PCL, the two-dimensional area formed after projection is automatically calculated, denoted as A, with the unit of cm², and the projection error is controlled to be less than or equal to 0.01cm² to avoid distortion of the density calculation due to projection deviation; the two-dimensional region formed after projection is divided into several uniform sub-regions with a fixed size of 2mm x 2mm, and the number of sub-regions is adaptively adjusted according to the size of the two-dimensional region formed after projection to ensure that the entire projection area is covered without omission; the total number of sub-regions determined as "overflow edge exists" is counted, denoted as N, and the distribution density is calculated by the ratio formula of "effective sub-region number / overflow edge projection area", and the specific formula is: M=N / A;

[0048] It needs to be further explained that the existence of the overflow edge is determined as follows: for each sub-region, the actual area of the overflow edge projection in the sub-region is calculated by the PCL algorithm, if the area is greater than or equal to 0.01 mm2, it is determined that the sub-region "exists overflow edge"; if the area is less than 0.01 mm2, it is determined that "no overflow edge";

[0049] S3, the state evaluation terminal calculates a state score based on the product quality data;

[0050] Further, in the above technical solution, the calculation method of the state score is: S=f(V,D,M), in the formula, V is the overflow edge volume, D is the average thickness, M is the distribution density, and f is a preset weighting function.

[0051] Further, the specific process of the state score is:

[0052] The product quality data is standardized to the [0,1] interval by min-max standardization to eliminate the interference of the order of magnitude on the score;

[0053] An analytic hierarchy process is used to construct a judgment matrix, and the weight coefficients corresponding to the overflow edge volume, the average thickness and the distribution density are determined in combination with the quality requirements of the injection molded product, such as that the appearance part pays more attention to M, and the structure part pays more attention to V and D;

[0054] The state score is calculated based on the calculation formula, and the formula is S=f(V,D,M)=100×((1-V std )×w V +(1-D std )×w D +(1-M std )×w M ), in the formula, V std , D std and M std are the normalized overflow edge volume, average thickness and distribution density respectively, w V , w D and w M are the weight coefficients corresponding to the overflow edge volume, the average thickness and the distribution density respectively;

[0055] S4, input the product quality data into the trimming strategy knowledge base, and output a trimming scheme;

[0056] Further, in the above technical solution, the determination process of the trimming scheme is: taking the product quality data as an index, traversing the trimming strategy knowledge base, and preferentially recommending a trimming scheme with a higher comprehensive energy efficiency index and a higher final state score in history.

[0057] It needs to be further explained that the specific implementation of the output trimming scheme is:

[0058] The product quality data is converted into a quality data index, which is matched with the trimming strategy knowledge base primary key, and the specific operation is as follows:

[0059] Data classification: according to the quality data interval division standard, judge the interval to which the current V, D, M belongs respectively, for example:

[0060] If the current V=3.2mm³, D=0.15mm, M=3.8%, the classification result is V∈[0,5), D∈[0,0.2), M∈[0,5);

[0061] The interval result is spliced into a string in the format of "V+interval&D+interval&M+interval" as the "quality data index" for retrieving the knowledge base. The index of the above example is "V[0,5)&D[0,0.2)&M[0,5)";

[0062] Further, in the preferred technical solution of the present application, the quality data interval division standard is as follows:

[0063] Overflow volume V (mm³): [0, 5], [5, 10], [10, 20], [20, 50], [50, 100];

[0064] Average thickness D (mm): [0, 0.2], [0.2, 0.5], [0.5, 1.0], [1.0, 2.0], [2.0, 5.0];

[0065] Distribution density M (%): [0, 5], [5, 15], [15, 30], [30, 50], [50, 100];

[0066] Precise matching retrieval: execute SQL query in the trimming strategy knowledge base to obtain all historical schemes that completely match the current index; sort the matched historical schemes according to priority: the first priority is the descending order of comprehensive energy efficiency index η; the second priority is the descending order of final state score S final ; that is, prefer to select a historical scheme with a higher comprehensive energy efficiency index η, and in the case of the same comprehensive energy efficiency index η, prefer to select a historical scheme with a higher final state score S final ;

[0067] Further, in the above technical solution, the determination process of the trimming scheme further includes a dynamic feedback mechanism: compare the trimming ability score with a preset threshold value, if it is lower than the threshold value, then match a conservative trimming scheme for the next stage; if it is higher than or equal to the threshold value, then match an aggressive trimming scheme.

[0068] It needs to be further explained that before determining the correction scheme, it is also necessary to judge whether this trimming is the first iteration, if it is the first iteration, the dynamic feedback mechanism is not triggered, and if it is not the first iteration, the dynamic feedback mechanism is triggered;

[0069] Further, if it is not the first iteration, the current trimming ability score C is compared with a preset threshold C0:

[0070] If C

[0071] If C≥C0, it means that the trimming ability meets the standard, so an aggressive scheme is selected from the preliminary screening candidate scheme to shorten the trimming time, and the characteristics of the aggressive scheme are: large trimming amount, high speed, and fast feed rate.

[0072] S5, execute the trimming scheme, the data acquisition terminal synchronously acquires power time series data during the trimming process, and after the trimming is completed, steps S1-S3 are executed again to obtain the product quality data and state score after trimming;

[0073] Further, in the trimming process, power data is collected in real time by a power sensor installed on the trimming device to form power time series data P(t), where t is a time variable. The data is used for subsequent calculation of trimming energy consumption E and evaluation of trimming ability.

[0074] S6, the ability score terminal calculates the trimming ability score based on the power time series data and the state score; the score reflects the quality improvement amplitude brought by unit energy consumption, and the higher the value, the higher the trimming efficiency;

[0075] Further, in the above technical solution, the calculation method of the trimming ability score is: C= (S after -S before ) / E, in the formula, S after is the state score after trimming, S before is the state score before trimming, and E is the trimming energy consumption of this stage.

[0076] Further, in the above technical solution, the trimming energy consumption is calculated based on the power time series data, and the calculation formula is E=∫P(t)dt, in the formula, P(t) is the power time series data, and t is the trimming duration, i.e. the time from the beginning to the end of this trimming.

[0077] S7, iteratively execute steps S4-S6 until the state score reaches or exceeds the preset target score, and output the injection molded product;

[0078] Further, in the above technical solution, the preset target score is determined according to the product manufacturing requirements, and specifically can be determined by calculating the quality score of the injection molded product that meets the minimum requirements.

[0079] It should be further explained that, in order to prevent infinite loop, the maximum iteration number is set, when the maximum iteration number is reached and the product quality score has not reached the preset target score, the injection product is marked as defective product and output.

[0080] S8、Comprehensive evaluation terminal based on state score, power timing data to calculate comprehensive energy efficiency index, and the finishing process parameters are stored in the strategy knowledge base.

[0081] Further, in the above technical solution, the finishing process parameters include finishing number, finishing scheme of each finishing, comprehensive energy efficiency index and state score after executing the last finishing;

[0082] Further, in the above technical solution, the calculation method of the comprehensive energy efficiency index is: η=S final / (E total +λ˙T total ), in the formula, S final is the final state score, E total is the total energy consumption, T total is the total time, i.e. the sum of each finishing time, λ is the time weight coefficient, which is set according to the actual process requirement, and in the preferred technical solution of the present application, λ=0.1.

[0083] Further, in the above technical solution, the calculation formula of the total energy consumption is: , in the formula, n is the finishing number, E i is the finishing energy consumption of the i-th finishing.

[0084] Finally, it should be pointed out that: the above only for the preferred embodiments of the present application, and does not limit the present application, although the present application has been described in detail with reference to the foregoing embodiments, for those skilled in the art, it still can modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement to part of the technical features, any modification, equivalent replacement, improvement, etc. within the spirit and principles of the present application, should be included in the protection scope of the present application.

Claims

1. A method for detecting the quality of flash trimming in injection molded products based on multi-source data fusion, characterized in that, It includes a data acquisition terminal, a feature extraction terminal, a status assessment terminal, a capability scoring terminal, a comprehensive assessment terminal, and a repair strategy knowledge base, and specifically includes the following steps: S1. The data acquisition terminal collects the shape data of the overflow of the injection molded product; S2. The feature extraction terminal extracts features from the morphological data and outputs product quality data. S3. The status assessment terminal calculates a status score based on product quality data. S4. Input product quality data into the repair strategy knowledge base and output repair plan; S5. Execute the adjustment plan. During the adjustment process, the data acquisition terminal synchronously collects power timing data. After the adjustment is completed, execute steps S1 to S3 again to obtain the adjusted product quality data and status score. S6. The capability scoring terminal calculates the trimming capability score based on power timing data and status score. S7. Iterate through steps S4 to S6 until the status score reaches or exceeds the preset target score, then terminate the iteration and output the injection molded product. S8. The comprehensive evaluation terminal calculates the comprehensive energy efficiency index based on the status score and power time series data, and stores the trimming process parameters into the trimming strategy knowledge base. The state score is calculated as follows: S = f(V, D, M), where V is the overflow volume, D is the average thickness, M is the distribution density, and f is a preset weighting function. The process of determining the repair plan also includes a dynamic feedback mechanism: the repair capability score is compared with a preset threshold. If it is lower than the threshold, a conservative repair plan is matched for the next stage; if it is higher than or equal to the threshold, an aggressive repair plan is matched. The comprehensive energy efficiency index is calculated as follows: η = S final / (E total +λ˙T total In the formula, S final For the final state score, E total For total energy consumption, T total Let λ be the total time and λ be the time weighting coefficient.

2. The method for detecting the quality of flash trimming in injection molded products based on multi-source data fusion as described in claim 1, characterized in that: The morphological data is the point cloud data of the overflow area of ​​the injection molded product; the product quality data includes the volume, average thickness and distribution density of the overflow; the power time series data is the real-time power of the trimming process; the trimming process parameters include the number of trimmings, the trimming scheme for each trimming, the comprehensive energy efficiency index and the status score after the last trimming.

3. The method for detecting the quality of flash trimming in injection molded products based on multi-source data fusion as described in claim 1, characterized in that: The process of determining the modification plan is as follows: using product quality data as an index, traversing the modification strategy knowledge base, and prioritizing the recommendation of modification plans with higher historical comprehensive energy efficiency index and higher final state score.

4. The method for detecting the quality of flash trimming in injection molded products based on multi-source data fusion as described in claim 1, characterized in that: The repair capability score is calculated as follows: C = (S after -S before ) / E, where S is in the formula after S is used to rate the condition after repair. before The score is the condition before repair, and E is the repair energy consumption in this stage.

5. The method for detecting the quality of flash trimming in injection molded products based on multi-source data fusion as described in claim 4, characterized in that: The energy consumption for the adjustment is calculated based on power time-series data, and the calculation formula is E=∫P(t)dt, where P(t) is the power time-series data and t is the adjustment duration.

6. The method for detecting the quality of flash trimming in injection molded products based on multi-source data fusion as described in claim 1, characterized in that: The formula for calculating the total energy consumption is: In the formula, n represents the number of trims, and E i Let be the maintenance energy consumption for the i-th maintenance.

Images