Intelligent identification and vibration control system for concrete test piece forming quality

Abstract

The present application relates to the technical fields of concrete specimen forming detection and intelligent vibration control, in particular to an intelligent identification and vibration control system for the forming quality of concrete specimens, comprising a data acquisition unit, a batch identification reference adaptive unit, a visual forming state judgment unit, a vibration response state judgment unit, a multi-modal fusion control unit and a vibration execution unit; the data acquisition unit acquires concrete specimen image data, vibration table vibration response data and current batch workability parameters; the batch identification reference adaptive unit corrects the visual identification reference according to the current batch workability parameters; the visual forming state judgment unit and the vibration response state judgment unit generate visual forming state results, vibration forming state results and corresponding confidence levels respectively; the multi-modal fusion control unit determines an under-vibration state, a qualified forming state or an over-vibration state in combination with the workability parameters, the corrected confidence levels and the continuous sampling trend.

Description

Technical Field

[0001] This invention relates to the field of concrete specimen molding inspection and intelligent vibration control technology, and in particular to an intelligent identification and vibration control system for concrete specimen molding quality. Background Technology

[0002] Concrete specimens are commonly used to test the compressive strength, flexural strength, and durability of concrete. The quality of their molding affects the accuracy of subsequent test results. During specimen preparation, the mixed concrete is typically poured into a mold and vibrated using a vibrating table or similar device to remove internal air bubbles, ensuring the concrete fills the mold and forms a dense, flat specimen structure.

[0003] Current concrete specimen molding processes largely rely on manual experience or fixed vibration parameters. Operators typically judge whether a specimen has been compacted based on visual phenomena such as surface bubbles, water bleeding, and surface smoothness. While some automated equipment can perform vibration molding according to preset vibration frequencies, amplitudes, and durations, it is difficult to dynamically adjust vibration parameters based on the actual molding state of the specimen during vibration, which can easily lead to under-vibration or over-vibration.

[0004] With the development of image recognition technology, existing technologies can identify features such as air bubbles, bleeding, corner filling, or surface flatness by collecting images of concrete specimen surfaces to help determine the quality of specimen molding. However, concrete surface images are easily affected by factors such as surface water film, strong reflection, changes in lighting, and occlusion. Relying solely on visual recognition results can easily reduce the stability of the molding status judgment.

[0005] During vibration molding, vibration response data such as the actual acceleration of the vibration table, table displacement, vibration frequency deviation, and motor current fluctuations can reflect the transmission of vibration effects to the mold and concrete specimen. Existing equipment typically collects this type of vibration response data for equipment operation monitoring or abnormal alarms, rarely using it to correct the molding state of concrete specimens, and it is difficult to establish an effective correlation with the visual molding state.

[0006] Furthermore, the workability of different batches of concrete varies. For example, parameters such as slump, spread, coarse aggregate size, water-cement ratio, and admixture type differ, leading to variations in air bubble escape, bleeding, flow filling, and surface smoothness. Existing visual recognition methods typically use fixed recognition benchmarks, making it difficult to adaptively correct for the workability parameters of the current batch of concrete, which can easily affect the consistency of judging the forming state of specimens from different batches.

[0007] Therefore, existing technologies still lack a technical solution that can correct the visual recognition benchmark based on the current batch's workability parameters during the vibration molding process of concrete specimens, and then use the correlation analysis between the visual molding state and the vibration response state for adaptive adjustment of vibration parameters. How to improve the accuracy of identifying the molding state of different batches of concrete specimens, and how to adjust the vibration frequency, amplitude, vibration time, or start / stop state in a timely manner according to the molding state, has become a technical problem that needs to be solved. Summary of the Invention

[0008] The purpose of this invention is to overcome the shortcomings of existing technologies and propose an intelligent identification and vibration control system for the molding quality of concrete specimens.

[0009] To achieve the above objectives, the present invention adopts the following technical solution: an intelligent identification and vibration control system for the molding quality of concrete specimens, including a data acquisition unit. The data acquisition unit is used to acquire the current batch working parameters before vibration molding, at the initial stage of vibration molding, or during vibration molding, and to acquire concrete specimen image data and vibration table vibration response data according to a preset sampling period during vibration molding.

[0010] The batch identification benchmark adaptive unit corrects the benchmark visual identification rules based on the deviation between the current batch workability parameters and the preset benchmark workability parameters to obtain a visual identification benchmark adapted to the current batch of concrete; the visual forming state judgment unit obtains the visual forming state result and visual confidence level based on the image data and the visual identification benchmark.

[0011] The vibration response state judgment unit extracts vibration forming response features based on the vibration response data, and obtains vibration forming state results and vibration confidence based on the comparison results of the vibration forming response features and the preset vibration response benchmark.

[0012] The multimodal fusion control unit performs correlation analysis based on the batch adjustment coefficient, visual forming status result, visual confidence level, vibration forming status result and vibration confidence level corresponding to the current batch working parameters, generates forming status evaluation result, and determines the forming status by combining the changes within the continuous sampling period.

[0013] The vibration execution unit adaptively adjusts the vibration frequency, amplitude, vibration time, or vibration start / stop state of the vibration table according to the molding state, so that the concrete specimen reaches the preset molding quality state.

[0014] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0015] This invention collects image data of concrete specimens, vibration response data of the vibration table, and workability parameters of the current batch during the vibration molding process of concrete specimens. Based on these batch workability parameters, the baseline visual recognition rules are revised to adapt the visual recognition baseline to the workability differences between different batches of concrete. Simultaneously, the molding state of the concrete specimen is determined by combining visual molding state results, visual confidence levels, vibration molding state results, and vibration confidence levels. The vibration frequency, amplitude, vibration time, or vibration start / stop state are adaptively adjusted according to the determined molding state. This reduces the influence of human experience, fixed vibration parameters, and single visual recognition on molding state judgment, improves the accuracy and stability of concrete specimen molding state identification, allows vibration parameters to adjust according to changes in the specimen molding state, reduces molding deviations caused by under-vibration or over-vibration, and improves the consistency of molding results across different batches of concrete specimens. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the system structure of the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0018] Example 1

[0019] This invention provides an intelligent identification and vibration control system for the molding quality of concrete specimens. The data acquisition unit collects vibration molding process data according to a preset sampling period T, where T represents the time interval between two adjacent samples, measured in seconds. The sampling period T is set based on the operating frequency of the vibration table, the image processing speed, and the specimen specifications, for example, 0.5-2 seconds. Within each sampling period, the data acquisition unit collects image data of the concrete specimen and vibration response data of the vibration table. The workability parameters for the current batch can be obtained before, during, or at the initial stage of vibration molding, and used as fixed input parameters for that batch when the current batch remains unchanged.

[0020] Image data of the concrete specimens is obtained by an image acquisition device positioned above or to the side of the mold. The acquisition area includes the surface area and corner areas of the concrete specimen. The surface area is used to identify changes in air bubbles, bleeding, reflection, and smoothness, while the corner area is used to identify the filling of the mold corners by the concrete. Vibration response data is acquired by sensors positioned on the vibration table, table surface, or load-bearing parts of the mold. This includes actual acceleration data, table surface displacement data, actual vibration frequency data, motor current data, and mold mass data. The units for actual acceleration are m / s², table surface displacement and amplitude are mm, vibration frequency is Hz, motor current is A, and mold mass is kg. When the weighing module outputs a gravity or load signal, it is first converted to a mass value using m=G / g before being included in the calculation.

[0021] Before extracting features such as bubbles, oozing, and flatness, the visual forming state judgment unit first determines the position of the test mold in the image: an effective search area is obtained by cropping according to a preset ratio to reduce interference from personnel, material buckets, and the ground at the edge of the image; a test mold target detection model is run within the effective search area to obtain the test mold bounding box. The test mold target detection model is a single-class or multi-class target detection model based on deep learning, and the detection category includes at least the test mold. During vibration, the test mold may translate or rotate slightly. Therefore, the test mold area is determined by the detection model in the first frame or re-detection frame, and the test mold position is updated by the inter-frame tracking algorithm in the middle frames. The detection model is run again at preset re-detection intervals for correction. If the overlap between the new detection box and the current tracking box is lower than a preset threshold, the new box is rejected and the test mold area from the previous sampling period is used. The stabilized test mold bounding box is reduced by a preset shrinkage ratio to avoid the mold frame and the edge of overflow slurry, obtaining the surface recognition area of ​​the specimen. Subsequent visual forming feature extraction is then performed within this area. Alternatively, brightness threshold segmentation and contour analysis can be used to locate the test mold within the effective search area.

[0022] The workability parameters of the current batch are obtained through manual input, data collection by testing equipment, or reading from the production management system. For ease of explanation, the "current batch workability parameters" referred to in this manual refer to the combination of material workability parameters and corresponding mold process parameters related to the molding identification of the current batch of concrete; among them, material workability parameters include slump, spread, coarse aggregate particle size, water-cement ratio, and admixture type, and mold process parameters include mold specifications. The batch identification benchmark adaptive unit corrects the benchmark visual identification conditions based on the deviation between the current batch workability parameters and the preset benchmark workability parameters, obtaining a visual identification benchmark adapted to the current batch of concrete. The visual identification benchmark is used to determine the correspondence between changes in air bubbles, bleeding, edge filling, surface smoothness, and surface reflectivity and the molding state of the concrete specimen.

[0023] The visual forming state judgment unit processes the image data within the k-th sampling period, identifies the surface and corner areas of the concrete specimen, and extracts dynamic visual forming features. These dynamic visual forming features include changes in the number of air bubbles, changes in bubble area, changes in bleeding area, changes in surface reflectivity, changes in corner filling, changes in surface smoothness, and changes in segregation evaluation.

[0024] Wherein, the number of bubbles N b (k) represents the number of connected regions in the image that satisfy the preset area range and have roundness or edge closure conditions for bubble recognition; bubble area A b (k) represents the sum of the areas of the connected regions of each bubble, in pixel area or mm² after calibration conversion; water seepage area A w (k) represents the area in the image whose brightness, color, or texture features satisfy the water leakage recognition criteria. The water leakage recognition criteria can be determined by a brightness threshold, a color difference threshold, a low-texture area threshold, or a reflective area continuity threshold; the surface reflective area A r (k) represents the area where the brightness value exceeds the preset brightness benchmark and the duration reaches the preset duration benchmark; the corner fill rate F c (k) is the ratio of the area covered by concrete in the corner region to the area of ​​the target corner region, which is a dimensionless value; surface flatness F p (k) is used to represent the degree of undulation of the concrete surface. It can be obtained by normalizing the height difference of the depth map, the change in surface contour, or the change in grayscale gradient. It is a dimensionless value.

[0025] Image calibration is performed by setting a calibration plate, calibration ruler, or a known-sized mold boundary in the sample acquisition area to obtain the conversion ratio between pixel length and actual length. When the installation height, angle, or lens parameters of the image acquisition device change, calibration is re-performed or the corresponding calibration parameters are called. Bubble area, bleed area, and surface reflective area are calculated based on the same target recognition area as the area value or area ratio. When pixel area is used in the judgment, each threshold uses pixel area units; when actual area is used, it is converted to mm² through the calibration ratio before being used in the judgment. Segregation evaluation changes are obtained from the changes in the proportion of exposed coarse aggregate area, the proportion of slurry aggregate area, and the degree of texture separation within adjacent sampling periods.

[0026] The visual forming state determination unit calculates the change in dynamic visual forming features based on the data changes between the k-th sampling period and the (k-1)-th sampling period, as follows:

[0027]

[0028]

[0029]

[0030]

[0031]

[0032]

[0033] The visual forming state determination unit calculates the change in dynamic visual forming features based on data changes within adjacent sampling periods; where, in the formula, This represents the number of bubbles in the k-th sampling period; This represents the bubble area during the k-th sampling period; This represents the area of ​​oozing water during the k-th sampling period; This represents the surface reflective area during the k-th sampling period; This represents the corner fill rate during the k-th sampling period; This represents the surface smoothness during the k-th sampling period; , , , , and These represent the changes in the corresponding visual forming features between adjacent sampling periods.

[0034] All the above-mentioned changes are used to characterize the change in the k-th sampling period relative to the (k-1)-th sampling period. The Δ sign represents the difference or absolute value of the difference between adjacent sampling periods, and does not represent the original feature value within the k-th sampling period. The absolute value of the change is used when determining convergence, while the positive or negative direction of the change is retained when determining the trend direction.

[0035] In addition to the aforementioned changes in bubble quantity and area based on connected component statistics, the visual forming state judgment unit can also extract texture-based dynamic visual forming features related to bubble escape from the image within the identification area on the specimen surface according to the sampling period. These features are then normalized and weighted to form a forming roughness index, used to characterize the overall change in surface roughness towards smoothness during vibration. The texture-based features include: bubble interface texture features, corresponding to the density of pits, pinholes, and bubble interfaces; bubble contour features, corresponding to the salience of bubble boundaries, circles, and pit edges; and surface light and dark undulation features, corresponding to the light and dark differences caused by the depth and uneven distribution of the slurry.

[0036] Each texture feature can be normalized by dividing it by the baseline value of the corresponding component in the initial stage of vibration, and then synthesized into forming roughness according to preset weights. After temporal smoothing of the forming roughness, its change trend is compared with the visual recognition baseline and preset threshold, and incorporated into the analysis framework of whether the change trend and plateau period are stable within the continuous sampling period mentioned above. This serves as a supplementary basis for the multimodal fusion control unit to determine whether bubble escape has converged and whether a qualified forming state has been reached. This method focuses on the overall decline and stabilization process of texture over time, and can be used alone or in combination with criteria such as changes in bubble number, bubble area, and surface smoothness to reduce the impact of strong reflection, shadows, and motion blur on the bubble count in a single frame.

[0037] Visual confidence Based on the initial visual confidence level The percentage of surface water film area, percentage of strongly reflective area, percentage of obstructed area, and illumination deviation are all corrected in one step, and the calculation method is as follows:

[0038] ;

[0039] in, It is obtained by normalizing the matching degree between dynamic visual shaping features and visual recognition benchmark; This represents the percentage of surface water film area. The percentage of highly reflective area. For the percentage of the area to be covered, The values ​​represent the normalized values ​​for illumination deviation, and a1, a2, a3, and a4 are preset correction coefficients. These preset correction coefficients can be determined based on historical sample data, experimental calibration data, or manual calibration results. Each visual interference factor participates in only one confidence level correction. When surface water film, strong reflections, abnormal illumination, or increased occlusion in the image, the visual confidence level is adjusted. The corresponding reduction.

[0040] The initial visual confidence level is obtained by comparing each dynamic visual forming feature with the visual recognition benchmark. Specifically, the changes in bubble volume, bleeding area, corner filling rate, surface smoothness, surface reflectivity, and segregation evaluation are compared with their corresponding corrected thresholds or scoring mapping relationships to obtain individual matching values ​​between 0 and 1. These individual matching values ​​are then weighted according to their preset weights in the current batch of concrete to obtain the initial visual confidence level. Both the individual matching values ​​and the initial visual confidence level are limited to the range of 0 to 1.

[0041] The vibration response state judgment unit extracts vibration molding response characteristics based on the vibration response data. Vibration molding response characteristics include actual acceleration variation characteristics, table displacement variation characteristics, deviation characteristics between actual vibration frequency and target vibration frequency, motor current fluctuation characteristics, and mold mass variation characteristics.

[0042] During the kth sampling period, the vibration response state judgment unit calculates the root mean square value of acceleration a based on the acceleration sensor data. rms (k), in m / s²; calculate the table displacement amplitude D(k) based on displacement sensor data, in mm; based on the actual vibration frequency f a (k) and target vibration frequency f t (k) Calculate the frequency deviation E f (k), in Hz; calculate motor current fluctuation value based on current sensor data. The unit is A; and the change in the mass of the test mold, Δm(k), is calculated based on the data from the weighing module, with the unit being kg. Wherein: .

[0043] Frequency deviation, motor current fluctuation, and mold mass change are all used as non-negative values ​​for confidence correction or anomaly judgment. When the original calculation yields a signed difference, the absolute value is taken first, or the deviation is normalized to a non-negative value according to the corresponding benchmark before being used in the judgment. The root mean square value of acceleration, table displacement amplitude, frequency deviation, motor current fluctuation, and mold mass change are calculated using data windows within the same sampling period and compared with the preset vibration response benchmark under the same mold specifications and the same target vibration frequency.

[0044] The vibration response state judgment unit will determine the root mean square value of acceleration a. rms (k), table displacement amplitude D(k), frequency deviation Motor current fluctuation value The change in the mass of the mold trial sample, Δm(k), is compared with the preset vibration response benchmark to obtain the vibration forming state result and the initial vibration confidence level C. rp (k), and based on vibration disturbance factors, the initial vibration confidence level C rp (k) After correction, the vibration confidence level is obtained. Vibration shaping results include under-vibration, acceptable, or over-vibration states. Vibration confidence level. A value between 0 and 1, used to indicate the degree to which the vibration response data supports the vibration shaping state results.

[0045] Vibration confidence Based on the initial vibration confidence level A one-time correction is made for frequency deviation, motor current fluctuation, and changes in trial mold quality. The calculation method is as follows:

[0046] ;in, The confidence level is obtained by normalizing the matching degree between the vibration forming response characteristics and the preset vibration response benchmark; E0 is the preset frequency deviation benchmark, in Hz; I0 is the preset current fluctuation benchmark, in A; M0 is the preset mass change benchmark, in kg; b1, b2, and b3 are preset correction coefficients. The preset frequency deviation benchmark, preset current fluctuation benchmark, and preset mass change benchmark can be determined based on the vibration table equipment parameters, mold specifications, historical qualified samples, or test calibration data. The same vibration interference factor only participates in the confidence level correction once. When the frequency deviation, motor current fluctuation, or mold mass change increases, the vibration confidence level... The corresponding reduction.

[0047] The corrected vibration confidence level is limited to the range of 0-1; when any correction term results in a confidence level less than 0, 0 is taken as the corrected vibration confidence level; when no corresponding vibration interference factor is detected, the correction term corresponding to that interference factor is taken as 0.

[0048] The multimodal fusion control unit receives visual shaping state results, visual confidence level Cv(k), vibration shaping state results, and vibration confidence level. The molding state evaluation results are generated by combining the current batch's working parameters. These molding state evaluation results are intermediate results generated by the multi-modal fusion control unit, including the undervibration score S. u (k) Qualified Molding Score q (k) and over-vibration score S o (k). The forming state is the final state determined based on the forming state evaluation results, including under-vibration state, qualified forming state, or over-vibration state; the current batch working parameters are used in the fusion weight calculation and state threshold correction through batch adjustment coefficients.

[0049] The multimodal fusion control unit is based on visual confidence level C. v (k) Vibration confidence level C r (k) and the visual-side batch adjustment coefficient λ corresponding to the current batch workability parameters. v (k) and vibration side batch adjustment coefficient λ r (k) determines the fusion weights of the visual forming state result and the vibration forming state result in the forming state determination. The calculation method is as follows: ; .

[0050] in, For visual fusion weights, For vibration fusion weights, both are values ​​between 0 and 1; and These are all coefficients determined based on the current batch's slump, spread, coarse aggregate size, water-cement ratio, admixture type, or mold specifications. When the value is 0, the multimodal fusion control unit uses a preset default weight. The preset default weight can be determined based on historical qualified samples, equipment calibration results, or manually set values.

[0051] The multimodal fusion control unit based on visual fusion weights and vibration fusion weight The visual state score and the vibration state score are fused to obtain the undervibration score for the current sampling period. Qualified molding score S q (k) and over-vibration score S o (k). For example, the visual lateral undersampling score is The undervibration score on the vibration side is S. ur (k) is the fused undervibration score. It can be represented as: .

[0052] To avoid confusion in the scoring symbols, the fused undervibration score is uniformly denoted as S in this paper. u (k), the qualified molding score after fusion is uniformly recorded as S. q (k), the fused over-vibration score is uniformly denoted as S. o (k); Visual lateral understrain score is denoted as S. uv (k), the undervibration score on the vibration side is denoted as S. ur (k) The same naming rules are used for the qualified molding score and over-vibration score on the visual side and the vibration side.

[0053] Qualified molding score S q (k) and over-vibration score S o (k) can be calculated in the same way. The multimodal fusion control unit is based on the undervibration score S. u (k) Qualified Molding Score q (k) and over-vibration score S o (k) Determine the molding status evaluation result for the current sampling period.

[0054] In one specific scoring method, when the visual forming state result is an undervibration visual state... The visual side score is set to 1, the visual side pass score is set to 0, and the visual side over-vibration score is set to 0. When the visual forming state result is a pass visual state or an over-vibration visual state, the corresponding visual side state score is generated in the same way. The vibration side state score is determined in the same way based on the vibration forming state result, or a score between 0 and 1 is obtained by normalizing the distance between the vibration forming response characteristics and the preset vibration response benchmark.

[0055] The multimodal fusion control unit further determines the molding state of the concrete specimen based on the changing trend of the molding state evaluation results within a continuous sampling period. The continuous sampling period consists of N consecutive sampling periods, where N is an integer greater than 1, and its value can be set according to the mold specifications, vibration molding duration, and image processing speed; for example, N can be 3-5. The changing trend is determined by dynamic visual molding characteristics, vibration molding response characteristics, and under-vibration score. Qualified molding score S q (k), Over-vibration score S o (k) Determine the magnitude of change within the continuous sampling period.

[0056] When the absolute value of the change in the number of bubbles or the change in the bubble area within a continuous sampling period is greater than the preset convergence benchmark, the corner filling rate is lower than the preset filling benchmark, or the surface flatness does not meet the preset molding requirements, the multimodal fusion control unit determines that the concrete specimen is in an under-vibration state.

[0057] When the absolute values ​​of the changes in the number of bubbles and the changes in the bubble area within a continuous sampling period are not greater than the preset convergence benchmark, the corner filling rate reaches the preset filling benchmark, and the change in the bleeding area is within the preset bleeding range, the multimodal fusion control unit determines that the concrete specimen is in a qualified molding state.

[0058] When the change in the bleeding area exceeds the preset bleeding range, the change in the surface reflective area exceeds the preset reflective range, or the segregation evaluation level increases within a continuous sampling period, the multimodal fusion control unit determines that the concrete specimen is in an over-vibration state. The segregation evaluation level can be determined by the proportion of exposed coarse aggregate area, the proportion of slurry aggregation area, or the degree of image texture separation.

[0059] The preset convergence benchmark, preset filling benchmark, preset molding requirements, preset bleeding range, and preset reflective range can be determined based on the mold specifications, concrete mix proportions, historical qualified samples, or test calibration data. Specifically, continuous sampling data should first be collected from qualified molded samples, under-vibrated samples, and over-vibrated samples. Then, the changes in bubble volume, corner filling rate, flatness, bleeding area change, and reflective area change should be extracted according to the same sampling period. For qualified molded samples, the mean ± standard deviation range, quantile range, or manually calibrated range can be used as the preset range, and these ranges should be grouped and stored according to mold specifications and concrete mix proportions.

[0060] When two or more of the following conditions are met simultaneously within the same continuous sampling period: under-vibration state condition, qualified forming state condition, or over-vibration state condition, the multi-modal fusion control unit first compares the under-vibration score S. u (k) Qualified Molding Score q (k) and over-vibration score S o(k) The average or weighted average of the scores over N consecutive sampling periods is used, with the state corresponding to the highest score being selected as the candidate state. When the difference between the highest and second-highest scores is less than a preset state distinction threshold, a risk priority rule is applied for adjudication, with the risk priority order being over-vibration state, under-vibration state, and qualified forming state. Alternatively, the state may be re-evaluated after extending the sampling period by more than one period. The preset state distinction threshold is determined based on the score distribution of historical qualified samples, under-vibration samples, and over-vibration samples.

[0061] Segregation assessment level can be based on the proportion of exposed coarse aggregate area. Percentage of slurry aggregation area Degree of separation of texture Calculation, for example ,in , , Preset separation weights; when When the separation falls into the first, second, or third preset interval, it is determined as a low separation level, a medium separation level, and a high separation level, respectively.

[0062] The vibration actuator generates vibration control commands based on the molding state determined by the multimodal fusion control unit and its changing trend within a continuous sampling period, and adjusts the vibration frequency, amplitude, vibration time, or vibration start / stop status of the vibration table according to the vibration control commands.

[0063] When a concrete specimen is determined to be under-vibrated, or the under-vibration score S is within a continuous sampling period u (k) Upon increasing the vibration level, the vibration actuator generates a vibration enhancement control command. This command includes increasing the vibration frequency, increasing the amplitude, or extending the vibration time. For example, within the allowable range of the vibration table, the vibration frequency is increased by a preset frequency adjustment amount Δf, the amplitude is increased by a preset amplitude adjustment amount ΔA, or the remaining vibration time is extended by a preset time adjustment amount Δt. Here, Δf is in Hz, ΔA is in mm, and Δt is in s.

[0064] When a concrete specimen is determined to be in a qualified molding state, or the qualified molding score S within a continuous sampling period is... q (k) When the vibration is within the preset acceptable range, the vibration actuator generates a vibration hold control command or a vibration stop control command. The vibration hold control command is used to maintain the current vibration frequency and amplitude, while the vibration stop control command is used to stop the vibration table from running.

[0065] When a concrete specimen is determined to be in an over-vibration state, or when the over-vibration score So(k) increases within a continuous sampling period, the vibration execution unit generates a vibration suppression control command or a vibration stop control command. Vibration suppression control commands include reducing the vibration frequency, decreasing the amplitude, or shortening the remaining vibration time. For example, within the allowable range of the vibration table, the vibration frequency is reduced by a preset frequency adjustment amount Δf, the amplitude is reduced by a preset amplitude adjustment amount ΔA, or the remaining vibration time is shortened by a preset time adjustment amount Δt.

[0066] When the vibration actuator executes the control command, it limits the adjusted vibration frequency to the preset lower frequency limit f. min and preset frequency upper limit f max Between these points, the amplitude will be limited to a preset lower amplitude limit A. min and preset amplitude upper limit A max Between, and limit the remaining vibration time to the preset minimum time t. min and the preset maximum time t max The start / stop switching is only performed when the forming state meets the same control condition within H consecutive sampling periods, in order to reduce frequent start / stop operations.

[0067] The preset frequency adjustment amount Δf, preset amplitude adjustment amount ΔA, and preset time adjustment amount Δt are determined by a control table or segmented mapping rules. The control table includes at least the molding state, state score, score change trend, mold specifications, target vibration frequency, allowable frequency range, allowable amplitude range, remaining vibration time range, and start / stop delay period. When the under-vibration score or over-vibration score exceeds the corresponding threshold by a larger margin, the adjustment amount increases according to the preset segments; when the score change tends to converge, the adjustment amount decreases according to the preset segments to avoid over-adjustment of vibration parameters.

[0068] This embodiment continuously acquires image data, vibration response data, and current batch workability parameters during the vibration molding process of concrete specimens. Through visual recognition benchmark correction, visual molding state judgment, vibration molding state judgment, multimodal fusion, and continuous sampling cycle trend judgment, it determines whether the concrete specimen is in an under-vibration state, a qualified molding state, or an over-vibration state, and adaptively adjusts the vibration table operating parameters based on the determined molding state. This reduces errors caused by manual experience judgment, fixed vibration time control, and single visual recognition, allowing the vibration control process to adjust according to changes in the actual molding state of the concrete specimen, thus improving the consistency of the concrete specimen molding process.

[0069] Example 2, based on Example 1, further explains how the batch identification benchmark adaptive unit corrects the visual identification benchmark according to the workability parameters of the current batch of concrete, and how the multimodal fusion control unit corrects the visual confidence and vibration confidence based on visual interference factors and vibration interference factors, and determines the fusion weight of the visual forming state result and the vibration forming state result in the forming state determination accordingly.

[0070] Specifically, the data acquisition unit obtains the workability parameters of the current batch before, during, or after vibration molding of concrete specimens. These parameters include slump (S), spread (F), coarse aggregate size (G), water-cement ratio (R), admixture type (P), and mold size (M). The units for slump (S), spread (F), coarse aggregate size (G), water-cement ratio (R), admixture type (P), and mold size (M) are all in mm. The mold size is used as a process parameter corresponding to the current batch molding identification for correction. The description of the current batch workability parameters is used to collectively refer to material workability parameters and corresponding process parameters. The current batch workability parameters can be obtained through manual input, importing concrete mix proportion data, importing test records, reading from the production management system, or using a slump test device or a spread test device.

[0071] The batch identification baseline adaptive unit stores preset baseline workability parameters. These preset baseline workability parameters are determined based on historical qualified molding samples, test calibration data, or standard specimen molding data, and include baseline slump S0, baseline spread F0, baseline coarse aggregate particle size G0, baseline water-cement ratio R0, baseline admixture type P0, and baseline mold size M0.

[0072] The batch identification baseline adaptive unit generates batch deviation data based on the deviation between the current batch's working parameters and the preset baseline working parameters. The batch deviation data includes the slump deviation δ. S Expansion deviation δ F , coarse aggregate particle size deviation δ G Water-to-binder ratio deviation δ R δ, a parameter representing the difference in admixture types P Difference parameter δ between the mold and the trial mold specifications M .

[0073] Batch deviation data are represented in normalized form: ; ; ; ; where δ S Indicates slump deviation; δ F Indicates the expansion deviation; δ GIndicates the coarse aggregate particle size deviation; δ R δ represents the water-cement ratio deviation; S represents the slump of the current batch of concrete, S0 represents the reference slump; F represents the spread of the current batch of concrete, F0 represents the reference spread; G represents the coarse aggregate size of the current batch of concrete, G0 represents the reference coarse aggregate size; R represents the water-cement ratio of the current batch of concrete, R0 represents the reference water-cement ratio; where δ S δ F δ G and δ R All values ​​are numerical. The admixture type difference parameter δ P Determined based on whether the admixture type is consistent, a preset category distance matrix, or a preset category weight table; when the current admixture type is consistent with the benchmark admixture type, δ P Take 0; if inconsistent, take a value between 0 and 1 based on the category difference. Trial mold specification difference parameter δ M δ is determined based on the normalized deviation between the current trial mold's main dimensions, volume, or molding surface area and the specifications of the benchmark trial mold. M A value between 0 and 1.

[0074] The batch identification baseline adaptive unit determines each identification baseline correction parameter based on the batch deviation data. The identification baseline correction parameters include the bubble identification correction parameter K. b The correction parameters for water seepage identification (Kw), segregation identification (Ks), edge filling identification (Kc), surface smoothness identification (Kp), and surface reflectivity change identification (Kr) are calculated as follows: ; ; ; ; ; Among them, c1, c2, c3, d1, d2, d3, s1, s2, s3, e1, e2, e3, p1, p2, p3, r1, r2, and r3 are preset correction coefficients. The preset correction coefficients are determined based on historical qualified samples, experimental calibration data, or manual calibration results. The batch identification benchmark adaptive unit can also generate identification benchmark correction parameters by using a lookup table, piecewise function, or mapping model.

[0075] The batch identification benchmark adaptive unit corrects the benchmark visual identification conditions according to the identification benchmark correction parameters to obtain a visual identification benchmark suitable for the current batch of concrete. The visual identification benchmark includes the correspondence between visual forming features and the forming state of concrete specimens. Visual forming features include bubble features, bleeding features, segregation features, edge and corner filling features, surface smoothness features, and surface reflectivity variation features; concrete specimen forming states include under-vibration state, qualified forming state, and over-vibration state.

[0076] Specifically, the bubble recognition correction parameter K b Used to adjust the bubble quantity benchmark, bubble area benchmark, or bubble change benchmark; Water seepage identification correction parameter K w Used to adjust the bleed area benchmark or bleed change benchmark; segregation identification correction parameter K s Used to adjust the benchmark for exposed area of ​​coarse aggregate, benchmark for aggregate aggregation area, or benchmark for texture separation; K parameter for edge and corner filling identification correction. c Used to adjust the corner fill rate benchmark; surface flatness identification correction parameter K p The surface flatness reference is used to adjust the surface reflectivity reference; the surface reflectivity change recognition correction parameter Kr is used to adjust the surface reflectivity area reference or reflectivity change amount reference. For example, the corrected bubble change amount reference can be expressed as K. b •B0, the corrected basal area of ​​the oozing water can be expressed as K w •W0, the revised segregation evaluation criterion can be expressed as K s ·S g 0, the corrected corner fill rate benchmark can be expressed as K c •C0, the corrected surface flatness criterion can be expressed as K p ·P p 0, the corrected standard for reflectance change can be expressed as Kr·R g 0, where B0, W0, S g 0, C0, P p 0 and R g 0 represents the corresponding baseline visual recognition condition.

[0077] Each identification benchmark correction parameter is set with a preset lower limit and a preset upper limit. The corrected threshold must not be less than the minimum value allowed for the corresponding physical quantity or scoring quantity. When the output of the linear weighting, lookup table, or mapping model exceeds the preset range, it is truncated according to the preset lower limit or preset upper limit to avoid negative thresholds or thresholds that do not conform to the actual forming state.

[0078] The aforementioned baseline visual recognition conditions can be stored as a threshold table or a scoring mapping table. The threshold table should include at least the thresholds for bubble change, bleeding change, edge filling rate, flatness, reflectivity change, and segregation evaluation for each mold specification, slump range, spread range, and coarse aggregate size range. The system should call the corresponding thresholds based on the current batch workability parameters or calculate the baseline thresholds by correcting the parameters.

[0079] The process of establishing a threshold table or scoring mapping table includes: collecting image data and vibration response data of qualified molding samples, under-vibration samples, and over-vibration samples respectively; extracting visual molding features under the same sampling period and the same image calibration conditions; grouping samples according to mold specifications, slump range, spread range, coarse aggregate particle size range, and water-cement ratio range; and forming corresponding thresholds for the characteristic distribution of each group of samples using the mean ± standard deviation, quantile range, or manually calibrated range.

[0080] For example, when the current batch of concrete has a high slump or large spread, the concrete's fluidity is enhanced, and surface bleeding and reflectivity may appear earlier. In this case, the batch identification baseline adaptive unit uses the bleeding identification correction parameter Kw and the surface reflectivity change identification correction parameter K to... r The reference for the water seepage area and the reference for the surface reflection area are corrected to reduce the impact of early water seepage or early reflection on the judgment of over-vibration state.

[0081] For example, when the coarse aggregate particle size of the current batch of concrete is large or the mold size is large, the difficulty of filling the corner areas increases. In this case, the batch identification benchmark adaptive unit corrects the corner filling rate benchmark by using the corner filling identification correction parameter Kc, thereby improving the role of corner filling features in the judgment of under-vibration state and enabling the system to identify the under-vibration risk caused by insufficient corner filling.

[0082] Based on a modified visual recognition benchmark, the visual forming state judgment unit identifies changes in the number of bubbles, bubble area, water seepage area, surface reflection, corner filling, and surface smoothness in the image data, obtaining visual forming state results and visual confidence levels. Since the visual recognition benchmark has been corrected based on the current batch workability parameters, the judgment effect of the same visual forming feature in different batches of concrete can be adjusted according to batch differences.

[0083] In this embodiment, the multimodal fusion control unit also identifies visual interference factors and vibration interference factors, and corrects the initial visual confidence and initial vibration confidence based on the identification results; when the confidence correction for the same interference factor has already been completed in Embodiment 1, this embodiment will not repeat the multiplication for the same interference factor.

[0084] Visual interference factors include surface water film, strong reflection, illumination changes, and image occlusion. The proportion of surface water film area is determined by using the target recognition area as the denominator, based on the area of ​​high-brightness, low-texture regions in the image; the proportion of strong reflection area is determined by using the target recognition area as the denominator, based on the area of ​​regions whose brightness values ​​exceed a preset brightness benchmark and whose saturation or texture changes meet the reflection conditions; illumination changes are determined based on the deviation between the overall brightness of the current image or the brightness of the background reference area and the benchmark brightness; image occlusion is determined based on the proportion of unrecognizable areas within the target recognition area.

[0085] The multimodal fusion control unit obtains the percentage of surface water film area in the kth sampling period. Percentage of highly reflective area , percentage of the area covered Normalized value of illumination deviation And based on the above-mentioned visual interference factors, the visual confidence level without visual interference correction was determined. Make corrections to obtain the corrected visual confidence level. The calculation method is as follows:

[0086] ;in, This represents the corrected visual confidence level within the k-th sampling period; This represents the visual confidence level that no visual interference correction was performed within the k-th sampling period; Indicates the percentage of surface water film area; Indicates the percentage of area with strong reflectivity; Indicates the percentage of the area obscured; This represents the normalized value of illumination deviation; a1, a2, a3, and a4 are preset visual interference correction coefficients.

[0087] in, , , , , and All values ​​are numerical, and the preset visual interference correction coefficient is determined based on historical samples, experimental calibration data, or manual calibration results. When the proportion of surface water film area, the proportion of strongly reflective area, the proportion of occluded area, or the normalized value of illumination deviation increases, the corrected visual confidence level... Reduce; the correction term corresponding to the same visual interference factor is only counted once in the same sampling period.

[0088] Vibration interference factors include vibration frequency fluctuations, abnormal table displacement, abnormal motor current, and abnormal mold quality. Vibration frequency fluctuations are determined based on the actual vibration frequency f. a (k) and the target vibration frequency f t The non-negative frequency deviation E between (k) f(k) is determined in Hz; abnormal table displacement is based on the non-negative table displacement deviation E between the table displacement amplitude D(k) and the preset displacement reference. d (k) is determined in mm; abnormal motor current is based on the motor current fluctuation value σ within the sampling period. I (k) is determined in A; the abnormality of the trial mold quality is determined based on the absolute value or normalized non-negative deviation of the change in trial mold quality Δm(k), in kg.

[0089] The multimodal fusion control unit is based on the frequency deviation E f (k) Table displacement deviation E d (k) Motor current fluctuation value σ I (k) and the change in mold mass Δm(k) are used to correct the initial vibration confidence level or the vibration confidence level without vibration disturbance correction:

[0090] Wherein, Cr'(k) is the corrected vibration confidence level, E0 is the preset frequency deviation benchmark (Hz), D0 is the preset displacement deviation benchmark (mm), I0 is the preset current fluctuation benchmark (A), M0 is the preset mass change benchmark (kg), and b1, b2, b3, and b4 are preset vibration interference correction coefficients. The preset frequency deviation benchmark, preset displacement deviation benchmark, preset current fluctuation benchmark, preset mass change benchmark, and preset vibration interference correction coefficients are determined based on the vibration table equipment parameters, mold specifications, historical qualified samples, or test calibration data. When the frequency deviation, table displacement deviation, motor current fluctuation value, or mold mass change increases, the corrected vibration confidence level Cr'(k) decreases; the same vibration interference factor is corrected only once within the same sampling period.

[0091] The multimodal fusion control unit determines the fusion weights of the visual forming state results and the vibration forming state results in the forming state determination based on the corrected visual confidence Cv'(k) and the corrected vibration confidence Cr'(k): ; .

[0092] Where Wv'(k) is the corrected visual fusion weight, and Wr'(k) is the corrected vibration fusion weight, both being values ​​between 0 and 1. When Cv'(k) + Cr'(k) is 0, the multimodal fusion control unit uses a preset default weight. The preset default weight is determined based on historical qualified samples, equipment calibration results, or manually set values.

[0093] When the visual side batch adjustment coefficient λ in this embodiment is compared with that in Embodiment 1 v (k) and vibration side batch adjustment coefficient λ r(k) When used together, the corrected visual confidence and vibration confidence are first multiplied by the corresponding batch adjustment coefficients, and then normalized to obtain the fusion weights; when batch adjustment coefficients are not enabled, λ v (k) and λ r (k) are all set to 1. This rule is used to maintain the consistency of the fusion weight caliber between Example 1 and Example 2.

[0094] The multimodal fusion control unit fuses the visual forming state results and vibration forming state results based on the modified visual fusion weight Wv'(k) and the modified vibration fusion weight Wr'(k), generating the forming state evaluation results of the concrete specimen. The forming state evaluation results include the under-vibration score S. u (k) Qualified Molding Score q (k) and over-vibration score S o (k) is used to further determine whether the concrete specimen is under-vibrated, properly formed, or over-vibrated.

[0095] When the image data contains a large area of ​​strong reflection or a surface water film, the corrected visual confidence level Cv'(k) decreases, the fusion weight of the visual forming state result in the forming state determination decreases, and the fusion weight of the vibration forming state result in the forming state determination increases accordingly. When the vibration table experiences frequency fluctuations, abnormal table displacement, or abnormal motor current, the corrected vibration confidence level Cr'(k) decreases, the fusion weight of the vibration forming state result in the forming state determination decreases, and the fusion weight of the visual forming state result in the forming state determination increases accordingly.

[0096] When using C in Example 1 v (k), C r (k) as the corrected confidence level, C in this embodiment v '(k), C r '(k) can be respectively equivalent to C v (k), C r (k) is used directly for fusion weight calculation; C is generated according to the formula in this embodiment only when the uncorrected confidence level is used as input. v '(k), C r '(k).

[0097] Through the above process, this embodiment can adaptively correct the visual recognition benchmark based on the current batch workability parameters, and correct the visual confidence and vibration confidence based on visual and vibration interference factors. This allows the multimodal fusion control unit to simultaneously consider material batch differences, visual recognition reliability, and vibration response reliability when determining the molding state. Therefore, it can reduce the impact of workability differences between different batches of concrete, surface water film, strong reflection, lighting variations, image occlusion, and abnormal vibration on the judgment of the molding state, improving the consistency and stability of concrete specimen molding state recognition.

[0098] Example 3, based on Example 1 or Example 2, further illustrates how the vibration response state judgment unit extracts vibration forming response features from vibration response data, how it generates vibration transmission state results based on actual acceleration change features, table displacement change features, and deviation features between actual vibration frequency and target vibration frequency, and how the vibration execution unit generates vibration control commands based on the determined forming state.

[0099] In this embodiment, the data acquisition unit collects vibration response data according to a preset sampling period T during the vibration molding process of the concrete specimen. The vibration response data is collected by sensors installed on the vibration table, table surface, motor drive end, or the load-bearing part of the mold, and includes actual acceleration data, table surface displacement data, actual vibration frequency data, motor current data, and mold mass data. The unit for actual acceleration is m / s², the unit for table surface displacement and amplitude is mm, the unit for actual vibration frequency and target vibration frequency is Hz, the unit for motor current is A, and the unit for mold mass is kg; when the weighing module outputs a load signal, it is converted to a mass value using m=G / g.

[0100] Actual vibration frequency f a (k) can be obtained from the feedback signal of the vibration table controller, or from the acceleration signal collected by the acceleration sensor through spectrum analysis; the target vibration frequency f t (k) is set by the vibration table control system.

[0101] The vibration response state determination unit extracts vibration shaping response characteristics based on the vibration response data. Specifically, within the k-th sampling period, the vibration response state determination unit calculates the root mean square value 'a' of the actual acceleration based on the acceleration sensor data. rms (k), calculate the displacement amplitude D(k) of the platform based on the displacement sensor data, and calculate the actual vibration frequency f. a (k) and target vibration frequency f t (k) Calculate the frequency deviation E f (k) Calculate the motor current fluctuation value σ based on the current sensor data. I(k), and calculate the change in the mass of the test mold Δm(k) based on the data from the weighing module, where: The actual root mean square acceleration value arms(k) is in m / s², the table displacement amplitude D(k) is in mm, and the frequency deviation E f (k) is in Hz, and the motor current fluctuation value σ I The unit of (k) is A, and the unit of the change in the mass of the test mold Δm(k) is kg.

[0102] The characteristics of actual acceleration variation include the root mean square value of actual acceleration a rms (k) and its change over a continuous sampling period; the table displacement variation characteristics include the table displacement amplitude D(k) and its change over a continuous sampling period; the frequency deviation characteristics include the actual vibration frequency f a (k) and the target vibration frequency f t The non-negative frequency deviation E between (k) f (k). The continuous sampling period consists of N consecutive sampling periods, where N is an integer greater than 1. Its value is set according to the vibration molding time, mold specifications and vibration table response speed, for example, N is 3-5.

[0103] The vibration response state judgment unit will determine the actual root mean square value of acceleration a. rms (k), table displacement amplitude D(k), non-negative frequency deviation E f (k) Motor current fluctuation value σ I The absolute values ​​or normalized non-negative deviations of (k) and the change in mold mass Δm(k) are compared with the preset vibration response benchmark to obtain the vibration forming state results and vibration confidence level C. r (k). The preset vibration response benchmark is determined based on the vibration table equipment parameters, mold specifications, historical qualified molded samples, or test calibration data, including the acceleration benchmark range, table displacement benchmark range, frequency deviation benchmark, motor current fluctuation benchmark, and mold mass change benchmark. Vibration confidence level C r (k) is a value between 0 and 1, used to represent the degree of support of the vibration response data for the vibration forming state results.

[0104] In this embodiment, the vibration response state judgment unit further generates a vibration transmission state result based on the actual acceleration change characteristics, table displacement change characteristics, and frequency deviation characteristics within the continuous sampling period. The vibration transmission state result includes the vibration transmission level L. v (k) or vibration deviation Q v (k) is used to reflect the correspondence between the actual acceleration variation characteristics, the table displacement variation characteristics, and the target vibration frequency. Among them, the vibration transmission level Lv(k) is a discrete level value, belonging to the result; the vibration deviation degree Q... v(k) is a numerical value.

[0105] The vibration response state judgment unit determines the vibration response state based on the target vibration frequency f. t (k) Determine the corresponding reference acceleration and the displacement of the reference platform Reference acceleration and the displacement of the reference platform The reference acceleration A can be determined by the target vibration frequency, mold specifications, and vibration table calibration data, or it can be obtained through vibration table no-load calibration, standard mold loading calibration, or historical qualified molded samples. g (k) can be represented as the target vibration frequency f t (k) and the trial mold specification M are functions Displacement of the reference platform It can be expressed as a function of the target vibration frequency ft(k) and the mold specification M. ;in, and It can be obtained by calibration function, lookup table data, or fitting with historical qualified samples.

[0106] The vibration response state judgment unit calculates the acceleration deviation Ea(k) and displacement deviation Ed(k): ; Where A0 is the preset acceleration deviation from the reference, in m / s²; D0 is the preset displacement deviation from the reference, in mm; E a (k) and E d (k) are all numerical values. The preset acceleration deviation from the reference A0 and the preset displacement deviation from the reference D0 are determined based on the vibration table equipment parameters, mold specifications, historical qualified molding samples, or test calibration data.

[0107] The vibration response state judgment unit judges the acceleration deviation E. a (k), Displacement deviation E d (k) and frequency deviation E f (k) Generation vibration deviation degree Qv(k): Where E0 is the preset frequency deviation benchmark, in Hz; g1, g2, and g3 are preset transmission correction coefficients, which are determined based on equipment calibration data, historical qualified samples, or manual calibration results. The vibration deviation degree Qv(k) is a numerical value; the larger Qv(k) is, the greater the deviation between the actual acceleration, table displacement, and target vibration frequency.

[0108] The above acceleration deviation E a (k), Displacement deviation E d (k) and frequency deviation E f(k) are all non-negative deviation quantities when used to generate the vibration deviation degree Qv(k); when a signed difference value is obtained by subtracting the reference value from the measured value, the absolute value is taken first or the value is normalized to a non-negative value according to the preset deviation reference before substitution, so as to avoid mutual cancellation of positive and negative deviations. When Qv(k) is used for confidence correction, it is first mapped or truncated to the range of 0-1.

[0109] The vibration response state judgment unit may further generate the vibration transmission level Lv(k) according to the vibration deviation degree Qv(k). For example, a first deviation threshold q1 and a second deviation threshold q2 are set, and 0<q1<q2; when Qv(k)≤q1, a low deviation level is generated; when q1<Qv(k)≤q2, a medium deviation level is generated; when Qv(k)>q2, a high deviation level is generated. q1 and q2 may be determined according to the vibration table equipment parameters, test mold specifications, historical qualified samples or test calibration data, for example, q1 is 0.3 and q2 is 0.6. The low deviation level indicates that the vibration transmission deviation degree is low, the medium deviation level indicates that there is a correctable deviation in vibration transmission, and the high deviation level indicates that the vibration transmission is obviously abnormal, which requires reducing the vibration confidence or correcting the vibration forming state result. The above example thresholds are only used to illustrate the level division method, and are determined according to the corresponding equipment and test mold calibration results in actual application.

[0110] The vibration response state judgment unit corrects at least one of the vibration forming state result and the vibration confidence according to the vibration transmission state result. Specifically, when the actual vibration frequency f a (k) is close to the target vibration frequency f t (k), but the actual acceleration root mean square value a rms (k) is lower than the reference acceleration A g (k), or the table top displacement amplitude D(k) is lower than the reference table top displacement D g (k), it indicates that the output frequency of the vibration table is close to the target vibration frequency, but there is a deviation in the transmission of vibration action to the table top, the test mold or the concrete test piece. The actual vibration frequency f a (k) is close to the target vibration frequency f t (k), which means that the frequency deviation E f (k) is not greater than the preset frequency proximity reference, and the preset frequency proximity reference is determined according to the vibration table equipment parameters, the target vibration frequency or test calibration data.

[0111] At this time, the vibration response state judgment unit increases the vibration deviation degree Qv(k) and reduces the vibration confidence Cr(k) to obtain the corrected vibration confidence , the corrected vibration confidence is determined according to the following formula: ; where, is the corrected vibration confidence, The vibration confidence level before correction is given, and Qv(k) represents the degree of vibration deviation. As the degree of vibration deviation Qv(k) increases, the corrected vibration confidence level... reduce.

[0112] Since Qv(k) is limited to a non-negative deviation, the corrected vibration confidence will not be abnormally amplified due to negative deviation; when the calculation result is less than 0, it is taken as 0, and when the calculation result is greater than 1, it is taken as 1.

[0113] It should be noted that, in different implementation methods, the corrected vibration confidence level... The vibration confidence level can be obtained from one or more of the vibration interference factors and vibration transmission state results. When considering both vibration interference factors and vibration transmission state results simultaneously, the multimodal fusion control unit or vibration response state judgment unit can sequentially correct the vibration confidence level, or it can generate a corrected vibration confidence level based on the corresponding correction coefficients of both factors. .

[0114] In another implementation, the vibration response state judgment unit corrects the vibration forming state result based on the vibration transmission state result. For example, when the vibration forming state result is initially determined to be a qualified forming state, but the vibration deviation Qv(k) reaches a preset deviation benchmark, and the table displacement amplitude D(k) within the continuous sampling period is lower than the benchmark table displacement Dg(k), the vibration response state judgment unit corrects the vibration forming state result to an under-vibration state. The preset deviation benchmark is determined based on historical qualified forming samples, equipment calibration data, or trial mold specifications.

[0115] When the vibration molding state result is initially determined to be a qualified molding state, but the actual root mean square value of acceleration arms(k) and the amplitude of table displacement D(k) are both higher than the corresponding reference range, the non-negative frequency deviation Ef(k) increases and reaches the preset frequency change condition, and at least one of the following—visual side water seepage area change, surface reflective area change, or segregation evaluation level—points to the risk of over-vibration, the vibration response state judgment unit will correct the vibration molding state result to an over-vibration state.

[0116] The preset frequency variation condition can be based on the non-negative frequency deviation E within the continuous sampling period. f The magnitude, number of changes, or slope of change of (k) are determined. For example, when the non-negative frequency deviation E within a continuous sampling period... f The increase in (k) reaches the preset frequency increment reference, or the non-negative frequency deviation E within multiple consecutive sampling periods. f (k) When all are higher than the preset frequency deviation benchmark, the non-negative frequency deviation E is determined. f (k) The preset frequency change condition is met.

[0117] The multimodal fusion control unit receives visual forming status results, visual confidence levels, corrected vibration forming status results, and corrected vibration confidence levels, and determines the forming status of the concrete specimen by combining these with the current batch workability parameters. Forming status includes under-vibration state, qualified forming state, and over-vibration state.

[0118] The vibration actuator generates vibration control commands based on the molding state determined by the multimodal fusion control unit and its changing trend within a continuous sampling period, and adjusts the vibration frequency, amplitude, vibration time, and vibration start / stop status of the vibration table according to the vibration control commands.

[0119] When a concrete specimen is determined to be under-vibrated, or when the under-vibration score increases within a continuous sampling period and reaches a preset under-vibration change condition, the vibration execution unit generates an enhanced vibration control command. This enhanced vibration control command includes increasing the vibration frequency, increasing the amplitude, or extending the vibration time. For example, within the allowable range of the vibration table, the vibration frequency may be increased by a preset frequency adjustment amount Δf, the amplitude by a preset amplitude adjustment amount ΔA, or the remaining vibration time by a preset time adjustment amount Δt. Here, Δf is in Hz, ΔA is in mm, and Δt is in seconds.

[0120] The preset undervibration change condition can be determined based on the magnitude, number of changes, or slope of the undervibration score within a continuous sampling period. For example, when the increase in the undervibration score within a continuous sampling period reaches the preset undervibration score increment benchmark, or when the undervibration score is higher than the preset undervibration score benchmark in multiple consecutive sampling periods, the undervibration score is determined to have reached the preset undervibration change condition.

[0121] When a concrete specimen is determined to be in a qualified molding state, or when the qualified molding score falls within the preset qualified range during a continuous sampling period, the vibration actuator generates a maintain vibration control command or a stop vibration control command. The maintain vibration control command is used to maintain the current vibration frequency and amplitude, while the stop vibration control command is used to stop the vibration table operation. The preset qualified range can be determined based on historical qualified molding samples, mold specifications, concrete mix proportions, or test calibration data.

[0122] When a concrete specimen is determined to be in an over-vibration state, or when the over-vibration score increases within a continuous sampling period and reaches the preset over-vibration change condition, the vibration execution unit generates a vibration suppression control command or a vibration stop control command. Vibration suppression control commands include reducing the vibration frequency, decreasing the amplitude, or shortening the remaining vibration time. For example, within the allowable range of the vibration table, the vibration frequency is reduced by a preset frequency adjustment amount Δf, the amplitude is reduced by a preset amplitude adjustment amount ΔA, or the remaining vibration time is shortened by a preset time adjustment amount Δt.

[0123] The preset over-vibration change condition can be determined based on the magnitude, number of changes, or slope of the over-vibration score within a continuous sampling period. For example, when the increase in the over-vibration score within a continuous sampling period reaches the preset over-vibration score increment benchmark, or when the over-vibration score is higher than the preset over-vibration score benchmark in multiple consecutive sampling periods, the preset over-vibration change condition is determined to be met.

[0124] The vibration actuator can store vibration enhancement control commands, vibration maintenance control commands, vibration suppression control commands, and vibration stop control commands as control tables. The control tables include forming state, change trend, frequency adjustment amount Δf, amplitude adjustment amount ΔA, time adjustment amount Δt, upper limit allowable, lower limit allowable, and start / stop delay period. During control, the control table is queried according to the current forming state and the corresponding control amount is output.

[0125] The frequency adjustment Δf, amplitude adjustment ΔA, and time adjustment Δt in the control table are set in segments according to the degree to which the state score exceeds the corresponding threshold. The higher the under-vibration score or the more persistent the under-vibration trend, the larger the Δf, ΔA, or Δt in the vibration enhancement control command; the higher the over-vibration score or the more persistent the over-vibration trend, the larger the Δf, ΔA, or Δt in the vibration suppression control command. Each adjustment is limited by the allowable parameter range of the vibration table, the mold specifications, and the preset start-stop delay period.

[0126] This embodiment enables the vibration response state judgment unit to not only obtain the vibration forming state result and vibration confidence level based on the vibration response data, but also generate the vibration transmission state result based on the actual acceleration change characteristics, table displacement change characteristics, and frequency deviation characteristics. The vibration transmission state result is then used to correct the vibration forming state result or vibration confidence level. Therefore, it can identify situations where the vibration table output frequency is close to the target vibration frequency, but the vibration effect is deviated from the transmission to the table, mold, or concrete specimen. This reduces misjudgments of the forming state caused by vibration transmission deviations, making the vibration control commands more consistent with the actual vibration state of the concrete specimen.

[0127] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. An intelligent identification and vibration control system for the molding quality of concrete specimens, characterized in that, It includes a data acquisition unit that acquires the current batch's workability parameters before, during, or after vibration molding, and collects concrete specimen image data and vibration table vibration response data according to a preset sampling cycle during vibration molding. The batch identification benchmark adaptive unit corrects the benchmark visual identification rules based on the deviation between the current batch workability parameters and the preset benchmark workability parameters, and obtains a visual identification benchmark that is suitable for the current batch of concrete. The visual forming state judgment unit obtains the visual forming state result and visual confidence level based on image data and visual recognition benchmark. The vibration response state judgment unit extracts vibration forming response features from the vibration response data, and obtains the vibration forming state result and vibration confidence level based on the comparison results between the vibration forming response features and the preset vibration response benchmark. The multimodal fusion control unit performs correlation analysis based on the batch adjustment coefficient, visual forming status result, visual confidence level, vibration forming status result and vibration confidence level corresponding to the current batch working parameters, generates forming status evaluation result, and determines the forming status by combining the changes within the continuous sampling period. The vibration actuator adaptively adjusts the vibration frequency, amplitude, vibration time, or vibration start / stop status of the vibration table according to the molding state, so that the concrete specimen reaches the preset molding quality state.

2. The intelligent identification and vibration control system for the molding quality of concrete specimens according to claim 1, characterized in that, The multimodal fusion control unit converts the visual forming state result and the vibration forming state result into at least one of the following: the concrete specimen is in an under-vibration state, a qualified forming state, or an over-vibration state. Based on the batch adjustment coefficient, visual confidence level, and vibration confidence level corresponding to the workability parameters of the current concrete batch, the unit determines the fusion weight of the visual forming state result and the vibration forming state result in the forming state judgment, and generates the forming state evaluation result of the concrete specimen based on the fusion weight.

3. The intelligent identification and vibration control system for the molding quality of concrete specimens according to claim 1, characterized in that, The batch identification benchmark adaptive unit generates identification benchmark correction parameters corresponding to at least one of the following: bubble identification, bleeding identification, segregation identification, edge filling identification, surface smoothness identification, and surface reflectivity change identification, based on the deviation between the workability parameters of the current concrete batch and the preset benchmark workability parameters. The identification benchmark correction parameters are determined by normalized deviation weighting, table lookup, piecewise function, or calibration model, and a visual identification benchmark adapted to the current concrete batch is generated or called based on the identification benchmark correction parameters.

4. The intelligent identification and vibration control system for the molding quality of concrete specimens according to claim 1, characterized in that, The visual recognition benchmark includes the correspondence between visual forming features and the forming state of concrete specimens. The correspondence includes visual feature thresholds, change thresholds, and state score mapping relationships. It is modified based on at least one of the following: slump, spread, coarse aggregate particle size, water-cement ratio, admixture type, and mold specifications of the current concrete batch. This is to adjust the degree of influence of bubble features, bleeding features, segregation features, edge filling features, surface smoothness features, and surface reflectivity change features in judging under-vibration state, qualified forming state, or over-vibration state.

5. The intelligent identification and vibration control system for the molding quality of concrete specimens according to claim 1 or 2, characterized in that, The multimodal fusion control unit corrects the initial visual confidence level or initial vibration confidence level based on the identified visual or vibration interference factors, obtaining corrected visual and vibration confidence levels. The multimodal fusion control unit multiplies the corrected visual confidence level by the visual-side batch adjustment coefficient and the corrected vibration confidence level by the vibration-side batch adjustment coefficient, then normalizes the multiplication results to determine the fusion weights of the visual forming state result and the vibration forming state result in the forming state determination. The visual interference factors include surface water film, strong reflection, lighting changes, or image occlusion; the vibration interference factors include vibration frequency fluctuations, abnormal table displacement, abnormal motor current, or abnormal mold quality.

6. The intelligent identification and vibration control system for the molding quality of concrete specimens according to claim 1, characterized in that, The multimodal fusion control unit determines the molding state of the concrete specimen based on the changing trend of the molding state evaluation results within a continuous sampling period. The changing trend is determined by the change amplitude of visual molding characteristics and vibration molding response characteristics. When the molding state evaluation results indicate that the bubbles have not converged, the corner filling has not met the preset molding requirements, or the surface flatness has not met the preset molding requirements, it is determined to be an under-vibration state. When the molding state evaluation results indicate that the bubble escape and convergence, the corner filling meets the preset molding requirements, and the bleeding change is within the preset range, it is determined to be a qualified molding state; when the molding state evaluation results indicate that the bleeding change exceeds the preset range, the surface reflection change exceeds the preset range, or the segregation evaluation level increases, it is determined to be an over-vibration state; among them, the preset molding requirements and preset range are determined by the mold specifications, concrete mix proportion, historical qualified molding samples, or test calibration data.

7. The intelligent identification and vibration control system for the molding quality of concrete specimens according to claim 1, characterized in that, The vibration execution unit generates vibration control commands based on the forming state determined by the multimodal fusion control unit and its changing trend within a continuous sampling period. It then adjusts at least one of the vibration frequency, amplitude, vibration time, and vibration start / stop state of the vibration table within the allowable parameter range of the vibration table according to these commands. Specifically, when the concrete specimen is determined to be in an under-vibration state or its changing trend points to an under-vibration state, an enhanced vibration control command is generated; when the concrete specimen is determined to be in a qualified forming state or its changing trend points to a qualified forming state, a sustained vibration control command or a stop vibration control command is generated; and when the concrete specimen is determined to be in an over-vibration state or its changing trend points to an over-vibration state, a suppressed vibration control command or a stop vibration control command is generated.

8. The intelligent identification and vibration control system for the molding quality of concrete specimens according to claim 1, characterized in that, The visual forming state judgment unit is used to analyze image data within a continuous sampling period. Before extracting dynamic visual forming features, it first determines the mold area in the image data and then determines the specimen surface recognition area based on the mold area. The mold area is determined by mold target detection, inter-frame tracking, or contour analysis, and the specimen surface recognition area is obtained by shrinking the mold area by a preset ratio. The visual forming state judgment unit extracts dynamic visual forming features from at least one of the surface area and corner area of ​​the concrete specimen, and obtains the visual forming state result and visual confidence based on the matching relationship between the dynamic visual forming features and the visual recognition benchmark. The dynamic visual forming features include at least one of the following: changes in the number of bubbles, changes in bubble area, changes in water bleeding area, changes in surface reflection, changes in corner filling, and changes in surface smoothness, obtained through image calibration, threshold segmentation, connected component analysis, texture analysis, or depth-height difference calculation; and forming roughness index obtained by normalizing and weighting bubble interface texture features, bubble contour features, and surface brightness and darkness undulation features. The forming roughness index is used to characterize the trend of the concrete specimen surface changing from rough to smooth.

9. The intelligent identification and vibration control system for the molding quality of concrete specimens according to claim 1, characterized in that, The vibration response state judgment unit extracts vibration forming response features based on the vibration response data, and compares the vibration forming response features with a preset vibration response benchmark to obtain the vibration forming state result and vibration confidence level; wherein, the vibration forming response features include at least one of the following: actual acceleration change features, table displacement change features, deviation features between actual vibration frequency and target vibration frequency, motor current fluctuation features, and mold mass change features.

10. The intelligent identification and vibration control system for the molding quality of concrete specimens according to claim 9, characterized in that, The vibration response state judgment unit generates a vibration transmission state result based on the actual acceleration change characteristics, table displacement change characteristics, and deviation characteristics between the actual vibration frequency and the target vibration frequency within a continuous sampling period. It then corrects at least one of the vibration forming state result and vibration confidence level based on the vibration transmission state result. The vibration transmission state result includes a vibration transmission level or vibration deviation degree. This deviation degree is obtained by weighting the non-negative deviation of the actual acceleration relative to the reference acceleration, the non-negative deviation of the table displacement relative to the reference table displacement, and the frequency deviation between the actual vibration frequency and the target vibration frequency. This weighted result reflects the correspondence between the actual acceleration change characteristics, the table displacement change characteristics, and the target vibration frequency. The reference acceleration and reference table displacement are determined based on the target vibration frequency, mold specifications, and vibration table calibration data.

Images