High-freedom multi-linear mixed adaptation glue spraying method and system, and storage medium

Abstract

The application provides a high-freedom-dimension multi-line mixed self-adaptive glue spraying method, system and storage medium; the method generates a spraying path meeting G2 continuous constraint in a world coordinate system, and takes line width W and cross section height H as outer ring control targets, and takes spraying speed v and spraying valve duty cycle D as inner ring interlocking quantities; in combination with a response surface / Gaussian process model, end point gradual change compensation, pattern gating and minimum opening and closing time τ_min constraint, stable forming of the line width and the cross section height is realized under different materials and environments; compared with a constant speed or open loop scheme, the application can inhibit corner glue accumulation and end point "size head", and shorten multi-section spraying beats.

Description

Technical Field

[0001] This invention relates to the field of intelligent automation control technology, specifically to a self-adaptive adhesive spraying method and system based on high-freedom-dimensional multi-linear mixing. Background Technology

[0002] In applications such as 3C electronics, automotive interiors, packaging / sealing, and precision assembly, adhesives or coatings are commonly deposited onto the substrate surface along a predetermined path to obtain a "coating path" with a target linewidth W and cross-sectional height H. Industrial equipment often uses spray valves (pneumatic or electromagnetically driven), screw valves, or time-pressure valves in conjunction with multi-axis motion platforms to complete the spraying process. To improve consistency and throughput, control systems are typically based on an open-loop / single-loop strategy of "fixed path + constant or limited speed": controlling the motion speed v and valve opening / closing (or duty cycle D), supplemented by process settings for temperature T, viscosity η, and supply pressure P. Some devices introduce online vision sensors or line laser sensors to measure W and H for simple deviation correction.

[0003] With the increase in multi-batch, small-volume production, material batch differences, ambient temperature drift, and equipment wear inevitably cause slow drift in (W,H) output. If the control side relies solely on fixed tables or limited calibration curves, it is often difficult to adapt in a timely manner, resulting in fluctuations in molding quality, decreased yield, and increased rework.

[0004] The shortcomings of existing technology: 1. Insufficient control coupling and inadequate closed-loop granularity Existing solutions mainly rely on single-variable adjustment of velocity v or valve opening / duty cycle D, without establishing a multi-factor mapping and online correction mechanism between (W,H) and (v, D, κ, T, η, P); when the environment and materials change, the width and height are prone to drift and converge slowly.

[0005] 2. Speed-valve co-planning without incorporating "curvature-driven" principles Conventional constant speed or planning based solely on acceleration and deceleration curves does not explicitly couple the path curvature κ(s) into the collaborative setting of v(s) and D(s); at small radius corners / sharp bends, due to the increased centrifugal effect and jet diffusion, corner build-up or edge undulations are easily formed.

[0006] 3. Insufficient dynamic compensation at endpoints / corners Valves and fluids exhibit response delays and wake effects. Without gradual and geometric compensation such as end pre-filling t_pre, shrinkage t_post, and line segment indentation δ_in / offset δ_off, "large head / tail / collapse" often occur at the end.

[0007] 4. Weak ability to mix and match multiple styles on the same line segment. Existing processes often use "segmentation + multiple program segments" to achieve combinations of solid lines, dashed lines, and dotted lines, which brings additional start / stop and alignment errors, and makes it difficult to maintain the continuity of style on a single line segment.

[0008] Therefore, existing technologies have shortcomings and need further improvement. Summary of the Invention

[0009] To address the problems existing in the prior art, this invention provides a self-adaptive spray adhesive method and system based on high-freedom-dimensional multi-linear mixing.

[0010] To achieve the above objectives, the specific solution of the present invention is as follows: This invention provides a highly flexible, multi-linear, hybrid, self-adaptive adhesive spraying method, which includes the following steps: Step S1: Generate a spraying path that satisfies G2 continuity based on the user's interactive selection of points or input of coordinates in the world coordinate system; Step S2: Using linewidth W and cross-sectional height H as the outer loop control targets, the deviation E between (W,H) and the target (W*,H*) is measured in real time by an online vision sensor or line laser sensor. The inner loop uses motion speed v and injection valve duty cycle D as interlocking adjustment values. Based on the data-driven mapping f:(v,κ,T,η,P)→(W,H), Δv and ΔD are calculated in real time. v(s) and D(s) are set according to the curvature-driven collaborative planning rules. The control commands are set to v_cmd(s)=sat(v(s)+Δv) and D_cmd(s)=sat(D(s)+ΔD), where sat(·) indicates amplitude saturation constraint based on the upper / lower limit of the device. in, Dynamic baseline velocity v(s) = min(v_max, sqrt(k_c / (|κ(s)|+ε))), Duty cycle D(s) = D0 + a1·v(s) + a2·|κ(s)| + a3· v / s; v_cmd and D_cmd represent the command output speed and duty cycle, respectively; sat(·) represents the amplitude saturation function. Step S3: Trigger S-shaped speed-duty cycle binding gradient control at the endpoints and corners, and superimpose the line segment indentation δ_in and line segment offset δ_off to reduce glue buildup and uneven end cross-section or abrupt line width changes. Step S4: Using a uniform style generator, solid lines, dashed lines, dotted lines, or dotted-dash lines are mixed and sprayed on a single line segment by section, and adjacent opening and closing actions meet the constraint of minimum opening and closing time τ_min, so as to maintain the stability of adhesive spraying formation when the production volume changes. Where: s is the arc length parameter; κ(s) represents the path curvature; v_max is the maximum allowable speed; k_c is the speed-curvature coupling coefficient; ε is a small positive number to avoid division by zero; D0, a1, a2, a3 are empirical parameters; T is the ambient or colloid temperature; η is the colloid viscosity; P is the colloid supply pressure; δ_in is the inward indentation of the line segment end; δ_off is the lateral offset of the line segment; τ_min is the minimum opening and closing time of the injection valve; E=[W W*, H H*], sqrt represents the square root, and v / s represents the first derivative of velocity with respect to arc length s, min takes the smaller of the two, and |·| is the absolute value.

[0011] Furthermore, the data-driven mapping f is modeled by response surface regression or Gaussian process regression, and the Jacobian matrix components are calculated at runtime through Jacobian approximation or local linearization. W / v, H / v, W / D, H / D) Using closed-loop correction of v and D, the model of the data-driven mapping f is updated online in an adaptive manner as temperature T, viscosity η, and adhesive supply pressure P change.

[0012] Furthermore, the endpoints and corner areas adopt a gradient window of length L_end, and set the pre-filling time t_pre and the glue removal time t_post, so that the speed and duty cycle change synchronously according to the same gradual law to avoid end point bulging or corner glue accumulation. t_pre is 5ms~33ms, t_post is 5ms~27ms, and L_end is no more than 20mm.

[0013] Furthermore, the uniform style generator is configured to generate at least one spraying style per segment on the same line segment, including: solid lines; dashed lines (short stroke length Ld and gap length Lg); dotted lines (dot spacing LP and opening / closing duty cycle τ); or dotted-dash lines (short stroke length Ld and opening / closing duty cycle τ); and a cosine transition window is applied at the style switching position to meet the constraint of the minimum valve opening / closing time τ_min.

[0014] Furthermore, the line sequence self-assignment module constructs a directed graph from the user-selected points. Using the weighted sum of inter-segment distance, start-up compensation amount, corner compensation amount, and fixture repositioning time as the optimization objective, it solves the shortest Hamiltonian path, or uses a minimum spanning tree combined with a local 2-opt approximation algorithm to determine the line segment spraying sequence, direction, and splicing point compensation, thereby shortening the cycle time and reducing switching defects.

[0015] Furthermore, curvature programming employs a clothoid curve to achieve a continuous G2 transition. Given the endpoint positions, endpoint tangent vectors, and endpoint curvatures, the parameter pairs (A, B) of the curvature function κ(s) = A·s + B are solved; and the error function M = ||path_gen|| is used. Minimize path_target||² to obtain the final spraying path. Where path_gen is the generated spiral path, and path_target is the target spraying path; And when |κ(s)|≥κ_th, in the interval [s0 Δ,s0+Δ] triggers corner protection and superimposes δ_in and δ_off, where κ_th≥1 / R_th, R_th is the minimum allowable radius of curvature and takes a value of 2mm~10mm, s0 is the arc length position that satisfies |κ(s)|=κ_th, and Δ is the half width of corner protection.

[0016] Furthermore, during the calibration phase, the system collects and records (W,H) multi-condition samples of v∈[v_min, v_max], κ∈[0,κ_max], T, η, and P, and constructs an initial model of the data-driven mapping f using response surface regression or Gaussian process regression. During the operation phase, the system performs fast queries using kd-trees or uniform grids and manages them by establishing a red-black tree index based on parameter version information to support rapid switching between different materials or nozzles. Where v is the velocity, κ is the curvature, T is the temperature, η is the viscosity, P is the glue supply pressure, W is the line width, and H is the cross-sectional height; Based on the above samples, an initial model for the data-driven mapping f is constructed using either a response surface regression model or a Gaussian process regression model.

[0017] The present invention also provides a high-freedom-dimensional multi-linear hybrid self-adaptive glue spraying system, which includes: a spray valve, a motion actuator and encoder, a glue supply pressure and temperature control module, an online vision sensor or a line laser sensor, a controller and a memory; The controller is configured to execute the aforementioned high-freedom-dimensional multilinear hybrid self-adaptive glue spraying method and communicate with the aforementioned components via a bus to achieve curvature-driven speed-duty cycle collaborative planning, endpoint / corner gradient control, and gated output of the unified style generator.

[0018] The controller includes: a first processing unit for path generation and curvature planning; a second processing unit for online correction of dual-loop adaptive control and data-driven mapping f; a third processing unit for line sequence self-allocation and splicing point compensation; and a fourth processing unit for pattern gating and minimum opening / closing time constraints; the memory stores the calibration dataset. The style dictionary S and parameter group {D0,a1,a2,a3,τ_min,k_c,ε,v_min,v_max,κ_max,R_th,Ld,Lg,LP,L_end,t_pre,t_post,δ_in,δ_off}.

[0019] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the processor performs the above-described high-freedom-dimensional multilinear hybrid self-adaptive adhesive spraying method.

[0020] The technical solution of this invention has the following beneficial effects: 1. Significantly improved dimensional stability of the forming process By employing a dual closed-loop strategy of "outer loop (W,H) target - inner loop (v,D) interlock", and incorporating temperature T, viscosity η, supply pressure P and path curvature κ(s) into the online correction of mapping f, the linewidth W and cross-sectional height H can be maintained stably and converged even when the material and environment drift slowly, thus reducing the impact of batch fluctuations on yield.

[0021] 2. Effective suppression of adhesive buildup and overflow in corner / small radius areas. Introducing a "curvature-driven" speed-duty cycle co-planning v(s) / D(s), and enabling corner protection windows and geometric compensation (δ_in, δ_off) when |κ(s)|≥ threshold, combined with endpoint gradient control (t_pre, t_post), can significantly reduce corner bulges, trailing, and "uneven head" phenomena, and improve appearance continuity.

[0022] 3. Endpoint quality is controllable and repeatability is good. The endpoints / joints adopt a uniform S-shaped gradient law to bind v and D, and superimpose the pre-filling and glue collection sequence to make the glue amount transition between the beginning and end smooth, reduce the sensitivity of under-spraying / over-spraying at the start and stop, and facilitate cross-tooling and cross-machine reproduction.

[0023] 4. Complex linear "one-piece" forming and stable gating The unified style generator completes parametric synthesis of styles such as solid lines, dashed lines, dotted lines, and dotted-dash lines on a single line segment, and applies a transition window to the valve opening and closing to meet the device constraint of minimum opening and closing time τ_min, thereby avoiding the additional start-stop, alignment errors and local glue build-up caused by "segmentation + multiple program segments". Attached Figure Description

[0024] Figure 1 This is a logic block diagram of the present invention; Figure 2 This is an overall flowchart of the present invention; Detailed Implementation The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention. It should also be noted that, for ease of description, only the parts related to the present invention are shown in the accompanying drawings, and not all of them.

[0025] Combination Figures 1-2 As shown, this invention provides a highly flexible, multi-linear, hybrid, self-adaptive adhesive spraying method, which includes the following steps: Step S1: Generate a spraying path that satisfies G2 continuity based on the user's interactive selection of points or input of coordinates in the world coordinate system; Step S2: Using linewidth W and cross-sectional height H as the outer loop control targets, the deviation E between (W,H) and the target (W*,H*) is measured in real time by an online vision sensor or line laser sensor. The inner loop uses motion speed v and injection valve duty cycle D as interlocking adjustment values. Based on the data-driven mapping f:(v,κ,T,η,P)→(W,H), Δv and ΔD are calculated in real time. v(s) and D(s) are set according to the curvature-driven collaborative planning rules. The control commands are set to v_cmd(s)=sat(v(s)+Δv) and D_cmd(s)=sat(D(s)+ΔD), where sat(·) indicates amplitude saturation constraint based on the upper / lower limit of the device. in, Dynamic baseline velocity v(s) = min(v_max, sqrt(k_c / (|κ(s)|+ε))), Duty cycle D(s) = D0 + a1·v(s) + a2·|κ(s)| + a3· v / s; v_cmd and D_cmd represent the command output speed and duty cycle, respectively; sat(·) represents the amplitude saturation function. Step S3: Trigger S-shaped speed-duty cycle binding gradient control at the endpoints and corners, and superimpose the line segment indentation δ_in and line segment offset δ_off to reduce glue buildup and uneven end cross-section or abrupt line width changes. Step S4: Using a uniform style generator, solid lines, dashed lines, dotted lines, or dotted-dash lines are mixed and sprayed on a single line segment by section, and adjacent opening and closing actions meet the constraint of minimum opening and closing time τ_min, so as to maintain the stability of adhesive spraying formation when the production volume changes. Where: s is the arc length parameter; κ(s) represents the path curvature; v_max is the maximum allowable speed; k_c is the speed-curvature coupling coefficient; ε is a small positive number to avoid division by zero; D0, a1, a2, a3 are empirical parameters; T is the ambient or colloid temperature; η is the colloid viscosity; P is the colloid supply pressure; δ_in is the inward indentation of the line segment end; δ_off is the lateral offset of the line segment; τ_min is the minimum opening and closing time of the injection valve; E=[W W*, H H*], sqrt represents the square root, and v / s represents the first derivative of velocity with respect to arc length s, min takes the smaller of the two, and |·| is the absolute value.

[0026] The data-driven mapping f is modeled by response surface regression or Gaussian process regression, and the Jacobian matrix components are calculated at runtime through Jacobian approximation or local linearization. W / v, H / v, W / D, H / D) Using closed-loop correction of v and D, the model of the data-driven mapping f is updated online in an adaptive manner as temperature T, viscosity η, and adhesive supply pressure P change.

[0027] The endpoints and corner areas use a gradient window of length L_end, and set the pre-filling time t_pre and the glue removal time t_post to make the speed and duty cycle change synchronously according to the same gradual law to avoid end point bulging or corner glue accumulation. t_pre is 5ms~33ms, t_post is 5ms~27ms, and L_end is no more than 20mm.

[0028] The uniform style generator is configured to generate at least one spraying style per segment on the same line segment, including: solid lines; dashed lines (short stroke length Ld and gap length Lg); dotted lines (dot spacing LP and opening / closing duty cycle τ); or dotted-dash lines (short stroke length Ld and opening / closing duty cycle τ); and a cosine transition window is applied at the style switching position to meet the constraint of the minimum valve opening / closing time τ_min.

[0029] The line sequence self-assignment module constructs a directed graph from user-selected points. It uses the weighted sum of inter-segment distance, start-up compensation, corner compensation, and fixture repositioning time as the optimization objective to solve for the shortest Hamiltonian path. Alternatively, it employs a minimum spanning tree combined with a local 2-opt approximation algorithm to determine the line segment spraying sequence, direction, and splicing point compensation, thereby shortening the cycle time and reducing switching defects.

[0030] Curvature programming employs a Clothoid curve to achieve a continuous G2 transition. Given the endpoint positions, endpoint tangent vectors, and endpoint curvatures, the parameter pairs (A, B) of the curvature function κ(s) = A·s + B are solved; the error function M = ||path_gen|| is used. Minimize path_target||² to obtain the final spraying path. Where path_gen is the generated spiral path, and path_target is the target spraying path; And when |κ(s)|≥κ_th, in the interval [s0 Δ,s0+Δ] triggers corner protection and superimposes δ_in and δ_off, where κ_th≥1 / R_th, R_th is the minimum allowable radius of curvature and takes a value of 2mm~10mm, s0 is the arc length position that satisfies |κ(s)|=κ_th, and Δ is the half width of corner protection.

[0031] During the calibration phase, the system collects and records (W,H) multi-condition samples of v∈[v_min, v_max], κ∈[0,κ_max], T, η, and P, and constructs an initial model of data-driven mapping f using response surface regression or Gaussian process regression. During the operation phase, the system performs fast queries using kd-trees or uniform grids and manages them by establishing a red-black tree index based on parameter version information to support rapid switching between different materials or nozzles. Where v is the velocity, κ is the curvature, T is the temperature, η is the viscosity, P is the glue supply pressure, W is the line width, and H is the cross-sectional height; Based on the above samples, an initial model for the data-driven mapping f is constructed using either a response surface regression model or a Gaussian process regression model.

[0032] The present invention also provides a high-freedom-dimensional multi-linear hybrid self-adaptive glue spraying system, which includes: a spray valve, a motion actuator and encoder, a glue supply pressure and temperature control module, an online vision sensor or a line laser sensor, a controller and a memory; The controller is configured to execute the aforementioned high-freedom-dimensional multilinear hybrid self-adaptive glue spraying method and communicate with the aforementioned components via a bus to achieve curvature-driven speed-duty cycle collaborative planning, endpoint / corner gradient control, and gated output of the unified style generator.

[0033] The controller includes: a first processing unit for path generation and curvature planning; a second processing unit for online correction of dual-loop adaptive control and data-driven mapping f; a third processing unit for line sequence self-allocation and splicing point compensation; and a fourth processing unit for pattern gating and minimum opening / closing time constraints; the memory stores the calibration dataset. The style dictionary S and parameter group {D0,a1,a2,a3,τ_min,k_c,ε,v_min,v_max,κ_max,R_th,Ld,Lg,LP,L_end,t_pre,t_post,δ_in,δ_off}.

[0034] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the processor performs the above-described high-freedom-dimensional multilinear hybrid self-adaptive adhesive spraying method.

[0035] Example 1: like Figure 1 As shown, the adhesive spraying technology proposed in this application includes six unique technologies, including custom multi-point trajectory technology, self-assigned line sequence technology, custom / self-adaptive dynamic spraying rate technology, multi-line shape self-selection mixing technology, spraying curvature planning technology, and multi-select line segment compensation scheme technology.

[0036] The system includes several key technologies: Custom multi-point trajectory technology, which allows users to select points on the world coordinate grid or by entering world coordinates to outline the entire sprayed graphic; self-assigned line sequence technology, which analyzes the order of point selection, closure relationships, and vector directions to assign key data such as line sequence and direction to each line segment; custom / adaptive dynamic spraying rate technology, which allows users to choose a suitable option from two schemes to determine whether the spraying speed is executed at a fixed value or dynamically adapted; multi-line type self-selection and mixing technology, which allows users to select or uniformly choose from several line types (straight line, dashed line, dotted line, dotted-dash line) provided by the system, enabling the system to more clearly classify and implement the attributes of each line segment; spraying curvature planning technology, which uses the curvature value (unit: °) entered by the user to automatically plan line segments that conform to the curvature value using geometric principles and polar coordinate system algorithms, enabling free drawing of curves and straight lines; and multi-segment compensation scheme technology, which includes three schemes and allows all to be enabled. One solution is a spray indentation technology, which allows the user to input independent indentation values ​​at both ends, affecting the entire network and simultaneously changing the indentation values ​​at both ends of all line segments, with multiple layers of protection. Another solution is a line segment offset compensation technology, which allows the user to input the line segment offset, affecting the entire network and simultaneously changing the starting spray position of all line segments, with multiple layers of protection. Finally, a line segment endpoint pulse width modulation (PWM) technology solution allows the user to configure the duty cycle of the adhesive at both ends of the line segment, thereby controlling the amount of adhesive at both ends and alleviating the problem of uneven adhesive application.

[0037] Therefore, considering all factors, the technical solution of this application adopts "measurable closed-loop control + curvature-driven speed and duty cycle co-planning + endpoint / corner gradient control + unified style generator", aiming to form clear and reproducible technical features and effects at the system and algorithm levels. The specific steps are as follows: A. Dual closed-loop self-adaptive glue spraying control (speed-glue quantity coordination) Outer Loop: Using line width W and cross-sectional height H as trajectory indicators, a target (W*, H*) is set. The (W, H) parameters after pixel coordinate transformation are obtained through online visual scanning, and the error E = [W] is calculated. W*, H H*].

[0038] Inner loop: Using encoder speed v and solenoid valve duty cycle D as interlocking quantities, a response surface model or Gaussian process regression f:(v, κ, T, η, P)→(W, H) is constructed, where κ is curvature, T is temperature, η is viscosity, and P is adhesive supply pressure. Δv and ΔD are solved using the Jacobian approximation of f to achieve real-time correction.

[0039] Start, stop, and finish: Set the t_pre pre-filling time and t_post finish time. When passing the endpoint, use "speed-bound duty cycle technology", that is, the S-shaped speed planning and duty cycle change proportionally to avoid "large and small ends".

[0040] B. Coordinated planning of speed and duty cycle for curvature spraying Dynamic baseline speed: v(s) = min(v_max, sqrt((k_c) / (|κ(s)|+ε))); The greater the curvature of the line segment, the lower the spraying speed.

[0041] Duty cycle: D(s) = D0 + a1·v(s) + a2·|κ(s)| + a3· v / s; The parameters {D0,a1,a2,a3} are obtained through calibration and fine-tuned in the closed loop.

[0042] Corner protection: When |κ(s)|>κ_th, introduce a corner window [s0] Δ,s0+Δ], linear or cosine gradually varying v and D, superimposed line segment inward contraction δ_in and offset δ_off, to ensure that the jet channels do not overlap.

[0043] C. Unified multi-line style generator (programmable implementation of mixed line segments) Define a style dictionary S={SOLID, DASHED(Ld, Lg), DOTTED(LP,τ), DASH_DOT(Ld, τ)}, where the parameters represent solid lines, dashed lines / intervals, dotted line periods / pulse widths, and dotted-dash line combinations, respectively.

[0044] A temporal gating mechanism G(u)∈{0,1} is generated on the parameterized coordinates u∈[0,1] of the line segment, and a transition window h(u)=0.5(1) is applied at the style transition point. cos(πu)) to suppress the glue buildup caused by step jump.

[0045] Mixed segment combinations: Given the segment division {[u_i, u_{i+1}]} and the pattern S_i, generate G_i(u) and concatenate them, while constraining continuity and minimum opening and closing time not less than τ_min.

[0046] D. Graph optimization implementation of line order self-assignment The user-selected points are used to form a directed graph. The weights of the points are composed of start compensation, inter-segment distance, corner compensation, and fixture repositioning time. The shortest Hamiltonian path is solved or approximated by MST+ local 2-opt. The line sequence, direction, and splicing point compensation are output.

[0047] E. Implementation details of curvature planning The G2 continuous curvature transition is achieved by using the Clothoid curve (Euler Spiral); given the endpoints and tangent vector, the (A,B) parameters are solved to make the curvature κ(s) = A·s+B.

[0048] In curvature programming, the polar coordinate / geometric method is only used for initial values, and the final value is expressed as the error function M = ||path_gen||. Minimize path_target‖^2.

[0049] F. Calibration and Data Structures Initial calibration sampling grid: v∈[v_min,v_max], κ∈[0,κ_max], collecting data under various operating conditions including temperature, viscosity, and glue supply pressure, with data collected in (W,H). f is constructed using response surface regression, and the query structure employs a kd-tree or uniform grid; red-black trees are used only as parameter version indexes.

[0050] G. Recommendations for some key parameters Endpoint gradient window values: L_end <= 20mm, t_pre = 5–33 ms, t_post = 5–27 ms; Minimum valve opening / closing cycle: τ_min ≥ 10 ms (depending on the frequency of the spray gun valve); Corner radius threshold: R_th = 2–10 mm (|κ| ≥ 1 / R_th triggers corner strategy).

[0051] Example 2: "Solid line + small radius transition" spray adhesive for sealing the frame of 3C electronic mobile phones 1) Equipment composition and calibration Equipment: spray valve (solenoid valve), multi-axis motion platform and encoder, pressure supply and temperature control components, line laser / online vision measurement W and H, and industrial controller and memory; the controller implements an adaptive glue spraying method (dual closed loop + curvature drive + gradient compensation + pattern gating).

[0052] Initial calibration: A multi-condition dataset {(v, κ, T, η, P) → (W, H)} was collected on a grid with v∈[v_min, v_max] and κ∈[0, κ_max], and f was constructed using response surface regression; online queries were accelerated using kd-trees or uniform grids.

[0053] 2) Path generation and curvature planning The operator interactively selects points in the world coordinate system to outline the border, and the system generates a spraying path that satisfies G2 continuity. Small radius corners use a clothoid curve to achieve a linear curvature transition, and the parameters are obtained by minimizing the error function E.

[0054] 3) Collaborative planning and dual closed-loop control With (W*, H*) as the outer loop target and (v, D) as the interlocking quantity of the inner loop, v(s) and D(s) are set according to the "curvature drive" principle and corrected by the measured (W, H) closed loop.

[0055] The f model is used to map (v, κ, T, η, P) to (W, H) and supports online correction.

[0056] 4) Gradation and geometric compensation at endpoints / corners In the corner area between the endpoint and the small R, an S-shaped gradient window L_end is used, and compensation for the inward shrinkage δ_in / offset δ_off at both ends is superimposed; in this embodiment, L_end=10 mm, t_pre=12 ms, and t_post=15 ms are preferably set, all of which fall within the recommended range; when R≤R_th (e.g., R_th=2–10 mm), the corner area strategy is automatically triggered.

[0057] 5) Results and Key Points In solid line mode, the glue path of the mobile phone frame is closed, with no obvious glue buildup / tailing at the corners, and the fluctuations of W and H are controlled; the entire process meets the minimum valve opening and closing time constraint (τ_min≥10 ms).

[0058] Example 3: "Dot-and-line mixed + line sequence optimization" spray adhesive for ventilation bonding of automotive interior parts 1) Tasks and Styles For a single long line segment, "dotted lines" and "dashed lines" should be mixed in different sections to balance breathability and bonding strength; the system calls a unified style generator on the same line segment to define a style dictionary. S={SOLID, DASHED(Ld, Lg), DOTTED(LP, τ), DASH_DOT(Ld, τ)}, generates a gated sequence G(u)∈{0,1} at parameterized coordinates u∈[0,1], and applies a transition window h(u)=0.5(1) at the style transition. cosπu), ensuring continuity and stability.

[0059] 2) Valve constraint and gating implementation All opening and closing operations are constrained by the minimum opening and closing time τ_min; when the style pitch approaches τ_min, the transition window suppresses local glue buildup and stringing caused by hard switching.

[0060] 3) Path / Speed / Duty Cycle Control The path is transitioned using Clothoid to ensure G2 continuity; v(s) and D(s) are co-planned using curvature κ(s), and then fine-tuned by the outer loop (W, H) to achieve consistent shaping of different style segments.

[0061] 4) Line sequence self-assignment and beat Multiple interior trim ribs require continuous adhesive spraying; the system constructs a directed graph of global line segments, using "inter-segment distance, start / turn angle compensation, and fixture repositioning time" as weights to solve for the shortest Hamiltonian path or approximate it with MST+2-opt, to obtain the optimal line sequence / direction / splitting point configuration, reducing idle travel and lowering splicing visibility.

[0062] 5) Results and Key Points In a single-segment path, a combination of dotted and dashed lines is used, resulting in a consistent appearance, stable valve opening and closing, and a significantly shorter cycle time compared to "segmentation + multiple program segments".

[0063] Example 4: Corner Protection + Endpoint PWM Spraying for High-Density Patterns with Small Radius Corners 1) Scene and threshold settings There are many corners and short line segment splices with R=3–5 mm on small structural components; the system sets the corner area trigger threshold R_th=2–10 mm (the corner area strategy is triggered when |κ|≥1 / R_th), and uses the endpoint gradient window L_end and t_pre / t_post.

[0064] 2) Corner protection and geometric / process compensation In each corner region [s0 Δ, s0+Δ]: a) Bind the S-shaped gradient of v(s) and D(s) to reduce transients; b) Geometric compensation correction of the indentation δ_in and offset δ_off of the superimposed line segment to ensure that the amount of glue at the beginning and end coincides with the path; c) Enable endpoint PWM technology to configure the duty cycle at both ends independently and fine-tune the "big head / small head".

[0065] 3) Path and Control The path uses Clothoid to generate a G2 continuous transition; v(s) and D(s) are set according to the curvature-driven principle, and (W, H) is used as the outer loop target for real-time correction; in this embodiment, L_end=12 mm, t_pre=10 ms, and t_post=20 ms are selected, all of which are within the recommended range.

[0066] 4) Data / Model Management Calibration and online self-adaptation use a response surface model f and are quickly queried via a kd tree / mesh; when switching materials or nozzles, only a small sample update is needed to restore the target (W, H) output.

[0067] 5) Results and Key Points The small R continuous corners have no obvious bulges or overflows, the endpoint dimensions are consistent, the short line segments are spliced ​​and the transition is smooth, satisfying the τ_min constraint.

[0068] The above description is only a preferred embodiment of the present invention and does not limit the patent scope of the present invention. All equivalent structural transformations made under the inventive concept of the present invention using the contents of the present invention specification and drawings, or direct / indirect applications in other related technical fields, are included within the protection scope of the present invention.

Claims

1. A highly flexible, multi-linear, hybrid, self-adaptive adhesive spraying method, characterized in that, The method includes the following steps: Step S1: Generate a spraying path that satisfies G2 continuity based on the user's interactive selection of points or input of coordinates in the world coordinate system; Step S2: Using linewidth W and cross-sectional height H as the outer loop control targets, the deviation E between (W,H) and the target (W*,H*) is measured in real time by an online vision sensor or line laser sensor. The inner loop uses motion speed v and injection valve duty cycle D as interlocking adjustment variables. A data-driven mapping f:(v,κ,T,η,P)→(W,H) is established using a response surface regression model or Gaussian process regression model. Where κ is the path curvature, T is the temperature, η is the viscosity, and P is the adhesive supply pressure; and during operation, the Jacobian matrix components are calculated through Jacobian approximation or local linearization. W / v, H / v, W / D, H / D) Correct v and D using closed-loop adjustment. Δv and ΔD are solved using the Jacobian approximation of f; v(s) and D(s) are set according to the curvature-driven co-programming rules. The control commands are set to v_cmd(s)=sat(v(s)+Δv) and D_cmd(s)=sat(D(s)+ΔD), where sat(·) indicates amplitude saturation constraint based on the upper / lower limit of the device. in, Dynamic baseline velocity v(s) = min(v_max, sqrt(k_c / (|κ(s)|+ε))), Duty cycle D(s) = D0 + a1·v(s) + a2·|κ(s)| + a3· v / s; v_cmd and D_cmd represent the command output speed and duty cycle, respectively; sat(·) represents the amplitude saturation function. Step S3: Trigger S-shaped speed-duty cycle binding gradient control at the endpoints and corners. The S-shaped speed-duty cycle binding gradient control means that the endpoints and corners adopt a gradient window of length L_end, and set the pre-filling time t_pre and the glue removal time t_post, so that the speed and duty cycle change synchronously according to the same gradual change law; and superimpose the line segment end indentation δ_in and the line segment lateral offset δ_off to reduce glue buildup and uneven end cross-section or sudden changes in line width. Step S4: Using a uniform style generator, solid lines, dashed lines, dotted lines, or dotted-dash lines are mixed and sprayed on a single line segment by section. The uniform style generator is configured to generate at least one spraying style on the same line segment by section, and a cosine transition window is applied at the style switching position to meet the constraint of the minimum opening and closing time τ_min of the valve, so as to maintain the stability of the spray adhesive forming when the production volume changes. Where: s is the arc length parameter; κ is the path curvature; v_max is the maximum permissible speed; k_c is the speed-curvature coupling coefficient; ε is a small positive number to avoid division by zero; D0, a1, a2, a3 are duty cycle calculation parameters obtained through calibration; δ_in is the inward indentation of the line segment end; δ_off is the lateral offset of the line segment; τ_min is the minimum opening and closing time of the injection valve; E=[W W*, H H*], sqrt represents the square root, and v / s represents the first derivative of velocity with respect to arc length s, min takes the smaller of the two, and |·| is the absolute value.

2. The method according to claim 1, characterized in that, The data-driven mapping f is updated online automatically when the temperature T, viscosity η, and adhesive supply pressure P change.

3. The method according to claim 1, characterized in that, The endpoints and corner areas use a gradient window of length L_end, and set the pre-filling time t_pre and the glue removal time t_post to make the speed and duty cycle change synchronously according to the same gradual law to avoid end point bulging or corner glue accumulation. t_pre is 5ms~33ms, t_post is 5ms~27ms, and L_end is no more than 20mm.

4. The method according to claim 1, characterized in that, The uniform style generator is configured to generate at least one spraying style per segment on the same line segment, including: solid lines; dashed lines, the dashed lines including a short stroke length Ld and a gap length Lg; dotted lines, the dotted lines including a dot spacing LP and an opening / closing duty cycle τ; or dotted-dash lines, the dotted-dash lines including a short stroke length Ld and an opening / closing duty cycle τ; wherein, a cosine transition window is applied at the style switching position to satisfy the constraint of the minimum valve opening / closing time τ_min.

5. The method according to claim 1, characterized in that, The line sequence self-assignment module constructs a directed graph from user-selected points. It uses the weighted sum of inter-segment distance, start-up compensation, corner compensation, and fixture repositioning time as the optimization objective to solve for the shortest Hamiltonian path. Alternatively, it employs a minimum spanning tree combined with a local 2-opt approximation algorithm to determine the line segment spraying sequence, direction, and splicing point compensation, thereby shortening the cycle time and reducing switching defects.

6. The method according to claim 1, characterized in that, Curvature programming employs a Clothoid curve to achieve a continuous G2 transition. Given the endpoint positions, endpoint tangent vectors, and endpoint curvatures, the parameter pairs (A, B) of the curvature function κ(s) = A·s + B are solved; the error function M = ||path_gen|| is used. Minimize path_target||² to obtain the final spraying path. Where path_gen is the generated spiral path, and path_target is the target spraying path; And when |κ(s)|≥κ_th, in the interval [s0 Δ,s0+Δ] triggers corner protection and superimposes δ_in and δ_off, where κ_th≥1 / R_th, R_th is the minimum allowable radius of curvature and takes a value of 2mm~10mm, s0 is the arc length position that satisfies |κ(s)|=κ_th, and Δ is the half width of corner protection.

7. The method according to claim 1, characterized in that, During the calibration phase, the system collects and records (W,H) multi-condition samples of v∈[v_min, v_max], κ∈[0,κ_max], T, η, and P, and constructs an initial model of data-driven mapping f using response surface regression or Gaussian process regression. During the operation phase, the system performs fast queries using kd-trees or uniform grids and manages them by establishing a red-black tree index based on parameter version information to support rapid switching between different materials or nozzles. Where v is the velocity, κ is the curvature, T is the temperature, η is the viscosity, P is the adhesive supply pressure, W is the line width, and H is the cross-sectional height.

8. A high-freedom-dimensional multi-linear hybrid self-adaptive spray adhesive system, characterized in that, The system includes: a spray valve, a motion actuator and encoder, a glue supply pressure and temperature control module, an online vision sensor or a line laser sensor, a controller and a memory; The controller is configured to perform the method described in any one of claims 1-7 and communicate with the aforementioned components via a bus to achieve curvature-driven speed-duty cycle co-planning, endpoint / corner gradient control, and gated output of the uniform style generator.

9. The system according to claim 8, characterized in that, The controller includes: a first processing unit for path generation and curvature planning; a second processing unit for online correction of dual-closed-loop adaptive control and data-driven mapping f; a third processing unit for line sequence self-allocation and splicing point compensation; and a fourth processing unit for pattern gating and minimum opening / closing time constraints. The memory stores the calibration dataset, the pattern dictionary S, and the parameter group {D0, a1, a2, a3, τ_min, k_c, ε, v_min, v_max, κ_max, R_th, Ld, Lg, LP, L_end, t_pre, t_ post, δ_in, δ_off}; where D0, a1, a2, a3 are the duty cycle calculation parameters obtained through calibration, τ_min is the minimum opening and closing time of the injection valve, k_c is the velocity-curvature coupling coefficient, ε is a small positive number to avoid division by zero, R_th is the minimum allowable radius of curvature, Ld is the short stroke length, Lg is the gap length, LP is the dot spacing, L_end is the gradient window length between the endpoint and the corner area, t_pre is the pre-filling time before spraying, t_post is the glue recovery time, δ_in is the inward shrinkage of the line segment end, and δ_off is the lateral offset of the line segment.

10. A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, causes the processor to perform the method according to any one of claims 1-7.

Images