Textile two-for-one twister inspection method based on multi-view real-time perception and adaptive path planning

Abstract

The invention relates to the technical field of equipment inspection, in particular to a textile two-for-one twister inspection method based on multi-view real-time perception and adaptive path planning, which comprises the following steps: step 1, building a virtual working map, and setting inspection points; setting a virtual wall in the virtual working map according to an obstacle identified by the robot in the inspection process, and updating the virtual wall at intervals; performing adaptive path planning according to the set virtual wall; and step 4, evaluating the plurality of adaptive paths generated in the step 3, selecting an optimal adaptive path, and carrying out routing inspection according to the optimal adaptive path. According to the method, the advanced visual technology, the artificial intelligence algorithm and the adaptive path planning strategy are integrated, so that efficient, safe and flexible intelligent routing inspection is realized.

Description

Technical Field

[0001] The present application relates to the technical field of equipment inspection, and in particular to a textile two-for-one twisting machine inspection method based on multi-perspective real-time perception and adaptive path planning. Background Art

[0002] The textile two-for-one twister is a key equipment in the textile industry, mainly used to twist single yarn into two or more strands. This process is essential to improve the strength, elasticity and uniformity of the yarn, and also affects the quality and performance of the final textile. In modern textile production, the stable operation of the two-for-one twister and product quality control are directly related to the efficiency and product quality of the entire production line.

[0003] Regular inspection is an important means to ensure the normal operation of the twisting machine. This work is mainly done manually. The staff needs to regularly inspect the workshop, check the operation of the machine, and promptly find and deal with abnormal situations, such as yarn breakage, winding rollers, and wire edge clamping. However, the noise in the textile workshop is relatively high, and there are many flying catkins in the air. On the one hand, it affects the line of sight and the accuracy of visual inspection is poor; on the other hand, the workshop is large, the inspection process is long, and the inspection efficiency is low. Summary of the invention

[0004] The technical problem to be solved by the present invention is that the existing inspection method for two-for-one twisting machines has low inspection efficiency.

[0005] To this end, the present invention provides a textile two-for-one twisting machine inspection method based on multi-perspective real-time perception and adaptive path planning.

[0006] The technical solution adopted by the present invention to solve its technical problem is:

[0007] A textile two-for-one twisting machine inspection method based on multi-view real-time perception and adaptive path planning includes the following steps:

[0008] Step 1: Build a virtual work map and set inspection points;

[0009] Step 2: According to the obstacles identified by the robot during the inspection process, a virtual wall is set in the virtual work map, and the virtual wall is updated at intervals;

[0010] Step 3: Perform adaptive path planning based on the set virtual wall;

[0011] Step 4: Evaluate the multiple adaptive paths generated in step 3, select the optimal adaptive path, and perform inspection according to the optimal adaptive path.

[0012] Furthermore, in step 2, the obstacle is an employee, and the properties of the virtual wall are dynamically calculated by real-time analysis of the employee's position, moving speed, and predicted trajectory. The properties of the virtual wall include position, length, and duration.

[0013] Furthermore, in step two, the safety zone is calculated considering the employee's moving speed and direction, by first setting a basic safety distance and then adjusting it according to the employee's current speed.

[0014] Furthermore, in step 2, the basic time limit of the virtual wall is set based on the movement pattern of the employees and the working status of the workshop.

[0015] Furthermore, the virtual wall foundation time is T Final , the initial foundation time of the virtual wall is preset to be T Base , virtual wall foundation time T Final The calculation formula is as follows:

[0016]

[0017] Among them, V staffmax V is the maximum moving speed of the employee; staff is the current employee moving speed, in m / s; k staff L is the factor affecting employee moving speed; workmax is the maximum workshop activity index; L work is the current workshop activity index; k work is the influencing factor of workshop activity.

[0018] Furthermore, in step three, an improved A* algorithm is used for path design, and combined with a genetic algorithm for global optimization.

[0019] Furthermore, the improved A* algorithm is expressed as: f(n)=g(n)+h(n)+d(n), where f(n) is the total estimated cost of node n; g(n) is the actual cost from the starting node to node n; h(n) is the estimated cost from node n to the target node, and d(n) is the newly added dynamic constraint cost item.

[0020] Furthermore, the dynamic constraint cost term d(n)=d Safety (n)+d Speed (n), where d Safety( n) is the safety distance constraint, d Speed (n) is the speed limit.

[0021] Furthermore, safety (n) = k_s*[max(0,D safety -D1)] 2, k_s is an adjustable weight factor, D safety is the predefined safety distance, D1 is the distance from node n to the nearest obstacle (including dynamic virtual wall); d speed (n) = k_s*[max(0,V current -V0)] 2 , k_v is another adjustable weight factor, V current is the current speed of the robot, and V0 is the speed limit of the area where node n is located.

[0022] Furthermore, in step 4, the candidate paths are evaluated by five key criteria: path length L, smoothness S, safety F, energy consumption E, and time efficiency T. The five criteria are standardized and the weighted sum method is used to calculate the comprehensive score of each candidate path. Finally, the optimal path is selected by comparing the scores.

[0023] The beneficial effect of the present invention is that the present invention proposes a textile two-for-one twister inspection method based on multi-perspective real-time perception and adaptive path planning, aiming to solve the problems of low efficiency, high safety risk, poor adaptability, etc. in traditional textile two-for-one twister inspection. By integrating advanced visual technology, artificial intelligence algorithms and adaptive path planning strategies, the present method realizes efficient, safe and flexible intelligent inspection.

[0024] Use multiple industrial cameras to capture the surrounding environment in real time, achieve multi-perspective real-time perception, and provide all-round real-time data for path planning. Based on the real-time perception results, dynamically generate virtual obstacles to ensure the safety of the inspection process. Finally, continuously optimize the inspection path based on real-time environmental changes and dynamic virtual walls.

[0025] This inspection method can adapt to the complex and changing factory environment in real time, greatly improve inspection efficiency and accuracy, and significantly reduce safety risks. It does not rely on fixed hardware facilities, but realizes intelligent inspection through the innovation of software algorithms, which has strong flexibility and scalability. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0027] Figure 1 It is a structural schematic diagram of the present invention.

[0028] Figure 2 It is a structural schematic diagram of the inspection robot used in the present invention.

[0029] Figure 3 This is a flow chart of dynamic virtual wall generation in the present invention.

[0030] Figure 4 It is a flow chart of the adaptive path planning algorithm in the present invention.

[0031] In the figure: 1. base; 2. controller; 3. bracket; 4. camera. DETAILED DESCRIPTION

[0032] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner, and therefore only show the components related to the present invention.

[0033] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise", "axial", "radial", "circumferential" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. In addition, features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, unless otherwise specified, "multiple" means two or more.

[0034] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0035] A textile two-for-one twisting machine inspection method based on multi-view real-time perception and adaptive path planning includes the following steps:

[0036] Step 1: Set up the working environment

[0037] S11 environment construction

[0038] When the factory is in normal operation, the laser radar and image acquisition equipment are used to deeply scan the entire working area. An accurate three-dimensional environmental map is established to provide a basis for subsequent path planning.

[0039] S12 Inspection Point Settings

[0040] Mark key inspection points on the constructed map to determine the target location of the inspection. The inspection points include the key parts of each two-for-one twister, and link the inspection points with the background production management system to ensure that all key equipment is covered. It can be flexibly adjusted according to production needs, and the coverage rate reaches 100%.

[0041] Step 2: Initial inspection

[0042] The inspection is carried out using a twister inspection robot, which includes a base 1, a controller 2, a bracket 3 and a visual system. The base 1 is equipped with a lithium battery pack, which can work continuously for 15 hours and is rechargeable. The bottom of the base 1 is equipped with universal wheels, which can move freely in a plane. The base 1 is also equipped with a laser radar for building the environmental map in step one.

[0043] The controller 2 is arranged on the base 1. The controller 2 is equipped with an embedded industrial control host with a Core 5th generation independent graphics card. The bracket 3 is connected to the base 1. The visual system includes four high-definition industrial cameras 4. The four high-definition industrial cameras 4 are grouped in pairs. The two groups of high-definition industrial cameras 4 are arranged on the bracket 3 in a vertical direction. The two cameras 4 in each group of high-definition industrial cameras 4 are located in the same plane, and the lenses of the two cameras 4 are set in opposite directions.

[0044] According to the marking order of the inspection points, the inspection is carried out in sequence. The four cameras 4 work simultaneously, and the twister image is collected every 50ms. Then the image is preprocessed: including denoising, enhancement and other operations, which takes about 10ms. The improved YOLOv5 algorithm is used to identify the obstacles captured in the image (because the obstacles that the robot will encounter in most cases in the textile twister workshop are only employees, and the focus is on identifying personnel), and the detection time is <20ms / frame.

[0045] It should be noted that the inspection robot maintains a speed of 0.5m / s; when an obstacle is detected or it approaches a detection point, the robot maintains a speed of 0.2m / s; when the robot reaches the detection point, it stops moving and takes all-round photos and analyses. If any abnormality is found, it will be reported immediately and a decision will be made as to whether to continue the inspection based on the severity.

[0046] Step 3: Generate dynamic virtual wall

[0047] S31 Setting Virtual Wall

[0048] When a person is detected between any two inspection points, the virtual wall is generated within 50ms, and the virtual wall status is updated every 500ms.

[0049] Specifically, in most cases, the only obstacle that the robot will encounter in a textile two-for-one twisting machine workshop is employees. Therefore, the dynamic virtual wall generation of this system is mainly aimed at the employees in the textile two-for-one twisting machine workshop.

[0050] The properties of the virtual wall (position, length, duration) are dynamically calculated by real-time analysis of the employee's position, movement speed, and predicted trajectory. After the camera 4 is calibrated, the real-time three-dimensional coordinates of the target person relative to the robot can be obtained through coordinate system conversion. The system uses a multi-view camera 4 to capture data. After image preprocessing, the YOLOV5 algorithm is applied for personnel detection. Subsequently, considering that the movement of personnel is nonlinear, a second-order Kalman filter for nonlinear motion is used for object tracking to accurately predict the employee's movement trajectory.

[0051] S32 Set safe area

[0052] The safety zone is calculated taking into account the speed and direction of the employee's movement. The system first sets a basic safety distance and then adjusts it based on the employee's current speed. For example, a fast-moving employee will trigger the system to generate a larger safety zone, while a stationary or slow-moving employee will correspond to a smaller safety zone.

[0053] S33 Setting basic time limit

[0054] Timeliness of the algorithm: The setting is based on the employee's movement pattern and the working status of the workshop. Usually, the system sets a basic timeliness for the detected employee, and then dynamically adjusts it according to the real-time situation. For example, during busy hours in the workshop, the timeliness of the virtual wall may be shortened to improve the responsiveness and flexibility of the system. The main purpose of setting the timeliness of the virtual wall is to ensure that the system can update the environmental information in a timely manner, while ensuring safety and not affecting the efficiency of path planning due to outdated virtual walls.

[0055] Specifically, the virtual wall foundation time is T Final First, the initial foundation time of the virtual wall is set to T Base , T Base It can be 60s, virtual wall foundation time T Final The calculation formula is as follows:

[0056]

[0057] Among them, V staffmax V is the maximum moving speed of the employee, which can generally be set to 2m / s; staff is the current employee moving speed, in m / s; k staff L is the factor affecting the employee's moving speed, which can be adjusted according to the actual situation; workmaxis the maximum workshop activity index, which is used to reflect the proportion of equipment in normal operation in the workshop. The value range is between 0 and 10, and the maximum workshop activity index is 10; L work is the current workshop activity index; k work This is the influencing factor of workshop activity and can be adjusted according to actual conditions.

[0058] When the workshop moves quickly for employees (V staff About 1.5m / s), high workshop activity (L work When k is about 7), staff Set to 0.5, k work Set to 0.3. In this case, due to the fast movement of employees and high activity in the workshop, the validity of the virtual wall is significantly shortened so that it can be updated more frequently to ensure the flexibility and security of the system. staff About 0.5m / s), low workshop activity (L work When k is about 3), staff Set to 0.7, k work Set to 0.5. In this case, due to the slow movement speed of employees and low workshop activity, the virtual wall has a longer validity period. When the workshop has the largest employee speed (V staff About 2m / s), the workshop activity is the highest (L work When k is about 10, staff Set to 0.3, k work It is set to 0.2. In this extreme case, the employee movement speed reaches the maximum value and the workshop activity is the highest, and the virtual wall has the shortest validity period to ensure that the system can respond to environmental changes in a timely manner.

[0059] It should be noted that when V staff When increasing, decreases, resulting in the entire aging T Final This is consistent with our expectation that the faster the workers move, the shorter the timeliness of the virtual wall should be, so that it can be updated more frequently;

[0060] Similarly, when L work When increasing, decrease, which will also lead to the aging T Final It is logical that the busier the shop floor is and the faster the environment changes, the more frequently we need to update the virtual wall.

[0061] Step 4: Adaptive path planning

[0062] After completing the virtual wall setting, adaptive path planning is performed. The virtual wall is updated every 500ms, and the adaptive path is completely replanned every 500ms. After the adaptive path is planned, inspections are performed according to the new adaptive path.

[0063] The planning principle of the adaptive path is to minimize the total inspection time while considering the detection coverage.

[0064] Specifically, an improved A* algorithm is used for path design, and a genetic algorithm is combined for global optimization. During the path planning process, the algorithm applies the improved A* algorithm to generate candidate paths. The key innovation here is how to deal with dynamic constraints, mainly including safety distance and speed limit. The safety distance constraint ensures that the robot maintains a sufficient distance from the employee, while the speed limit reduces the robot's movement speed when approaching the employee. These constraints are implemented by modifying the cost function of the A* algorithm, so that areas close to virtual walls or areas that need to slow down are given higher costs in path selection.

[0065] The traditional A* algorithm is a heuristic search algorithm that finds the optimal path by comprehensively considering the actual cost and estimated cost from the starting point to the target point. Its cost function is generally defined as:

[0066] f(n)=g(n)+h(n)

[0067] Where f(n) is the total estimated cost of node n; g(n) is the actual cost from the starting node to node n; h(n) is the estimated cost from node n to the target node

[0068] In the improved A* algorithm, in order to handle dynamic constraints (safe distance and speed limit), the cost function needs to be adjusted and the g(n) part needs to be modified. The modified cost function can be expressed as:

[0069] f(n)=g(n)+h(n)+d(n)

[0070] Where d(n) is the newly added dynamic constraint cost term. d(n) = d Safety (n)+d Speed (n)

[0071] d Safety( n) is the safety distance constraint, d safety (n) = k_s*[max(0,D safety -D1)] 2 , where k_s is an adjustable weight factor, D safety is the predefined safety distance, and D1 is the distance from node n to the nearest obstacle (including dynamic virtual wall).

[0072] dSpeed (n) is the speed limit, dspeed(n) = k_s*[max(0,Vcurrent-V0)] 2 , where k_v is another adjustable weight factor, V current is the current speed of the robot, and V0 is the speed limit of the area where node n is located.

[0073] The algorithm generates multiple candidate paths and then evaluates the quality of these paths based on factors such as path length, smoothness, and safety. The evaluation process not only considers static factors, but also predicts the possible movements of the employee to select the optimal path. The selected path is then smoothed using Bezier curves to ensure smooth motion to ensure smoother movement of the robot.

[0074] Specifically, the implementation process of the path evaluation method is:

[0075] Candidate paths are evaluated using five key criteria: path length L, smoothness S, safety F, energy consumption E, and time efficiency T.

[0076] In general, the shorter the path length L, the better; the curvature change of the path, that is, the smoother the smoothness S, the better; the smaller the distance to the obstacle (including dynamic virtual walls), the better the safety F; considering the smaller the slope change of the path and the fewer turns, the lower the energy consumption E; the speed limit and expected driving time are used to judge the time efficiency T.

[0077] In order to make the different standards comparable, each standard is normalized:

[0078] x Norm =(xx Min ) / (x Max -x Min )

[0079] Where x represents each standard, x Norm It is a standardized value, ranging between [0,1].

[0080] The weighted sum method is used to calculate the comprehensive score of each candidate path, and finally the optimal path is selected by comparing the scores:

[0081] Score = w L *L Norm +w S *S Norm +w F *F Norm +w E *E Norm +w T *T Norm

[0082] where w L , w S , w F , w E ,,w T is the weight of each criterion.

[0083] The adaptive path planning algorithm takes less than 200ms for a single planning, and the system performs a complete path replanning every 500ms to ensure that it can quickly respond to any changes in the environment. This high-frequency update allows the robot to flexibly avoid moving employees while maintaining an efficient inspection route.

[0084] In summary, the adaptive path planning algorithm used in this system is tailored for the special environment of the textile two-for-one machine workshop. The core of the algorithm is to respond to the dynamically changing environment in real time, especially the movement of employees, to ensure that the robot can complete the inspection task safely and efficiently.

[0085] In this application, the inspection robot uses multi-perspective real-time perception and adaptive path planning, allowing the robot to continuously optimize the movement path, thereby reducing unnecessary movement and waiting time. At the same time, the robot can work 24 hours a day, eliminating the fatigue factor of manual inspection, thereby greatly improving the inspection efficiency of the twister, and at the same time being able to detect and prevent problems early, which can reduce equipment maintenance costs.

[0086] In addition, the multi-perspective real-time perception system can detect dynamic obstacles in the surrounding area in a timely manner. Combined with the adaptive path planning algorithm, it ensures that the minimum safe distance between the robot and the worker is maintained at more than 1.5 meters through dynamic obstacle avoidance, effectively reducing the risk of work-related injuries and greatly improving the safety of the working environment.

[0087] In this application, real-time environmental perception technology, dynamic virtual wall generation technology, and adaptive path planning enable the system to respond quickly to environmental changes and adapt to dynamic changes in factory layout in a short period of time, such as temporary equipment movement or the setting of maintenance areas, and adjust inspection strategies without human intervention.

[0088] The adaptive path planning algorithm ensures that all key points are checked regularly, while dynamic adjustments can cope with temporarily inaccessible areas, expanding the coverage of the inspection machine.

[0089] On the other hand, dynamic path planning ensures the efficient movement of the robot and reduces unnecessary energy consumption. At the same time, a centralized mobile sensing platform is more energy-efficient than a distributed fixed sensor network.

[0090] Based on the above ideal embodiments of the present invention, the relevant staff can make various changes and modifications without departing from the technical concept of the present invention through the above description. The technical scope of the present invention is not limited to the contents of the specification, but must be determined according to the scope of the claims.

Claims

1. A textile two-for-one twisting machine inspection method based on multi-view real-time perception and adaptive path planning, characterized in that: The following steps are included: Step 1: Build a virtual work map and set inspection points; Step 2: According to the obstacles identified by the robot during the inspection process, a virtual wall is set in the virtual work map, and the virtual wall is updated at intervals; Step 3: Perform adaptive path planning based on the set virtual wall; Step 4: Evaluate the multiple adaptive paths generated in step 3, select the optimal adaptive path, and perform inspection according to the optimal adaptive path.

2. The textile two-for-one twisting machine inspection method based on multi-perspective real-time perception and adaptive path planning according to claim 1 is characterized in that: In step 2, the obstacle is an employee, and the properties of the virtual wall are dynamically calculated by real-time analysis of the employee's position, movement speed, and predicted trajectory. The properties of the virtual wall include position, length, and duration.

3. The textile two-for-one twisting machine inspection method based on multi-perspective real-time perception and adaptive path planning according to claim 1 is characterized in that: In step two, the safety zone is calculated considering the worker’s movement speed and direction, by first setting a base safety distance and then adjusting it based on the worker’s current speed.

4. The textile two-for-one twisting machine inspection method based on multi-view real-time perception and adaptive path planning according to claim 1 is characterized in that: In step 2, the basic time limit of the virtual wall is set based on the employee's mobility pattern and the working status of the workshop.

5. The textile two-for-one twisting machine inspection method based on multi-view real-time perception and adaptive path planning according to claim 4 is characterized in that: The virtual wall foundation time is T Final , the initial foundation time of the virtual wall is preset to be T Base , virtual wall foundation time T Final The calculation formula is as follows: Among them, V staffmax V is the maximum moving speed of the employee; staff is the current employee moving speed, in m / s; k staff L is the factor affecting employee moving speed; workmax is the maximum workshop activity index; L work is the current workshop activity index; k work is the influencing factor of workshop activity.

6. The textile two-for-one twisting machine inspection method based on multi-view real-time perception and adaptive path planning according to claim 1 is characterized in that: In step three, the improved A* algorithm is used for path design, and combined with the genetic algorithm for global optimization.

7. The textile two-for-one twisting machine inspection method based on multi-view real-time perception and adaptive path planning according to claim 6 is characterized in that: The improved A* algorithm is expressed as: f(n)=g(n)+h(n)+d(n), where f(n) is the total estimated cost of node n; g(n) is the actual cost from the starting node to node n; h(n) is the estimated cost from node n to the target node, and d(n) is the newly added dynamic constraint cost item.

8. The textile two-for-one twisting machine inspection method based on multi-view real-time perception and adaptive path planning according to claim 7 is characterized in that: The dynamic constraint cost term d(n)=d Safety (n)+d Speed (n), where d Safety( n) is the safety distance constraint, d Speed (n) is the speed limit.

9. The textile two-for-one twisting machine inspection method based on multi-view real-time perception and adaptive path planning according to claim 8, characterized in that: d safety (n) = k_s*[max(0,D safety -D1)] 2 , k_s is an adjustable weight factor, D safety is the predefined safety distance, D1 is the distance from node n to the nearest obstacle (including dynamic virtual wall); d speed (n) = k_s*[max(0,V current -V0)] 2 , k_v is another adjustable weight factor, V current is the current speed of the robot, and V0 is the speed limit of the area where node n is located.

10. According to the textile two-for-one twister inspection method based on multi-perspective real-time perception and adaptive path planning described in claim 1, in step four, candidate paths are evaluated by five key criteria: path length L, smoothness S, safety F, energy consumption E and time efficiency T. The five criteria are standardized and a weighted summation method is used to calculate the comprehensive score of each candidate path. Finally, the optimal path is selected by comparing the scores.

Images