Assembly detection method and device of electric actuator, terminal and storage medium

Abstract

The invention discloses an assembly detection method and device for an electric actuator, a terminal and a storage medium, and the method comprises the steps: collecting the initial recognition information of a to-be-assembled part, storing the initial recognition information in a data processing center, and carrying out the recognition and classification of the to-be-assembled part according to the initial recognition information; if the to-be-assembled part is an assembly part, secondary identification and weight detection are sequentially conducted on the assembly part, and if the secondary identification result and the weight detection result are both qualified, the assembly part is conveyed to an assembly center; if the to-be-assembled part is an outer part, the outer part is sequentially subjected to ultrasonic detection and three-time recognition, and if the results of the ultrasonic detection and the three-time recognition are qualified, the outer part is conveyed to an assembly center; and the assembled electric actuator is recognized for four times, and if the result of the four times of recognition is qualified, the assembling result of the electric actuator is qualified. According to the method, assembly errors can be effectively avoided, the assembly efficiency is improved, and the problems of potential quality hazards and wrong assembly in the assembly process are solved.

Description

Technical Field

[0001] The present application relates to the field of industrial automatic control systems, and particularly to an assembly detection method, device, terminal and storage medium for an electric actuator. Background Art

[0002] The electric actuator is an important part of modern industrial automation systems and plays a key role in multiple industries such as electric power, petroleum, and metallurgy. With the rapid development of industrial production and technological progress, the performance and reliability of electric actuators directly affect the operating efficiency and product quality of the entire production line. In order to ensure its long-term stable operation, especially its adaptability in harsh environments, higher requirements are put forward for its assembly quality and process.

[0003] Currently, the common quality control means during the assembly process of electric actuators mainly include manual visual inspection, the application of traditional mechanical measuring tools, and the assistance of simple electronic detection equipment, etc. Although these methods can meet the basic requirements to a certain extent, they have obvious limitations and uncertainties. For example, manual inspection is easily affected by subjective factors, resulting in low accuracy; while simply relying on traditional instruments with fixed patterns lacks flexibility and cannot cope with the changes of complex situations.

[0004] However, it is difficult for the above-mentioned various methods to effectively solve the problem of total quality management during the assembly process. Especially when facing multi-process joint operations, how to accurately judge the status of each component has become a key point that urgently needs to be broken through. Therefore, there is an urgent need for a new solution that can accurately distinguish different types of parts and can continuously implement multiple verification steps to ensure that the final product meets the design specifications. Summary of the Invention

[0005] The present invention aims to significantly improve the assembly quality of electric actuators through a series of innovative detection logics, effectively avoid assembly errors, greatly improve the assembly efficiency, and solve the quality hidden dangers and misassembly problems during the assembly process.

[0006] In a first aspect, the present application provides an assembly detection method for an electric actuator, adopting the following technical solution: An assembly detection method for an electric actuator, applied to an assembly device, the assembly device including a conveyor belt and an assembly center, specifically including the following steps: Collect the initial identification information of the parts to be assembled through the identification mechanism on the conveyor belt and store it in the data processing center, and identify and classify the parts to be assembled according to the initial identification information, and the results of the classification and identification include assembled parts and appearance parts; If the component to be assembled is an assembly part, performing secondary identification and weight detection on the assembly part in sequence, and if the results of the secondary identification and the weight detection are both qualified, the assembly part is transported to the assembly center; If the component to be assembled is an appearance part, the appearance part is subjected to ultrasonic testing and three recognitions in sequence. If the results of the ultrasonic testing and the three recognitions are both qualified, the appearance part is transported to the assembly center; When the assembly center assembles the assembly parts and the appearance parts into a complete machine, the assembled electric actuator is identified four times. If the results of the four identifications are qualified, the assembly result of the electric actuator is qualified.

[0007] By adopting the above technical solution, the electric actuator assembly process is comprehensively inspected through four different identification processes. Through multi-level and multi-dimensional intelligent identification and detection methods, the assembly quality and production efficiency of the electric actuator are significantly improved, and rework time and costs are reduced.

[0008] Preferably, the identification mechanism on the conveyor belt collects initial identification information of the components to be assembled and stores it in a data processing center, and the components to be assembled are identified and classified based on the initial identification information. The classification and identification results include assembly parts and appearance parts, and specifically include the following steps: Performing initial identification of the components to be assembled by an identification mechanism on the conveyor belt, the identification mechanism comprising a high-definition camera, an RFID reader / writer, and a MEMS sensor, collecting an appearance image of the components to be assembled by the high-definition camera, collecting basic identity information of the components to be assembled by the RFID reader / writer, and collecting physical parameters of the components to be assembled by the MEMS sensor, and aggregating the appearance image, the basic identity information, and the physical parameters into initial identification information of the components to be assembled; The initial recognition information is fused and analyzed through a pre-built deep learning algorithm model to obtain an analysis result, and the category of the parts to be assembled is determined based on the analysis result. The categories of the parts to be assembled include assembly parts and appearance parts.

[0009] By adopting the above technical solution, comprehensive initial identification information is formed by collecting the appearance images, basic identity information and physical parameters of the parts to be assembled, thereby improving the integrity and reliability of data collection; secondly, the fusion analysis of the received information can accurately distinguish between assembly parts and appearance parts, ensuring the pertinence and effectiveness of subsequent processes. This not only improves the accuracy of the initial classification, but also lays a solid data foundation for the entire assembly and inspection process, thereby effectively reducing human intervention and improving the level of automation and production efficiency.

[0010] Preferably, if the component to be assembled is an assembled part, the assembled part is subjected to secondary identification and weight detection in sequence. If the results of the secondary identification and the weight detection are both qualified, the assembled part is conveyed to the assembly center, which specifically includes the following steps: If the component to be assembled is an assembled part, perform secondary identification on the assembled part and obtain a secondary identification result; The process of the secondary identification includes extracting the contour features of the assembled part through a preset contour detection algorithm, converting the contour features into quantifiable contour feature parameters, sending the contour feature parameters to the data processing center for comparison with preset standard contour feature data, and calculating the deviation values of various parameters; establishing a historical detection database for the assembled parts, extracting the historical contour feature data of the current batch to which the assembled part belongs from the historical detection database, and analyzing the processing error pattern of the assembled parts in this batch; based on the deviation values and the processing error pattern, combining with a preset Bayesian inference algorithm, obtaining the secondary identification result; If the posterior probability in the Bayesian inference algorithm is greater than a preset determination threshold, determine that the secondary identification result is qualified; if it is not greater than the determination threshold, the secondary identification result is unqualified; If the secondary identification result is unqualified, send a first problem signal and control the assembled part to be conveyed to the defective product warehouse.

[0011] By adopting the above technical solution, the macroscopic characteristics of the assembled part including the contour are identified, the extraction and quantification of the contour features of the assembled part are carried out, the actual dimensional deviation of the assembled part is accurately evaluated. By introducing historical detection data and processing error patterns, the accuracy of the identification is further improved. Combining with the Bayesian inference algorithm to obtain the posterior probability, qualified and unqualified products are clearly distinguished. For unqualified products, an alarm is sent in time and transferred to the defective product warehouse, effectively preventing unqualified parts from entering the subsequent processes, and significantly improving the product quality and assembly reliability.

[0012] Preferably, it further includes the following steps: If the secondary identification result is qualified, perform weight detection on the assembled part to obtain weighing data, extract historical weighing data from the historical detection database, calculate the historical average weight of the current batch to which the assembled part belongs, compare the weighing data with the historical weighing data, and calculate the weight deviation between the current weight and the historical average weight; If the weight deviation does not exceed the preset weight range, the result of the weight detection is qualified, and the assembled part is conveyed to the assembly center through the conveyor belt; If the weight deviation exceeds the weight range, the result of the weight detection is unqualified, send a second problem signal, and control the assembled part to be conveyed to the defective product warehouse.

[0013] By adopting the above technical solutions, accurate weight detection of the assembled parts is achieved, which can effectively judge whether there are problems such as missing or redundant parts in the assembled parts, prevent defective products from flowing into the next process, significantly improve the quality control level in the assembly process, and reduce the rework cost caused by assembly errors.

[0014] Preferably, if the part to be assembled is an appearance part, the appearance part is sequentially subjected to ultrasonic detection and three identifications, which specifically include the following steps: If the part to be assembled is an appearance part, the ultrasonic coating thickness gauge in the assembly equipment is used to perform ultrasonic detection on the appearance part to obtain the outer coating thickness, and the outer coating thickness is sent to the data processing center for comparison with the preset standard thickness range; If the outer coating thickness is within the standard range, the result of the ultrasonic detection is qualified, and the appearance part is subjected to three identifications to judge whether the color difference and smoothness of the appearance part are qualified; if the outer coating thickness exceeds the standard range, the result of the ultrasonic detection is unqualified, a third problem signal is sent, and the appearance part is controlled to be transferred to the defective product warehouse.

[0015] By adopting the above technical solutions, first, the ultrasonic coating thickness gauge is used to accurately measure the outer coating thickness of the appearance part to detect the spraying quality; through the detailed analysis of the color characteristic parameters and surface roughness parameters of the appearance part, it is ensured that the color difference and smoothness of the product reach the predetermined standards, thereby ensuring the consistency and reliability of the final product quality, improving the accuracy and efficiency of detection, reducing the number of reworks caused by coating quality problems, and significantly improving the production efficiency and finished product rate.

[0016] Preferably, the three identifications include the following steps: A standard color card database is established, the original images of the appearance part at multiple angles are collected, each original image is preprocessed to obtain the corresponding processed image, each processed image is divided into multiple uniform sub-regions, the color characteristic parameters of each sub-region are extracted, and combined with the parameters of the corresponding standard color in the standard color card database, the average color difference and the maximum color difference between the appearance part and the standard color are obtained. If the average color difference is less than the preset first color difference threshold and the maximum color difference is less than the preset second color difference threshold, the color difference of the appearance part is qualified; otherwise, it is determined that the color difference is unqualified. If the color difference of the appearance part is qualified, the three-dimensional contour data of the surface of the appearance part is obtained, and the surface roughness parameters of the appearance part are calculated in combination with the preset surface roughness evaluation algorithm, and the surface characteristic parameters of the appearance part are extracted from the three-dimensional contour data; If the roughness parameter and the surface feature parameter meet the preset smoothness standard, the smoothness of the appearance part is qualified, and the result of the three - time identification is qualified; otherwise, the smoothness of the appearance part is unqualified.

[0017] By adopting the above - mentioned technical solution, the color consistency of the appearance part is ensured to reach the preset threshold by extracting the color feature parameter, thus guaranteeing the visual quality of the product; the surface of the appearance part is subjected to contour recognition, and the surface roughness parameter is calculated in combination with the surface roughness evaluation algorithm to further verify whether the smoothness of the appearance part meets the standard requirements, which can effectively prevent product quality problems caused by unqualified color difference and smoothness, and improve the product quality and user experience.

[0018] Preferably, the assembled electric actuator is identified four times. If the results of the four - time identification are qualified, the assembly result of the electric actuator is qualified. The specific steps are as follows: The assembled electric actuator is scanned as a whole based on machine vision technology, and the area to be detected for screws is located by combining the pre - constructed three - dimensional coordinate system and the preset template matching algorithm; The number of screws in the area to be detected is detected and counted based on image recognition technology, and the screw heights of all the screws are obtained. The number of screws and the screw heights are sent to the data processing center for comparison with the preset screw standard values. If both the number of screws and the screw heights match the screw standard values, the result of the four - time identification is qualified, and the assembly result of the electric actuator is qualified.

[0019] By adopting the above - mentioned technical solution, through machine vision technology, the pre - constructed three - dimensional coordinate system and the template matching algorithm, the area of the screws to be detected is accurately located, reducing the errors caused by manual operation, improving the detection accuracy, using image recognition technology to comprehensively count the number of screws, and measuring the heights of all the screws to ensure that each screw meets the standard requirements, avoiding assembly problems caused by missing screws or inconsistent heights. The whole detection process realizes automation, reduces manual intervention, reduces the error rate, and improves the overall intelligent level of the production line.

[0020] In the second aspect, the present application provides an assembly detection device for an electric actuator, adopting the following technical solution: An assembly detection device for an electric actuator includes the following modules: A component classification module, which is used to collect the primary identification information of the components to be assembled through the identification mechanism on the conveyor belt and store it in the data processing center, and identify and classify the components to be assembled according to the primary identification information. The results of the classification and identification include assembled parts and appearance parts; An assembly detection module, which is used to, if the to-be-assembled part is an assembly, perform secondary identification and weight detection on the assembly in sequence. If the results of the secondary identification and the weight detection are both qualified, transfer the assembly to the assembly center; An appearance part detection module, which is used to, if the to-be-assembled part is an appearance part, perform ultrasonic detection and three-time identification on the appearance part in sequence. If the results of the ultrasonic detection and the three-time identification are both qualified, transfer the appearance part to the assembly center; A whole machine detection module, which is used to, when the assembly center assembles the assembly and the appearance part into a whole machine, perform four-time identification on the assembled electric actuator. If the result of the four-time identification is qualified, the assembly result of the electric actuator is qualified.

[0021] By adopting the above technical means, through multiple identifications and detections, combining machine vision, ultrasonic detection and weighing technology, the whole assembly process is comprehensively monitored, ensuring that each link meets the expected standards, effectively avoiding quality problems caused by assembly errors, and improving the reliability of the final product.

[0022] In a third aspect, the present application provides an intelligent terminal, adopting the following technical solution: An intelligent terminal includes a memory and a processor. At least one instruction, at least one program, a code set or an instruction set is stored in the memory, and the at least one instruction, at least one program, the code set or the instruction set is loaded and executed by the processor to implement the assembly detection method of the electric actuator as described above.

[0023] In a fourth aspect, the present application provides a computer-readable storage medium, adopting the following technical solution: A computer-readable storage medium stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, at least one program, the code set or the instruction set is loaded and executed by a processor to implement the assembly detection method of the electric actuator as described above.

[0024] In summary, the present application at least includes the following beneficial effects: 1. The present application realizes a multi-level detection mechanism through four-time identification, ensuring the assembly quality. By comparing the contour features and checking the weight of the assembly, performing ultrasonic coating thickness testing and color difference and smoothness evaluation on the appearance part, it effectively avoids missed detection or misjudgment caused by insufficient single detection means.

[0025] 2. By gradually detecting multiple key indicators such as the type, macroscopic characteristics, weight, coating thickness, color difference, and smoothness of the assembled components, a fusion determination of multi-source data is formed, and correlation analysis is carried out in combination with historical data, ensuring the quality controllability of each link, improving the comprehensiveness and accuracy of detection, effectively avoiding the occurrence of assembly errors, improving assembly efficiency, reducing production costs, and enhancing the overall quality and market competitiveness of the product. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 is one of the flowcharts of the assembly detection method for the electric actuator in this embodiment; Figure 2 is the second flowchart of the assembly detection method for the electric actuator in this embodiment; Figure 3 is the structural diagram of the assembly detection device for the electric actuator in this embodiment. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0027] The present application provides an assembly detection method, device, terminal, and storage medium for an electric actuator. To make the purpose, technical solution, and advantages of the present application clearer, the following will further elaborate on the embodiments of the present application.

[0028] The following further describes in detail an embodiment of the assembly detection method for an electric actuator of the present application with reference to the accompanying drawings of the specification.

[0029] An assembly detection method for an electric actuator of the present application is applied to an assembly device. In this embodiment, the assembly device includes a conveyor belt, a rotating mechanism, an identification mechanism, a mechanical gripper, a rotating mechanism, a high-precision weighing machine, an ultrasonic coating thickness gauge, and an assembly center.

[0030] As Figure 1 shown, it includes the following steps: S1. Collect the initial identification information of the components to be assembled through the identification mechanism on the conveyor belt and store it in the data processing center. Identify and classify the components to be assembled according to the initial identification information. The classification results include assembled parts and appearance parts, and specifically include the following steps: S11. Conduct an initial identification of the components to be assembled through the identification mechanism on the conveyor belt. The identification mechanism includes a high-definition camera, an RFID reader, and a MEMS sensor.

[0031] Use the mechanical gripper to place each component to be assembled on the conveyor belt. The conveyor belt transports the components to be assembled to the position of the rotating mechanism, and the rotating mechanism rotates the components to be assembled at different angles; The high-definition camera continuously identifies and takes pictures of the components to be assembled, and collects the appearance images of the components to be assembled through the high-definition camera; Each component to be assembled is equipped with an RFID tag, which stores basic information about the component to be assembled, including component type and batch. The basic identity information of the component to be assembled is collected through an RFID reader. MEMS sensors are used to collect the physical parameters of the components to be assembled during the transmission process, including acceleration and vibration physical parameters.

[0032] S12, summarizing the appearance image, basic identity information, and physical parameters into initial identification information of the component to be assembled and transmitting it to a data processing center; S13. Through the pre-built deep learning algorithm model, the initial recognition information is integrated and analyzed to obtain analysis results. The categories of the parts to be assembled are determined based on the analysis results. The categories of the parts to be assembled include assembly parts and appearance parts.

[0033] In a specific feasible solution, when the analysis results of the deep learning algorithm model show that the appearance image has a regular metal mechanical structure, the RFID tag information indicates that it is a mechanical component, and the MEMS sensor data is consistent with the vibration characteristics of the mechanical component during transmission, it can be determined that the component to be assembled is an assembly part.

[0034] S2. If the component to be assembled is an assembly, the assembly is sequentially subjected to secondary identification and weight inspection. If both the secondary identification and weight inspection results are qualified, the assembly is transported to the assembly center. The specific steps include the following: S21. If the component to be assembled is an assembly part, perform secondary recognition on the assembly part and obtain a secondary recognition result.

[0035] The secondary identification process includes: S211 , extracting the contour features of the assembly part through a preset contour detection algorithm, and converting the contour features into quantifiable contour feature parameters, specifically including the following steps.

[0036] 3D laser scanning technology is used to obtain point cloud data of the assembly, and the edge contour of the assembly is detected using the Canny edge detection algorithm. The least squares method is used for contour fitting, and the detected discrete edge points are fitted into a continuous contour curve. The curvature of each point on the contour curve is obtained to quantify the degree of contour curvature. Determine the coordinates of the key points of the contour, including the inflection points and extreme points of the contour, and convert the assembly contour into quantifiable contour feature parameters through the contour curvature and key point coordinates.

[0037] S212, sending the contour feature parameters to the data processing center for comparison with preset standard contour feature data, and calculating the deviation value of each parameter, which specifically includes the following steps.

[0038] Establish a historical inspection database for assembled parts. The historical inspection database stores the historical inspection data of each batch of assembled parts, including profile feature parameters, dimension data, and weight data.

[0039] In this embodiment, the profile feature parameters obtained from the current inspection are sent to the data processing center, compared with the standard profile data in the design model, and the deviation values of each parameter are calculated.

[0040] S213. Extract the historical profile feature data of the batch to which the current assembled part belongs from the historical inspection database, and analyze the machining error rules of the assembled parts in different set profile regions.

[0041] In a specific implementable manner, the average error value and the error fluctuation range in a certain key profile region of this batch are determined through statistical analysis. If the profile deviation of the currently inspected assembled part in this region exceeds the normal fluctuation range, even if it is still within the design tolerance, a further detailed inspection process needs to be initiated.

[0042] S214. Based on the deviation values and machining error rules, combined with the preset Bayesian inference algorithm, obtain the secondary recognition result, which specifically includes the following steps: Calculate the prior probability: Based on the historical inspection database, statistically obtain the probability distributions of various deviation values and machining error rules in the cases where the previous assembled parts meet the design requirements (i.e., pass the inspection) and do not meet the design requirements (fail the inspection).

[0043] In a specific implementable manner, through the analysis of a large amount of historical data, determine the probability that the deviation value of a certain type of assembled part in a specific profile region is within a certain range under the production condition of passing the inspection, and the probability that the assembled parts of this batch present specific machining error rules, and summarize them into the prior probability.

[0044] Obtain new evidence: Use the deviation values calculated in the above steps and the machining error rules analyzed related to the batch to which it belongs as new inspection evidence.

[0045] In a specific implementable manner, the deviation value of a certain assembled part detected currently in a specific profile region; by comparing with the historical inspection data, it is found that its machining error rule is similar to that of a certain batch but there are certain deviations.

[0046] Calculate the total probability of the current deviation value and machining error rule through the total probability formula. Update the probability according to the Bayesian formula: The Bayesian formula is Among them, P(A) is the prior probability, that is, the initial probability of the assembled part being qualified or unqualified; P(B|A) is the probability of the current deviation value and machining error law occurring under the condition that the assembled part is qualified or unqualified; P(B) is the total probability of the current deviation value and machining error law occurring.

[0047] Through Bayes' formula, the prior probability and new evidence are combined to calculate the posterior probability P(A|B) of the component being qualified or unqualified under the current detection situation.

[0048] S215. Set a determination threshold. In this embodiment, the determination threshold is 0.8.

[0049] If the posterior probability in the Bayesian inference algorithm is greater than the determination threshold, it is determined that the secondary recognition result is qualified; If it is not greater than the determination threshold, the secondary recognition result is unqualified, and it is determined that the assembled part is a defective product.

[0050] S22. If the secondary recognition result is unqualified, send a first problem signal, and control the conveyor belt to convey the assembled part to the defective product library for re-inspection or re-assembly.

[0051] S23. If the secondary recognition result is qualified, the mechanical gripper places the assembled part on the conveyor belt and conveys it to the position of the high-precision weighing machine. The high-precision weighing machine performs weight detection on the assembled part to obtain weighing data, and the high-precision weighing machine transmits the weighing data to the data processing center for review.

[0052] Specifically, extract historical weighing data from the historical detection database, calculate the historical average weight of the batch to which the current assembled part belongs, compare the weighing data with the historical weighing data, and calculate the weight deviation between the current weight and the historical average weight.

[0053] S24. If the weight deviation does not exceed the preset weight range, the result of the weight detection is qualified, and the assembled part is conveyed to the assembly center through the conveyor belt; If the weight deviation exceeds the weight range, the result of the weight detection is unqualified, send a second problem signal, and control the conveyor belt to convey the assembled part to the defective product library.

[0054] S3. If the part to be assembled is an appearance part, perform ultrasonic detection and three-time recognition on the appearance part in sequence. If the results of the ultrasonic detection and the three-time recognition are both qualified, convey the appearance part to the assembly center, which specifically includes the following steps: S31. If the part to be assembled is an appearance part, use the mechanical gripper to place the appearance part on the conveyor belt. The conveyor belt conveys the appearance part to the rotating mechanism, and the rotating mechanism rotates the appearance part at different angles. At the same time, use the ultrasonic coating thickness gauge in the assembly equipment to perform ultrasonic detection on the appearance part and obtain the outer coating thickness.

[0055] In this embodiment, the appearance part is a painted part, and the outer coating is paint.

[0056] S32. Send the thickness of the outer coating to the data processing center for comparison with the preset standard thickness range.

[0057] S33. If the thickness of the outer coating is within the standard range, the result of the ultrasonic detection is qualified; If the thickness of the outer coating exceeds the standard range, the result of the ultrasonic detection is unqualified, send a third problem signal, and control the conveyor belt to convey the appearance part to the defective product warehouse.

[0058] S34. When the result of the ultrasonic detection is qualified, perform three identifications on the appearance part to determine whether the color difference and smoothness of the appearance part are qualified.

[0059] S341. Establish a standard color card database. The establishment of the standard color card database requires using a spectrophotometer to measure a standard paint sample to obtain its precise color values within the visible light spectrum range.

[0060] S342. Place the appearance part on the conveyor belt through a mechanical gripper. The conveyor belt conveys the appearance part to a rotating mechanism, and the rotating mechanism rotates the appearance part at different angles. Under fixed lighting conditions, use a high-definition camera to collect the original images of the appearance part at multiple angles, and perform preprocessing on each original image to obtain the corresponding processed images.

[0061] In this embodiment, the preprocessing includes denoising, gray correction, and color space conversion. Convert the original image from the RGB color space to the CIELAB color space consistent with the standard color card database to obtain the corresponding processed images of each original image for subsequent accurate comparison.

[0062] S343. Divide each processed image into multiple uniform sub-regions, extract the color feature parameters of each sub-region, and combine them with the parameters of the corresponding standard color in the standard color card database to obtain the average color difference and the maximum color difference between the appearance part and the standard color.

[0063] Specifically, in this embodiment, for each sub-region, extract its color feature parameters and compare them with the parameters of the corresponding standard color in the standard color card database. Use the CIEDE2000 color difference formula to calculate the color difference between each sub-region color and the standard color. When all sub-regions are calculated, statistically obtain the average color difference and the maximum color difference between the appearance part and the standard color.

[0064] If the average color difference is less than the preset first color difference threshold and the maximum color difference is less than the preset second color difference threshold, the color difference of the appearance part is qualified; otherwise, determine that the color difference is unqualified; S344. If the color difference of the appearance part is qualified, obtain the three-dimensional contour data of the surface of the appearance part, which specifically includes the following steps.

[0065] Through the Scale-Invariant Feature Transform (SIFT) algorithm, find the same feature points in different processed images, and then calculate the coordinates of these feature points in three-dimensional space according to the principle of triangulation. Continuously repeat this process to obtain a sufficient number of three-dimensional coordinate points, and then construct a three-dimensional point cloud model of the surface of the appearance part. Perform filtering processing on the constructed three-dimensional point cloud model, and use a surface reconstruction algorithm to convert the point cloud model into a continuous three-dimensional surface model, thereby obtaining the three-dimensional contour data.

[0066] S345. Combine the preset surface roughness evaluation algorithm to calculate the roughness parameters of the surface of the appearance part. In this embodiment, the surface roughness evaluation algorithm used is the root mean square roughness (Ra).

[0067] Use the least squares method to fit a reference plane on the continuous three-dimensional surface model to serve as a reference for measuring the surface roughness. Calculate the vertical distance from each point on the three-dimensional surface model to the reference plane, and use this vertical distance as the height deviation value of the point.

[0068] In a specific implementation, the coordinates of a certain point on the three-dimensional surface model are (x i , y i , z i ), and the reference plane equation is Ax + By + Cz + D = 0. Then the distance from this point to the reference plane is: Among them, d i is the height deviation value, and A, B, C, and D are the parameters for determining the reference plane, which are obtained by least squares fitting.

[0069] According to the height deviation values of all points, calculate the root mean square roughness, Among them, n is the total number of points participating in the calculation on the surface model, d i is the height deviation of the i-th point, and Ra is the root mean square roughness of the surface of the appearance part, with the unit of micrometer.

[0070] S346. Extract the surface feature parameters of the appearance part from the three-dimensional contour data.

[0071] In this embodiment, the surface feature parameters include the number of wave peaks and wave valleys, and these parameters can comprehensively reflect the smoothness of the surface of the appearance part.

[0072] S35. If the roughness parameter and the surface feature parameter meet the preset smoothness standard, the smoothness of the appearance part is qualified, and the results of the three identifications are qualified; otherwise, the smoothness of the appearance part is unqualified.

[0073] In a specific implementable manner, the preset smoothness standard is: if the root mean square roughness Ra is less than the set roughness threshold, and the numbers of peaks and valleys are both within their respective preset standard ranges, it is determined that the smoothness of the appearance part is qualified, that is, the results of the three identifications are qualified; otherwise, it is determined that the smoothness is unqualified.

[0074] S4. When the assembly center performs the overall assembly of the assembled part and the appearance part, the assembled electric actuator is identified four times. If the results of the four identifications are qualified, the assembly result of the electric actuator is qualified, and the specific steps are as follows.

[0075] S41. Based on machine vision technology, the assembled electric actuator is scanned as a whole to obtain an overall machine image, and the area to be detected for screws is located by combining the pre-constructed three-dimensional coordinate system and the preset template matching algorithm.

[0076] In this embodiment, a unified spatial reference framework is set for the overall machine by using the three-dimensional coordinate system; The overall machine image is compared with the standard template stored in the data processing center by using the template matching algorithm. The standard template contains screw distribution information, and the screw distribution information includes the theoretical positions and angle information of each screw in the overall machine; In the actual matching process, the similarity between each area in the overall machine image and the standard template is calculated, and the area to be detected for screws is accurately located. Specifically, when the similarity exceeds the preset similarity threshold, it can be determined that this area is the area where the screw is located, which can ensure the accuracy and reliability of the positioning.

[0077] S42. Based on image recognition technology, the area to be detected is detected to count the number of screws and obtain the screw heights of all screws. Specifically, In this embodiment, the contour of the screw is recognized by the edge detection algorithm, and then the connected region analysis method is used to make the contour of each screw an independent connected region, so as to accurately count the number of screws.

[0078] The screw heights of all screws are measured by using a laser ranging sensor, and the distance from the screw head to the sensor is accurately calculated according to the reflection of the laser beam, and this distance is set as the screw height.

[0079] S43. The number of screws and the screw heights are sent to the data processing center for comparison with the preset screw standard values. The screw standard values include the standard number of screws and the standard screw heights.

[0080] S44. If both the number of screws and the screw height conform to the screw standard values, that is, the number of screws is the same as the standard number of screws, and the screw height of each screw is the same as the corresponding standard screw height, the results of the four identifications are qualified, and the assembly result of the electric actuator is qualified. If not, the whole machine is reassembled.

[0081] In another feasible specific implementation, while judging whether the screw data of the whole machine conforms to the screw standard values, it is also possible to detect detailed features such as the markings on the screw head and the integrity of the thread through an image analysis algorithm to judge whether the screws are installed correctly. For example, through character recognition technology, judge whether the markings on the screw head are consistent with the design requirements; by analyzing the edge features and texture information of the thread, judge whether there are defects, deformations, etc. in the thread to detect the integrity of the thread. If both the markings on the screw head and the integrity of the thread conform to the design requirements, it is determined that the results of the four identifications are qualified, and the above method can comprehensively judge whether the screws are installed correctly.

[0082] S5. Combine a fault prediction model based on a long short-term memory network (LSTM) to deeply analyze the operation data of the electric actuator and predict the types and probabilities of faults that the electric actuator will occur in a future period of time.

[0083] In another feasible specific implementation, collect the multimodal operation data of the complete electric actuator. Specifically, use a current sensor to monitor the current of the motor and a vibration sensor to monitor the vibration signal of the whole machine.

[0084] Obtain the historical operation data of the electric actuator under normal operation conditions, and train the LSTM model through the historical operation data. The training process will not be elaborated here.

[0085] Use the trained LSTM model and multimodal operation data for prediction. When the model detects that both the vibration signal and the current fluctuation near a certain screw exceed the predefined normal range, predict the risk that the screw in this area may become loose during future operation.

[0086] If the risk predicted by the model exceeds the set standard risk value, even if the current number and height of the screws are both detected to be qualified, the assembly equipment will be controlled to reinforce the screws at this part to ensure the assembly quality of the whole machine.

[0087] Based on the above same inventive concept, the embodiments of the present application also disclose an assembly detection device for an electric actuator, and the architecture is as Figure 3 shown. The device includes the following modules: An assembly detection device for an electric actuator includes the following modules: The component classification module is used to collect the initial identification information of the components to be assembled through the identification mechanism on the conveyor belt and store it in the data processing center, identify and classify the components to be assembled based on the initial identification information, and the classification results include assembled parts and appearance parts; The assembled part detection module is used to, if the component to be assembled is an assembled part, conduct secondary identification and weight detection on the assembled part in sequence. If the results of the secondary identification and weight detection are both qualified, transfer the assembled part to the assembly center; The appearance part detection module is used to, if the component to be assembled is an appearance part, conduct ultrasonic detection and tertiary identification on the appearance part in sequence. If the results of the ultrasonic detection and tertiary identification are both qualified, transfer the appearance part to the assembly center; The whole machine detection module is used to, when the assembly center assembles the assembled parts and appearance parts into a whole machine, conduct quaternary identification on the assembled electric actuator. If the result of the quaternary identification is qualified, the assembly result of the electric actuator is qualified.

[0088] In a specific feasible implementation scheme, the component classification module includes the following units: The first component classification unit is used to conduct initial identification on the components to be assembled through the identification mechanism on the conveyor belt. The identification mechanism includes a high-definition camera, an RFID reader, and a MEMS sensor. Collect the appearance image of the component to be assembled through the high-definition camera, collect the basic identity information of the component to be assembled through the RFID reader, and collect the physical parameters of the component to be assembled through the MEMS sensor. Summarize the appearance image, basic identity information, and physical parameters into the initial identification information of the current component to be assembled; The second component classification unit is used to conduct fusion analysis on the initial identification information through a pre-constructed deep learning algorithm model to obtain an analysis result, and judge the category of the component to be assembled based on the analysis result. The categories of the components to be assembled include assembled parts and appearance parts.

[0089] In a specific feasible implementation scheme, the assembled part detection module includes the following units: The first assembled part detection unit is used to, if the component to be assembled is an assembled part, conduct secondary identification on the assembled part and obtain a secondary identification result; The process of secondary identification includes extracting the contour features of the assembled part through a preset contour detection algorithm, converting the contour features into quantifiable contour feature parameters, sending the contour feature parameters to the data processing center for comparison with the preset standard contour feature data, and calculating the deviation values of each parameter; establishing a historical detection database for the assembled parts, extracting the historical contour feature data of the current batch to which the assembled part belongs from the historical detection database, and analyzing the processing error law of the assembled parts in this batch; based on the deviation values and the processing error law, combining with the preset Bayesian inference algorithm, obtaining the secondary identification result; The second assembly detection unit is used to determine that the secondary recognition result is qualified if the posterior probability in the Bayesian inference algorithm is greater than the preset determination threshold, and determine that the secondary recognition result is unqualified if it is not greater than the determination threshold; if the secondary recognition result is unqualified, send a first problem signal and control the assembly to be transferred to the defective product warehouse.

[0090] The third assembly detection unit is used to, if the secondary recognition result is qualified, perform a weight detection on the assembly to obtain weighing data, extract historical weighing data from the historical detection database, calculate the historical average weight of the batch to which the current assembly belongs, compare the weighing data with the historical weighing data, and calculate the weight deviation between the current weight and the historical average weight. The fourth assembly detection unit is used to, if the weight deviation does not exceed the preset weight range, the result of the weight detection is qualified, and transfer the assembly to the assembly center through the conveyor belt; if the weight deviation exceeds the weight range, the result of the weight detection is unqualified, send a second problem signal, and control the assembly to be transferred to the defective product warehouse.

[0091] In a specific feasible implementation, the appearance part detection module includes the following units: The first appearance part detection unit is used to, if the part to be assembled is an appearance part, perform ultrasonic detection on the appearance part through the ultrasonic coating thickness gauge in the assembly equipment to obtain the outer coating thickness, and send the outer coating thickness to the data processing center for comparison with the preset standard thickness range. The second appearance part detection unit is used to, if the outer coating thickness is within the standard range, the result of the ultrasonic detection is qualified, and perform three-time recognition on the appearance part to determine whether the color difference and smoothness of the appearance part are qualified; if the outer coating thickness exceeds the standard range, the result of the ultrasonic detection is unqualified, send a third problem signal, and control the appearance part to be transferred to the defective product warehouse.

[0092] In a specific feasible implementation, the second appearance part detection unit includes the following sub-units: The first appearance part detection sub-unit is used to establish a standard color card database, collect the original images of the appearance part from multiple angles, preprocess each original image to obtain the corresponding processed image, divide each processed image into multiple uniform sub-regions, extract the color feature parameters of each sub-region, and combine the parameters of the corresponding standard color in the standard color card database to obtain the average color difference and the maximum color difference between the appearance part and the standard color. If the average color difference is less than the preset first color difference threshold and the maximum color difference is less than the preset second color difference threshold, the color difference of the appearance part is qualified; otherwise, it is determined that the color difference is unqualified. The second appearance part detection subunit is used to, if the color difference of the appearance part is qualified, obtain the three-dimensional contour data of the surface of the appearance part, calculate the surface roughness parameters of the appearance part in combination with a preset surface roughness evaluation algorithm, and extract the surface feature parameters of the appearance part from the three-dimensional contour data; The third appearance part detection subunit is used to, if the roughness parameters and the surface feature parameters meet the preset smoothness standard, then the smoothness of the appearance part is qualified, and the result of the three-time recognition is qualified; otherwise, the smoothness of the appearance part is unqualified.

[0093] In a specific feasible implementation, the whole machine detection module includes the following units: The first whole machine detection unit is used to perform an overall scan on the assembled electric actuator based on machine vision technology, and locate the area to be detected for screws in combination with a pre-constructed three-dimensional coordinate system and a preset template matching algorithm; The second whole machine detection unit is used to detect and count the number of screws in the area to be detected based on image recognition technology, and obtain the screw heights of all the screws, and send the number of screws and the screw heights to the data processing center for comparison with the preset screw standard values; The third whole machine detection unit is used to, if both the number of screws and the screw heights match the screw standard values, then the result of the four-time recognition is qualified, and the assembly result of the electric actuator is qualified.

[0094] It can be seen from the above function introduction that an assembly detection device for an electric actuator in this application builds a set of systems, improves the intelligent and automated levels, enhances, and promotes the sustainable, rapid and healthy development of the economic society.

[0095] Based on the above same inventive concept, an embodiment of this application also discloses a computer-readable storage medium, in which at least one instruction, at least one program, a code set or an instruction set is stored, and at least one instruction, at least one program, a code set or an instruction set can be loaded and executed by a processor to implement the assembly detection method for an electric actuator provided by the above method embodiment.

[0096] Also based on the above same inventive concept, an embodiment of this application also discloses a computer-readable storage medium, in which at least one instruction, at least one program, a code set or an instruction set is stored, and at least one instruction, at least one program, a code set or an instruction set is loaded and executed by a processor to implement the assembly detection method for an electric actuator as described above.

[0097] Those of ordinary skill in the art can understand that all or part of the steps to implement the above embodiments can be completed by hardware, or can be completed by instructing relevant hardware through a program. The program can be stored in the computer-readable storage medium, and the computer-readable storage medium includes, for example: various media that can store program codes such as USB flash drives, mobile hard disks, read-only memories (ROM), random access memories (RAM), magnetic disks, or optical discs.

[0098] The above are only alternative embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application shall be included within the protection scope of the present application.

Claims

1. An assembly detection method for an electric actuator, characterized in that, Applied to assembly equipment, the assembly equipment includes a conveyor belt and an assembly center, specifically including the following steps: The identification mechanism on the conveyor belt collects initial identification information of the components to be assembled and stores it in a data processing center, and classifies the components to be assembled based on the initial identification information, wherein the classification and identification results include assembly parts and appearance parts; If the component to be assembled is an assembly part, performing secondary identification and weight detection on the assembly part in sequence, and if the results of the secondary identification and the weight detection are both qualified, the assembly part is transported to the assembly center; If the component to be assembled is an appearance part, the appearance part is subjected to ultrasonic testing and three recognitions in sequence. If the results of the ultrasonic testing and the three recognitions are both qualified, the appearance part is transported to the assembly center; When the assembly center assembles the assembly parts and the appearance parts into a complete machine, the assembled electric actuator is identified four times. If the results of the four identifications are qualified, the assembly result of the electric actuator is qualified.

2. The assembly inspection method of the electric actuator according to claim 1, wherein The identification mechanism on the conveyor belt collects the initial identification information of the components to be assembled and stores it in a data processing center, and the components to be assembled are identified and classified based on the initial identification information. The classification and identification results include assembly parts and appearance parts, which specifically include the following steps: Performing initial identification of the components to be assembled by an identification mechanism on the conveyor belt, the identification mechanism comprising a high-definition camera, an RFID reader / writer, and a MEMS sensor, collecting an appearance image of the components to be assembled by the high-definition camera, collecting basic identity information of the components to be assembled by the RFID reader / writer, and collecting physical parameters of the components to be assembled by the MEMS sensor, and aggregating the appearance image, the basic identity information, and the physical parameters into initial identification information of the components to be assembled; The initial recognition information is fused and analyzed through a pre-built deep learning algorithm model to obtain an analysis result, and the category of the parts to be assembled is determined based on the analysis result. The categories of the parts to be assembled include assembly parts and appearance parts.

3. The assembly detection method of the electric actuator according to claim 1, characterized in that, If the component to be assembled is an assembly, performing secondary identification and weight detection on the assembly in sequence, and if the results of the secondary identification and the weight detection are both qualified, transporting the assembly to the assembly center, specifically comprising the following steps: If the component to be assembled is an assembly part, performing secondary identification on the assembly part and obtaining a secondary identification result; The process of the secondary identification includes: extracting the contour features of the assembled part through a preset contour detection algorithm, converting the contour features into quantifiable contour feature parameters, sending the contour feature parameters to the data processing center for comparison with the preset standard contour feature data, and calculating the deviation values of various parameters; establishing a historical detection database for the assembled parts, extracting the historical contour feature data of the current batch to which the assembled part belongs from the historical detection database, and analyzing the processing error law of the assembled parts in this batch; based on the deviation values and the processing error law, combining with the preset Bayesian inference algorithm, obtaining the secondary identification result; If the posterior probability in the Bayesian inference algorithm is greater than the preset determination threshold, it is determined that the secondary identification result is qualified; if it is not greater than the determination threshold, the secondary identification result is unqualified; If the secondary identification result is unqualified, send a first problem signal and control the assembled part to be transferred to the defective product warehouse.

4. The assembly detection method of the electric actuator according to claim 3, characterized in that It also includes the following steps: If the secondary identification result is qualified, conduct a weight detection on the assembled part to obtain weighing data, extract the historical weighing data from the historical detection database, calculate the historical average weight of the current batch to which the assembled part belongs, compare the weighing data with the historical weighing data, and calculate the weight deviation between the current weight and the historical average weight; If the weight deviation does not exceed the preset weight range, the result of the weight detection is qualified, and the assembled part is transferred to the assembly center through the conveyor belt; If the weight deviation exceeds the weight range, the result of the weight detection is unqualified, send a second problem signal, and control the assembled part to be transferred to the defective product warehouse.

5. The assembly detection method of the electric actuator according to claim 1, characterized in that, If the to-be-assembled part is an appearance part, conduct ultrasonic detection and tertiary identification on the appearance part in sequence, specifically including the following steps: If the to-be-assembled part is an appearance part, conduct ultrasonic detection on the appearance part through the ultrasonic coating thickness gauge in the assembly equipment to obtain the outer coating thickness, and send the outer coating thickness to the data processing center for comparison with the preset standard thickness range; If the outer coating thickness is within the standard range, the result of the ultrasonic detection is qualified, conduct tertiary identification on the appearance part to determine whether the color difference and smoothness of the appearance part are qualified; if the outer coating thickness exceeds the standard range, the result of the ultrasonic detection is unqualified, send a third problem signal, and control the appearance part to be transferred to the defective product warehouse.

6. The assembly inspection method of the electric actuator according to claim 5, characterized in that, The tertiary identification includes the following steps: Establish a standard color card database, collect the original images of the appearance part from multiple angles, preprocess each original image to obtain the corresponding processed image, divide each processed image into multiple uniform sub-regions, extract the color feature parameters of each sub-region, and combine with the parameters of the corresponding standard color in the standard color card database to obtain the average color difference and the maximum color difference between the appearance part and the standard color. If the average color difference is less than the preset first color difference threshold and the maximum color difference is less than the preset second color difference threshold, the color difference of the appearance part is qualified; Otherwise, it is determined that the color difference is unqualified; If the color difference of the appearance part is qualified, obtain the three-dimensional contour data of the surface of the appearance part, calculate the surface roughness parameters of the appearance part in combination with a preset surface roughness evaluation algorithm, and extract the surface feature parameters of the appearance part from the three-dimensional contour data; If the roughness parameters and the surface feature parameters meet the preset smoothness standard, the smoothness of the appearance part is qualified, and the result of the three identifications is qualified; otherwise, the smoothness of the appearance part is unqualified.

7. The assembly detection method of the electric actuator according to claim 1, characterized in that, Perform four identifications on the assembled electric actuator. If the results of the four identifications are qualified, the assembly result of the electric actuator is qualified. The specific steps are as follows: Based on machine vision technology, perform an overall scan on the assembled electric actuator, and locate the area to be detected for screws in combination with a pre-constructed three-dimensional coordinate system and a preset template matching algorithm; Detect and count the number of screws in the area to be detected based on image recognition technology, and obtain the screw heights of all the screws. Send the number of screws and the screw heights to the data processing center for comparison with the preset screw standard values; If both the number of screws and the screw heights match the screw standard values, the result of the four identifications is qualified, and the assembly result of the electric actuator is qualified.

8. An assembly detection device for an electric actuator, characterized in that, It includes the following modules: The component classification module is used to collect the initial identification information of the components to be assembled through the identification mechanism on the conveyor belt and store it in the data processing center, and identify and classify the components to be assembled according to the initial identification information. The results of the classification identification include assembled parts and appearance parts; The assembled part detection module is used to, if the component to be assembled is an assembled part, perform secondary identification and weight detection on the assembled part in sequence. If the results of the secondary identification and the weight detection are both qualified, transfer the assembled part to the assembly center; The appearance part detection module is used to, if the component to be assembled is an appearance part, perform ultrasonic detection and three identifications on the appearance part in sequence. If the results of the ultrasonic detection and the three identifications are both qualified, transfer the appearance part to the assembly center; The whole machine detection module is used to, when the assembly center performs whole machine assembly on the assembled part and the appearance part, perform four identifications on the assembled electric actuator. If the results of the four identifications are qualified, the assembly result of the electric actuator is qualified.

9. An intelligent terminal, characterized in that, It includes a memory and a processor. At least one instruction, at least one program, a code set or an instruction set is stored in the memory. The at least one instruction, at least one program, a code set or an instruction set is loaded and executed by the processor to implement the assembly detection method of the electric actuator according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, At least one instruction, at least one program, a code set or an instruction set is stored in the readable storage medium. The at least one instruction, at least one program, a code set or an instruction set is loaded and executed by the processor to implement the assembly detection method of the electric actuator according to any one of claims 1 to 7.

Images