Electric tricycle frame welding quality evaluation method combined with vibration detection

Abstract

The invention discloses an electric tricycle frame welding quality evaluation method combined with vibration detection, and relates to the technical field of fault diagnosis, and the method comprises the steps: reading historical operation records of similar tricycles, collecting a welding fault event set, carrying out fault tracing, and determining a high-frequency fault source. A vibration monitoring array is arranged, a vibration test is performed on a target tricycle in a predetermined scene, and a plurality of vibration signal distributions are generated. And the welding fault probability is analyzed in combination with industrial big data, and a welding quality evaluation coefficient is obtained according to probability matching and serves as an evaluation result of the frame welding quality. The technical problems that in the prior art, welding quality detection cannot be dynamically combined with the actual use condition of the vehicle for comprehensive evaluation, so that prediction precision is low, and the evaluation process is tedious are solved, and the technical effect of improving convenience, accuracy and practicability of welding quality evaluation is achieved.

Description

Technical Field

[0001] The present application relates to the technical field of fault diagnosis, and in particular to a method for evaluating the welding quality of an electric tricycle frame in combination with vibration detection. Background Art

[0002] With the widespread use of electric tricycles, the quality of frame welding directly affects the safety, stability and durability of the entire vehicle. Traditional frame welding quality assessment methods mainly rely on manual inspection, visual inspection or simple mechanical testing. These methods cannot fully reflect the performance of the frame in actual use, and lack dynamic assessment capabilities, resulting in insufficient fault prediction capabilities. At the same time, the existing technology fails to make full use of the historical operation data and dynamic data such as vibration signals of the tricycle for comprehensive analysis, and cannot achieve accurate welding quality prediction and evaluation. Therefore, there is an urgent need for a new method that combines vibration monitoring with industrial big data to provide a more accurate and efficient frame welding quality assessment solution, thereby improving the accuracy and practicality of detection and meeting the safety requirements of electric tricycles in different usage scenarios.

[0003] In the current relevant technologies, there is a technical problem that welding quality detection cannot be dynamically combined with the actual use of the vehicle for comprehensive evaluation, resulting in low prediction accuracy and cumbersome evaluation process. Summary of the invention

[0004] This application provides an electric tricycle frame welding quality assessment method combined with vibration detection. By reading the historical operation records of similar tricycles, a welding fault event set is collected, fault tracing is performed, and high-frequency fault sources are determined. A vibration monitoring array is deployed, and a vibration test is performed on the target tricycle under a predetermined scenario to generate multiple vibration signal distributions. The welding failure probability is analyzed in combination with industrial big data, and the welding quality evaluation coefficient is obtained according to the probability matching as the evaluation result of the frame welding quality, thereby achieving the technical effect of improving the convenience, accuracy, and practicality of welding quality evaluation.

[0005] The present application provides a method for evaluating the welding quality of an electric tricycle frame combined with vibration detection, comprising: The historical operation records of similar tricycles of the target tricycle are read, and a welding fault event set is obtained based on the historical operation records; fault tracing is performed based on the welding fault event set to determine a number of high-frequency fault sources, wherein the high-frequency fault source is one of the predetermined welding parts of the target tricycle frame; a vibration monitoring array is arranged based on the several high-frequency fault sources, and under a predetermined test scenario, multiple vibration tests are performed on the target tricycle based on the vibration monitoring array to generate multiple vibration signal distributions; the target tricycle is used as an equipment constraint, and the multiple vibration signal distributions are used as feature constraints, and welding fault analysis is performed in combination with industrial big data to determine the welding failure probability; a welding quality evaluation coefficient is obtained based on the welding failure probability matching as a frame welding quality evaluation result of the target tricycle.

[0006] The electric tricycle frame welding quality assessment method proposed in this application in combination with vibration detection is firstly to read the historical operation records of similar tricycles, collect welding fault event sets, trace faults, and determine high-frequency fault sources. A vibration monitoring array is deployed to perform vibration tests on the target tricycle under predetermined scenarios to generate multiple vibration signal distributions. The welding failure probability is analyzed in combination with industrial big data, and the welding quality evaluation coefficient is obtained according to the probability matching as the evaluation result of the frame welding quality, achieving the technical effect of improving the convenience, accuracy, and practicality of welding quality evaluation. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] In order to more clearly illustrate the technical solution of the embodiment of the present invention, the accompanying drawings of the embodiment of the present invention will be briefly introduced below. A flow chart is used in the present application to illustrate the operations performed by the system according to the embodiment of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps can be processed in reverse order or simultaneously as needed. At the same time, other operations can also be added to these processes, or one or more operations can be removed from these processes.

[0008] Figure 1 A schematic flow chart of a method for evaluating welding quality of an electric tricycle frame combined with vibration detection provided in an embodiment of the present application.

[0009] Figure 2 A schematic diagram of a process flow of constructing a predetermined test scenario for an electric tricycle frame welding quality assessment method combined with vibration detection provided in an embodiment of the present application. DETAILED DESCRIPTION

[0010] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below.

[0011] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limiting the present application. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of this application.

[0012] In the following description, reference is made to "some embodiments", which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict, and the terms "first\second" involved are merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "including" and "having" and any variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by technicians in the technical field of this application. The terms used herein are for the purpose of describing the embodiments of the present application only.

[0013] The present application embodiment provides a method for evaluating the welding quality of an electric tricycle frame in combination with vibration detection, such as Figure 1 As shown, the method includes: Step S100, read the historical operation records of similar tricycles of the target tricycle, and collect a welding fault event set based on the historical operation records. Specifically, first determine the data source for accessing and reading the historical operation records of similar products of the target tricycle, configure the authority and compatibility of the data reading tool or interface, use professional software or scripts to extract records according to rules and formats, check data integrity and accuracy and handle problems. Then, in a large number of records, identify welding fault related records by keyword search, screening condition setting or fault diagnosis model, confirm its authenticity and relevance in combination with driving data before and after the fault, collect detailed fault vehicle information, time and place of occurrence, driving conditions, fault phenomenon description and maintenance processing information, etc., and form a welding fault event set after sorting and summarizing, which provides an important data basis for subsequent analysis of fault causes, determination of high-frequency fault sources, and formulation of quality assessment and improvement strategies to ensure the welding quality of electric tricycle frames.

[0014] Step S200, performing fault tracing based on the welding fault event set to determine a plurality of high-frequency fault sources, wherein the high-frequency fault source is one of the predetermined welding locations of the target tricycle frame. Specifically, the welding fault event set is analyzed to extract the details of the frame welding fault location. The predetermined welding positions that take into account all factors such as structural mechanics, welding technology and overall stability of the frame are determined based on the target tricycle frame structure design drawing and relevant technical information, such as the connection points between the frame crossbeam and the longitudinal beam, the welding points between the frame and the wheel axle seat, etc. For each predetermined welding position, professional data statistical tools or methods are used to construct a statistical model to carefully count the number of failures in the welding fault event set. Multiple rounds of verification are performed to ensure that the data is accurate and complete to avoid repeated statistics or omissions. The predetermined threshold is determined based on factors such as the data volume of the welding fault event set, industry standards, and quality requirements for this model of tricycle. Special treatment is given to parts with greater safety impact, such as the connection between the frame and the steering system. The failure frequency of each predetermined welding position is compared with the predetermined threshold one by one. Those that exceed the threshold are high-frequency fault sources, which will become the focus of subsequent research, providing a key basis for exploring frame welding quality issues and formulating detection and improvement strategies to improve frame welding quality and ensure safe and stable operation of tricycles.

[0015] In a possible implementation, fault tracing is performed based on the welding fault event set to determine several high-frequency fault sources, wherein the high-frequency fault source is one of the predetermined welding parts of the target tricycle frame, and step S200 further includes step S210, obtaining the predetermined welding parts of the target tricycle frame. Specifically, according to the design blueprint, engineering drawings and related technical documents of the target tricycle frame, the predetermined welding parts of the frame are comprehensively sorted out and determined. The predetermined welding parts are key connection points pre-set based on mechanical principles, material properties, overall strength and stability requirements and other factors in the process of structural design and manufacturing process planning of the frame. For example, the intersection welding of the longitudinal and transverse beams of the main frame of the frame plays an important role in transmitting and dispersing various complex loads during the driving process of the vehicle; the welding parts where the frame is connected to the suspension system are of key significance for ensuring the smooth driving and handling performance of the vehicle under different road conditions; there are also welding points where the frame is connected to the power system and the transmission system, and the welding quality of these parts directly affects the efficiency and stability of power transmission. By accurately determining these predetermined welding parts, the foundation for subsequent fault statistical analysis is laid.

[0016] Step S220, based on the predetermined welding position, the multiple welding fault events in the welding fault event set are respectively statistically analyzed for the fault positions, and multiple fault frequencies of the multiple welding positions are obtained. Specifically, after the predetermined welding position is clarified, each welding fault event in the welding fault event set is analyzed. The specific position where the frame welding fault occurs involved in each fault event is compared and matched with the predetermined welding position one by one, and then the number of failures at each predetermined welding position is statistically counted. The fault frequency information corresponding to each predetermined welding position is recorded by means of data statistical software or by establishing a special statistical table. For example, when 100 welding fault events are counted, if it is found that there are faults at the welding point between the front cross beam and the longitudinal beam of the frame in 20 events, then the fault frequency of this part is 20 times. Through such rigorous and meticulous statistical work, multiple fault frequency data corresponding to each of the multiple welding positions can be obtained, thereby clearly presenting the fault occurrence of different predetermined welding positions during the historical operation process.

[0017] Step S230, select the fault part whose fault frequency exceeds the predetermined threshold value as the high-frequency fault source, and obtain the several high-frequency fault sources. Specifically, according to the preset fault frequency threshold value, the fault frequency of each welding part is screened and judged. The setting of the predetermined threshold value is not arbitrary, but is based on the analysis of many factors, such as the overall scale of the welding fault event set, the average fault frequency level of the welding parts of similar products in the industry, the quality standards and safety requirements of the tricycle of this model, etc. For example, if the welding fault event set is large in scale and covers a rich variety of operating condition information, then the threshold value can be relatively increased; on the contrary, if the amount of event set data is relatively small, the threshold value should be appropriately lowered to ensure that representative high-frequency fault sources can be accurately screened out. When the fault frequency of a certain welding part exceeds the predetermined threshold value, it is determined as a high-frequency fault source. For example, if the threshold value is set to 15 times, if the fault frequency of the connection part between the rear suspension of the frame and the frame reaches 20 times, then the part is identified as a high-frequency fault source. It is possible to accurately identify several high-frequency fault sources, which will become the focus of subsequent in-depth research on frame welding quality issues, formulation of targeted detection plans, and improvement of welding processes. It is of great significance to improve the overall welding quality and reliability of the electric tricycle frame.

[0018] Step S300, a vibration monitoring array is arranged according to the several high-frequency fault sources, and in a predetermined test scenario, multiple vibration tests are performed on the target tricycle based on the vibration monitoring array to generate multiple vibration signal distributions. Specifically, a vibration monitoring array arrangement scheme is planned based on the specific location of the determined high-frequency fault source on the target tricycle frame, and high-precision vibration sensors that take into account factors such as measurement accuracy, frequency response range, and adaptability to frame vibration characteristics are installed at each high-frequency fault source and its surrounding key areas. During installation, the specifications are followed to ensure that the sensor fits tightly and firmly with the frame, and a predetermined test scenario is constructed to simulate different road conditions such as cement roads, gravel roads, climbing roads, rural dirt roads, different driving speeds from low speed to high speed, and different loads from light load to full load or even overload, etc., which may be encountered in actual use. After the scenario is set up, the tricycle is started, and multiple vibration tests are carried out based on the vibration monitoring array. During the test, the sensor The device collects vibration signals containing amplitude, frequency, phase and other information from various parts of the frame in real time and transmits them to the data acquisition system for storage and record. After completing multiple tests and collecting a large amount of data, the frame is mapped to a two-dimensional coordinate space, and the coordinate positions of each predetermined welding part and key structural point are determined by the frame geometry and structural layout. The vibration signal data processing at the corresponding position is visualized, and its characteristics are represented by colors, symbols or lines to generate vibration signal distribution, thereby intuitively presenting the differences in vibration conditions of different parts of the frame under different test scenarios, providing intuitive and rich data basis for subsequent analysis of vibration characteristics, evaluation of welding part reliability, and judgment of potential risks of high-frequency fault sources, so as to locate frame welding quality problems and formulate improvement measures.

[0019] In one possible implementation, Figure 2 As shown, a vibration monitoring array is arranged according to the several high-frequency fault sources, and in a predetermined test scenario, multiple vibration tests are performed on the target tricycle based on the vibration monitoring array to generate multiple vibration signal distributions. Step S300 further includes step S310, based on the historical operation records of similar tricycles, multiple driving record sets are obtained, wherein the driving records include road characteristics and driving distance. Specifically, the historical operation records of similar tricycles are analyzed, and multiple driving record sets are screened and extracted from the data. Each driving record contains road characteristics, such as the flatness, slope, curve conditions and road surface material of the road, which will have a significant impact on the force and vibration of the tricycle frame. At the same time, the driving distance is also an important component, which reflects the vehicle's operating mileage under different road conditions, and indirectly reflects the accumulated fatigue degree of the frame under the corresponding working conditions. By sorting out the driving record set, the foundation for subsequent analysis and processing is laid.

[0020] Step S320, clustering the multiple driving record sets according to the predetermined road type to obtain multiple road types and multiple total driving distances. Specifically, clustering operations are performed on the multiple driving record sets according to the predetermined road type. The predetermined road type can be set according to actual needs and experience, for example, it can be divided into urban flat roads, rural bumpy roads, mountain rugged roads, etc. In the clustering process, driving records with similar road characteristics are classified into one category to obtain multiple road types. At the same time, for all driving records under each road type, the driving distances are counted and summed to obtain multiple total driving distances. For example, under the urban flat road type, the driving distances of all driving records belonging to this type are summarized to obtain the total driving distance, and the data will be used for subsequent calculation of the test compensation coefficient.

[0021] Step S330, multiple test compensation coefficients are calculated based on the multiple total driving distances, the initial test times are compensated, and multiple optimized test times are obtained after rounding, wherein the test compensation coefficient is the ratio of the total driving distance to the sum of the multiple total driving distances, and the initial test times is 10. Specifically, multiple test compensation coefficients are calculated based on the multiple total driving distances obtained. The specific calculation method is that the ratio of the total driving distance of a certain road type to the sum of the total driving distances of all road types is the test compensation coefficient of the road type. Based on the initial test times of 10 times, the initial test times are compensated using the test compensation coefficient. For example, if the test compensation coefficient of a certain road type is 0.3, then the compensated test times corresponding to the road type are 10×0.3=3 times. After calculating the compensated test times of all road types, rounding operations are performed to finally obtain multiple optimized test times. Through calculation, the test times can be reasonably allocated according to the proportion of different road types in historical operation, so that the test resources can be configured more scientifically.

[0022] Step S340, configure the test space based on the multiple road types, and construct the predetermined test scenario in combination with the multiple optimized test times. Specifically, the test space is configured based on the multiple road types obtained above, and the test space is the actual test scenario, such as building an urban road simulation scenario, a rural road simulation scenario, a mountain road simulation scenario, etc., and the scale and complexity of the scenario can be determined according to actual conditions and needs. Then, a predetermined test scenario is constructed in combination with multiple optimized test times. According to the principle that the greater the proportion, the more tests, the corresponding optimized test times are arranged in different test spaces. For example, if the total driving distance of the mountain road type accounts for a large proportion, and the number of optimized tests is large, then more tests will be arranged in the mountain road simulation scenario. In order to improve the accuracy and practicality of the test, ensure that the test scenario can comprehensively and accurately reflect the various situations that the tricycle may encounter in actual operation, and provide reliable test data support for the frame welding quality assessment.

[0023] In a possible implementation, a vibration monitoring array is arranged according to the several high-frequency fault sources, and in a predetermined test scenario, multiple vibration tests are performed on the target tricycle based on the vibration monitoring array to generate multiple vibration signal distributions. Step S300 further includes step S350, in which multiple tests are performed on the target tricycle in a predetermined test scenario, and in each test process, vibration signals are collected through the vibration monitoring array to obtain multiple vibration signal sets, wherein the vibration signal is the maximum value of the vibration characteristic during the test process, including the maximum vibration amplitude and the maximum vibration frequency. Specifically, multiple test processes are performed on the target tricycle in the constructed predetermined test scenario. In each test process, the vibration monitoring array previously arranged is fully utilized to carry out vibration signal collection. Since the vibration condition of the frame is in dynamic change during the test process, in order to ensure that the collected vibration signal can accurately reflect the key vibration characteristics of the frame under different working conditions, it is set to perform multiple collection operations in each test. The vibration signal collected each time contains rich information, such as vibration characteristics such as vibration amplitude and vibration frequency. Then, the maximum value of the vibration characteristics is screened out from the vibration signals collected multiple times, including the maximum vibration amplitude and the maximum vibration frequency, and the vibration signal corresponding to the maximum value of the characteristics is determined as the monitoring result of this test. After multiple tests in this way, multiple vibration signal sets can be obtained, and each signal set represents the signal data of the frame in the most representative vibration state in a test.

[0024] Step S360, construct a two-dimensional coordinate space based on the target tricycle frame and the predetermined welding position, map the multiple vibration signal sets to the two-dimensional coordinate space, and obtain the distribution of the multiple vibration signals. Specifically, a two-dimensional coordinate space is constructed based on the structural design of the target tricycle frame and the predetermined welding position information determined in advance. In the two-dimensional coordinate space, based on the geometric shape and layout of the frame, each predetermined welding position and other key structural points are accurately mapped to the corresponding coordinate position. Subsequently, the multiple vibration signal sets obtained previously are mapped one by one to the two-dimensional coordinate space according to their corresponding frame position information. In the mapping process, the characteristics of the vibration signal are represented by specific visualization means, such as using different colors, symbols or line thicknesses, such as using the depth of color to represent the size of the maximum vibration amplitude, and using the density of lines to represent the maximum vibration frequency. The distribution of multiple vibration signals can be presented intuitively in the two-dimensional coordinate space, clearly showing the differences in vibration characteristics of different parts of the frame under the predetermined test scenarios, providing an extremely intuitive and effective data presentation form and analysis basis for in-depth analysis of the vibration characteristics of the frame, evaluation of the reliability of welding parts, and subsequent further exploration of frame welding quality issues.

[0025] Step S400, taking the target tricycle as the equipment constraint, taking the multiple vibration signal distributions as the feature constraints, and combining industrial big data to perform welding fault analysis, and determine the welding fault probability. Specifically, the target tricycle is analyzed as the equipment constraint, and the multiple vibration signal distributions obtained from the previous test are used as feature constraints. The relevant industrial big data of electric tricycles are collected and carefully sorted, and after cleaning, screening and integration, they are adapted to the equipment and feature constraints, so as to lay a solid data foundation for welding fault analysis. According to the vibration signal distribution, key features such as the vibration amplitude trend and frequency stability of the predetermined welding part are extracted, and the correlation analysis with industrial big data is carried out. With the help of data mining and statistical methods, the intrinsic connection between the vibration characteristics and the fault type and degree is explored, and a mapping relationship model between the vibration characteristics and the fault type and degree is constructed to support the fault probability assessment. The Bayesian network, neural network and other algorithms are combined with the equipment constraints and the prior probability of industrial big data, and the welding fault probability is calculated by fully considering the influence weight of the vibration characteristics and the complex interaction. After analyzing each part and the fault type, the current welding fault probability of the target tricycle frame is determined, which provides a key decision basis for subsequent strategy formulation to ensure the quality and reliability of frame welding.

[0026] In a possible implementation, the target tricycle is used as the equipment constraint, the multiple vibration signal distributions are used as the feature constraint, and the welding failure analysis is performed in combination with industrial big data to determine the welding failure probability. Step S400 further includes step S410, randomly selecting a first vibration signal distribution from the multiple vibration signal distributions, and obtaining the first road type of the first vibration signal distribution. Specifically, among the multiple vibration signal distributions that have been obtained, one of them is selected by random sampling and defined as the first vibration signal distribution. The test scenario information corresponding to the first vibration signal distribution is analyzed to extract the first road type to which it belongs. Road type information will serve as an important scenario constraint condition for subsequent analysis, because different road types will cause the target tricycle frame to be subjected to different stresses and vibration modes, thereby affecting the reliability of the welding part and the probability of failure.

[0027] Step S420, taking the target tricycle as the equipment constraint, the first vibration signal distribution as the feature constraint, and the first road type as the scene constraint, similar equipment retrieval is performed in combination with industrial big data, and the operating status of multiple similar equipment within the historical time range is obtained to obtain multiple operating status sets. Specifically, the model, structure, technical parameters, etc. of the target tricycle itself are used as equipment constraints, and the vibration characteristics contained in the selected first vibration signal distribution, such as the maximum vibration amplitude of a specific part, the maximum vibration frequency, and the distribution pattern of the vibration signal in the two-dimensional coordinate space of the frame are used as feature constraints, and the first road type determined previously is used as a scene constraint, and the multi-dimensional constraint information is matched and retrieved with the huge industrial big data. Multiple similar equipment similar to the target tricycle in equipment characteristics, vibration characteristics, and operating scenarios are searched in the industrial big data. For the retrieved similar equipment, detailed operating status information within the historical time range is further obtained, including the time, mileage, working conditions, whether a welding failure occurs, and the specific circumstances of the failure, etc. of each operation, and the relevant information is summarized and sorted to obtain multiple operating status sets.

[0028] Step S430, performing welding failure frequency statistics based on the multiple operating state sets to obtain a first welding failure probability, wherein the welding failure probability is the ratio of the number of welding failures to the total number of operations. Specifically, based on the multiple operating state sets collected, detailed frequency statistics are performed on the welding failure situations therein. The records of welding failures in each operating state set are sorted out, and the total number of welding failures that occur in all similar devices under the corresponding conditions is counted. The total number of operations of similar equipment within the historical time range is calculated, including all operation records of normal operation and failure. According to the definition of welding failure probability, that is, the ratio of the number of welding failures to the total number of operations, the first welding failure probability under the current first vibration signal distribution and the first road type constraint is calculated. The first welding failure probability preliminarily reflects the possibility of welding failures under similar conditions.

[0029] Step S440, the welding failure probability is calculated based on the first welding failure probability. Specifically, the first welding failure probability is corrected and improved by analyzing various factors to calculate the final welding failure probability. Factors such as the degree of difference between different similar equipment and the target tricycle, the integrity and accuracy of industrial big data, and the changes in the current test environment and the historical operating environment need to be considered. Through comprehensive evaluation and analysis of these factors, the first welding failure probability is weighted or adjusted using a more complex calculation model to obtain a welding failure probability value that is more accurate, reliable and can truly reflect the welding failure risk of the target tricycle frame. The final welding failure probability will provide an extremely critical basis for subsequent frame welding quality assessment, maintenance decisions, and the formulation of quality improvement measures.

[0030] In a possible implementation, the welding failure probability is calculated according to the first welding failure probability, and step S440 further includes step S441, analyzing and obtaining multiple welding failure probabilities of the multiple vibration signal distributions in sequence. Specifically, multiple vibration signal distributions are analyzed one by one to determine their corresponding welding failure probabilities. For each vibration signal distribution, the complete analysis process described above is repeated. That is, first clarify the corresponding road type, then take the target tricycle as the equipment constraint, take the vibration signal distribution as the feature constraint, combine industrial big data to perform similar equipment retrieval, obtain the historical operating state set of similar equipment, and then perform welding failure frequency statistics based on the operating state set, and calculate the welding failure probability under the vibration signal distribution according to the ratio of the number of welding failures to the total number of operations. This cycle is repeated until the analysis of all vibration signal distributions is completed, thereby obtaining multiple welding failure probabilities. Considering the impact of the frame state differences reflected by different vibration signal characteristics on the welding failure probability, the welding failure risk of the target tricycle frame under different working conditions is evaluated from multiple angles.

[0031] Step S442, average calculation is performed on the multiple welding failure probabilities, and the average calculation result is set as the welding failure probability. Specifically, after obtaining multiple welding failure probability values, the probability values ​​are averaged. All welding failure probabilities are added together, and then divided by the number of probability values, and the average value obtained is the final set welding failure probability. By taking the average value, the welding failure tendency reflected by different vibration signal distributions can be comprehensively balanced, and the deviation that may be caused by the analysis result of a single vibration signal distribution can be avoided, so as to obtain a more representative and stable welding failure probability value. The welding failure probability can more comprehensively and objectively reflect the overall welding failure possibility of the target tricycle frame, and provide a reliable quantitative basis for subsequent frame quality assessment, maintenance plan formulation, and safety performance assurance and other related decisions, which is helpful to take more accurate and effective measures based on comprehensive consideration of various factors to ensure the welding quality of the target tricycle frame and the safe operation of the vehicle.

[0032] Step S500, obtain the welding quality evaluation coefficient according to the welding failure probability matching, as the frame welding quality evaluation result of the target tricycle. Specifically, analyze and organize historical data, comprehensively consider the operating performance, maintenance records and quality evaluation information of many electric tricycles under different welding failure probabilities, so as to build an accurate mapping relationship between welding failure probability and welding quality evaluation coefficient, such as the evaluation coefficient is 90 when the welding failure probability is 1%, and 80 when it is 1.5%, etc. This mapping relationship fully analyzes the influence of the failure probability on various factors of the vehicle to ensure that it can accurately reflect the actual level of frame welding quality. Then, based on the determined welding failure probability of the target tricycle, this mapping relationship model is substituted and the corresponding welding quality evaluation coefficient is determined using a search, calculation or specific matching algorithm. For example, if the welding failure probability is 1.2%, the corresponding evaluation coefficient is matched to about 85. The welding quality evaluation coefficient obtained by this matching is used as the target tricycle frame welding quality assessment result, providing a clear quantitative basis for subsequent quality improvement, vehicle maintenance plan formulation and product quality monitoring, helping relevant personnel understand the frame welding quality status and make reasonable decisions to ensure the overall performance, safety and reliability of the electric tricycle.

[0033] In a possible implementation, a welding quality evaluation coefficient is obtained according to the welding failure probability matching, as the frame welding quality evaluation result of the target tricycle, and step S500 further includes step S510, performing image acquisition on multiple welding parts of the target tricycle to obtain multiple welding image sets. Specifically, multiple welding parts of the target tricycle are imaged. Use high-resolution and high-precision image acquisition equipment, such as professional industrial cameras or high-definition cameras, to ensure that the detailed features of each welding part can be clearly captured. During the acquisition process, the welding parts should be photographed from multiple angles and different distances to obtain all-round image information. At the same time, different lighting conditions are considered to avoid interference with the clarity and accuracy of the image due to factors such as shadows or reflections. After multiple shots of each welding part, all the acquired images are classified and sorted according to the welding part, thereby obtaining multiple welding image sets, which provide a basic data source for subsequent welding defect identification.

[0034] Step S520, based on the convolutional neural network, the welding defects of the plurality of welding image sets are respectively identified, a plurality of welding defect feature sets are determined, and a plurality of welding defect feature distributions are generated. Specifically, the welding defects of the plurality of welding image sets are respectively identified based on the convolutional neural network deep learning model. The convolutional neural network is based on the convolutional layer, the pooling layer and the fully connected layer structure, and can automatically learn the feature patterns in the image. The welding image set is input into the pre-trained convolutional neural network model, and the model identifies the welding defects such as pores, cracks, incomplete penetration, slag inclusion, etc. by analyzing and processing the image pixel information layer by layer. While identifying the defects, a plurality of welding defect features related to the defects are extracted, including the shape, size, position, color contrast and texture features of the defects, etc., to form a plurality of welding defect feature sets. Then, the feature set is mapped to a two-dimensional coordinate space or organized in a specific data structure to generate a plurality of welding defect feature distributions, which intuitively display the distribution and feature performance of welding defects in the welding parts of the frame.

[0035] Step S530, predicting the welding quality of the frame according to the multiple welding defect characteristic distributions, and outputting the welding quality prediction coefficient. Specifically, predicting the welding quality of the frame according to the generated multiple welding defect characteristic distributions. Various types of characteristic information in the welding defect characteristic distribution are used as input data and input into the constructed quality prediction model. The model is trained based on a large amount of historical welding data and the corresponding actual welding quality data, and learns the inherent correlation between welding defect characteristics and welding quality. For example, when the welding defect characteristic distribution shows that there are more and larger porosity defects in a certain welding part, the model infers that the welding strength of this part may be low based on historical experience data, thereby affecting the welding quality of the entire frame. Through the calculation and analysis of the model, the welding quality prediction coefficient is output. The welding quality prediction coefficient reflects the expected level of frame welding quality in the form of a numerical value. The higher the value, the better the welding quality, and vice versa, it means that there are more quality risks.

[0036] Step S540, a welding quality coefficient is obtained by weighted calculation based on the welding quality evaluation coefficient and the welding quality prediction coefficient, and the welding quality coefficient is used as the frame welding quality evaluation result of the target tricycle. Specifically, a welding quality coefficient is obtained by weighted calculation based on the obtained welding quality evaluation coefficient and the welding quality prediction coefficient. Determining a reasonable weighted weight is the key to this step. The setting of the weight needs to comprehensively consider the accuracy, reliability of the welding failure probability analysis method and the welding image defect recognition method, as well as the relative importance of the two in the overall welding quality evaluation. For example, if in a specific application scenario or product type, the welding failure probability analysis based on historical operation data is considered more authoritative, then the weight of the corresponding welding quality evaluation coefficient can be appropriately increased; conversely, if the welding image defect recognition technology shows higher accuracy and effectiveness in recent practice, the weight of the welding quality prediction coefficient is correspondingly increased. By performing weighted sum calculation on the two according to the set weights, the final welding quality coefficient is obtained, and the welding quality coefficient is used as the frame welding quality evaluation result of the target tricycle. The comprehensive evaluation results combine the advantages of two different evaluation methods, and can more comprehensively and accurately reflect the welding quality status of the target tricycle frame, providing a favorable decision-making basis for subsequent frame quality improvement, vehicle safety performance improvement and production process optimization.

[0037] In a possible implementation, the frame welding quality is predicted according to the multiple welding defect characteristic distributions, and the welding quality prediction coefficient is output. Step S530 further includes step S531, querying the historical welding log, collecting the sample welding defect characteristic distribution set, and statistically analyzing the sample welding quality coefficients under different sample welding defect characteristic distributions to obtain the sample welding quality coefficient set, wherein the sample welding quality coefficient is set based on the average value of the normal use time of the vehicle corresponding to the sample welding defect characteristic distribution. Specifically, query and sort the historical welding log. From the log records, collect and extract the sample welding defect characteristic distribution set. Each sample welding defect characteristic distribution records the defect characteristic information of a specific vehicle welding part in detail, such as the type, size, shape, position and distribution pattern of the defect on the frame. The sample welding quality coefficient is set according to the average value of the normal use time of the vehicle corresponding to the sample welding defect characteristic distribution. The setting process is based on a reasonable assumption: to a certain extent, the defect characteristic distribution of the vehicle welding part is related to the normal use time of the vehicle, that is, the better the defect characteristic distribution, the longer the normal use time of the vehicle is, and the higher the corresponding welding quality coefficient. Through statistics and analysis of a large amount of historical data, the sample welding quality coefficients under different sample welding defect characteristic distributions are summarized to obtain a sample welding quality coefficient set, which provides a key data basis for subsequent neural network training and ensures that the training data can fully reflect the intrinsic relationship between welding defect characteristics and welding quality.

[0038] Step S532, taking the sample welding defect characteristic distribution set as input and the sample welding quality coefficient set as output, supervise the BP neural network until convergence, and obtain the welding quality prediction plug-in. Specifically, taking the collected sample welding defect characteristic distribution set as input data and the corresponding sample welding quality coefficient set as output data, supervise the BP neural network. BP neural network is a multi-layer feedforward neural network widely used in fields such as function approximation and pattern recognition. During the training process, the input layer of the network receives the sample welding defect characteristic distribution data, and the data is subjected to complex nonlinear transformation and processing by the neurons of the hidden layer, and finally the predicted welding quality coefficient is output at the output layer. The goal of the training is to continuously adjust the weight and threshold of the network so that the output value of the network is as close as possible to the real sample welding quality coefficient. The supervised training method is adopted, that is, after each sample data is input, the output of the network is compared with the corresponding real sample welding quality coefficient, the error is calculated, and the network parameters are adjusted according to the error back propagation algorithm. The training process continues until the error of the network converges to an acceptable range, and the trained BP neural network obtained at this time is the welding quality prediction plug-in. The welding quality prediction plug-in has the ability to predict the welding quality coefficient based on the input welding defect characteristic distribution data, providing an efficient and accurate tool for frame welding quality prediction.

[0039] Step S533, input the multiple welding defect feature distributions into the welding quality prediction plug-in to predict the frame welding quality, and calculate the mean of multiple prediction results to obtain the welding quality prediction coefficient. Specifically, the multiple welding defect feature distributions obtained by analyzing the welding image of the target tricycle are sequentially input into the trained welding quality prediction plug-in. The plug-in analyzes and calculates each welding defect feature distribution according to the model and parameters learned inside it, and outputs the corresponding frame welding quality prediction result. Since multiple predictions are made, multiple prediction results are obtained. In order to obtain a more representative and stable welding quality prediction coefficient, the mean of multiple prediction results is calculated. All prediction results are added and then divided by the number of prediction results. The average value obtained is the final welding quality prediction coefficient. The welding quality prediction coefficient can comprehensively reflect the overall level of the welding quality of the target tricycle frame. Based on the actual characteristic distribution of welding defects, it provides an important quantitative basis for subsequent frame quality assessment, maintenance decision-making and production process improvement, which helps to improve the management level and control accuracy of the welding quality of the target tricycle frame.

[0040] The embodiment of the present application adopts the method of reading the historical operation records of similar tricycles, collecting welding fault event sets, tracing fault sources, and determining high-frequency fault sources. A vibration monitoring array is deployed to perform vibration tests on the target tricycle under predetermined scenarios to generate multiple vibration signal distributions. The welding failure probability is analyzed in combination with industrial big data, and the welding quality evaluation coefficient is obtained according to the probability matching as the evaluation result of the frame welding quality, thereby achieving the technical effect of improving the convenience, accuracy, and practicality of welding quality evaluation.

[0041] The above specific implementation manner does not constitute a limitation to the protection scope of the present application. It should be understood by those skilled in the art that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application should be included in the protection scope of the present application. In some cases, the actions or steps recorded in the present application can be performed in an order different from that in the embodiment and can still achieve the desired results. In addition, the process depicted in the accompanying drawings does not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A method for evaluating welding quality of an electric tricycle frame combined with vibration detection, characterized in that: Methods include: Reading historical operation records of similar tricycles of the target tricycle, and acquiring a welding fault event set according to the historical operation records; Performing fault tracing based on the welding fault event set to determine a number of high-frequency fault sources, wherein the high-frequency fault source is one of the predetermined welding parts of the target tricycle frame; Arranging a vibration monitoring array according to the plurality of high-frequency fault sources, and performing multiple vibration tests on the target tricycle based on the vibration monitoring array in a predetermined test scenario to generate multiple vibration signal distributions; Taking the target tricycle as the equipment constraint and the distribution of the multiple vibration signals as the feature constraint, welding failure analysis is performed in combination with industrial big data to determine the welding failure probability; A welding quality evaluation coefficient is obtained according to the welding failure probability matching as a frame welding quality evaluation result of the target tricycle.

2. The electric tricycle frame welding quality assessment method combined with vibration detection according to claim 1 is characterized in that: Based on the welding fault event set, fault tracing is performed to determine several high-frequency fault sources, including: Obtaining a predetermined welding position of a target tricycle frame; Based on the predetermined welding position, performing fault position statistics on a plurality of welding fault events in the welding fault event set, and obtaining a plurality of fault frequencies of the plurality of welding positions; Fault locations whose fault frequencies exceed a predetermined threshold are selected as high-frequency fault sources to obtain the plurality of high-frequency fault sources.

3. The electric tricycle frame welding quality assessment method combined with vibration detection according to claim 1 is characterized in that: Configure the scheduled test scenarios, including: Based on the historical operation records of similar tricycles, multiple driving record sets are obtained, wherein the driving records include road characteristics and driving distances; Clustering the plurality of driving record sets according to predetermined road types to obtain a plurality of road types and a plurality of total driving distances; Calculating multiple test compensation coefficients according to the multiple total driving distances, compensating the initial test times, and obtaining multiple optimized test times after rounding, wherein the test compensation coefficient is a ratio of the total driving distance to the sum of the multiple total driving distances, and the initial test times is 10; The test space is configured based on the multiple road types, and the predetermined test scenario is constructed in combination with the multiple optimized test times.

4. The electric tricycle frame welding quality assessment method combined with vibration detection according to claim 1 is characterized in that: Based on the vibration monitoring array, multiple vibration tests are performed on the target tricycle to generate multiple vibration signal distributions, including: Conduct multiple tests on the target tricycle under a predetermined test scenario, and during each test, collect vibration signals through the vibration monitoring array to obtain multiple vibration signal sets, wherein the vibration signal is the maximum value of the vibration characteristics during the test, including the maximum vibration amplitude and the maximum vibration frequency; A two-dimensional coordinate space is constructed based on the target tricycle frame and the predetermined welding position, and the multiple vibration signal sets are mapped to the two-dimensional coordinate space to obtain the distribution of the multiple vibration signals.

5. The electric tricycle frame welding quality assessment method combined with vibration detection according to claim 3 is characterized in that: Taking the target tricycle as the equipment constraint and the distribution of the multiple vibration signals as the feature constraint, welding failure analysis is performed in combination with industrial big data to determine the welding failure probability, including: randomly selecting a first vibration signal distribution from the plurality of vibration signal distributions, and obtaining a first road type of the first vibration signal distribution; Taking the target tricycle as the device constraint, the first vibration signal distribution as the feature constraint, and the first road type as the scene constraint, similar devices are retrieved in combination with industrial big data to obtain the operating status of multiple similar devices within a historical time range, and obtain multiple operating status sets; Performing welding failure frequency statistics based on the multiple operating state sets to obtain a first welding failure probability, wherein the welding failure probability is a ratio of the number of welding failures to the total number of operations; The welding failure probability is calculated according to the first welding failure probability.

6. The electric tricycle frame welding quality assessment method combined with vibration detection according to claim 5 is characterized in that: Calculating the welding failure probability according to the first welding failure probability includes: Analyzing in sequence to obtain a plurality of welding failure probabilities distributed by the plurality of vibration signals; A mean value is calculated for the multiple welding failure probabilities, and a mean value calculation result is set as the welding failure probability.

7. The electric tricycle frame welding quality assessment method combined with vibration detection according to claim 1 is characterized in that: The method also includes: Capturing images of multiple welding positions of the target tricycle to obtain multiple welding image sets; Based on a convolutional neural network, the multiple welding image sets are respectively subjected to welding defect recognition, multiple welding defect feature sets are determined, and multiple welding defect feature distributions are generated; Predicting the welding quality of the frame according to the distribution of the plurality of welding defect characteristics, and outputting a welding quality prediction coefficient; A welding quality coefficient is obtained by weighted calculation based on the welding quality evaluation coefficient and the welding quality prediction coefficient, and the welding quality coefficient is used as a frame welding quality evaluation result of the target tricycle.

8. The electric tricycle frame welding quality assessment method combined with vibration detection according to claim 7 is characterized in that: The frame welding quality is predicted according to the plurality of welding defect characteristic distributions, and a welding quality prediction coefficient is output, including: Query historical welding logs, collect sample welding defect characteristic distribution sets, and count sample welding quality coefficients under different sample welding defect characteristic distributions to obtain a sample welding quality coefficient set, wherein the sample welding quality coefficient is set based on the average value of the normal use time of the vehicle corresponding to the sample welding defect characteristic distribution; Taking the sample welding defect feature distribution set as input and the sample welding quality coefficient set as output, the BP neural network is supervised and trained until convergence, thereby obtaining a welding quality prediction plug-in; The multiple welding defect feature distributions are input into the welding quality prediction plug-in to predict the frame welding quality, and the welding quality prediction coefficient is obtained after the mean value of multiple prediction results is calculated.

Images