Digital disordered measurement method for complex parts

Abstract

The invention discloses a digital disordered measurement method for complex parts, and the method comprises the steps: obtaining part data, measuring tool data and operator data, combining a part type recognition model and a measuring point position recognition model, precisely recognizing the types of the parts and the positions of measuring points, and carrying out the recognition of the types of the parts and the positions of the measuring points. The part data comprises geometrical characteristics, sizes, surface textures and scanning image information of the part, the measuring tool data is used for determining proper measuring tool types and measuring positions, the operator data is used for evaluating operator qualification and generating corresponding use records, and the weight of a training model is iterated for multiple times to obtain the corresponding use records. The method has a tracing function in the data transmission process, can trace a specific operator, a measuring tool and a measuring point when the measurement is abnormal, guarantees the accuracy and traceability of data, and has the advantages that compared with a traditional method, the method is simple in operation and convenient to use. The accuracy, the automation degree and the data reliability of the measurement process can be improved.

Description

Technical Field

[0001] The present invention relates to a digital disordered measurement method for complex parts. Background Art

[0002] With the continuous development of the automotive fixture industry, there has emerged a need to detect parts according to different measurement points. In the past disordered detection work of parts, at least one person was required to record while detecting, and the recording was carried out on paper or ordinary electronic devices. It was difficult to ensure the accuracy of the data and difficult to trace, and it took a lot of time, and the labor cost was very high. The meaning of disordered detection is that the number of measurement points of small parts may not be many, but medium and large parts often have dozens or hundreds of measurement points at different positions, and different parts have different measurement points.

[0003] With the advent of the digital age, some data recording methods and systems have emerged. However, due to disordered detection, there are still problems such as difficult measurement point identification and difficult data transmission. A detection device can adapt to the detection of one part, but cannot adapt to the detection of other parts. And often, one failure in identification or data transmission will lead to the failure of this detection and requires starting over. In this case, the overall efficiency of the digital measurement method is often lower than that of manual recording.

[0004] Therefore, how to prevent the measurement efficiency from being lowered due to data transmission failure has become a technical problem to be solved. Summary of the Invention

[0005] The purpose of the present invention is to solve the above-mentioned deficiencies of the prior art, and provide a digital disordered measurement method for complex parts. By obtaining various types of data and based on each model, the problem of measurement data transmission failure in the measurement of complex disordered parts is effectively solved. The purpose of the present invention is achieved as follows:

[0006] The present invention provides a digital disordered measurement method for complex parts, including: obtaining part data, gauge data, and operator data;

[0007] Input the part data into a part type recognition model and a measurement point position recognition model respectively. Obtain the output result of the part type recognition model through the part type recognition model, and obtain the output result of the measurement point position recognition model through the measurement point position recognition model. Determine the type of recognition label according to the output result of the part type recognition model, and determine the position of the recognition label through the output result of the measurement point position recognition model. Among them, the part type recognition model is obtained by training the weights of the model through multiple iterations based on a set of existing part data, and the measurement point position recognition model is obtained by training the weights of the model through multiple iterations based on a set of existing part data;

[0008] Input the measuring tool data into the measuring tool determination model. The measuring tool determination model obtains the output result of the measuring tool type based on the recognized label type, and the measuring tool determination model obtains the output result of the measuring tool position based on the recognized label position. Among them, the measuring tool determination model is trained by iteratively adjusting the weights of the model based on the recognized label position and the recognized label type.

[0009] Combine the operator data with the output result of the measuring tool type, and combine the operator data with the output result of the measuring tool position to obtain the part measurement result.

[0010] Furthermore, the operator data includes identity information. The step of combining the operator data with the output result of the measuring tool type and combining the operator data with the output result of the measuring tool position to obtain the part measurement result includes: calculating the operator qualification information based on the identity information; combining the operator qualification information with the output result of the measuring tool type to generate a measuring tool type usage record; combining the operator qualification information with the output result of the measuring tool position to generate a measuring tool position usage record.

[0011] Furthermore, the steps of calculating the operator qualification information based on the identity information, combining the operator qualification information with the output result of the measuring tool type to generate a measuring tool type usage record, and combining the operator qualification information with the output result of the measuring tool position to generate a measuring tool position usage record include: obtaining the identity information of the operator, where the identity information includes the operator's name, employee number, qualification certification level, and operation history record; evaluating the operator's qualification information based on the operator's identity information, where the qualification information includes whether the operator has passed relevant operation training, whether the operator holds a valid operation certificate, and the operator's operation experience and historical performance; combining the operator's qualification information with the output result of the measuring tool type to generate a measuring tool type usage record; combining the operator's qualification information with the output result of the measuring tool position to generate a measuring tool position usage record.

[0012] Furthermore, the part type recognition model is trained based on the following steps: obtaining the feature data sets of different part types in the part data, where the feature data sets include part geometric features, part dimensions, part surface textures, and part scan image information; dividing the collected feature data sets into a training set and a validation set; using the supervised learning method, inputting the labeled data in the training set into the part type recognition model, and iteratively adjusting the weights of the model through the cross-entropy loss function and the gradient descent method; performing cross-validation on the part type recognition model based on the validation set to obtain the part type recognition model.

[0013] Furthermore, the measuring point position recognition model is trained based on the following steps: obtaining the measuring point position feature data sets of different parts in the part data, where the measuring point position feature data sets include: surface point cloud data, scanned image information, and preliminary position annotations; dividing the collected measuring point position feature data sets into a training set and a validation set; using a supervised learning method, inputting the labeled data in the measuring point position feature data set into the measuring point position recognition model, and iteratively determining the weights of the model through a cross-entropy loss function and the gradient descent method; performing cross-validation on the measuring point position recognition model based on the validation set to obtain the part type recognition model.

[0014] Furthermore, the measuring tool determination model is trained based on the following steps: obtaining the recognition label position data set and the recognition label type data set; dividing the collected recognition label position data set into a position training set and a position validation set, and dividing the collected recognition label type data set into a type training set and a type validation set; using a supervised learning method, inputting the labeled data in the position training set and the type training set into the measuring tool determination model, and determining the parameters of the model by iteratively weighting the model multiple times, using a cross-entropy loss function and the gradient descent method; performing cross-validation on the measuring tool determination model based on the position validation set and the type validation set to evaluate its accuracy and obtain the measuring tool determination model.

[0015] Furthermore, the obtaining of part data includes: obtaining geometric feature data, where the geometric feature data includes the external dimensions, contour information, and surface curvature of the part; obtaining dimension data, where the dimension data includes the key dimensions, calibration point positions, hole diameters, and edge distances of the part; obtaining surface texture data, where the surface texture data includes the texture images, surface roughness, and surface reflectivity of the part surface; obtaining scanned image data, where the scanned image data includes surface point cloud data obtained by laser scanning, optical scanning, or a 3D scanner.

[0016] Compared with the prior art, the beneficial effects of the present invention are: by obtaining various types of data and based on each model, it is possible to accurately identify the part type and the measuring point position, and intelligently match the appropriate measuring tool type and position according to the requirements of different measuring points, and the point position recognition and data reading are more accurate and fast; by combining the operator data with the measuring tool type and position data, when the data transmission fails, it is possible to trace the specific personnel and data, thereby improving the accuracy and traceability of the measurement process. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 is a schematic flow chart of a digital disordered measurement method for complex parts. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] To deepen the understanding of the present invention, the present invention will be further described in detail below in conjunction with embodiments and the accompanying drawings. The embodiments are only used to explain the present invention and do not constitute a limitation on the protection scope of the present invention.

[0019] An embodiment of the present invention provides a digital disordered measurement method for complex parts, including:

[0020] Step 11: Obtain part data, measuring tool data, and operator data; provide a basic input for subsequent model recognition and data processing;

[0021] Step 12: Input the part data into the part type recognition model and the measuring point position recognition model respectively. Obtain the output result of the part type recognition model through the part type recognition model, and obtain the output result of the measuring point position recognition model through the measuring point position recognition model. Determine the recognition label type according to the output result of the part type recognition model, and determine the recognition label position through the output result of the measuring point position recognition model; the accurate recognition of the part type and the measuring point position ensures the efficiency and accuracy of the subsequent measuring tool selection and data acquisition process;

[0022] Among them, the part type recognition model is obtained by training the weights of the model through multiple iterations based on the existing set of part data, and the measuring point position recognition model is obtained by training the weights of the model through multiple iterations based on the existing set of part data;

[0023] Step 13: Input the measuring tool data into the measuring tool determination model. The measuring tool determination model obtains the measuring tool type output result according to the recognition label type, and the measuring tool determination model obtains the measuring tool position output result according to the recognition label position; the function of the measuring tool determination model is to intelligently select the appropriate measuring tool type and position according to the type and measuring point position of the part, which avoids mistakes and inefficiencies in the traditional manual selection of measuring tools and improves the measurement accuracy and efficiency;

[0024] Among them, the measuring tool determination model is obtained by training the weights of the model through multiple iterations based on the recognition label position and the recognition label type;

[0025] Step 14: Combine the operator data with the measuring tool type output result, and combine the operator data with the measuring tool position output result to obtain the part measurement result; this step solves the problem that data cannot be traced in the traditional measurement process. By combining the operator data with the measuring tool information, it is possible to trace back to the specific operator, the measuring tool used and its position when the data transmission fails or the measurement result is abnormal, thus ensuring the high traceability, accuracy and effectiveness of the measurement data.

[0026] Furthermore, the operator data includes identity information; the operator data is merged with the measuring tool type output result, and the operator data is merged with the measuring tool position output result. The obtained part measurement result includes: calculating the operator qualification information based on the identity information; combining the operator qualification information with the measuring tool type output result to generate a measuring tool type usage record; combining the operator qualification information with the measuring tool position output result to generate a measuring tool position usage record.

[0027] Even further, calculating the operator qualification information based on the identity information; combining the operator qualification information with the measuring tool type output result to generate a measuring tool type usage record; combining the operator qualification information with the measuring tool position output result to generate a measuring tool position usage record includes: obtaining the identity information of the operator, where the identity information includes the operator's name, employee number, qualification certification level, and operation history record; evaluating the operator's qualification information based on the operator's identity information, where the qualification information includes whether the operator has passed the relevant operation training, whether holds a valid operation certificate, and the operator's operation experience and historical performance; combining the operator's qualification information with the measuring tool type output result to generate a measuring tool type usage record; combining the operator's qualification information with the measuring tool position output result to generate a measuring tool position usage record.

[0028] Further, the part type recognition model is trained based on the following steps: obtaining the feature data sets of different part types in the part data, where the feature data sets include part geometric features, part dimensions, part surface textures, and part scan image information; dividing the collected feature data sets into a training set and a validation set; using the supervised learning method, inputting the labeled data in the training set into the part type recognition model, and iteratively adjusting the weights of the model through the cross-entropy loss function and the gradient descent method; performing cross-validation on the part type recognition model based on the validation set to obtain the part type recognition model.

[0029] Further, the measuring point position recognition model is trained based on the following steps: obtaining the measuring point position feature data sets of different parts in the part data, where the measuring point position feature data sets include: surface point cloud data, scan image information, and preliminary position annotations; dividing the collected measuring point position feature data sets into a training set and a validation set; using the supervised learning method, inputting the labeled data in the measuring point position feature data set into the measuring point position recognition model, and iteratively determining the weights of the model through the cross-entropy loss function and the gradient descent method; performing cross-validation on the measuring point position recognition model based on the validation set to obtain the part type recognition model.

[0030] Further, the measuring tool determination model is trained based on the following steps: obtaining a recognition label position data set and a recognition label type data set; dividing the collected recognition label position data set into a position training set and a position validation set, and dividing the collected recognition label type data set into a type training set and a type validation set; using a supervised learning method, inputting the labeled data in the position training set and the type training set into the measuring tool determination model, and determining the parameters of the model by iteratively adjusting the weights of the model multiple times, using a cross-entropy loss function and a gradient descent method; performing cross-validation on the measuring tool determination model based on the position validation set and the type validation set, evaluating its accuracy, and obtaining the measuring tool determination model.

[0031] Further, obtaining part data includes: obtaining geometric feature data, where the geometric feature data includes the external dimensions, contour information, and surface curvature of the part; obtaining dimension data, where the dimension data includes the key dimensions, calibration point positions, hole diameters, and edge distances of the part; obtaining surface texture data, where the surface texture data includes the texture image, surface roughness, and surface reflectivity of the part surface; obtaining scanned image data, where the scanned image data includes surface point cloud data obtained through laser scanning, optical scanning, or a 3D scanner.

[0032] Embodiment 1:

[0033] In the field of automobile manufacturing, a 3D laser scanner is used to scan a part to obtain the surface point cloud data of the part; at the same time, the geometric features of the part are extracted, including external dimensions, contour information, surface curvature, etc., and in addition, dimension data such as key dimensions, calibration point positions, hole diameters, and edge distances are also included. Combining the surface texture image and surface roughness information of the part, a complete part data is finally obtained; the above part data is input into a part type recognition model and a measuring point position recognition model. The part type recognition model recognizes different part types and obtains the model weights through iterative training. The measuring point position recognition model identifies the positions of each measuring point based on the surface point cloud data and scanned image information of the part; according to the output result of the part type recognition model, the recognition label type is determined; according to the output result of the measuring point position recognition model, the recognition label position is determined. The recognition label type and position provide a basis for subsequent measuring tool selection; the type and position of the recognition label are input into the measuring tool determination model, which accurately identifies the suitable measuring tool type and measuring tool position based on the parameters trained based on the label position and type; the identity information of the operator is obtained, including name, employee number, and operation history record. According to the identity information of the operator, their qualification information is evaluated to determine whether the operator meets the operation requirements; the operator data is combined with the measuring tool type output result and the measuring tool position output result to generate a complete part measurement result, and a measuring tool type usage record and a measuring tool position usage record are recorded to ensure data traceability.

[0034] Embodiment 2:

[0035] In the field of aviation manufacturing, precise digital unordered measurement of aircraft components is crucial for ensuring product quality. The following are the specific implementation steps: In this embodiment, an aviation manufacturer uses a high-precision three-dimensional laser scanner to scan the aircraft fuselage parts to obtain the surface point cloud data of the parts. In addition, geometric feature data of the parts is extracted, including external dimensions, contour information, surface curvature, etc. The dimension data includes key dimensions of the parts, hole diameters, calibration point positions, edge distances, etc. In addition, surface texture data is obtained, including surface roughness and surface reflectivity. Finally, the scanned image of the part is also obtained. The measuring tool data comes from measuring devices, including the specifications and usage status of measuring tools such as digital calipers and micrometers. The operator data includes the operator's identity information, operation history, qualification certificates, etc.

[0036] The obtained part data is respectively input into the part type recognition model and the measuring point position recognition model. The part type recognition model identifies the type of the part, such as fuselage, wing surface, or engine component, etc., based on the geometric features and scanned image of the part. The measuring point position recognition model identifies the specific positions on the part surface that need to be measured, such as rivet positions, joint points, center of holes, etc., based on the surface point cloud data and scanned image.

[0037] After receiving the recognized label type and position data, the measuring tool determination model selects the most suitable measuring tool and position according to the recognized part type and measuring point position; for example, for the rivet position on the fuselage, the measuring tool determination model will select a suitable micrometer and digital caliper and determine their usage positions; for the hole diameter of the engine component, an internal diameter gauge or other specific measuring tools may be selected. This process obtains accurate measuring tool selection and measuring position through model weight iteration.

[0038] In terms of operator data, by obtaining the operator's identity information, such as name, employee number, and operation history records, the model will evaluate whether the operator has passed relevant training and holds a valid operation certificate. Subsequently, the operator data is combined with the output results of the measuring tool type and position to generate a measurement record, including the type and position of the measuring tool used by the operator, and record the usage history of the measuring tool. In this way, when problems occur, it is possible to trace back to specific operators, measuring tools, and operation processes, thereby ensuring the reliability and accuracy of the measurement results.

[0039] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not restrictive. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the purpose and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

Claims

1. A digital disordered measurement method for complex parts, characterized in that: include: Obtain part data, gage data, and operator data; Input the part data into the part type recognition model and the measuring point position recognition model respectively, obtain the part type recognition model output result through the part type recognition model, obtain the measuring point position recognition model output result through the measuring point position recognition model, determine the type of recognition label according to the part type recognition model output result, and determine the recognition label position according to the measuring point position recognition model output result, wherein the part type recognition model is based on a set of existing part data and is obtained through multiple iterative model weight training, and the measuring point position recognition model is based on a set of existing part data and is obtained through multiple iterative model weight training; The gage data is input into a gage determination model, wherein the gage determination model obtains a gage type output result according to the identification tag type, and the gage determination model obtains a gage position output result according to the identification tag position, wherein the gage determination model is obtained by weight training of multiple iterations of the model based on the identification tag position and the identification tag type; Combine the operator data with the gage type output, and combine the operator data with the gage position output to get the part measurement results.

2. A digital disordered measurement method for complex parts according to claim 1, characterized in that: The operator data includes identity information; The combining of the operator data with the output result of the gauge type and the combining of the operator data with the output result of the gauge position to obtain the part measurement result includes: calculating the operator qualification information based on the identity information; The operator qualification information is combined with the output result of the measuring tool type to generate the measuring tool type usage record; the operator qualification information is combined with the output result of the measuring tool position to generate the measuring tool position usage record.

3. A digital disordered measurement method for complex parts according to claim 2, characterized in that: The method of calculating the operator's qualification information based on the identity information; combining the operator's qualification information with the output result of the measuring tool type to generate a measuring tool type usage record; combining the operator's qualification information with the output result of the measuring tool position to generate a measuring tool position usage record includes: obtaining the operator's identity information, the identity information including the operator's name, employee number, qualification certification level and operation history record; based on the operator's identity information, evaluating the operator's qualification information, the qualification information including whether the operator has passed relevant operation training, whether he holds a valid operation certificate, and the operator's operation experience and historical performance; combining the operator's qualification information with the output result of the measuring tool type to generate a measuring tool type usage record; combining the operator's qualification information with the output result of the measuring tool position to generate a measuring tool position usage record.

4. A digital disordered measurement method for complex parts according to claim 1, characterized in that: The part type recognition model is trained based on the following steps: obtaining feature data sets of different part types in the part data, wherein the feature data sets include part geometric features, part dimensions, part surface textures, and part scanned image information; dividing the collected feature data sets into a training set and a validation set; using a supervised learning method, inputting the labeled data in the training set into the part type recognition model, and iterating the weights of the model through a cross entropy loss function and a gradient descent method; cross-validating the part type recognition model based on the validation set to obtain the part type recognition model.

5. A digital disordered measurement method for complex parts according to claim 1, characterized in that: The measuring point position recognition model is trained based on the following steps: obtaining a measuring point position feature data set of different parts in the part data, wherein the measuring point position feature data set includes: surface point cloud data, scanned image information, and preliminary position annotations; dividing the collected measuring point position feature data set into a training set and a validation set; using a supervised learning method, inputting the annotated data in the measuring point position feature data set into the measuring point position recognition model, and iteratively determining the weight of the model through a cross entropy loss function and a gradient descent method; cross-validating the measuring point position recognition model based on the validation set to obtain a part type recognition model.

6. A digital disordered measurement method for complex parts according to claim 1, characterized in that: The measuring tool determination model is trained based on the following steps: obtaining an identification label position data set and an identification label type data set; dividing the collected identification label position data set into a position training set and a position verification set, and dividing the collected identification label type data set into a type training set and a type verification set; Using the supervised learning method, the labeled data in the location training set and the type training set are input into the gage determination model, and the model weights are iterated multiple times, and the model parameters are determined using the cross entropy loss function and the gradient descent method; the gage determination model is cross-validated based on the location validation set and the type validation set to evaluate its accuracy and obtain the gage determination model.

7. A digital disordered measurement method for complex parts according to claim 1, characterized in that: The acquiring of part data comprises: acquiring geometric feature data, wherein the geometric feature data comprises the outer dimensions, contour information and surface curvature of the part; acquiring dimensional data, wherein the dimensional data comprises the key dimensions, calibration point positions, apertures and edge distances of the part; acquiring surface texture data, wherein the surface texture data comprises the texture image, surface roughness and surface reflectivity of the part surface; and acquiring scanned image data, wherein the scanned image data comprises surface point cloud data obtained by laser scanning, optical scanning or a three-dimensional scanner.

Images