Acquisition and optimization application method based on aero-engine part machining data

Abstract

The invention provides a collection and optimization application method based on aero-engine part machining data, and relates to the technical field of aero-engines, and the method specifically comprises the steps: carrying out the space-time mapping processing of time domain data or carding data and the geometric space position of part machining, and storing the machining process data into a technological process database, a processing quality-effect prediction mapping relation model is constructed, a mapping relation between the technological parameters and processing quality-effect prediction is established, meanwhile, the technological parameters and processing quality-effect data of the technological process database serve as boundary input, and a processing quality-effect prediction model is obtained after training; constructing a process optimization model and performing optimization; the working condition data contained in the technological process database is used as original data, the machining quality-effect prediction model and the technological optimization model are embedded into the technological process database, and the technological process data are continuously accumulated and iterated, so that the accuracy of machining quality-effect prediction is improved, technological parameters are optimized, and the machining quality and efficiency are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of aero-engines, and in particular to a method for collecting and optimizing the application of aero-engine part processing data. Background Art

[0002] Aircraft engines are the heart of fighter planes. They are a major national strategic need and the forefront of international engineering science. They are called the "crown jewel" of modern industry because of their wide technical span, long R&D cycle and large capital demand. At present, the processing level of difficult-to-process materials such as titanium alloys, high-temperature alloys and intermetallic compounds in my country is relatively backward, especially the lack of systematic sorting, summarization and evaluation of processing data knowledge, and the incomplete enabling tools such as data knowledge reuse reasoning and optimization application. The process parameters finally applied to parts processing vary from person to person, resulting in poor stability and consistency in the processing of typical parts such as aircraft engine blades, integral blade disks, casings, turbine disks, etc., which has brought huge challenges to the stability of mass production model manufacturing and the progress of research and development models. Summary of the invention

[0003] In view of the deficiencies in the prior art, the present invention provides a method for collecting and optimizing the application of aero-engine part processing data, which realizes the collection and spatiotemporal mapping processing of processing data, so that the processing data corresponds one-to-one with the geometric space position of part processing, thereby allowing the processing results to be correlated with the part structure for analysis, and completing the optimization of process parameters for similar parts or feature processing through a large amount of accumulated data and an optimization algorithm for process parameters. At the same time, a processing quality and efficiency prediction model is constructed to realize the application prediction of parameters such as processing efficiency, processing accuracy, surface quality, tool life, processing deformation, processing vibration, surface integrity, etc., thereby guiding actual processing, effectively improving production and processing efficiency and saving processing costs.

[0004] A method for collecting and optimizing the processing data of aircraft engine parts includes the following steps:

[0005] Step 1: Summarize the processing data acquisition method, define the data type to be collected, and apply the corresponding acquisition method to obtain the processing data according to the data type;

[0006] The data types are divided into combing data and collection data; the combing data is data that can be directly obtained, and the collection data is data that needs to be detected and collected and cannot be directly obtained;

[0007] The collection methods are divided into two categories. The collection scheme of collection data is as follows: based on the standard communication interface of CNC machine tools or third-party interfaces, the spindle power or torque collection system is embedded to complete the collection of spindle power or torque data; the spindle motor current detection technology is used to realize the detection of cutting force and tool wear and damage; the vibration and deformation data in the processing are detected by the embedded non-disturbance sensor attached to the spindle and the workbench, the on-machine probe, and the motion axis vibration monitoring of the CNC system itself; the three-dimensional morphology measuring instrument and the residual stress test and analysis system are used to measure and collect the three-dimensional morphology of the processed surface and the surface residual stress data; the collection method of combing data is as follows: according to the collection strategy and specifications, the type of data to be combed is defined, so as to collect data according to the rules;

[0008] Step 2: Perform time-space mapping processing on the time domain data or combing data in the collected data and the geometric space position of the part processing, and store the processing data in the process database;

[0009] The time-space mapping processing method is based on the tool trajectory of typical part CNC machining, introduces a voxel model to discretize the machining process in time and space; adopts a short-time domain processing method to characterize the machining process signal as a corresponding short-time domain signal feature; establishes a knowledge association between the unit voxel working condition and the short-time machining process signal, forms a time-space mapping model corresponding to the machining process data and the part machining position, so as to map the machining process data to the part space, and store the machining process data in a process database.

[0010] The data stored in the process database are the working condition data of typical parts processing sites, including process routes, processing stages, processing methods, machine tool equipment, clamping methods, selected tools, programming strategies, cutting parameters, cooling and lubrication, spindle power, cutting force, cutting heat, surface morphology, tool wear, deformation, vibration, and surface integrity.

[0011] Step 3: Construct a processing quality and efficiency prediction mapping relationship model, establish a mapping relationship between process parameters and processing quality and efficiency prediction, and use the process parameters and processing quality and efficiency data of the process database as boundary inputs to continuously train the iterative model to obtain a processing quality and efficiency prediction model;

[0012] The construction of the processing quality and efficiency prediction mapping relationship model is specifically as follows: the process parameters are used as the input of the network structure, and the surface roughness, tool life, and workpiece fatigue life evaluation parameters are used as the output, specifically including a surface roughness prediction model, a tool life application prediction model, and a workpiece fatigue life prediction model, to achieve the mapping of process parameters and processing quality and efficiency;

[0013] The surface roughness prediction model is based on an artificial neural network, takes cutting feed, feed speed, and cutting depth as inputs of the network structure, and takes surface roughness as output, and establishes an N-Net multi-hidden layer neural network part cutting surface roughness prediction model that integrates local weights;

[0014] The tool life application prediction model is to train the neural network model by collecting data during the use of the tool, including cutting speed, cutting depth, feed speed, tool material information, cutting environment temperature and humidity, optimize the model parameters, and evaluate the performance of the model through a cross-validation method;

[0015] The workpiece fatigue life prediction model uses the morphology of the metal surface detected by white light interferometer, and after completing the filtering process of the detection point cloud data, completes the removal of sparse outliers and the smoothing of the point cloud data; secondly, based on the morphological feature image defects, grooves, and cracks, it identifies and finally realizes the fatigue life prediction;

[0016] Step 4: Build a process optimization model, conduct sample tests, and verify the optimization results of process parameters. If the processing effect meets the requirements, the process optimization model is built. If the processing effect does not meet the requirements, the process optimization model is iterated repeatedly until the processing effect meets the requirements.

[0017] The process optimization model is based on neural networks and genetic algorithms, with processing quality and efficiency as the optimization goal and process data matching the processing quality and efficiency evaluation requirements as the basic constraint condition;

[0018] Step 5: Use the working condition data contained in the process database as the original data, and embed the processing quality and efficiency prediction model and the process optimization model into the process database;

[0019] Step 6: Apply the processing quality and efficiency prediction model and process optimization model of the process database to continuously accumulate and iterate the process data, thereby improving the accuracy of processing quality and efficiency prediction, optimizing process parameters, and improving processing quality and efficiency.

[0020] The processing quality and efficiency prediction model: enter the processing quality and efficiency prediction model of the process database, select the indicators to be predicted, input the initial conditions according to the requirements of the processing quality and efficiency prediction model, and output the prediction results; wherein the initial conditions include feed speed, cutting speed, cutting depth, part material, part structure, tool material, tool type, cutting method, and processing equipment, and the prediction result is the surface roughness Ra value;

[0021] The process optimization model: enter the process optimization model of the process database, select the initial conditions of CNC machining, output the optimized process parameters or process scheme, and choose whether to store the optimized process parameters or process scheme in the process database according to the prompt of the process database. The initial conditions include part materials, part features, processing equipment, processing methods, tool materials, tool categories, and processing process information;

[0022] The beneficial effects of adopting the above technical solution are:

[0023] The present invention provides a collection and optimization application method based on aero-engine parts processing data. In the traditional CNC processing process, the selection of process parameters generally relies on the experience of technicians and the accumulated technical data. This method has the problem that the rationality of the selection of process parameters is limited by the ability of technicians and has poor flexibility. The present invention can realize the effective collection of processing condition data, and complete the optimization of process parameters for similar parts or feature processing through a large amount of accumulated data and the optimization algorithm of process parameters. At the same time, the constructed processing quality and efficiency prediction module can realize the application prediction of parameters such as processing efficiency, processing accuracy, surface quality, tool life, processing deformation, processing vibration, surface integrity, etc., so as to guide the actual processing, effectively improve the production and processing efficiency and save processing costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 This is an overall flow chart of the method for collecting and optimizing the application of aerospace engine parts processing data according to the present invention;

[0025] Figure 2 Processing data acquisition scheme for the present invention

[0026] Figure 3 Processing data spatiotemporal mapping solution

[0027] Figure 4 Data structure of the prediction model for machining quality effect DETAILED DESCRIPTION

[0028] The specific implementation of the present invention is further described in detail below in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0029] A method for collecting and optimizing the processing data of aircraft engine parts, such as Figure 1 As shown, taking the processing data processing of the ring casing as an example, the following steps are included:

[0030] Step 1: Summarize the processing data acquisition methods, define the data type to be collected, and apply the corresponding acquisition methods to obtain the processing data according to the data type;

[0031] The data types are divided into combing data and collection data; the combing data are data that can be directly obtained, including: materials (brand, heat treatment status), blank form, typical structural characteristics, process route, processing stage, benchmark selection, allowance allocation, process (machine tool equipment, positioning and clamping method, fixture, tool, processing surface, tool path, cutting parameters), etc. The collection data are data that need to be detected and collected and cannot be directly obtained, including: cutting force / heat, tool wear, processing efficiency, processing accuracy, surface quality, processing deformation, chatter, surface integrity (surface roughness, surface microhardness, surface residual stress, burns), time domain data, etc.

[0032] In this embodiment, a collection module is formulated to collect the annular casing material (brand, heat treatment status), blank form, typical structural characteristics, process route, processing stage, reference selection, allowance allocation, process (machine tool equipment, positioning and clamping method, fixture, tool, processing surface, tool path, cutting parameters), etc.

[0033] The collection methods are divided into two categories. The collection scheme of collection data is as follows: based on the standard communication interface of CNC machine tools or third-party interfaces, the spindle power or torque collection system is embedded to complete the collection of spindle power or torque data; the spindle motor current detection technology is used to realize the detection of cutting force and tool wear and damage; the vibration and deformation data in the processing are detected by the embedded non-disturbance sensor attached to the spindle and the workbench, the on-machine probe, and the motion axis vibration monitoring of the CNC system itself; the three-dimensional morphology measuring instrument and the residual stress test and analysis system are used to measure and collect the three-dimensional morphology of the processed surface and the surface residual stress data; the collection method of combing data is as follows: according to the collection strategy and specifications, the type of data to be combed is defined, so as to collect data according to the rules;

[0034] Step 2: Perform time-space mapping processing on the time domain data or combing data in the acquisition data and the geometric space position of the part processing, and store the processing data in the process database, such as Figure 2 As shown;

[0035] The time-space mapping processing method is as follows Figure 3 As shown, based on the tool trajectory of typical part CNC machining, the voxel model is introduced to discretize the machining process in time domain and space; the short-time domain processing method is used to characterize the machining process signal as the corresponding short-time domain signal feature; the knowledge association between the unit voxel working condition and the short-time machining process signal is established to form a space-time mapping model corresponding to the machining process data and the part machining position, so as to map the machining process data to the part space, and store the machining process data in the process database.

[0036] The data stored in the process database are the working condition data of typical parts processing sites, including process routes, processing stages, processing methods, machine tool equipment, clamping methods, selected tools, programming strategies, cutting parameters, cooling and lubrication, spindle power, cutting force, cutting heat, surface morphology, tool wear, deformation, vibration, and surface integrity.

[0037] Step 3: Construct a processing quality and efficiency prediction mapping relationship model to establish the mapping relationship between process parameters and processing quality and efficiency prediction, such as Figure 4 As shown, the process parameters and processing quality and efficiency data of the process database are used as boundary inputs, and the model is continuously trained and iterated to obtain a processing quality and efficiency prediction model; ultimately, the accuracy of the prediction model is guaranteed.

[0038] The specific construction of the processing quality and efficiency prediction mapping relationship model is as follows: the process parameters are used as the input of the network structure, and the surface roughness, tool life, and workpiece fatigue life evaluation parameters are used as the output, specifically including a surface roughness prediction model, a tool life application prediction model, and a workpiece fatigue life prediction model, to achieve the mapping of process parameters and processing quality and efficiency. At the same time, the process parameters and processing quality and efficiency data of the process database are used as boundary inputs, and the model is continuously trained to ultimately ensure the accuracy of the prediction model.

[0039] The surface roughness prediction model is based on an artificial neural network, takes cutting feed, feed speed, and cutting depth as inputs of the network structure, and takes surface roughness as output, and establishes an N-Net multi-hidden layer neural network part cutting surface roughness prediction model that integrates local weights;

[0040] The tool life application prediction model collects data during the use of the tool, including cutting speed, cutting depth, feed speed, tool material information, cutting environment temperature and humidity, trains the neural network model, optimizes the model parameters, and evaluates the performance of the model through cross-validation methods to continuously improve the prediction accuracy of the model.

[0041] The workpiece fatigue life prediction model uses the morphology of the metal surface detected by white light interferometer, and after completing the filtering process of the detection point cloud data, completes the removal of sparse outliers and the smoothing of the point cloud data; secondly, based on the morphological feature image defects, grooves, and cracks, it identifies and finally realizes the fatigue life prediction;

[0042] Step 4: Develop a process optimization algorithm, build a process optimization model, conduct sample tests, and verify the optimization results of process parameters. If the processing effect meets the requirements, the process optimization model is built. If the processing effect does not meet the requirements, the process optimization model is iterated repeatedly until the processing effect meets the requirements.

[0043] The process optimization model is based on neural networks and genetic algorithms, with processing quality and efficiency as the optimization goal and process data matching the processing quality and efficiency evaluation requirements as the basic constraint condition;

[0044] This embodiment takes surface quality and processing efficiency as optimization goals, and takes processing equipment, processing materials, service environment, processing requirements and other parameters as basic constraints;

[0045] Step 5, using the working condition data contained in the process database as the original data, and embedding the processing quality and efficiency prediction model and the process optimization model into the process database;

[0046] Step 6: Apply the processing quality and efficiency prediction model and process optimization model of the process database to continuously accumulate and iterate the process data, thereby improving the accuracy of processing quality and efficiency prediction, optimizing process parameters, and improving processing quality and efficiency.

[0047] The processing quality and efficiency prediction model: enter the processing quality and efficiency prediction model of the process database, select the indicators to be predicted, input the initial conditions according to the requirements of the processing quality and efficiency prediction model, and output the prediction results; wherein the initial conditions include feed speed, cutting speed, cutting depth, part material, part structure, tool material, tool type, cutting method, and processing equipment, and the prediction result is the surface roughness Ra value;

[0048] The process optimization model: enter the process optimization model of the process database, select the initial conditions of CNC machining, output the optimized process parameters or process scheme, and choose whether to store the optimized process parameters or process scheme in the process database according to the prompt of the process database. The initial conditions include part materials, part features, processing equipment, processing methods, tool materials, tool categories, and processing process information;

[0049] The above description is only a preferred embodiment of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present disclosure is not limited to the technical solutions formed by a specific combination of the above-mentioned technical features, but should also cover other technical solutions formed by any combination of the above-mentioned technical features or their equivalent features without departing from the above-mentioned inventive concept. For example, the above-mentioned features are replaced with the technical features with similar functions disclosed in the embodiments of the present disclosure (but not limited to) to form a technical solution.

Claims

1. A method for collecting and optimizing the processing data of aircraft engine parts, characterized in that: The following steps are involved: Step 1: Summarize the processing data acquisition method, define the data type to be collected, and apply the corresponding acquisition method to obtain the processing data according to the data type; Step 2: Perform time-space mapping processing on the time domain data or combing data in the collected data and the geometric space position of the part processing, and store the processing data in the process database; Step 3: Construct a processing quality and efficiency prediction mapping relationship model, establish a mapping relationship between process parameters and processing quality and efficiency prediction, and use the process parameters and processing quality and efficiency data of the process database as boundary inputs to continuously train the iterative model to obtain a processing quality and efficiency prediction model; Step 4: Build a process optimization model, conduct sample tests, and verify the optimization results of process parameters. If the processing effect meets the requirements, the process optimization model is built. If the processing effect does not meet the requirements, the process optimization model is iterated repeatedly until the processing effect meets the requirements. Step 5: Use the working condition data contained in the process database as the original data, and embed the processing quality and efficiency prediction model and the process optimization model into the process database; Step 6: Apply the processing quality and efficiency prediction model and process optimization model of the process database to continuously accumulate and iterate the process data, thereby improving the accuracy of processing quality and efficiency prediction, optimizing process parameters, and improving processing quality and efficiency.

2. The method for collecting and optimizing the processing data of aircraft engine parts according to claim 1 is characterized in that: The data types in step 1 are divided into combing data and collection data; the combing data is data that can be directly obtained, and the collection data is data that needs to be detected and collected and cannot be directly obtained; The collection methods are divided into two categories. The collection scheme for collection data is as follows: based on the standard communication interface of CNC machine tools or a third-party interface, a spindle power or torque collection system is embedded to complete the collection of spindle power or torque data; the spindle motor current detection technology is used to realize cutting force and tool wear and damage detection; the vibration and deformation data during processing are detected by means of embedded non-disturbance sensors attached to the spindle and the workbench, on-machine probes, and the vibration monitoring of the motion axis of the CNC system itself; a three-dimensional morphology measuring instrument and a residual stress test and analysis system are used to measure and collect the three-dimensional morphology of the machined surface and the surface residual stress data; the collection method for combing data is as follows: according to the collection strategy and specifications, the type of data to be combed is defined, so that data collection is performed according to the rules.

3. The method for collecting and optimizing the processing data of aircraft engine parts according to claim 1 is characterized in that: The time-space mapping processing method described in step 2 is based on the tool trajectory of typical parts CNC machining, introduces a voxel model to discretize the machining process in time and space; adopts a short-time domain processing method to characterize the machining process signal as a corresponding short-time domain signal feature; The knowledge association between unit element working conditions and short-term machining process signals is established to form a spatiotemporal mapping model corresponding to the machining process data and the part machining position, so as to map the machining process data to the part space and store the machining process data in the process database.

4. The method for collecting and optimizing the processing data of aircraft engine parts according to claim 3 is characterized in that: The data stored in the process database are the working condition data of typical parts processing sites, including process routes, processing stages, processing methods, machine tool equipment, clamping methods, selected tools, programming strategies, cutting parameters, cooling and lubrication, spindle power, cutting force, cutting heat, surface morphology, tool wear, deformation, vibration, and surface integrity.

5. The method for collecting and optimizing the processing data of aircraft engine parts according to claim 3 is characterized in that: The construction of the processing quality and efficiency prediction mapping relationship model described in step 3 is specifically as follows: taking the process parameters as the input of the network structure, and taking the surface roughness, tool life, and workpiece fatigue life evaluation parameters as the output, specifically including the surface roughness prediction model, the tool life application prediction model, and the workpiece fatigue life prediction model, to achieve the mapping of process parameters and processing quality and efficiency.

6. The method for collecting and optimizing the processing data of aircraft engine parts according to claim 5 is characterized in that: The surface roughness prediction model is based on an artificial neural network, takes cutting feed, feed speed, and cutting depth as inputs of the network structure, and takes surface roughness as output, and establishes an N-Net multi-hidden layer neural network part cutting surface roughness prediction model that integrates local weights; The tool life application prediction model is to train the neural network model by collecting data during the use of the tool, including cutting speed, cutting depth, feed speed, tool material information, cutting environment temperature and humidity, optimize the model parameters, and evaluate the performance of the model through a cross-validation method; The workpiece fatigue life prediction model uses the morphology of the metal surface detected by white light interferometer. After completing the filtering processing of the detection point cloud data, the sparse outliers are removed and the point cloud data is smoothed; secondly, based on the morphological feature image defects, grooves, and cracks are identified, and finally the fatigue life prediction is realized.

7. The method for collecting and optimizing the processing data of aircraft engine parts according to claim 1 is characterized in that: The process optimization model described in step 4 is based on a neural network and a genetic algorithm, with processing quality and efficiency as the optimization target and process data matching the processing quality and efficiency evaluation requirements as the basic constraint condition.

8. The method for collecting and optimizing the processing data of aircraft engine parts according to claim 1 is characterized in that: The processing quality and efficiency prediction model described in step 6: enter the processing quality and efficiency prediction model of the process database, select the indicators to be predicted, input the initial conditions according to the requirements of the processing quality and efficiency prediction model, and output the prediction results; The initial conditions include feed speed, cutting speed, cutting depth, part material, part structure, tool material, tool type, cutting method, and processing equipment, and the predicted result is the surface roughness Ra value; The process optimization model: enters the process optimization model of the process database, selects the initial conditions of CNC machining, outputs the optimized process parameters or process scheme, and at the same time, according to the prompt of the process database, selects whether to store the optimized process parameters or process scheme in the process database; The initial conditions include part material, part features, processing equipment, processing method, tool material, tool type, and processing procedure information.

Images