CN116485032A - Aviation product processing quality prediction method considering multidimensional influence factors - Google Patents
Aviation product processing quality prediction method considering multidimensional influence factors Download PDFInfo
- Publication number
- CN116485032A CN116485032A CN202310528162.1A CN202310528162A CN116485032A CN 116485032 A CN116485032 A CN 116485032A CN 202310528162 A CN202310528162 A CN 202310528162A CN 116485032 A CN116485032 A CN 116485032A
- Authority
- CN
- China
- Prior art keywords
- quality
- processing
- prediction model
- data
- quality prediction
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000012545 processing Methods 0.000 title claims abstract description 163
- 238000000034 method Methods 0.000 title claims abstract description 54
- 238000012549 training Methods 0.000 claims abstract description 44
- 238000012360 testing method Methods 0.000 claims abstract description 28
- 230000004927 fusion Effects 0.000 claims abstract description 15
- 238000007781 pre-processing Methods 0.000 claims abstract description 5
- 238000013441 quality evaluation Methods 0.000 claims description 20
- 239000002994 raw material Substances 0.000 claims description 17
- 230000006870 function Effects 0.000 claims description 15
- 230000004913 activation Effects 0.000 claims description 14
- 230000008569 process Effects 0.000 claims description 13
- 238000003672 processing method Methods 0.000 claims description 13
- 101001095088 Homo sapiens Melanoma antigen preferentially expressed in tumors Proteins 0.000 claims description 9
- 102100037020 Melanoma antigen preferentially expressed in tumors Human genes 0.000 claims description 9
- 238000001514 detection method Methods 0.000 claims description 8
- 238000005516 engineering process Methods 0.000 claims description 5
- 238000005299 abrasion Methods 0.000 claims description 4
- 238000005520 cutting process Methods 0.000 claims description 4
- 238000005286 illumination Methods 0.000 claims description 4
- 239000004973 liquid crystal related substance Substances 0.000 claims description 4
- 238000005461 lubrication Methods 0.000 claims description 4
- 230000007246 mechanism Effects 0.000 claims description 4
- 238000005070 sampling Methods 0.000 claims description 4
- 230000010354 integration Effects 0.000 abstract description 4
- 238000004519 manufacturing process Methods 0.000 description 10
- 238000007689 inspection Methods 0.000 description 9
- 238000004364 calculation method Methods 0.000 description 8
- 238000011156 evaluation Methods 0.000 description 7
- 238000007726 management method Methods 0.000 description 6
- 238000003326 Quality management system Methods 0.000 description 4
- 238000013528 artificial neural network Methods 0.000 description 3
- 210000004027 cell Anatomy 0.000 description 3
- 238000011425 standardization method Methods 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 230000000875 corresponding effect Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000007613 environmental effect Effects 0.000 description 2
- 230000006996 mental state Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000003449 preventive effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06395—Quality analysis or management
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/04—Manufacturing
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Abstract
An aviation product processing quality prediction method considering multidimensional influence factors relates to the field of processing quality prediction. The method aims to solve the problems that the existing aviation product processing quality prediction method is low in prediction precision, lacks complete data of effective association integration and cannot uniformly predict the processing quality of aviation products. The invention comprises the following steps: acquiring and preprocessing related data of the processing quality of the aviation product; sliding grouping the preprocessed data, and dividing the data after sliding grouping into a training set and a testing set; constructing a quality prediction model, respectively learning quality data features from transverse and longitudinal influence factors, and fusing transverse and longitudinal features through a model fusion layer to obtain a multi-input quality prediction model; evaluating the trained quality prediction model by using the trained quality prediction model, wherein the quality prediction model with the best performance is a multi-input quality prediction model; and obtaining the processing quality of the aviation product by using the multi-input quality prediction model. The method is used for predicting the processing quality of the aviation product.
Description
Technical Field
The invention relates to the field of processing quality prediction, in particular to an aviation product processing quality prediction method considering multidimensional influence factors.
Background
The aviation industry is taken as the key development field of manufacturing industry, is a specific embodiment of national comprehensive national force and core competitiveness, is a knowledge-intensive, technology-intensive and multidisciplinary integrated industry, is a typical representative of multi-variety, small-batch and discrete manufacturing, and is very important in quality management of complex aviation products because of extremely high complexity, numerous parts and components and high requirements on safety and reliability.
The traditional quality management is simply based on quality detection, and cannot timely and accurately control the quality in advance, and in recent years, with rapid development and application of technologies such as big data, artificial intelligence and the like, preventive control of product quality by predicting the processing quality through deep mining of processing process data by an intelligent algorithm gradually becomes a research hotspot. However, the following problems still exist in the current research on the prediction of the processing quality of complex aviation products:
1. the processing quality of the aviation product is influenced by factors such as processing personnel, processing equipment, raw materials, processing methods, processing environment, measuring modes and the like, meanwhile, a quality detection result shows a time sequence change trend in a time dimension, the processing quality can be influenced by the influence of various transverse factors or the longitudinal time sequence change trend, and the conventional complex aviation product processing quality prediction model does not consider transverse and longitudinal multi-dimensional factors, so that the prediction precision is low.
2. The processing quality of the complex aviation product is affected by multidimensional factors, related data are distributed in a plurality of systems, the data volume is huge and complex, and the data island can not be formed by direct intercommunication due to different data acquisition, uploading and management modes of each system, so that the existing complex aviation product processing quality prediction method is lack of complete data of effective association integration.
3. The production types of the complex aviation products are various small batches, the products are various, the size of each product to be processed is complicated, and the batch is small, but the existing complex aviation product processing quality prediction method cannot unify the sizes of the products with different orders of magnitude, so that the processing quality of the aviation products cannot be predicted uniformly.
Disclosure of Invention
The invention aims to solve the problems that the existing aviation product processing quality prediction method is low in prediction precision, lacks complete data of effective association integration and cannot uniformly predict the aviation product processing quality, and provides an aviation product processing quality prediction method considering multidimensional influence factors.
The aviation product processing quality prediction method considering multidimensional influence factors specifically comprises the following steps: acquiring processing quality related data of the aviation product to be predicted, and inputting the processing quality related data of the aviation product to be predicted into a multi-input quality prediction model to acquire the processing quality of the aviation product;
the aerospace product processing quality association data comprises: a lateral influencing factor, a longitudinal influencing factor;
the lateral influencing factors include: processing personnel, processing equipment, raw materials, a processing method, a processing environment and a measuring mode;
the longitudinal influencing factors are historical quality detection results arranged according to the processing time;
the multi-input quality prediction model is obtained by:
step one, acquiring processing quality associated data of aviation products;
step two, preprocessing the processing quality related data of the aviation product;
the pretreatment comprises the following steps: performing feature coding on the aviation product processing quality associated data; performing standardization on the aviation product processing quality associated data tag column to obtain a quality evaluation index;
step three, sliding grouping is carried out on the preprocessed complex aviation product processing quality related data, and the preprocessed aviation product processing quality related data after sliding grouping is divided into a training set and a testing set;
step four, constructing a quality prediction model, training the quality prediction model by utilizing a training set to obtain a trained quality prediction model, then evaluating the trained quality prediction model, and taking the trained quality prediction model with the best performance as a multi-input quality prediction model;
in the process of training the quality prediction model by using the training set, the quality prediction model learns the characteristics from the transverse influence factors and the longitudinal influence factors, and fuses the two characteristics.
Further, the lateral influencing factors include: the processing personnel, processing equipment, raw materials, processing method, processing environment and measuring mode are as follows:
the processing personnel include: age, team, continuous working time, work age, technical grade of personnel;
the processing apparatus includes: manufacturer, model and machine tool of machine tool, cutter, clamp, and use duration of machine tool, abrasion degree of cutter, lubrication degree of cutter, cutting parameters of cutter, clamping mode of clamp and clamping force;
the processing method comprises the following steps: processing technology, processing pressure and processing temperature;
the raw materials comprise: raw material batch, supplier, quality grade;
the processing environment includes: ambient temperature, humidity, noise, illumination;
the measuring mode comprises the following steps: measuring tool, measuring method, measuring accuracy.
Further, the quality evaluation index has the following formula:
further, in the third step, the preprocessed complex aviation product processing quality related data is subjected to sliding grouping, and the preprocessed aviation product processing quality related data after sliding grouping is divided into a training set and a testing set, which comprises the following steps:
firstly, grouping the data processed in the second step by setting a sliding window, wherein the sliding window has a size of timetable, namely, the previous timetable label column data is used as a longitudinal influence factor, the timetable+1 characteristic column data is used as a transverse influence factor, and the transverse and longitudinal influence factors are used as inputs to predict the timetable+1 label column data, and the specific formula is as follows:
x1 t ＝[F t (1),F t (2)…F t (m)…F t (M)]
x2 t ＝[Y t-1 ,Y t-2 …Y t-timestep ]
x t ＝[x1 t ,x2 t ]
wherein x1 t Representing transverse feature input of sample t, F t (M) represents the mth lateral influence factor of the sample t, M is the total number of lateral influence factors, x2 t Represents the longitudinal characteristic input of the sample t, yt-timer represents the quality evaluation index of the previous timer at the moment of the sample t, x t Is the input of a quality prediction model; then, randomly disturbing the preprocessed aviation product processing quality related data after sliding grouping, and carrying out layered sampling and dividing a training set and a testing set in the range of a quality evaluation index value;
wherein each set of data in the preprocessed aviation product processing quality associated data after sliding grouping is a sample.
Further, the quality prediction model includes: the device comprises a feature learning layer, a feature fusion layer, a full connection layer, a Dropout layer and an output layer;
the feature learning layer includes: a lateral influence factor characteristic acquisition unit and a longitudinal influence factor characteristic acquisition unit;
the lateral influence factor characteristic acquisition unit: learning quality data features from transverse influencing factors in the training set and the testing set; the lateral influence factor characteristic acquisition unit includes: LSTM layer, full connection layer, dropout layer;
the LSTM layer comprises a plurality of LSTM networks;
the longitudinal influence factor feature acquisition unit: learning quality data features from transverse influencing factors in the training set and the testing set; the longitudinal influence factor characteristic acquisition unit includes: attention mechanisms, LSTM layers, full connection layers, dropout layers;
the feature fusion layer: fusing the features learned from the training set and the features learned from the testing set;
the full connection layer: connecting the features output by the feature fusion layer into global features;
the Dropout layer: preventing the quality prediction model from being over fitted;
the output layer: and outputting the predicted processing quality of the aviation product.
Further, the output of the LSTM layer is of the formula:
h t ＝o t ·Activation(c t )
c t ＝f t ·c t-1 +i t ·C t
C t ＝Activation(ω c ·(h t-1 ,x t )+b c )
f t ＝RecurrentActivation(ω f ·(h t-1 ,x t )+b f )
i t ＝RecurrentActivation(ω i ·(h t-1 ,x t )+b i )
o t ＝RecurrentActivation(ω o ·(h t-1 ,x t )+b o )
wherein f t Is a forgetful door, i t Is an input door o t Is an output door C t Is a process value for updating the state of the cell c t Is the state of sample t, h t Is the final output of the LSTM layer, and the RecurrentActivate and the Activate are nonlinear Activation functions, omega f 、ω i 、ω o 、ω c Representing weights, b f 、b i 、b o 、b c Indicating the deviation.
Further, the training quality prediction model employs the following loss function:
where t is the sample number, n is the total number of samples, y t Representing the true value of the sample,representing the sample prediction value.
Further, in the fourth step, the trained quality prediction model is evaluated, and the average absolute percentage error MAPE is adopted for evaluation, wherein the average absolute percentage error MAPE is as follows:
wherein, the liquid crystal display device comprises a liquid crystal display device,and (3) representing the average value of the true values of the samples, wherein the trained quality prediction model corresponding to the MAPE minimum value is the trained quality prediction model with the best performance.
The beneficial effects of the invention are as follows:
according to the complex aviation product processing quality prediction method considering the multidimensional influence factors under the data driving, the processing quality influence factors are subdivided into transverse factors, namely processing personnel, processing equipment, raw materials, processing methods, processing environments and measuring modes, and longitudinal factors, namely historical quality detection results arranged according to processing time, the influence of the multidimensional influence factors on the processing quality in different dimensions is comprehensively considered and fused into a prediction model, so that the processing quality prediction effect and precision are improved, and the processing cost is reduced.
According to the invention, the complex aviation product is considered to be influenced by multidimensional factors, and related data are distributed in a plurality of systems, so that the related data are correlated and integrated through business logic among the systems after the related data are acquired from different systems, so that the quality problem is accurately analyzed, and the complete data of effective correlation and integration are obtained.
According to the invention, the complex aviation product production type is considered to be multiple varieties and small batches, the variety of products is numerous, the size of each product to be processed is complicated, and the batch is smaller, so that the unified quality evaluation index is required to expand the data volume. Because the standard values of different sizes differ more, and the measured values are not in the same order of magnitude, the quality inspection result is mapped between [ -1,1] by the data standardization method considering the size characteristics, so that the quality evaluation index is unified, the physical characteristics of the processing size are reserved, and the processing quality of products in different sizes is unified by unifying various size data under the same evaluation standard.
According to the complex aviation product processing quality prediction model taking multidimensional influence factors into consideration under data driving, quality data features are respectively learned from transverse and longitudinal influence factors, transverse and longitudinal features are fused through the model fusion layer, meanwhile, the processing quality influence factors such as people, machines, materials, methods, rings and measurement and time sequence features of quality data are considered, prediction accuracy is effectively improved, processing quality of complex aviation products can be improved, production cost is reduced, and the complex aviation product processing quality prediction model has high practical value.
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a schematic diagram of a multi-input quality prediction model;
FIG. 3 is a diagram showing the comparison of predicted values and actual values of a training set according to an embodiment of the present invention;
FIG. 4 is a graph showing the comparison of predicted values and actual values of a test set according to an embodiment of the present invention.
Detailed Description
The first embodiment is as follows: as shown in fig. 1, the method for predicting the processing quality of an aviation product according to the embodiment, which considers multidimensional influencing factors, includes: acquiring complex aviation product processing quality related data to be predicted, and inputting the complex aviation product processing quality related data to be predicted into a multi-input quality prediction model to acquire the processing quality of the complex aviation product;
the aerospace product processing quality association data comprises: a lateral influencing factor, a longitudinal influencing factor;
the lateral influencing factors include: processing personnel, processing equipment, raw materials, a processing method, a processing environment and a measuring mode;
the longitudinal influencing factors are the historical quality detection results arranged according to the processing time.
The multi-input quality prediction model is obtained by:
step one, acquiring processing quality associated data of complex aviation products;
the method for acquiring the processing quality associated data of the complex aviation product, which is a multidimensional factor affecting the processing quality, from various digitization systems of a workshop comprises the following steps:
a lateral influencing factor comprising: processing personnel, processing equipment, raw materials, a processing method, a processing environment and a measuring mode;
the processing personnel specifically comprise the age, team, continuous working time, working age and technical grade of the personnel, so that the influence of the mental state and the proficiency of the processing personnel is reflected;
the processing equipment comprises a machine tool, a cutter, a manufacturer and a model of a clamp, the use time of the machine tool, the abrasion degree of the cutter, the lubrication degree of the cutter, the cutting parameters of the cutter, the clamping mode of the clamp, the clamping force and the like, so that the influence of the working parameters and the working state of the equipment is reflected;
the processing method comprises a processing technology, processing pressure and processing temperature, so that the influence of processing parameters is reflected;
the raw materials comprise batches, suppliers and quality grades thereof, so that the influence of the quality of the raw materials is reflected;
the processing environment comprises environment temperature, humidity, noise and illumination, so that the influence of environmental factors is reflected;
the measuring mode comprises a measuring tool, a measuring method and measuring precision, so that the influence of measuring errors is reflected.
Longitudinal influencing factors: historical quality detection results arranged according to processing time;
the processing quality of complex aviation products is affected by multidimensional factors, related data are distributed in a plurality of systems, and data islands can not be formed by direct intercommunication of different data acquisition, uploading and management modes of each system, so that the related data are acquired from different systems and then are associated and integrated through business logic among the systems to accurately analyze the quality problem.
Step two, preprocessing the complex aviation product processing quality associated data:
step two, performing feature coding on the complex aviation product processing quality associated data;
the feature format of the original data related to the processing quality of the complex aviation product is complex, the number of non-numerical discrete features is large, and the non-numerical discrete features are processed by adopting a tag coding method.
Step two, standardizing a label column of the complex aviation product processing quality related data, namely a quality inspection result column, and obtaining a quality evaluation index:
the complex aviation products are small in production type and large in quantity, the sizes of the products are large in variety and each product is required to be processed, the quantity of the products is small, and the data of different products and sizes are required to be predicted uniformly by using the same quality evaluation index so as to enlarge the data quantity. Because the standard values of different sizes differ more, the quality inspection results are not in the same order of magnitude, and the quality inspection results need to be mapped into the same range in a reasonable way to unify quality evaluation indexes. The invention adopts a data standardization method considering the size characteristics to process and standardize the quality inspection result, and takes the standardized result as a quality evaluation index;
the data normalization method considering the size characteristics ensures that the result is mapped between [ -1,1] and accords with the physical characteristics of the processing size, and the calculation formula is as follows:
step three, sliding grouping is carried out on the preprocessed complex aviation product processing quality associated data, and the preprocessed complex aviation product processing quality associated data after sliding grouping is divided into a training set and a testing set:
firstly, grouping the data processed in the second step by setting a sliding window, wherein the sliding window has a size of timetable, namely, the previous timetable label column data is used as a longitudinal influence factor, the timetable+1 characteristic column data is used as a transverse influence factor, and the transverse and longitudinal influence factors are used as inputs to predict the timetable+1 label column data, and the specific formula is as follows:
x1 t ＝[F t (1),F t (2)…F t (m)…F t (M)]
x2 t ＝[Y t-1 ,Y t-2 …Y t-timestep ]
x t ＝[x1 t ,x2 t ]
wherein x1 t Representing transverse feature input of sample t, F t (M) represents the mth lateral influence factor of the sample t, M is the total number of lateral influence factors, x2 t Represents the longitudinal characteristic input of the sample t, yt-timer represents the quality evaluation index of the previous timer at the moment of the sample t, x t Is the input of a quality prediction model;
then, randomly disturbing the data after grouping, and carrying out hierarchical sampling and dividing a training set and a testing set in a range where the quality evaluation index value is located, wherein the ratio of the training set to the testing set is 7:3.
wherein, a group of data in the training set and the testing set is a sample;
step four, constructing a quality prediction model, training the quality prediction model by using a training set to obtain a trained quality prediction model, then evaluating the trained quality prediction model, and taking the trained quality prediction model with the best performance as a multi-input quality prediction model, wherein the method comprises the following steps of:
step four, constructing a quality prediction model;
the quality prediction model comprises: the device comprises a feature learning layer, a feature fusion layer, a full connection layer, a Dropout layer and an output layer;
the feature learning layer includes: a lateral influence factor characteristic acquisition unit and a longitudinal influence factor characteristic acquisition unit;
the lateral influence factor characteristic acquisition unit: learning quality data features from transverse influencing factors in the training set and the testing set; the lateral influence factor characteristic acquisition unit includes: LSTM layer, full connection layer, dropout layer;
the LSTM layer comprises a plurality of LSTM networks;
the longitudinal influence factor feature acquisition unit: learning quality data features from transverse influencing factors in the training set and the testing set; the longitudinal influence factor characteristic acquisition unit includes: attention mechanisms, LSTM layers, full connection layers, dropout layers;
the feature fusion layer: fusing the features learned from the training set and the features learned from the testing set; the feature fusion layer is realized by adopting a Concate function, and the following formula is adopted:
merged＝Concatenate([output1,output2])
wherein output1 is a learned quality data transverse feature, output2 is a learned quality data longitudinal feature, and Concate is a fusion layer function;
the full connection layer is a Dense layer: connecting all the extracted local features into global features;
dropout layer: preventing overfitting of the network by modifying the number of hidden layer neurons;
the output layer is a Dense layer: for outputting the result.
The long-short-term memory neural network, namely an LSTM layer, is composed of chain units, each unit comprises a forgetting gate, an input gate and an output gate, and the calculation formula is as follows:
f t ＝RecurrentActivation(ω f ·(h t-1 ,x t )+b f )
i t ＝RecurrentActivation(ω i ·(h t-1 ,x t )+b i )
o t ＝RecurrentActivation(ω o ·(h t-1 ,x t )+b o )
C t ＝Activation(ω c ·(h t-1 ,x t )+b c )
c t ＝f t ·c t-1 +i t ·C t
h t ＝o t ·Activation(c t )
wherein f t Indicating forgetful door, i t Represents an input gate, o t Indicating the output gate, C t To update the process value of the cell state c t Representing the state of sample t, h t Representing the final output of the LSTM layer, the RecurrentActivate and Activate are nonlinear Activation functions, ω f 、ω i 、ω o 、ω c Representing weights, b f 、b i 、b o 、b c Representing weights and deviations;
wherein h is t Outputting the sample predicted value after passing through the full connection layer
Training a quality prediction model by using a training set to obtain a trained quality prediction model;
the training quality prediction model adopts the following loss function:
with mean square errorAs a loss function of the model, t is the sample number and n is the total number of samples.
Step four, performing performance evaluation on the trained quality prediction model, and taking the trained quality prediction model with the best performance as a multi-input quality prediction model:
the average absolute percentage error MAPE is used as an evaluation index to evaluate the model, and the performance of the quality prediction model is verified;
selecting a trained quality prediction model corresponding to the MAPE minimum value as a multi-input quality prediction model; the calculation formula is as follows:
wherein y is t Representing the true value of the sample,representing a sample prediction value, +.>Representing the mean of the true values.
Examples:
the invention is illustrated by taking an aircraft casing product as an example, and fig. 1 is a flow chart for predicting the processing quality of a complex aircraft product.
Step one, acquiring big data related to the processing quality of complex aviation products. The method comprises the steps of obtaining multidimensional factors influencing the processing quality from various digitization systems of workshops, wherein the multidimensional factors comprise transverse influencing factors, namely processing personnel, processing equipment, raw materials, a processing method, a processing environment, a measuring mode and longitudinal influencing factors, namely historical quality detection results arranged according to processing time. Wherein the processing personnel specifically comprise the age, team, continuous working time, working age and technical grade of the personnel, so as to reflect the influence of the mental state and the proficiency of the processing personnel; the processing equipment comprises a machine tool, a cutter, a manufacturer and a model of a clamp, the use time of the machine tool, the abrasion degree of the cutter, the lubrication degree of the cutter, the cutting parameters of the cutter, the clamping mode of the clamp, the clamping force and the like, so that the influence of the working parameters and the working state of the equipment is reflected; the raw materials comprise batches, suppliers and quality grades thereof, so that the influence of the quality of the raw materials is reflected; the processing method comprises a processing technology, processing pressure and processing temperature, so that the influence of processing parameters is reflected; the processing environment comprises environment temperature, humidity, noise and illumination, so that the influence of environmental factors is reflected; the measuring mode comprises a measuring tool, a measuring method and measuring precision, so that the influence of measuring errors is reflected.
The enterprise workshop various digitization systems comprise a Manufacturing Execution System (MES), a Quality Management System (QMS), a resource management System (SAP) and workshop detail manufacturing data and a process system (MDC), wherein product processing related data are derived from the manufacturing execution system, the Quality Management System (QMS), the resource management System (SAP) and the workshop detail manufacturing data and the process system (MDC), the product processing related data comprise a quality inspection table, a business flow water meter, a work flow water meter, an order and delivery situation table, quality inspection information, business information, work information and order information of processed products are obtained, data in different systems are associated and integrated according to keywords such as part numbers, size numbers and work time, product processing quality associated big data are obtained, and the data are ordered according to the time of reporting as unique ids.
And step two, preprocessing data.
The feature codes adopt a label coding method to process non-numerical discrete features, such as W1, W2 and the like of different processing workers.
The data standardization method considering the size characteristics is adopted to standardize the quality inspection result, and the standardized result is used as a quality evaluation index.
If in the embodiment, a piece of data is a quality inspection result of 9.12, the size reference value is 9.00, the size lower limit is 8.80, and the size upper limit is 9.20, the following calculation is performed to obtain a standardized result:
dimensional tolerance=9.20-8.80=0.40
Step three, grouping the data processed in the step two by setting a sliding window, wherein the size of the sliding window is timetep, namely, the previous timetep tag column data is used as a longitudinal influence factor, the timetep+1th feature column data is used as an influence factor of a transverse influence factor, and the transverse and longitudinal influence factors are used as input to predict the timetep+1th tag column data, and the specific formula is as follows:
x1 t ＝[F t (1),F t (2)…F t (m)…F t (M)]
x2 t ＝[Y t-1 ,Y t-2 …Y t-timestep ]
x t ＝[x1 t ,x2 t ]
wherein x1 t Representing transverse feature input of sample t, F t (M) represents the mth lateral influence factor of the sample t, M is the total number of lateral influence factors, x2 t Represents the longitudinal characteristic input of the sample t, yt-timer represents the quality evaluation index of the previous timer at the moment of the sample t, x t Is the input of a quality prediction model;
randomly disturbing the data after grouping, and carrying out hierarchical sampling to divide a training set and a testing set according to the range of the quality evaluation index value;
in this embodiment, the transverse feature input includes a processing worker, a processing machine tool, a processing order, and a processing process, and the longitudinal feature input is a quality evaluation index of the first 50 times, that is, timetep=50.
Step four, designing a neural network to construct a multi-input quality prediction model, wherein the model structure is as shown in fig. 2, and comprises the following steps: the device comprises a feature learning layer, a feature fusion layer, a full connection layer, a Dropout layer and an output layer;
the feature learning layer includes: a lateral influence factor characteristic acquisition unit and a longitudinal influence factor characteristic acquisition unit;
the lateral influence factor characteristic acquisition unit: learning quality data features from transverse influencing factors in the training set and the testing set; the lateral influence factor characteristic acquisition unit includes: LSTM layer, full connection layer, dropout layer;
the LSTM layer comprises a plurality of LSTM networks;
the longitudinal influence factor feature acquisition unit: learning quality data features from transverse influencing factors in the training set and the testing set; the longitudinal influence factor characteristic acquisition unit includes: attention mechanisms, LSTM layers, full connection layers, dropout layers;
the feature fusion layer: fusing the features learned from the training set and the features learned from the testing set;
the long-short-term memory neural network of the quality prediction model, namely an LSTM layer, is composed of chain units, each unit comprises a forgetting gate, an input gate and an output gate, and the calculation formula is as follows:
f t ＝RecurrentActivation(ω f ·(h t-1 ,x t )+b f )
i t ＝RecurrentActivation(ω i ·(h t-1 ,x t )+b i )
o t ＝RecurrentActivation(ω o ·(h t-1 ,x t )+b o )
C t ＝Activation(ω c ·(h t-1 ,x t )+b c )
c t ＝f t ·c t-1 +i t ·C t
h t ＝o t ·Activation(c t )
wherein f t Indicating forgetful door, i t Represents an input gate, o t Indicating the output gate, C t To update the process value of the cell state c t Representing the state of sample t, h t Represents the final output, ω, of the LSTM layer f 、ω i 、ω o 、ω c Representing weights, b f 、b i 、b o 、b c The weight and the deviation are represented, the RecurrentActivate and the Activate are nonlinear Activation functions, the RecurrentActivate is an Activation function of a forgetting gate, an input gate and an output gate, the RecurrentActivate is set as Hard_sigmoid in the embodiment, the Activate is an Activation function of updating the state of a unit and final output, and the RecurrentActivate is set as the Relu, omega and b in the embodiment, the weight and the deviation are represented.
And fifthly, training a quality prediction model. Design the loss function to mean square errorAs a function of the loss of the model. The parameter settings of the quality prediction model are shown in table 1 below.
TABLE 1 multiple input quality prediction model training parameters
And step six, performing performance evaluation on the quality prediction model to obtain a multi-input quality prediction model.
The model was trained 1000 times according to the parameters of table 1, and the model was evaluated using the mean absolute percentage error MAPE as an evaluation index, verifying the performance of the quality prediction model, and the calculation formula was as follows:
wherein n represents the total number of samples in the test set, y t Representing the true value of the sample,representing a sample prediction value, +.>Representing the mean of the true values. The comparison of the true value and the predicted value of the training set and the test set is shown in fig. 3 and 4, the performance index results are shown in the following table 2, and the prediction precision of the quality prediction model in the test set reaches 90.69%.
Table 2 evaluation index results
The above examples of the present invention are only for describing the calculation model and calculation flow of the present invention in detail, and are not limiting of the embodiments of the present invention. Other variations and modifications of the above description will be apparent to those of ordinary skill in the art, and it is not intended to be exhaustive of all embodiments, all of which are within the scope of the invention.
Claims (9)
1. A method for predicting the processing quality of an aviation product by considering multidimensional influence factors is characterized by comprising the following specific processes: acquiring processing quality related data of the aviation product to be predicted, and inputting the processing quality related data of the aviation product to be predicted into a multi-input quality prediction model to acquire the processing quality of the aviation product;
the aerospace product processing quality association data comprises: a lateral influencing factor, a longitudinal influencing factor;
the lateral influencing factors include: processing personnel, processing equipment, raw materials, a processing method, a processing environment and a measuring mode;
the longitudinal influencing factors are historical quality detection results arranged according to the processing time;
the multi-input quality prediction model is obtained by:
step one, acquiring processing quality associated data of aviation products;
step two, preprocessing the processing quality related data of the aviation product;
the pretreatment comprises the following steps: performing feature coding on the aviation product processing quality associated data; performing standardization on the aviation product processing quality associated data tag column to obtain a quality evaluation index;
step three, sliding grouping is carried out on the preprocessed complex aviation product processing quality related data, and the preprocessed aviation product processing quality related data after sliding grouping is divided into a training set and a testing set;
step four, constructing a quality prediction model, training the quality prediction model by utilizing a training set to obtain a trained quality prediction model, then evaluating the trained quality prediction model, and taking the trained quality prediction model with the best performance as a multi-input quality prediction model;
in the process of training the quality prediction model by using the training set, the quality prediction model learns the characteristics from the transverse influence factors and the longitudinal influence factors, and fuses the two characteristics.
2. The method for predicting the processing quality of an aerospace product by considering multidimensional influencing factors as recited in claim 1, wherein the method comprises the following steps of: the lateral influencing factors include: the processing personnel, processing equipment, raw materials, processing method, processing environment and measuring mode are as follows:
the processing personnel include: age, team, continuous working time, work age, technical grade of personnel;
the processing apparatus includes: manufacturer, model and machine tool of machine tool, cutter, clamp, and use duration of machine tool, abrasion degree of cutter, lubrication degree of cutter, cutting parameters of cutter, clamping mode of clamp and clamping force;
the processing method comprises the following steps: processing technology, processing pressure and processing temperature;
the raw materials comprise: raw material batch, supplier, quality grade;
the processing environment includes: ambient temperature, humidity, noise, illumination;
the measuring mode comprises the following steps: measuring tool, measuring method, measuring accuracy.
3. The method for predicting the processing quality of an aerospace product by considering multidimensional influencing factors as recited in claim 1, wherein the method comprises the following steps of: the quality evaluation index has the following formula:
4. a method for predicting the processing quality of an aerospace product by taking into account multi-dimensional influencing factors as defined in claim 3, wherein: in the third step, the preprocessed complex aviation product processing quality related data is subjected to sliding grouping, and the preprocessed aviation product processing quality related data after sliding grouping is divided into a training set and a testing set, and the method comprises the following steps:
firstly, grouping the data processed in the second step by setting a sliding window, wherein the sliding window has a size of timetable, namely, the previous timetable label column data is used as a longitudinal influence factor, the timetable+1 characteristic column data is used as a transverse influence factor, and the transverse and longitudinal influence factors are used as inputs to predict the timetable+1 label column data, and the specific formula is as follows:
x1 t ＝[F t (1)，F 1 (2)...F t (m)...F t (M)]
x2 t ＝[Y t-1 ，Y t-2 ...Y t-timestep ]
x t ＝[x1 t ，x2 t ]
wherein x1 t Representing transverse feature input of sample t, F t (M) represents the mth lateral influence factor of the sample t, M is the total number of lateral influence factors, x2 t Represents the longitudinal characteristic input of the sample t, yt-timer represents the quality evaluation index of the previous timer at the moment of the sample t, x t Is the input of a quality prediction model;
then, randomly disturbing the preprocessed aviation product processing quality related data after sliding grouping, and carrying out layered sampling and dividing a training set and a testing set in the range of a quality evaluation index value;
wherein each set of data in the preprocessed aviation product processing quality associated data after sliding grouping is a sample.
5. The method for predicting the processing quality of an aerospace product by considering multidimensional influencing factors as recited in claim 4, wherein the method comprises the following steps of: the quality prediction model comprises: the device comprises a feature learning layer, a feature fusion layer, a full connection layer, a Dropout layer and an output layer;
the feature learning layer includes: a lateral influence factor characteristic acquisition unit and a longitudinal influence factor characteristic acquisition unit;
the lateral influence factor characteristic acquisition unit: learning quality data features from transverse influencing factors in the training set and the testing set; the lateral influence factor characteristic acquisition unit includes: LSTM layer, full connection layer, dropout layer;
the LSTM layer comprises a plurality of LSTM networks;
the longitudinal influence factor feature acquisition unit: learning quality data features from transverse influencing factors in the training set and the testing set; the longitudinal influence factor characteristic acquisition unit includes: attention mechanisms, LSTM layers, full connection layers, dropout layers;
the feature fusion layer: fusing the features learned from the training set and the features learned from the testing set;
the full connection layer: connecting the features output by the feature fusion layer into global features;
the Dropout layer: preventing the quality prediction model from being over fitted;
the output layer: and outputting the predicted processing quality of the aviation product.
6. The method for predicting the processing quality of an aerospace product by considering multidimensional influencing factors as recited in claim 5, wherein the method comprises the following steps of: the output of the LSTM layer is of the formula:
h t ＝o t ·Activation(c t )
c t ＝f t ·c t-1 +i t ·C t
C t ＝Activation(ω c ·(h t-1 ，x t )+b c )
f t ＝RecurrentActivation(ω f ·(h t-1 ，x t )+b f )
i t ＝RecurrentActivation(ω i ·(h t-1 ，x t )+b i )
o t ＝RecurrentActivation(ω o ·(h t-1 ，x t )+b o )
wherein f t Is a forgetful door, i t Is an input door o t Is an output door C t Is a process value for updating the state of the cell c t Is the state of sample t, h t Is the final transport of the LSTM layerGo out, recurrentActivate and Activate are nonlinear Activation functions, ω f 、ω i 、ω o 、ω c Representing weights, b f 、b i 、b o 、b c Indicating the deviation.
7. The method for predicting the processing quality of an aerospace product by considering multidimensional influencing factors as recited in claim 6, wherein the method comprises the following steps of: the training quality prediction model adopts the following loss function:
where t is the sample number, n is the total number of samples, y t Representing the true value of the sample,representing the sample prediction value.
8. The method for predicting the processing quality of an aerospace product by considering multidimensional influencing factors as recited in claim 7, wherein the method comprises the following steps of: the feature fusion layer is realized by adopting a Concate function.
9. The method for predicting the processing quality of an aerospace product by considering multidimensional influencing factors as recited in claim 8, wherein the method comprises the following steps of: and in the fourth step, evaluating the trained quality prediction model by adopting an average absolute percentage error MAPE, wherein the average absolute percentage error MAPE is as follows:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing the mean of the true values of the samples, trained quality predictions corresponding to MAPE minimaThe model is the best-performance quality prediction model with good training.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310528162.1A CN116485032A (en) | 2023-05-11 | 2023-05-11 | Aviation product processing quality prediction method considering multidimensional influence factors |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310528162.1A CN116485032A (en) | 2023-05-11 | 2023-05-11 | Aviation product processing quality prediction method considering multidimensional influence factors |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116485032A true CN116485032A (en) | 2023-07-25 |
Family
ID=87216320
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310528162.1A Pending CN116485032A (en) | 2023-05-11 | 2023-05-11 | Aviation product processing quality prediction method considering multidimensional influence factors |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116485032A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117592865A (en) * | 2023-12-21 | 2024-02-23 | 中国人民解放军军事科学院系统工程研究院 | Equipment spare part quality state prediction method and device |
-
2023
- 2023-05-11 CN CN202310528162.1A patent/CN116485032A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117592865A (en) * | 2023-12-21 | 2024-02-23 | 中国人民解放军军事科学院系统工程研究院 | Equipment spare part quality state prediction method and device |
| CN117592865B (en) * | 2023-12-21 | 2024-04-05 | 中国人民解放军军事科学院系统工程研究院 | Equipment spare part quality state prediction method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Mittag et al. | Statistical methods of quality assurance | |
| Evans et al. | The management and control of quality | |
| Topalidou et al. | Review of multinomial and multiattribute quality control charts | |
| Donauer et al. | Identifying nonconformity root causes using applied knowledge discovery | |
| Brundage et al. | Smart manufacturing through a framework for a knowledge-based diagnosis system | |
| CN109871002B (en) | Concurrent abnormal state identification and positioning system based on tensor label learning | |
| CN116485032A (en) | Aviation product processing quality prediction method considering multidimensional influence factors | |
| Tan et al. | Quality analytics in a big data supply chain: commodity data analytics for quality engineering | |
| CN112712314A (en) | Logistics data recommendation method based on sensor of Internet of things | |
| CN110516920B (en) | Gyroscope quality grade evaluation method based on index fusion | |
| Lyu et al. | A data-driven approach for identifying possible manufacturing processes and production parameters that cause product defects: A thin-film filter company case study | |
| CN117393076B (en) | Intelligent monitoring method and system for heat-resistant epoxy resin production process | |
| Diallo et al. | Data-based fault diagnosis model using a Bayesian causal analysis framework | |
| CN117270482A (en) | Automobile factory control system based on digital twin | |
| KR102437917B1 (en) | Equipment operation system | |
| Moore et al. | Acquisition challenge: the importance of incompressibility in Comparing Learning Curve Models | |
| Hsu et al. | Data mining for yield enhancement in TFT-LCD manufacturing: an empirical study | |
| Lv et al. | Data mining for material feeding optimization of printed circuit board template production | |
| Xia et al. | Degradation prediction and rolling predictive maintenance policy for multi-sensor systems based on two-dimensional self-attention | |
| Xu et al. | A Data Quality Assessment and Control Method in Multiple Products Manufacturing Process | |
| Turkoglu et al. | Application of data mining in failure estimation of cold forging machines: An industrial research | |
| Sethi | Using Machine Learning Methods to Predict Order Lead Times | |
| Lundén | Implementing data analytics for improved quality in manufacturing: a case study | |
| CN113721560B (en) | Open collaborative manufacturing cloud architecture system oriented to bottom data and implementation method | |
| Zyryanova et al. | Quality assessment of regulatory legal acts using the Mamdani algorithm |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |