CN113516285B - Product quality analysis and prediction method of automatic assembly detection production line in production - Google Patents
Product quality analysis and prediction method of automatic assembly detection production line in production Download PDFInfo
- Publication number
- CN113516285B CN113516285B CN202110518133.8A CN202110518133A CN113516285B CN 113516285 B CN113516285 B CN 113516285B CN 202110518133 A CN202110518133 A CN 202110518133A CN 113516285 B CN113516285 B CN 113516285B
- Authority
- CN
- China
- Prior art keywords
- quality
- station
- predicted
- data
- product
- 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.)
- Active
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 61
- 238000004519 manufacturing process Methods 0.000 title claims abstract description 58
- 238000000034 method Methods 0.000 title claims abstract description 41
- 238000004458 analytical method Methods 0.000 title claims abstract description 15
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 128
- 230000004927 fusion Effects 0.000 claims abstract description 71
- 238000012549 training Methods 0.000 claims abstract description 23
- 238000007637 random forest analysis Methods 0.000 claims abstract description 22
- 238000007781 pre-processing Methods 0.000 claims abstract description 7
- 238000007689 inspection Methods 0.000 claims description 9
- 238000012417 linear regression Methods 0.000 claims description 7
- 238000010276 construction Methods 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 4
- 238000004140 cleaning Methods 0.000 claims description 4
- 238000010219 correlation analysis Methods 0.000 claims description 4
- 238000000605 extraction Methods 0.000 claims description 4
- 238000012821 model calculation Methods 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- 238000011144 upstream manufacturing Methods 0.000 claims description 3
- 238000005516 engineering process Methods 0.000 abstract description 4
- 238000012797 qualification Methods 0.000 abstract description 3
- 239000000047 product Substances 0.000 description 49
- 238000012360 testing method Methods 0.000 description 5
- 239000011265 semifinished product Substances 0.000 description 4
- 238000003066 decision tree Methods 0.000 description 3
- 230000007547 defect Effects 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 125000004122 cyclic group Chemical group 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000002790 cross-validation Methods 0.000 description 1
- 230000001351 cycling effect Effects 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 238000011282 treatment Methods 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
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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—Information and communication technology [ICT] specially adapted for implementation of business processes of 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
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Human Resources & Organizations (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Development Economics (AREA)
- Health & Medical Sciences (AREA)
- Marketing (AREA)
- Tourism & Hospitality (AREA)
- General Health & Medical Sciences (AREA)
- General Business, Economics & Management (AREA)
- Entrepreneurship & Innovation (AREA)
- Molecular Biology (AREA)
- Evolutionary Biology (AREA)
- Educational Administration (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Quality & Reliability (AREA)
- Biomedical Technology (AREA)
- Software Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Game Theory and Decision Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Operations Research (AREA)
- Primary Health Care (AREA)
- Manufacturing & Machinery (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- General Factory Administration (AREA)
Abstract
The invention discloses a product quality analysis and prediction method for an automatic assembly and detection production line in production, which comprises the steps of collecting quality data of production stations (working procedures) in the production line, preprocessing the collected data to form a quality data set, constructing a random forest, XGBoost and LSTM algorithm submodel, using the formed quality data set training algorithm model, and fusing and predicting by a Stacking algorithm fusion technology. And applying a prediction model in actual production, and pre-warning to staff through message pushing when quality problems occur. The invention predicts the quality of the production process of the product by analyzing in advance, thereby ensuring that the assembly production of the next station (process) is carried out under the condition that the quality of the previous station (process) meets the requirement, avoiding the loss of the subsequent parts to be assembled, improving the utilization rate of the parts in the assembly detection production process of the product, practically reducing unnecessary scrapping and improving the one-time offline qualification rate of the product production.
Description
Technical Field
The invention relates to the technical field of automatic assembly, in particular to a product quality analysis and prediction method of an automatic assembly detection production line in production.
Background
On the automatic assembly and detection production lines of gas meters, refrigerator compressors, automobile door locks and the like, the number of assembly stations and detection stations on each automatic assembly and detection production line is up to more than 10. Therefore, the matching precision requirement for each station is higher. When one or more stations deviate, the corresponding detection stations cannot be found in time or all stations are near the upper and lower detection limits and are judged to be qualified, so that the conditions of poor assembly semi-finished products or finished products and the like are often caused, and the once-off qualification rate is low.
When the defect of the semi-finished product or the finished product occurs, the defect of the semi-finished product or the finished product is caused seriously, and when the defect of the semi-finished product or the finished product is not caused seriously, the repair is usually required, the repair is time-consuming and labor-consuming, and the parts assembled subsequently are easy to damage, so that the utilization rate of the parts in the process of product assembly detection production is low.
Disclosure of Invention
The invention aims at solving the technical problems of the prior art, and provides a product quality analysis and prediction method of an automatic assembly and detection production line in production.
In order to solve the technical problems, the invention adopts the following technical scheme:
a product quality analysis and prediction method for an automatic assembly detection production line in production comprises the following steps.
Step 1, determining a predicted output station: the prediction output station is arranged at a repair station of the automatic assembly and detection production line.
Step 2, predicting the product mark of the input station: and placing the product main body to be assembled on the following tools of the automatic assembly and detection production line, wherein each following tool is internally provided with an RFID chip, and the RFID chip is internally provided with product information corresponding to the product main body to be assembled.
Step 3, collecting quality data: n accompanying tools which are respectively provided with a product main body to be assembled are conveyed to a prediction output station from a feeding station of the product main body to be assembled after sequentially passing through each intermediate station; the method comprises the steps of feeding a main body of a product to be assembled, each intermediate station and a prediction output station; the upper computer respectively reads the RFID chips in the follower fixture, and stores the quality detection data of each intermediate station and the predicted output station in the corresponding RFID chips; wherein n is more than or equal to 10.
Step 4, determining a predicted input station and key quality characteristics: carrying out data preprocessing on all quality data acquired in the step 3, and determining a predicted input station and key quality characteristics; the prediction input station is arranged at a key station which is positioned at the upstream of the repair station and is used for controlling the quality of the repair station; the data preprocessing comprises data cleaning, quality feature extraction and correlation analysis; the key quality features are key features which influence the quality of the repair station in all quality data of the predicted input station.
Step 5, constructing a quality data set: constructing a quality data set according to the predicted input station determined in the step 4 and all the quality data acquired in the step 3; the quality data sets comprise a predicted input station quality data set and a predicted output station quality data set; the predicted input station quality data set comprises n product information of the product main bodies to be assembled and n groups of key quality characteristic data of the predicted input stations; the predicted output station quality dataset includes n groups of predicted output station quality detection data.
And 6, constructing a Stacking fusion algorithm model, which specifically comprises the following steps.
Step 61, constructing three algorithm models: the three algorithm models comprise a random forest, XGBoost and LSTM algorithm models; randomly sampling from a predicted input station quality data set to serve as input quantities of three algorithm models respectively, taking quality data corresponding to the input quantities in a predicted output station quality data set as output quantities of the three algorithm models respectively, and training the three algorithm models respectively until convergence; and further obtaining the random forest, XGBoost and LSTM algorithm model.
Step 62, three model output data are obtained: the whole predicted input station quality data set in the step 5 is respectively used as input quantities in the random forest, XGBoost and LSTM algorithm model constructed in the step 61, and three model output data are obtained; the three model output data comprise predicted output station quality detection data of n groups of random forest algorithms, predicted output station quality detection data of n groups of XGBoost algorithms and predicted output station quality detection data of n groups of LSTM algorithms.
Step 63, constructing a Stacking fusion algorithm model: taking the three model output data obtained in the step 62 as the input quantity of the Stacking fusion algorithm model, taking the quality detection data of n groups of prediction output stations in the step 5 as the output quantity of the Stacking fusion algorithm model, and training the Stacking fusion algorithm model until convergence, so as to obtain the required Stacking fusion algorithm model; the method comprises the steps that a random forest algorithm model, an XGBoost algorithm model, an LSTM algorithm model and a Stacking fusion algorithm model jointly form the Stacking fusion algorithm model, the input quantity of the Stacking fusion algorithm model is product information and key quality characteristic data of a prediction input station, and the output quantity of the Stacking fusion algorithm model is quality detection data of the prediction output station.
Step 7, product quality analysis and prediction: and after the pallet on which the predicted product to be analyzed is placed is conveyed to the predicted output station from the feeding station, the upper computer reads the RFID chip in the corresponding pallet to obtain the information of the predicted product to be analyzed and the key quality characteristic data of the predicted input station, and uses the information as the input quantity of the Stacking fusion algorithm model to perform Stacking fusion algorithm model calculation so as to obtain the quality detection data of the predicted product to be analyzed at the predicted output station.
In step 63, the construction process of the Stacking fusion algorithm model includes the following steps:
step 63A, weighted average: and carrying out weighted average on the three model output data input into the Stacking fusion algorithm model to form n groups of the fused key quality characteristic data of the prediction input station.
Step 63B, linear regression: and D, taking the key quality characteristic data of the n groups of fused predicted input stations formed in the step 63A as the fused input quantity in the Stacking fusion algorithm model, performing Stacking calculation to obtain n groups of Stacking fusion algorithm model output quantity containing undetermined coefficients, performing linear regression analysis on the Stacking fusion algorithm model output quantity containing undetermined coefficients and the quality detection data of the n groups of predicted output stations in the step 5, and solving to obtain undetermined coefficients in the Stacking fusion algorithm model, thereby obtaining the required Stacking fusion algorithm model.
In step 63A, the weight value of each model is the accuracy of the corresponding model when the three models are weighted and averaged.
In step 61, in the construction process of the random forest algorithm model and the XGBoost algorithm model, parameter adjustment is performed through a grid search method.
And each intermediate station and each forecast output station are provided with a card reader connected with the upper computer and used for reading the RFID chip.
In the step 7, the upper computer can compare and judge the quality detection data of the output predicted product to be analyzed in the predicted output station with the preset quality data; when the predicted product to be analyzed is judged to have the quality problem, the upper computer can push the quality detection data of the predicted product to be analyzed at the predicted output station and the comparison judgment result to corresponding production line management personnel for early warning.
The automatic assembly detection production line is an automatic assembly detection production line of a gas meter, a refrigerator compressor or an automobile door lock.
In the step 6, the Stacking fusion algorithm model is a 5-fold Stacking fusion algorithm model.
The invention has the following beneficial effects:
1. the invention collects quality data of each production station (working procedure) in a production line, firstly uses a data preprocessing technology to carry out data cleaning, feature extraction, correlation analysis and other treatments on the data to form a quality data set, then constructs a random forest, XGBoost and LSTM algorithm submodel, uses the formed quality data set training algorithm model, uses 3 algorithm submodels to predict through a Stacking algorithm fusion technology, uses weighted average to combine the predicted data to form a new data set, uses the new data set training linear regression model, and finally completes fusion of the algorithm model. And applying a prediction model in actual production, and pre-warning to staff through message pushing when quality problems occur.
2. The invention predicts the quality of the production process of the product by analyzing in advance, thereby ensuring that the assembly production of the next station (process) is carried out under the condition that the quality of the previous station (process) meets the requirement, avoiding the loss of the subsequent parts to be assembled, improving the utilization rate of the parts in the assembly detection production process of the product, practically reducing unnecessary scrapping and improving the one-time off-line qualification rate of the product production.
Drawings
FIG. 1 is a flow chart of a method for predicting product quality analysis in production of an automated assembly inspection line according to the present invention.
Detailed Description
The invention will be described in further detail with reference to the accompanying drawings and specific preferred embodiments.
In the description of the present invention, it should be understood that the terms "left", "right", "upper", "lower", etc. indicate orientations or positional relationships based on the orientations or positional relationships shown in the drawings, are merely for convenience in describing the present invention and simplifying the description, and do not indicate or imply that the apparatus or element being referred to must have a specific orientation, be configured and operated in a specific orientation, and "first", "second", etc. do not indicate the importance of the components, and thus are not to be construed as limiting the present invention. The specific dimensions adopted in the present embodiment are only for illustrating the technical solution, and do not limit the protection scope of the present invention.
As shown in FIG. 1, the method for predicting the product quality analysis of the automatic assembly inspection production line in production comprises the following steps.
Step 1, determining a predicted output station: the prediction output station is arranged at a repair station of the automatic assembly and detection production line; further, a station with a high repair rate is preferable, and the repair station may be provided for each repair station.
The automatic assembly and detection production line is preferably an automatic assembly and detection production line for a refrigerator compressor, a gas meter, an automobile door lock and the like.
Step 2, predicting the product mark of the input station: and placing the product main body to be assembled on the following tools of the automatic assembly and detection production line, wherein each following tool is internally provided with an RFID chip, and the RFID chip is internally provided with product information corresponding to the product main body to be assembled.
Step 3, quality data acquisition
N accompanying tools which are respectively provided with a product main body to be assembled are conveyed to a prediction output station from a feeding station of the product main body to be assembled after sequentially passing through each intermediate station; and a loading station, each intermediate station and a prediction output station of the product main body to be assembled.
And each intermediate station and each forecast output station are provided with a card reader connected with the upper computer and used for reading the RFID chip. The method is characterized in that the method is directly communicated with the PLC, data of all process equipment in the production process on the production line are collected in real time and at high speed, the collected data are stored in an upper computer without using third party software such as OPC, and the communication is more stable and reliable. And the assembly detection station is collected in real time, and quality result data is collected on products entering the repair station through the RFID chip.
The upper computer respectively reads the RFID chips in the follower fixture, and stores the quality detection data of each intermediate station and the predicted output station in the corresponding RFID chips; wherein n is more than or equal to 10.
Step 4, determining the predicted input station and key quality characteristics
Carrying out data preprocessing such as data cleaning, quality feature extraction, correlation analysis and the like on all the quality data acquired in the step 3, and determining a predicted input station and key quality features; the prediction input station is arranged at a key station which is positioned at the upstream of the repair station and is used for controlling the quality of the repair station; the key quality features are key features which influence the quality of the repair station in all quality data of the predicted input station.
Step 5, constructing a quality data set: constructing a quality data set according to the predicted input station determined in the step 4 and all the quality data acquired in the step 3; the quality data sets comprise a predicted input station quality data set and a predicted output station quality data set; the predicted input station quality data set comprises n product information of the product main bodies to be assembled and n groups of key quality characteristic data of the predicted input stations; the predicted output station quality dataset includes n groups of predicted output station quality detection data.
And 6, constructing a Stacking fusion algorithm model, which specifically comprises the following steps.
Step 61, constructing three algorithm models
The three algorithm models include random forest, XGBoost and LSTM algorithm models.
Randomly sampling from a predicted input station quality data set to serve as input quantities of three algorithm models respectively, taking quality data corresponding to the input quantities in a predicted output station quality data set as output quantities of the three algorithm models respectively, and training the three algorithm models respectively until convergence; and further obtaining the random forest, XGBoost and LSTM algorithm model.
In the construction process of the random forest algorithm model and the XGBoost algorithm model, parameter adjustment is carried out through a grid search method, an optimal parameter combination is obtained, and the 3 models are used as a first layer model for model fusion.
The random forest algorithm is an integrated algorithm model formed by combining a plurality of decision trees. Wherein the training data set of each decision tree is randomly sampled from the overall quality data set, and the training features are randomly extracted from the sample data set, and then each decision tree is trained.
The XGBoost is an improved algorithm based on the GBDT algorithm. L1 and L2 regularization terms are added in the XGBoost algorithm, so that overfitting is reduced. The second-order Taylor expansion is used for approximate calculation of the loss function, so that the training speed of the model is increased.
The LSTM is a long-short-term memory neural network and is a special structure type of a cyclic neural network. By means of special structural units: input gate, output gate and forget gate, the previous information in a sequence is extracted and passed to the following neurons. The forgetting door can effectively prevent the information explosion problem.
Step 62, three model output data are obtained: the whole predicted input station quality data set in the step 5 is respectively used as input quantities in the random forest, XGBoost and LSTM algorithm model constructed in the step 61, and three model output data are obtained; the three model output data comprise predicted output station quality detection data of n groups of random forest algorithms, predicted output station quality detection data of n groups of XGBoost algorithms and predicted output station quality detection data of n groups of LSTM algorithms.
Step 63, constructing a Stacking fusion algorithm model: and (3) taking the three model output data obtained in the step (62) as the input quantity of the Stacking fusion algorithm model, taking the quality detection data of the n groups of prediction output stations in the step (5) as the output quantity of the Stacking fusion algorithm model, and training the Stacking fusion algorithm model until convergence, so as to obtain the required Stacking fusion algorithm model.
The method comprises the steps that a random forest algorithm model, an XGBoost algorithm model, an LSTM algorithm model and a Stacking fusion algorithm model jointly form the Stacking fusion algorithm model, the input quantity of the Stacking fusion algorithm model is product information and key quality characteristic data of a prediction input station, and the output quantity of the Stacking fusion algorithm model is quality detection data of the prediction output station.
The construction process of the Stacking fusion algorithm model preferably comprises the following steps:
step 63A, weighted average: and carrying out weighted average on the three model output data input into the Stacking fusion algorithm model to form n groups of the fused key quality characteristic data of the prediction input station.
When the three models are weighted and averaged, the weight value of each model is preferably the accuracy of the corresponding model.
Step 63B, linear regression: and D, taking the key quality characteristic data of the n groups of fused predicted input stations formed in the step 63A as the input quantity of the fused second layer in the Stacking fusion algorithm model, performing Stacking calculation to obtain n groups of Stacking fusion algorithm model output quantity containing undetermined coefficients, performing linear regression analysis on the Stacking fusion algorithm model output quantity containing undetermined coefficients and the quality detection data of the n groups of predicted output stations in the step 5, and solving to obtain undetermined coefficients in the Stacking fusion algorithm model, thereby obtaining the required Stacking fusion algorithm model.
The specific implementation method of the Stacking fusion algorithm model is as follows:
1. the quality dataset is divided into a training set train and a test set test.
2. The training set train is divided into k subsets train_c= { train_1, train_2, …, train_k } and the model to be fused is stored in the set and labeled model_c= { model_1, model_2, …, model_n }.
3. And (3) circulating the train_C, taking each subset train_i (1 < =i < =k) as a test set, and combining other subsets as training sets.
4. And (3) cycling the model_C, taking each sub-model model_j (1 < = j < = n), training and predicting the model_j by using the training set and the test set generated in the step (3), storing the prediction result as { pred_1, pred_2, …, pred_k }, and combining the prediction result into a new data set pred_model_j.
5. And (5) carrying out weighted average on all pred_model_j data sets generated in the step (5) to obtain a new prediction result set pred.
6. And constructing and training a fusion model by using the pred data set as a training set and the test as a prediction set, namely completing model fusion.
Further, the Stacking fusion algorithm model is preferably a 5-fold Stacking fusion algorithm model. The training data set is divided into 5 sub data sets by using a Stacking algorithm fusion and a 5-fold cross validation technology, 4 parts are used as training sets and 1 part is used as a prediction set each time, cyclic training and prediction are performed, namely each algorithm predicts the 5 sub data sets to obtain a complete prediction set, and then weighted average is performed on the 3 prediction sets according to the accuracy of random forests, XGBoost and LSTM models to serve as the training data set fused by the next layer of algorithm.
Step 7, product quality analysis and prediction: and after the pallet on which the predicted product to be analyzed is placed is conveyed to the predicted output station from the feeding station, the upper computer reads the RFID chip in the corresponding pallet to obtain the information of the predicted product to be analyzed and the key quality characteristic data of the predicted input station, and uses the information as the input quantity of the Stacking fusion algorithm model to perform Stacking fusion algorithm model calculation so as to obtain the quality detection data of the predicted product to be analyzed at the predicted output station.
Then, the upper computer can compare and judge the quality detection data of the output predicted product to be analyzed at the predicted output station with preset quality data; when the predicted product to be analyzed is judged to have the quality problem, the upper computer can push the quality detection data of the predicted product to be analyzed at the predicted output station and the comparison judgment result to corresponding production line management personnel for early warning.
The preferred embodiments of the present invention have been described in detail above, but the present invention is not limited to the specific details of the above embodiments, and various equivalent changes can be made to the technical solution of the present invention within the scope of the technical concept of the present invention, and all the equivalent changes belong to the protection scope of the present invention.
Claims (8)
1. A product quality analysis and prediction method of an automatic assembly detection production line in production is characterized in that: the method comprises the following steps:
step 1, determining a predicted output station: the prediction output station is arranged at a repair station of the automatic assembly and detection production line;
step 2, predicting the product mark of the input station: placing a product main body to be assembled on a follower tool of an automatic assembly detection production line, wherein each follower tool is internally provided with an RFID chip, and product information corresponding to the product main body to be assembled is arranged in the RFID chip;
step 3, collecting quality data: n accompanying tools which are respectively provided with a product main body to be assembled are conveyed to a prediction output station from a feeding station of the product main body to be assembled after sequentially passing through each intermediate station; in a feeding station, each intermediate station and a forecast output station of a product main body to be assembled, an upper computer respectively reads RFID chips in a follower fixture, and quality detection data of each intermediate station and the forecast output station are stored in corresponding RFID chips; wherein n is more than or equal to 10;
step 4, determining a predicted input station and key quality characteristics: carrying out data preprocessing on all quality data acquired in the step 3, and determining a predicted input station and key quality characteristics; the prediction input station is arranged at a key station which is positioned at the upstream of the repair station and is used for controlling the quality of the repair station; the data preprocessing comprises data cleaning, quality feature extraction and correlation analysis; the key quality features are key features which influence the quality of the repairing station in all quality data of the input station;
step 5, constructing a quality data set: constructing a quality data set according to the predicted input station determined in the step 4 and all the quality data acquired in the step 3; the quality data sets comprise a predicted input station quality data set and a predicted output station quality data set; the predicted input station quality data set comprises n product information of the product main bodies to be assembled and n groups of key quality characteristic data of the predicted input stations; the predicted output station quality data set comprises n groups of quality detection data of predicted output stations;
step 6, constructing a Stacking fusion algorithm model, which specifically comprises the following steps:
step 61, constructing three algorithm models: the three algorithm models comprise a random forest, XGBoost and LSTM algorithm models; randomly sampling from a predicted input station quality data set to serve as input quantities of three algorithm models respectively, taking quality data corresponding to the input quantities in a predicted output station quality data set as output quantities of the three algorithm models respectively, and training the three algorithm models respectively until convergence; further obtaining the random forest, XGBoost and LSTM algorithm models;
step 62, three model output data are obtained: the whole predicted input station quality data set in the step 5 is respectively used as input quantities in the random forest, XGBoost and LSTM algorithm model constructed in the step 61, and three model output data are obtained; the three model output data comprise predicted output station quality detection data of n groups of random forest algorithms, predicted output station quality detection data of n groups of XGBoost algorithms and predicted output station quality detection data of n groups of LSTM algorithms;
step 63, constructing a Stacking fusion algorithm model: taking the three model output data obtained in the step 62 as the input quantity of the Stacking fusion algorithm model, taking the quality detection data of n groups of prediction output stations in the step 5 as the output quantity of the Stacking fusion algorithm model, and training the Stacking fusion algorithm model until convergence, so as to obtain the required Stacking fusion algorithm model; the method comprises the steps that a random forest algorithm model, an XGBoost algorithm model, an LSTM algorithm model and a Stacking fusion algorithm model jointly form the Stacking fusion algorithm model, the input quantity of the Stacking fusion algorithm model is product information and key quality characteristic data of a prediction input station, and the output quantity of the Stacking fusion algorithm model is quality detection data of the prediction output station;
step 7, product quality analysis and prediction: and after the pallet on which the predicted product to be analyzed is placed is conveyed to the predicted output station from the feeding station, the upper computer reads the RFID chip in the corresponding pallet to obtain the information of the predicted product to be analyzed and the key quality characteristic data of the predicted input station, and uses the information as the input quantity of the Stacking fusion algorithm model to perform Stacking fusion algorithm model calculation so as to obtain the quality detection data of the predicted product to be analyzed at the predicted output station.
2. The method for predicting product quality in production of an automated assembly inspection line of claim 1, wherein: in step 63, the construction process of the Stacking fusion algorithm model includes the following steps:
step 63A, weighted average: carrying out weighted average on three model output data input into a Stacking fusion algorithm model to form n groups of fused key quality characteristic data of the prediction input station;
step 63B, linear regression: and D, taking the key quality characteristic data of the n groups of fused predicted input stations formed in the step 63A as the fused input quantity in the Stacking fusion algorithm model, performing Stacking calculation to obtain n groups of Stacking fusion algorithm model output quantity containing undetermined coefficients, performing linear regression analysis on the Stacking fusion algorithm model output quantity containing undetermined coefficients and the quality detection data of the n groups of predicted output stations in the step 5, and solving to obtain undetermined coefficients in the Stacking fusion algorithm model, thereby obtaining the required Stacking fusion algorithm model.
3. The method for predicting product quality in production of an automated assembly inspection line of claim 2, wherein: in step 63A, the weight value of each model is the accuracy of the corresponding model when the three models are weighted and averaged.
4. The method for predicting product quality in production of an automated assembly inspection line of claim 1, wherein: in step 61, in the construction process of the random forest algorithm model and the XGBoost algorithm model, parameter adjustment is performed through a grid search method.
5. The method for predicting product quality in production of an automated assembly inspection line of claim 1, wherein: and each intermediate station and each forecast output station are provided with a card reader connected with the upper computer and used for reading the RFID chip.
6. The method for predicting product quality in production of an automated assembly inspection line of claim 1, wherein: in the step 7, the upper computer can compare and judge the quality detection data of the output predicted product to be analyzed in the predicted output station with the preset quality data; when the predicted product to be analyzed is judged to have the quality problem, the upper computer can push the quality detection data of the predicted product to be analyzed at the predicted output station and the comparison judgment result to corresponding production line management personnel for early warning.
7. The method for predicting product quality in production of an automated assembly inspection line of claim 1, wherein: the automatic assembly detection production line is an automatic assembly detection production line of a gas meter, a refrigerator compressor or an automobile door lock.
8. The method for predicting product quality in production of an automated assembly inspection line of claim 1, wherein: in the step 6, the Stacking fusion algorithm model is a 5-fold Stacking fusion algorithm model.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110518133.8A CN113516285B (en) | 2021-05-12 | 2021-05-12 | Product quality analysis and prediction method of automatic assembly detection production line in production |
| PCT/CN2021/130621 WO2022237105A1 (en) | 2021-05-12 | 2021-11-15 | Quality analysis and prediction method for product during production of automatic assembly and detection production line |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110518133.8A CN113516285B (en) | 2021-05-12 | 2021-05-12 | Product quality analysis and prediction method of automatic assembly detection production line in production |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113516285A CN113516285A (en) | 2021-10-19 |
| CN113516285B true CN113516285B (en) | 2024-02-13 |
Family
ID=78064228
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110518133.8A Active CN113516285B (en) | 2021-05-12 | 2021-05-12 | Product quality analysis and prediction method of automatic assembly detection production line in production |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN113516285B (en) |
| WO (1) | WO2022237105A1 (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113516285B (en) * | 2021-05-12 | 2024-02-13 | 中船重工鹏力（南京）智能装备***有限公司 | Product quality analysis and prediction method of automatic assembly detection production line in production |
| CN113655768B (en) * | 2021-10-21 | 2022-02-15 | 深圳市信润富联数字科技有限公司 | Assembly yield control method, equipment and computer readable storage medium |
| CN115511367B (en) * | 2022-10-21 | 2023-07-25 | 上海数瞳信息科技有限公司 | Intelligent quality management system of production line |
| CN116119284B (en) * | 2022-12-16 | 2023-11-24 | 工业富联(杭州)数据科技有限公司 | Material assembling method, device, equipment and medium based on artificial intelligence |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106845796A (en) * | 2016-12-28 | 2017-06-13 | 中南大学 | One kind is hydrocracked flow product quality on-line prediction method |
| CN109348497A (en) * | 2018-09-30 | 2019-02-15 | 南昌航空大学 | Wireless sensor network link quality prediction method |
| CN109472057A (en) * | 2018-10-16 | 2019-03-15 | 浙江大学 | Based on product processing quality prediction meanss and method across the implicit parameters memorizing of work step |
| CN109711714A (en) * | 2018-12-24 | 2019-05-03 | 浙江大学 | Product quality prediction technique is assembled in manufacture based on shot and long term memory network in parallel |
| CN110288199A (en) * | 2019-05-29 | 2019-09-27 | 北京航空航天大学 | The method of product quality forecast |
| CN111814385A (en) * | 2020-05-28 | 2020-10-23 | 平安科技（深圳）有限公司 | Method, device and computer equipment for predicting quality of workpiece |
| CN112463643A (en) * | 2020-12-16 | 2021-03-09 | 郑州航空工业管理学院 | Software quality prediction method |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TWI267012B (en) * | 2004-06-03 | 2006-11-21 | Univ Nat Cheng Kung | Quality prognostics system and method for manufacturing processes |
| US10902368B2 (en) * | 2014-03-12 | 2021-01-26 | Dt360 Inc. | Intelligent decision synchronization in real time for both discrete and continuous process industries |
| EP4055454A4 (en) * | 2019-11-06 | 2024-01-31 | Nanotronics Imaging, Inc. | Systems, methods, and media for manufacturing processes |
| CN111461555B (en) * | 2020-04-02 | 2023-06-09 | 中国电子产品可靠性与环境试验研究所（（工业和信息化部电子第五研究所）（中国赛宝实验室）） | Production line quality monitoring method, device and system |
| CN113516285B (en) * | 2021-05-12 | 2024-02-13 | 中船重工鹏力（南京）智能装备***有限公司 | Product quality analysis and prediction method of automatic assembly detection production line in production |
-
2021
- 2021-05-12 CN CN202110518133.8A patent/CN113516285B/en active Active
- 2021-11-15 WO PCT/CN2021/130621 patent/WO2022237105A1/en active Application Filing
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106845796A (en) * | 2016-12-28 | 2017-06-13 | 中南大学 | One kind is hydrocracked flow product quality on-line prediction method |
| CN109348497A (en) * | 2018-09-30 | 2019-02-15 | 南昌航空大学 | Wireless sensor network link quality prediction method |
| CN109472057A (en) * | 2018-10-16 | 2019-03-15 | 浙江大学 | Based on product processing quality prediction meanss and method across the implicit parameters memorizing of work step |
| CN109711714A (en) * | 2018-12-24 | 2019-05-03 | 浙江大学 | Product quality prediction technique is assembled in manufacture based on shot and long term memory network in parallel |
| CN110288199A (en) * | 2019-05-29 | 2019-09-27 | 北京航空航天大学 | The method of product quality forecast |
| CN111814385A (en) * | 2020-05-28 | 2020-10-23 | 平安科技（深圳）有限公司 | Method, device and computer equipment for predicting quality of workpiece |
| CN112463643A (en) * | 2020-12-16 | 2021-03-09 | 郑州航空工业管理学院 | Software quality prediction method |
Non-Patent Citations (2)
| Title |
|---|
| 基于RFID优化发动机装配过程质量控制;陈海;;装备制造技术;20161015(第10期);全文 * |
| 基于RFID的变速器装配线质量追溯***研究;景熠;王旭;李文川;;现代科学仪器;20111015(第05期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| WO2022237105A1 (en) | 2022-11-17 |
| CN113516285A (en) | 2021-10-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113516285B (en) | Product quality analysis and prediction method of automatic assembly detection production line in production | |
| CN107657288B (en) | Power dispatching flow data anomaly detection method based on isolated forest algorithm | |
| CN112508053A (en) | Intelligent diagnosis method, device, equipment and medium based on integrated learning framework | |
| CN112765768B (en) | Discrete workshop digital traceability method based on Internet of things | |
| CN106897774B (en) | Multiple soft measurement algorithm cluster modeling methods based on Monte Carlo cross validation | |
| CN104778298A (en) | Gaussian process regression soft measurement modeling method based on EGMM (Error Gaussian Mixture Model) | |
| CN102522709B (en) | Decision-making method and decision-making system for state overhaul of transformers | |
| CN111343147B (en) | Network attack detection device and method based on deep learning | |
| CN102270302A (en) | Grey support vector machine-based multi-stress accelerated life testing forecasting method | |
| CN111898673A (en) | Dissolved oxygen content prediction method based on EMD and LSTM | |
| CN112241606A (en) | Cooperative decision-making method for operation and maintenance of ship intelligent equipment based on CPS decision-making module | |
| CN113256036B (en) | Power supply cost analysis and prediction method based on Prophet-LSTNet combined model | |
| CN111784022A (en) | Short-time adjacent fog prediction method based on combination of Wrapper method and SVM method | |
| CN113989550A (en) | Electric vehicle charging pile operation state prediction method based on CNN and LSTM hybrid network | |
| CN115983048A (en) | Condenser quality detection method and system based on digital twinning | |
| CN114548494A (en) | Visual cost data prediction intelligent analysis system | |
| CN116862209B (en) | New energy automobile charging facility management method and system | |
| CN115759395A (en) | Training of photovoltaic detection model, detection method of photovoltaic power generation and related device | |
| CN114372640A (en) | Wind power prediction method based on fluctuation sequence classification correction | |
| CN108345726A (en) | Input variable automatic selecting method based on mutual information and soft instrument error | |
| CN111062118B (en) | Multilayer soft measurement modeling system and method based on neural network prediction layering | |
| CN114545276A (en) | Power battery service life prediction method based on capacity test and Internet of vehicles big data | |
| CN117371623B (en) | Electric energy meter running state early warning method and system | |
| Wang et al. | MSCNN-BLSTM based Prediction of the Remaining Useful Life of Aeroengine | |
| Wang et al. | A method of automatic test case generation based on ct-lssvm algorithm in fao systems |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |