CN119066369A - A method, medium and system for estimating filter replacement time in ABS production process - Google Patents
A method, medium and system for estimating filter replacement time in ABS production process Download PDFInfo
- Publication number
- CN119066369A CN119066369A CN202411554028.XA CN202411554028A CN119066369A CN 119066369 A CN119066369 A CN 119066369A CN 202411554028 A CN202411554028 A CN 202411554028A CN 119066369 A CN119066369 A CN 119066369A
- Authority
- CN
- China
- Prior art keywords
- filter screen
- data
- time
- estimating
- monitoring
- 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
- 238000004519 manufacturing process Methods 0.000 title claims abstract description 105
- 238000000034 method Methods 0.000 title claims abstract description 64
- 238000012544 monitoring process Methods 0.000 claims abstract description 103
- 230000008859 change Effects 0.000 claims abstract description 28
- 238000010801 machine learning Methods 0.000 claims abstract description 27
- 238000007781 pre-processing Methods 0.000 claims abstract description 14
- 238000012549 training Methods 0.000 claims abstract description 10
- 230000000903 blocking effect Effects 0.000 claims description 53
- 239000000463 material Substances 0.000 claims description 33
- 239000012530 fluid Substances 0.000 claims description 27
- 238000013528 artificial neural network Methods 0.000 claims description 21
- 238000000605 extraction Methods 0.000 claims description 13
- 238000004422 calculation algorithm Methods 0.000 claims description 9
- 230000007246 mechanism Effects 0.000 claims description 9
- 238000003860 storage Methods 0.000 claims description 9
- 230000002159 abnormal effect Effects 0.000 claims description 7
- 241000270322 Lepidosauria Species 0.000 claims description 5
- 230000006870 function Effects 0.000 claims description 5
- 238000012216 screening Methods 0.000 claims description 4
- 238000012795 verification Methods 0.000 claims description 4
- 239000004676 acrylonitrile butadiene styrene Substances 0.000 description 44
- 229920000122 acrylonitrile butadiene styrene Polymers 0.000 description 44
- XECAHXYUAAWDEL-UHFFFAOYSA-N acrylonitrile butadiene styrene Chemical compound C=CC=C.C=CC#N.C=CC1=CC=CC=C1 XECAHXYUAAWDEL-UHFFFAOYSA-N 0.000 description 44
- 230000008569 process Effects 0.000 description 12
- 238000004458 analytical method Methods 0.000 description 7
- 230000004927 fusion Effects 0.000 description 5
- 238000003062 neural network model Methods 0.000 description 5
- 238000010606 normalization Methods 0.000 description 5
- 238000005457 optimization Methods 0.000 description 5
- 238000012545 processing Methods 0.000 description 4
- 238000011049 filling Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000007405 data analysis Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000009826 distribution Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000007774 longterm Effects 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000013450 outlier detection Methods 0.000 description 2
- 238000007619 statistical method Methods 0.000 description 2
- 238000010200 validation analysis Methods 0.000 description 2
- 241000669618 Nothes Species 0.000 description 1
- 230000005856 abnormality Effects 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 230000001174 ascending effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 230000003111 delayed effect Effects 0.000 description 1
- 229920006351 engineering plastic Polymers 0.000 description 1
- 238000001125 extrusion Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 239000012535 impurity Substances 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 239000000155 melt Substances 0.000 description 1
- 238000000465 moulding Methods 0.000 description 1
- 238000005293 physical law Methods 0.000 description 1
- 230000002028 premature Effects 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 239000011347 resin Substances 0.000 description 1
- 229920005989 resin Polymers 0.000 description 1
- 230000000452 restraining effect Effects 0.000 description 1
- 230000000630 rising effect Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000001960 triggered effect Effects 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/10—Pre-processing; Data cleansing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/10—Pre-processing; Data cleansing
- G06F18/15—Statistical pre-processing, e.g. techniques for normalisation or restoring missing data
-
- 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
- G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
-
- 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
- G06N3/0464—Convolutional networks [CNN, ConvNet]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
本发明提供了一种ABS生产过程滤网更换时刻预估方法、介质及系统，属于ABS生产技术领域，包括：收集多条ABS生产线上多个滤网监测点的历史参数数据；对收集的历史参数数据进行预处理,得到标准化的监测数据；针对每个监测点,统计其历史参数数据的变化趋势,并标记已知的滤网更换时刻,作为历史标记数据；基于历史标记数据,对每个监测点建立和训练滤网堵塞机器学习模型；利用所述滤网堵塞机器学习模型对当前滤网的实时数据进行在线预测,估计当前的滤网堵塞程度；设定滤网堵塞程度的阈值,当任一监测点的预测结果超过阈值时,判定该监测点对应的滤网需要更换。解决了现有技术根据工作人员经验判断更换滤网时刻，存在局限性的技术问题。
The present invention provides a method, medium and system for estimating the time of filter replacement in an ABS production process, which belongs to the field of ABS production technology, and includes: collecting historical parameter data of multiple filter monitoring points on multiple ABS production lines; preprocessing the collected historical parameter data to obtain standardized monitoring data; for each monitoring point, counting the change trend of its historical parameter data, and marking the known filter replacement time as historical marking data; based on the historical marking data, establishing and training a filter blockage machine learning model for each monitoring point; using the filter blockage machine learning model to perform online prediction of the real-time data of the current filter, and estimating the current filter blockage degree; setting a threshold value of the filter blockage degree, when the prediction result of any monitoring point exceeds the threshold value, determining that the filter corresponding to the monitoring point needs to be replaced. The present invention solves the technical problem that the prior art judges the time of replacing the filter based on the experience of the staff, and has limitations.
Description
Technical Field
The invention belongs to the technical field of ABS production, and particularly relates to a method, medium and system for estimating a filter screen replacement time in an ABS production process.
Background
ABS (AcrylonitrileButadieneStyrene) resin is engineering plastic widely applied to the fields of automobiles, electronic appliances, daily necessities and the like, and is known for its excellent mechanical properties, chemical resistance and processability. In the ABS production process, the melt is filtered by a filter screen in the process links of extrusion, molding and the like so as to remove impurities and improve the product quality. However, the filter screen gradually blocks in the long-term use process, so that the pressure is increased, the flow is reduced, and the production stability and the product performance are seriously affected. Therefore, how to accurately predict the blocking condition of the filter screen and replace the filter screen in time is always one of the key difficulties in ABS production management. At present, the replacement of a filter screen on an ABS production line is mainly judged by experience, production management staff generally estimates the blocking degree of the filter screen by observing the changes of technological parameters such as pressure, temperature and the like, and arranges the replacement time of the filter screen according to a production plan. The method has certain limitations that on one hand, a single parameter is difficult to accurately reflect the overall blocking condition of the filter screen, misjudgment is easy to occur, and on the other hand, resource waste is caused by premature replacement, and production interruption is caused by delayed replacement.
Disclosure of Invention
In view of the above, the invention provides a method, medium and system for estimating the replacement time of a filter screen in the ABS production process, which can solve the technical problems of most of the prior art that the blocking degree of the filter screen is estimated according to experience of production personnel, and the replacement time of the filter screen is arranged according to production plan, and the limitation exists.
The invention is realized in the following way:
The first aspect of the invention provides a method for estimating the replacement time of a filter screen in an ABS production process, which comprises the following steps:
s10, collecting historical parameter data of a plurality of filter screen monitoring points on a plurality of ABS production lines, wherein the parameter data comprise filter screen pressure, filter screen deformation, material fluid temperature and material fluid flow;
s20, preprocessing the collected historical parameter data, including removing abnormal values, interpolating missing values and normalizing, so as to obtain standardized monitoring data;
S30, counting the change trend of historical parameter data of each monitoring point, and marking known filter screen replacement time as historical marking data;
S40, establishing and training a filter screen blocking machine learning model for each monitoring point based on the historical mark data;
s50, online predicting real-time data of the current filter screen by using the filter screen blocking machine learning model, and estimating the current filter screen blocking degree;
and S60, setting a threshold value of the blocking degree of the filter screen, and judging that the filter screen corresponding to any monitoring point needs to be replaced when the prediction result of the monitoring point exceeds the threshold value.
The specific implementation mode of collecting historical parameter data of a plurality of filter screen monitoring points on a plurality of ABS production lines in the step S10 comprises the steps of collecting parameter data such as historical filter screen pressure, filter screen deformation, material fluid temperature, material fluid flow and the like of each monitoring point from a plurality of monitoring sensors arranged on the ABS production lines, wherein the data cover key indexes of filter screen operation in the ABS production process, and provide basis for subsequent predictive modeling.
The specific implementation mode of preprocessing the collected historical parameter data in the step S20 comprises the steps of firstly detecting and removing abnormal values of the data to remove abnormal data points deviating from a normal range, then filling the missing data points by adopting methods such as interpolation to ensure the integrity of the data, and finally normalizing the index data to convert the index data into standardized monitoring data so as to meet the requirements of subsequent modeling and analysis.
The specific implementation mode of counting historical parameter data change trend and marking known filter screen replacement time in the step S30 comprises the steps of carrying out statistical analysis on standardized monitoring data collected by each monitoring point to find change rules of indexes such as pressure, deformation and the like of the filter screen in the using process, and marking the historically known filter screen replacement time by combining production reality to provide a reference basis for training of a subsequent machine learning model.
The specific implementation mode of the filter screen blockage machine learning model established and trained based on the historical marking data in the step S40 comprises the steps of selecting a neural network embedded with a physical model as a prediction model, wherein the model comprises key components such as a physical model module, a time sequence module, an attention module, a feature extraction module, a weight real-time updating module and the like, and establishing a foundation for subsequent online prediction through learning and modeling a filter screen blockage rule by the structured neural network.
The specific implementation mode of the filter screen blockage machine learning model in step S50 for carrying out online prediction on the real-time data of the current filter screen comprises the steps of receiving the parameter data such as pressure, deformation, temperature, flow and the like from each monitoring point in real time, and predicting the blockage degree of the current filter screen by utilizing a neural network model in combination with the trained physical model constraint so as to provide a basis for judging the subsequent filter screen replacement time.
The specific implementation mode of setting the filter screen blocking degree threshold and judging the filter screen replacing time in the step S60 comprises the steps of setting the filter screen blocking degree threshold to 70% -80% according to production practice and experience, and judging that the filter screen corresponding to any monitoring point needs to be replaced when the predicted blocking degree of the monitoring point exceeds the threshold so as to ensure the stability and the product quality of ABS production.
The filter screen at least comprises an extruder head filter screen, a die head filter screen and a filter screen of an ABS production line, wherein the filter screens play a key role in filtering in the ABS production process, and the blocking condition of the filter screens needs to be monitored and predicted in real time.
Wherein, the filter screen at least comprises an extruder head filter screen, a die head filter screen and a filter screen of an ABS production line.
The filter screen blocking machine learning model is a neural network embedded with a physical model, and the physical model is specifically based on an Ergun equation between a filter screen pressure difference and parameters of flow, temperature and viscosity, and is used for restraining the output of the neural network to conform to a physical rule.
Furthermore, the neural network further comprises a time sequence module, an attention module, a feature extraction module and a weight real-time updating module, wherein the weight real-time updating module is used for optimizing weights according to new filter screen replacement records, and the weight real-time updating module has a verification data screening function and is used for automatically selecting proper real-time monitoring data to be used for verifying and updating the weights of the filter screen blocking machine learning model.
Furthermore, the physical model adopts an Ergun equation, and specifically constructs physical constraint through a relation among the pressure difference of a filter screen, the flow, the viscosity of fluid and the blockage rate of the filter screen.
Further, the timing module is used for capturing the timing variation characteristics of each parameter, and the specific structure is a convolution LSTM network.
Further, the attention module is used for capturing the correlation between different characteristics, and the specific structure is a multi-head attention mechanism.
Further, the feature extraction module is used for extracting features from different parameters, and the specific structure is divided into two parts:
1) Extracting characteristics of the filter screen pressure, the material fluid temperature and the material fluid flow by using a one-dimensional convolution network;
2) Extracting characteristics of the deformation data of the filter screen by using a two-dimensional convolution network;
The weight real-time updating module is specifically structured to update model weights on line through Reptile th element learning algorithm.
A second aspect of the present invention provides a computer readable storage medium, where the computer readable storage medium stores program instructions, where the program instructions are configured to execute the above-mentioned method for estimating a screen replacement time in an ABS production process when the program instructions are executed.
A third aspect of the present invention provides a system for estimating a replacement time of a filter screen in an ABS production process, wherein the system includes the above computer-readable storage medium.
Compared with the prior art, the method, medium and system for predicting the replacement moment of the filter screen in the ABS production process have the beneficial effects that the scheme of the invention fully utilizes historical monitoring data collected on the production line, including a plurality of key indexes such as the pressure, deformation, material temperature and flow of the filter screen, and an intelligent filter screen blockage prediction system is established through the steps of data preprocessing, feature learning, dynamic modeling and the like. Compared with the prior art, the method has the following main advantages:
1. The method collects a plurality of monitoring indexes including the pressure, deformation, temperature, flow and the like of the filter screen, and comprehensively reflects the running state of the filter screen in the ABS production process. The fusion of the multi-source data is beneficial to improving the accuracy and the robustness of prediction, and the problem that a single index is difficult to accurately describe the blocking condition of the filter screen is solved.
2. The method combines the physical model with the neural network, on one hand, the output of the neural network is constrained by utilizing the physical rules such as the Ergun equation and the like to ensure that the prediction result accords with the physical rules, and on the other hand, the complex nonlinear relation hidden in the historical data can be excavated through the strong learning capacity of the neural network. The modeling strategy combining physical constraint and data driving ensures that the prediction model has good interpretability and can fully exert the prediction capability of data.
3. The method adopts a convolution LSTM network to capture the change characteristics of each monitoring parameter along with time, and can effectively model the dynamic evolution rule of the filter screen blocking condition. Compared with the existing method only focusing on single-moment prediction, the modeling mode based on time sequence data can better reflect the continuity and the dynamics of the actual operation of the filter screen.
4. The method has the capacity of on-line self-adaptive optimization, and realizes the real-time updating of the weights by adopting Reptile-unit learning algorithm on the basis of a trained prediction model. When a new filter screen replacement record appears, the model can automatically adjust parameters according to the latest sample, and continuously optimize the prediction performance, so that the model is suitable for the change of the production environment and provides guarantee for long-term stable application.
In summary, the invention solves the technical problems that most of the prior art estimates the blocking degree of the filter screen according to experience of production personnel, and arranges the filter screen replacement time according to production plan, and has limitation.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the description of the embodiments of the present invention will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of a method provided by the present invention;
FIG. 2 is a graph of screen pressure versus time with the abscissa being the date;
FIG. 3 is a graph of screen deformation over time with the abscissa being the date;
FIG. 4 is a graph of material temperature over time with the abscissa being the date;
FIG. 5 is a graph of material flow over time with date on the abscissa;
FIG. 6 is a graph of predicted screen plugging results with the abscissa being the date;
FIG. 7 is a graph comparing the tendency of blockage at various monitoring points, with the abscissa being the date.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.
As shown in FIG. 1, the invention provides a flowchart of a method for estimating the replacement time of a filter screen in an ABS production process, which comprises the following steps:
S10, collecting historical parameter data of a plurality of filter screen monitoring points on a plurality of ABS production lines, wherein the parameter data comprise filter screen pressure, filter screen deformation, material fluid temperature and material fluid flow;
s20, preprocessing the collected historical parameter data, including removing abnormal values, interpolating missing values and normalizing, so as to obtain standardized monitoring data;
S30, counting the change trend of historical parameter data of each monitoring point, and marking known filter screen replacement time as historical marking data;
S40, establishing and training a filter screen blocking machine learning model for each monitoring point based on the historical mark data;
S50, online predicting real-time data of the current filter screen by using a filter screen blocking machine learning model, and estimating the current filter screen blocking degree;
and S60, setting a threshold value of the blocking degree of the filter screen, and judging that the filter screen corresponding to any monitoring point needs to be replaced when the prediction result of the monitoring point exceeds the threshold value.
The following describes in detail the specific embodiments of the above steps:
step S10, collecting historical parameter data of a plurality of filter screen monitoring points on a plurality of ABS production lines:
The main purpose of this step is to collect and sort the relevant historical parameter data of the operation of the filter screen during the ABS production process. Specifically, it is desirable to collect historical monitoring data including a plurality of indicators of screen pressure, screen deformation, material fluid temperature, material fluid flow, etc. The data are derived from a plurality of monitoring points on the ABS production line, so that sufficient sample data are obtained, and a foundation is laid for the establishment of a subsequent prediction model.
Step S20, preprocessing the collected historical parameter data:
The raw monitoring data collected may have some outliers or missing values, and therefore, the data needs to be preprocessed. First, outlier detection and removal of the data is required to remove those outlier data points that deviate significantly from the normal range. Then, for the missing data points, interpolation or other methods can be adopted to fill in, so as to ensure the integrity of the data. And finally, carrying out normalization processing on each index data, and converting the index data into standardized monitoring data, so that preparation is made for subsequent modeling and analysis.
Step S30, counting the change trend of the historical parameter data, and marking the known filter screen replacement time:
Standardized monitoring data collected for each monitoring point requires careful analysis of its historical trend. Through statistical analysis of the data, the change rule of indexes such as pressure, deformation and the like of the filter screen in the using process can be found. Meanwhile, in combination with production practice, historically known filter screen replacement time is required to be marked, and a reference basis is provided for subsequent machine learning model training.
Step S40, based on the historical marking data, establishing and training a filter screen blockage machine learning model:
With the historical data and the marking information accumulated in the previous steps, a machine learning predictive model of filter screen plugging can be initially built and trained. A neural network embedded with a physical model is chosen as the predictive model. Specifically, this neural network model includes the following key modules:
1) And the physical model module is used for establishing the relation among the pressure difference, the flow, the fluid viscosity and the blockage rate of the filter screen based on the Ergun equation and is used for restricting the output of the neural network to conform to the physical rule.
2) And the time sequence module is used for capturing the change characteristics of each parameter along with time by adopting a convolution LSTM network.
3) And the attention module adopts a multi-head attention mechanism to learn the correlation among different features.
4) And the characteristic extraction module is used for extracting the characteristics of different types of input data (pressure, temperature, flow and deformation) by adopting different convolution networks.
5) And the weight real-time updating module adopts Reptile element learning algorithm to optimize the weight parameters of the neural network in real time according to the new filter screen replacement record.
Through the neural network model with rich structure and the combination of the supervised learning of the historical mark data, the filter screen blocking law can be well learned and modeled, and a foundation is laid for the follow-up online prediction.
Step S50, online prediction is carried out on real-time data of the current filter screen by utilizing a filter screen blocking machine learning model:
The trained filter screen blockage prediction model can be used for on-line prediction of real-time monitoring data of the filter screen on the current production line. Specifically, the model can receive parameter data such as pressure, deformation, temperature, flow and the like from each monitoring point in real time, and predicts the blocking degree of the current filter screen by combining physical model constraint. The prediction result provides an important basis for judging the subsequent filter screen replacement time.
Step S60, setting a threshold value of the blocking degree of the filter screen, and judging the replacing time of the filter screen:
After obtaining the predicted result of the clogging degree of the filter screen, a reasonable clogging degree threshold value needs to be set. When the predicted blocking degree of any monitoring point exceeds the threshold value, the filter screen corresponding to the monitoring point can be judged to need to be replaced. The threshold value is set by combining production practice and experience, and can be generally set to be 70% -80% of blocking degree. When this threshold is reached, screen changes should be made immediately to ensure the stability of ABS production and product quality.
In general, the method for estimating the replacement moment of the filter screen in the ABS production process fully utilizes historical monitoring data on a production line, and establishes a powerful online prediction system by combining machine learning and a physical model. The system can monitor the blocking state of the filter screen in real time, predict the preferable time for replacing the filter screen in advance, and greatly improve the automation level and the operation efficiency of ABS production.
A second aspect of the present invention provides a computer readable storage medium, where the computer readable storage medium stores program instructions, where the program instructions are configured to execute the above-mentioned method for estimating a screen replacement time in an ABS production process when the program instructions are executed.
A third aspect of the present invention provides a system for estimating a replacement time of a filter screen in an ABS production process, wherein the system includes the above computer-readable storage medium.
For a better understanding and implementation of the present invention, the following provides example 1 of specific steps of a computer readable program of the present invention running in a computer readable medium or a computer-like device, in this example 1, for step S10, historical parameter data of a plurality of screen monitoring points on a plurality of ABS production lines are collected, and the following variables may be defined:
indicating the number of ABS production lines, Indicating the number of screen monitoring points on each production line. For the firstFirst on strip production lineThe following historical parameter data are collected at each monitoring point:
The pressure of the filter screen;
Deformation of the filter screen;
The temperature of the material fluid;
material fluid flow;
Wherein the method comprises the steps of Representing a time series. By collecting the historical parameter data, the comprehensive information of the running state of the filter screen in the ABS production process is obtained.
When preprocessing the collected historical parameter data in step S20, it is first necessary to detect and remove an abnormal value. Assume that a certain monitoring point is at the momentPressure data of (2)The presence of an abnormality can be detected by:
;
Wherein the method comprises the steps of AndRespectively representAnd standard deviation. If it isThen judgeIs an outlier and is culled.
For missing data points, interpolation can be used to fill in. Assume that in a time intervalInternal presence ofThe missing values of (2) can be padded using a cubic spline interpolation method:
;
Wherein the method comprises the steps of Is the coefficient of a cubic polynomial by the known methodAndAnd determining derivative conditions.
Finally, to meet the requirement of subsequent modeling, normalization processing is required for each index data. Assume thatThe normalized filter screen pressure, deformation, temperature and flow data are:
;
;
;
;
In the formula, AndThe maximum and minimum of the corresponding terms, respectively.
Through the preprocessing step, standardized monitoring data are obtained, and a foundation is laid for subsequent analysis and modeling.
When the trend of the historical parameter data is counted and the known screen replacement time is marked in step S30, the following method may be adopted:
For each monitoring point First, theThe production line can calculate the change rate of each parameter along with time:
;
by analyzing the change rates, the change rule of each index of the filter screen in the use process can be found. At the same time, the historically known filter screen replacement time needs to be marked in combination with the actual production record Providing a reference basis for subsequent machine learning model training.
In step S40, a neural network embedded with a physical model is selected as a prediction model when the filter screen plugging machine learning model is built and trained based on the history flag data. Specifically, the model includes the following key components:
1. physical model module:
establishing a filter screen differential pressure by using an Ergun equation Flow rate and flow rateViscosity of fluidBlockage rateRelationship between:
;
Wherein, Is the aperture of the filter screen,Is the fluid density. The physical model is used for constraining the output of the neural network to conform to the physical rule.
2. A time sequence module:
The convolution LSTM network is adopted to capture the change characteristics of each parameter along with time. Specifically, for the first Strip production line NoEach monitoring point is input with the following sequenceWhereinIs the time window length. The convolutional LSTM network can extract these timing features to provide support for subsequent occlusion prediction.
3. Attention module:
and a multi-head attention mechanism is adopted to learn the correlation among different characteristics. For the first Strip production line NoThe monitoring points are at the momentCharacteristic representation of (a)The attention mechanism can be expressed as:
;
;
Wherein, Is a weight matrix that can be learned. The attention mechanism can capture the correlation between different characteristics and improve the expression capacity of the model.
4. And the feature extraction module is used for:
for different types of input data, different convolution networks are used for feature extraction.
For screen pressureTemperature of material fluidAnd flow rateWaiting for one-dimensional data, and extracting features by using a one-dimensional convolution network;
For deformation of the filter screen And (5) waiting for two-dimensional data, and extracting features by using a two-dimensional convolution network.
By the differentiated feature extraction method, the characteristics of different types of data can be better utilized.
5. And the weight real-time updating module is used for:
In order to enable the model to dynamically adjust parameters along with time and adapt to the change of the production environment, a Reptile-unit learning algorithm is adopted to optimize and update the model weight in real time. Specifically, when a new screen change record appears, it is used as a validation set, and model parameters are fine-tuned by Reptile algorithm to better fit the new sample distribution. Meanwhile, the module also has the function of automatically screening proper verification data, and the online learning effect is further improved.
Through the design of the neural network model and the combination of the constraint of the physical model, the historical monitoring data can be fully utilized, the filter screen blocking rule can be learned and modeled, and powerful support is provided for subsequent online prediction.
In step S50, when the filter screen blocking machine learning model is used for online prediction of the real-time data of the current filter screen, the pressure from each monitoring point is received in real timeDeformation ofTemperature (temperature)And flow rateAnd isoparametric data. The blocking degree of the current filter screen can be predicted by combining the neural network model trained in the step S40:
;
Wherein, Representing parameters of the model. The prediction result provides an important basis for judging the subsequent filter screen replacement time.
When the threshold value of the clogging degree of the screen is set and the screen replacement timing is judged in step S60, the threshold value of the clogging degree of the screen may be set according to production practice and experienceIs set to 70% -80%. When any monitoring pointIs the first of (2)Predicted degree of jam for strip production lineWhen the threshold value is exceeded, the filter screen corresponding to the monitoring point can be judged to need to be replaced:
;
by the method based on real-time monitoring and prediction, the filter screen blocking seedling can be found in time, decision support is provided for production management staff, and the stability of ABS production and the product quality are ensured.
In general, the method for estimating the replacement moment of the filter screen in the ABS production process fully utilizes historical monitoring data on a production line, combines physical model constraint and strong neural network prediction capability, and establishes an intelligent online monitoring and prediction system. The system can monitor the running state of the filter screen in real time, predicts the preferable time for replacing the filter screen in advance, and greatly improves the automation level and the operation efficiency of ABS production.
Specifically, the principle of the invention is to build an intelligent prediction system based on the combination of machine learning and physical model. The key of the system is that historical monitoring data collected on a production line is fully utilized, and a machine learning model capable of accurately predicting the blocking degree of the filter screen is finally obtained through the steps of data preprocessing, feature learning, dynamic modeling and the like. This predictive model contains the following key innovation points:
1. physical model constraint when the machine learning prediction model is established, the method is integrated with the physical models such as an Ergun equation and the like and is used for constraining the output of the neural network. The Ergun equation describes the relationship between the filter screen pressure differential, flow, fluid viscosity and plug rate, conforming to the laws of physics. The physical model is embedded into the neural network, so that the prediction result can be ensured not to violate the basic physical law, and the interpretation and reliability of the model are improved.
2. Modeling time sequence characteristics, namely, the blocking of a filter screen is a dynamic change process, and the change characteristics of each monitoring parameter along with time need to be considered. The method adopts a convolution LSTM network to capture the time sequence characteristics, and the network structure can effectively extract time dependency information contained in an input sequence and provide support for dynamic prediction. By means of the time sequence modeling mode, the blocking trend of the filter screen can be predicted more accurately, and the method is not limited to a state at a certain moment.
3. The multisource characteristic fusion is that the running state of the filter screen is the result of the combined action of a plurality of factors, and the blocking degree of the single monitoring index is difficult to comprehensively reflect. The method collects various heterogeneous data such as filter screen pressure, deformation, material temperature, flow and the like, and adopts a differential characteristic extraction method (a one-dimensional convolution network and a two-dimensional convolution network) to fully utilize the information of the multi-source data. The fusion of the multi-source features can provide richer and more comprehensive input and improve the generalization capability of the prediction model.
4. In the actual production, the running state of the filter screen can be changed to a certain extent due to the continuous change of the process conditions. In order to enable the prediction model to adapt to the changes, the method adopts Reptile-element learning algorithm to realize real-time optimization and update of model parameters. When a new filter screen replacement record appears, the model automatically adjusts the weight according to the latest sample, so that the prediction result can continuously track the actual running condition of the filter screen, and the accuracy and stability of prediction are improved.
In summary, the technical principle of the method is mainly characterized in that 1) the interpretability of the model is improved by combining with the constraint of a physical model, 2) a dynamic change rule is captured by adopting time sequence feature modeling, 3) the prediction capability is enhanced by utilizing multi-source feature fusion, and 4) the adaptability of the model is enhanced by self-adaption on-line optimization.
In order to better understand the invention, an embodiment 2 of a specific application scenario is designed below, wherein an enterprise has 3 ABS production lines, and each production line is provided with 5 filter screen monitoring points which are respectively positioned at key positions such as an extruder head, a die head, a filter and the like. The enterprise is always seeking to improve the automation level and the operation efficiency of the ABS production process, hopes to find the filter screen blocking seedling in time, and provides decision support for filter screen replacement. Based on the method, the enterprise decides to adopt the method for estimating the replacement time of the filter screen in the ABS production process, and the method is implemented as follows:
1. data collection
Firstly, the enterprise collects historical operation data of 5 filter screen monitoring points on 3 ABS production lines in the last 2 years, wherein the historical operation data comprise parameters such as filter screen pressure, filter screen deformation, material fluid temperature, material fluid flow and the like. Taking the 3 rd monitoring point of the 1 st production line as an example, the collected part of historical data is shown in table 1:
table 1 historical data table
| Time of | Filter screen pressure (MPa) | Deformation of filter screen (mm) | Material temperature (°c) | Material flow (L/min) |
| 2022-01-0108:00 | 2.1 | 0.3 | 210 | 80 |
| 2022-01-0108:15 | 2.2 | 0.4 | 211 | 79 |
| 2022-01-0108:30 | 2.3 | 0.4 | 209 | 78 |
| ... | ... | ... | ... | ... |
| 2023-12-3123:45 | 3.2 | 0.8 | 215 | 72 |
Fig. 2-5 are graphs showing the trend of the screen monitoring data over time, showing the trend of four key parameters of screen pressure, deformation, material temperature and material flow over the period of 2022, 1/12/31/1/2022. Four sub-graphs are included in the graph, showing the variation of these parameters over time, respectively. The pressure and deformation of the filter screen can be observed to be in an ascending trend, the temperature of the material is relatively stable, and the flow rate of the material is in a descending trend. These trends intuitively reflect the process of gradual clogging of the screen. Through the collection of the historical data, enterprises accumulate rich filter screen running state samples, and a foundation is laid for subsequent predictive modeling.
2. Data preprocessing
The collected raw monitoring data inevitably has some abnormal values and missing values, so that necessary pretreatment is required. For the data of the 3 rd monitoring point of the 1 st production line, the specific preprocessing steps are as follows:
(1) Outlier detection and removal:
taking the pressure data of the filter screen as an example, calculating the standard fraction: ;
Wherein, As the mean value of the sample,Is the standard deviation of the samples. If it isThen the data point is determined to be an outlier and is culled. In this way, a total of 18 outlier pressure data points are removed.
(2) Filling up the missing value:
and filling the missing data points by adopting a cubic spline interpolation method. Taking the loss of the flow data in a certain period as an example, a loss interval is set as Then the following interpolation function can be constructed:
;
Wherein the method comprises the steps of For the undetermined coefficients, they can be determined by known boundary conditions. In this way, a total of 27 missing traffic data points are filled.
(3) Data normalization:
In order to eliminate the influence of dimension on subsequent modeling, the minimum-maximum normalization processing is carried out on each monitoring index data. Taking the filter screen pressure as an example, normalized pressure data The calculation is as follows:
;
the normalization processing mode of other indexes is similar. Through the preprocessing step, the enterprise obtains standardized monitoring data, and a foundation is laid for subsequent analysis and modeling.
3. Historical data analysis
After the data preprocessing is completed, the enterprise starts to analyze the change trend of the historical monitoring data of each monitoring point. Taking the 3 rd monitoring point of the 1 st production line as an example, the analysis result is as follows:
(1) The pressure of the filter screen gradually rises from the initial 2.1MPa to 3.2MPa, and the rising amplitude reaches 52%.
(2) The deformation of the filter screen is gradually increased from 0.3mm to 0.8mm, and the increase is 167%.
(3) The temperature of the material is basically kept stable and fluctuates between 209 ℃ and 215 ℃.
(4) The material flow rate is gradually reduced from the initial 80L/min to 72L/min, and the reduction is 10%.
Meanwhile, the enterprise marks 3 times of screen replacement time in the history of the monitoring point in the production record, which respectively occur on the day of 2022, the day of 2023, the day of 2021, the day of 10, and the day of 2023, the day of 7, and the day of 5. Through analysis of the historical data, enterprises master basic change rules of the running state of the monitoring point filter screen, and an important basis is provided for the establishment of a subsequent prediction model.
4. Predictive model training
Based on the foregoing historical data analysis results, the enterprise began to set up a predictive model of screen plugging. Considering that screen plugging is a complex process involving multiple physical factors, enterprises decide to use a neural network embedded with a physical model as a predictive model, which specifically includes the following key modules:
(1) Physical model module-Enterprise describes the pressure difference of the filter screen by adopting an Ergun equation Flow rate and flow rateViscosity of fluidBlockage rateThe relation between the two is taken as a constraint condition: Wherein, the method comprises the steps of, Is the aperture of the filter screen,Is the fluid density. The physical model is embedded into the neural network, so that the output result is ensured to accord with the physical rule.
(2) And the time sequence module is used for capturing the change characteristics of each monitoring parameter along with time, and an enterprise adopts a convolution LSTM network as a time sequence characteristic extraction module. The input of the module is the pastSequences of monitored data within time steps, e.g.WhereinNormalized pressure, deformation, temperature and flow data, respectively. The convolution LSTM network can effectively extract these timing characteristics and provide support for subsequent occlusion prediction.
(3) Attention module in order to capture the correlation between different monitoring indexes, the enterprise integrates a multi-head attention mechanism in the model. Specifically, for time of dayCharacteristic representation of timeThe attention module calculates as follows:
;
;
Wherein, Is a weight matrix that can be learned. The attention mechanism can adaptively give different weights to different features, and the capturing capacity of the model on key features is enhanced.
(4) And the characteristic extraction module is used for extracting different characteristics of enterprises according to different types of monitoring data. For one-dimensional pressure, temperature and flow data, a one-dimensional convolution network is used for extracting features, and for two-dimensional deformation data, a two-dimensional convolution network is used for extracting features. By means of the differentiated feature extraction, the characteristics of various data can be better utilized, and high-quality input features are provided for subsequent blockage prediction.
(5) And the weight real-time updating module is used for enabling the model to dynamically adjust parameters along with time, adapting to the change of the production environment, and enabling the enterprise to integrate Reptile-element learning algorithm in the prediction model to realize real-time optimization of the weight. Specifically, whenever a new screen change record occurs, the enterprise will use it as a validation set to fine tune the model parameters through Reptile algorithm to better fit the new sample distribution. Meanwhile, the module also has the function of automatically screening proper verification data, and the online learning effect is further improved.
With the model design, the enterprise trains and verifies the model by using the historical data of the 3 rd monitoring point of the 1 st production line. In the training process, the enterprise adopts a layered sampling strategy, so that the proportion of samples in various blocking states is balanced, and the generalization of the model is improved. And finally, obtaining a machine learning model capable of accurately predicting the blocking degree of the filter screen of the monitoring point through repeated iterative optimization.
5. On-line prediction and decision-making, namely, a trained prediction model is provided, and enterprises start to apply in real time on a production line. Taking the 3 rd monitoring point of the 1 st production line as an example, the specific process is as follows:
(1) And acquiring real-time data, namely acquiring the data of the pressure, deformation, material temperature and material flow of the filter screen of the monitoring point by a sensor arranged on the production line in real time, and transmitting the data to a prediction model.
(2) The blockage degree prediction is that a prediction model receives the monitoring data in real time and predicts the current blockage degree of the filter screen of the monitoring point according to the mechanisms such as physical constraint, time sequence characteristic, multi-source characteristic fusion and the like introduced above。
(3) Decision support, namely, enterprises threshold the blocking degree of a filter screen according to production practice experienceSet at 80%. Predictive occlusion degree at any one monitoring pointWhen this threshold is exceeded, a screen change decision is triggered.
As shown in fig. 6, the screen clogging degree prediction results during 2023, 11, 1, to 2023, 12, 31 are shown. In the figure, a blue solid line indicates an actual clogging degree, a red dotted line indicates a predicted clogging degree, a green dotted line indicates a replacement threshold value (0.8), and a yellow hatched area indicates an early warning area. It can be seen from the graph that the predicted result is very close to the actual blocking degree, when the predicted blocking degree exceeds the threshold value, the filter screen is required to be replaced by entering the early warning area, in the practical application, the enterprise finds that the predicted blocking degree of the 3 rd monitoring point of the 1 st production line reaches 82% after 2 hours and exceeds 80% of the threshold value, and after receiving the early warning signal, production management staff immediately arranges the filter screen corresponding to the monitoring point to be replaced, so that the production interruption and the product quality problem caused by the blocking of the filter screen are avoided.
Through continuous real-time monitoring and prediction, enterprises find out that certain differences exist in the filter screen blocking condition of each monitoring point. For example, the 4 th monitoring point of the 2 nd production line has slower changes of the pressure and deformation of the filter screen due to the adjustment of the process parameters, and the predicted blocking degree does not exceed the threshold. The 1 st monitoring point of the 3 rd production line shows a faster blocking trend, and the enterprise correspondingly changes the filter screen earlier. Figure 7 compares the plugging trend of three different monitoring points (line 1-point 3, line 2-point 4, line 3-point 1) throughout 2023. As can be seen from the graph, there is a significant difference in the rate of occlusion at the different monitoring points. The line 3-point 1 is plugged at the fastest rate, while the line 2-point 4 is plugged at the slowest rate. The black dashed line indicates the replacement threshold (0.8). The comparison graph is helpful for production management personnel to know the blocking characteristics of different monitoring points, so that a more targeted maintenance strategy is formulated.
In general, by adopting the method, enterprises can realize intelligent monitoring and prediction of the state of the filter screen in the ABS production process, and timely and accurate decision support is provided for filter screen replacement. The production stability and the product quality are improved, and the manpower and the resource consumption for replacing the filter screen are greatly saved. Meanwhile, a large amount of monitoring and predicting data accumulated by the method also provides valuable data support for further optimizing ABS production process parameters.
The foregoing is merely illustrative of the present invention, and the present invention is not limited thereto, and any person skilled in the art will readily recognize that variations or substitutions are within the scope of the present invention.
Claims (10)
1. The method for estimating the replacement time of the filter screen in the ABS production process is characterized by comprising the following steps of:
s10, collecting historical parameter data of a plurality of filter screen monitoring points on a plurality of ABS production lines, wherein the parameter data comprise filter screen pressure, filter screen deformation, material fluid temperature and material fluid flow;
s20, preprocessing the collected historical parameter data, including removing abnormal values, interpolating missing values and normalizing, so as to obtain standardized monitoring data;
S30, counting the change trend of historical parameter data of each monitoring point, and marking known filter screen replacement time as historical marking data;
S40, establishing and training a filter screen blocking machine learning model for each monitoring point based on the historical mark data;
s50, online predicting real-time data of the current filter screen by using the filter screen blocking machine learning model, and estimating the current filter screen blocking degree;
and S60, setting a threshold value of the blocking degree of the filter screen, and judging that the filter screen corresponding to any monitoring point needs to be replaced when the prediction result of the monitoring point exceeds the threshold value.
2. The method for estimating a replacement time of a filter screen in an ABS production process according to claim 1, wherein the filter screen at least comprises an extruder head filter screen, a die head filter screen, and a filter screen of an ABS production line.
3. The method for estimating a replacement moment of a filter screen in an ABS production process according to claim 1, wherein the filter screen plugging machine learning model is a neural network embedded with a physical model, and the physical model is specifically based on an Ergun equation between a filter screen pressure difference and flow, temperature and viscosity parameters, and is used for constraining output of the neural network to conform to a physical rule.
4. The method for estimating a replacement moment of a filter screen in an ABS production process according to claim 3, wherein the neural network further comprises a timing module, an attention module, a feature extraction module, and a weight real-time update module, wherein the weight real-time update module is configured to optimize weights according to new filter screen replacement records, and the weight real-time update module has a verification data screening function, and is configured to automatically select appropriate real-time monitoring data for verifying and updating weights of the filter screen blocking machine learning model.
5. The method for estimating a replacement time of a filter screen in an ABS production process according to claim 4, wherein the physical model uses an Ergun equation, and specifically constructs physical constraints by a relation among a filter screen pressure difference, a flow rate, a fluid viscosity and a filter screen blockage rate.
6. The method for estimating a replacement time of a filter screen in an ABS production process according to claim 4, wherein the timing module is configured to capture a timing variation characteristic of each parameter, and the specific structure is a convolution LSTM network.
7. The method for estimating a replacement time of a filter screen in an ABS manufacturing process according to claim 4, wherein the attention module is configured to capture a correlation between different features, and the specific structure is a multi-head attention mechanism.
8. The method for estimating a replacement time of a filter screen in an ABS production process according to claim 4, wherein the feature extraction module is configured to extract features from different parameters, and the specific structure is divided into two parts:
1) Extracting characteristics of the filter screen pressure, the material fluid temperature and the material fluid flow by using a one-dimensional convolution network;
2) Extracting characteristics of the deformation data of the filter screen by using a two-dimensional convolution network;
The weight real-time updating module is specifically structured to update model weights on line through Reptile th element learning algorithm.
9. A computer readable storage medium, wherein program instructions are stored in the computer readable storage medium, and the program instructions are executed to perform a method for estimating a moment of replacement of a filter screen in an ABS production process according to any one of claims 1 to 8.
10. A screen change time estimation system for an ABS production process, comprising the computer readable storage medium of claim 9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202411554028.XA CN119066369A (en) | 2024-11-04 | 2024-11-04 | A method, medium and system for estimating filter replacement time in ABS production process |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202411554028.XA CN119066369A (en) | 2024-11-04 | 2024-11-04 | A method, medium and system for estimating filter replacement time in ABS production process |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN119066369A true CN119066369A (en) | 2024-12-03 |
Family
ID=93633823
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202411554028.XA Pending CN119066369A (en) | 2024-11-04 | 2024-11-04 | A method, medium and system for estimating filter replacement time in ABS production process |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN119066369A (en) |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109615226A (en) * | 2018-12-12 | 2019-04-12 | 焦点科技股份有限公司 | A method for abnormal monitoring of operational indicators |
| WO2021072890A1 (en) * | 2019-10-18 | 2021-04-22 | 平安科技（深圳）有限公司 | Traffic abnormality monitoring method and apparatus based on model, and device and storage medium |
| WO2023093010A1 (en) * | 2021-11-26 | 2023-06-01 | 中国华能集团清洁能源技术研究院有限公司 | Wind power prediction method and device based on deep learning fusion model |
| CN116432529A (en) * | 2023-04-13 | 2023-07-14 | 华晨宝马汽车有限公司 | Method for determining drying temperature of automobile coating and application |
| CN117390428A (en) * | 2023-10-24 | 2024-01-12 | 江苏科跃过滤设备有限公司 | A filter adaptive adjustment method and system based on big data |
| CN117743961A (en) * | 2023-12-12 | 2024-03-22 | 上海迈内能源科技有限公司 | Cable joint fault detection method and system based on long-short-term memory LSTM |
| CN118014185A (en) * | 2023-12-22 | 2024-05-10 | 绍兴市麦芒智能科技有限公司 | Municipal pipe network health degree on-line monitoring system based on big data |
| CN118014115A (en) * | 2023-11-09 | 2024-05-10 | 福建亿把首信息科技有限公司 | Production line loss monitoring method based on long-short-term memory network |
| CN118469352A (en) * | 2024-07-12 | 2024-08-09 | 北京爱朗格瑞科技有限公司 | Deep learning driven power grid line loss dynamic prediction and optimization control method |
| WO2024207627A1 (en) * | 2023-04-03 | 2024-10-10 | 山东大学 | Multivariate time series anomaly detection method for intelligent internet of things system |
| CN118861642A (en) * | 2024-08-19 | 2024-10-29 | 江苏理工学院 | A water quality monitoring and control method and system |
-
2024
- 2024-11-04 CN CN202411554028.XA patent/CN119066369A/en active Pending
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109615226A (en) * | 2018-12-12 | 2019-04-12 | 焦点科技股份有限公司 | A method for abnormal monitoring of operational indicators |
| WO2021072890A1 (en) * | 2019-10-18 | 2021-04-22 | 平安科技（深圳）有限公司 | Traffic abnormality monitoring method and apparatus based on model, and device and storage medium |
| WO2023093010A1 (en) * | 2021-11-26 | 2023-06-01 | 中国华能集团清洁能源技术研究院有限公司 | Wind power prediction method and device based on deep learning fusion model |
| WO2024207627A1 (en) * | 2023-04-03 | 2024-10-10 | 山东大学 | Multivariate time series anomaly detection method for intelligent internet of things system |
| CN116432529A (en) * | 2023-04-13 | 2023-07-14 | 华晨宝马汽车有限公司 | Method for determining drying temperature of automobile coating and application |
| CN117390428A (en) * | 2023-10-24 | 2024-01-12 | 江苏科跃过滤设备有限公司 | A filter adaptive adjustment method and system based on big data |
| CN118014115A (en) * | 2023-11-09 | 2024-05-10 | 福建亿把首信息科技有限公司 | Production line loss monitoring method based on long-short-term memory network |
| CN117743961A (en) * | 2023-12-12 | 2024-03-22 | 上海迈内能源科技有限公司 | Cable joint fault detection method and system based on long-short-term memory LSTM |
| CN118014185A (en) * | 2023-12-22 | 2024-05-10 | 绍兴市麦芒智能科技有限公司 | Municipal pipe network health degree on-line monitoring system based on big data |
| CN118469352A (en) * | 2024-07-12 | 2024-08-09 | 北京爱朗格瑞科技有限公司 | Deep learning driven power grid line loss dynamic prediction and optimization control method |
| CN118861642A (en) * | 2024-08-19 | 2024-10-29 | 江苏理工学院 | A water quality monitoring and control method and system |
Non-Patent Citations (1)
| Title |
|---|
| 刘巍;刘威;谷建伟;: "基于机器学习方法的油井日产油量预测", 石油钻采工艺, no. 01, 20 January 2020 (2020-01-20) * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109365769B (en) | A prediction method for mold breakout based on mixed model judgment | |
| KR101948604B1 (en) | Method and device for equipment health monitoring based on sensor clustering | |
| CN114015825B (en) | Method for monitoring abnormal state of blast furnace heat load based on attention mechanism | |
| CN103116306B (en) | Automatic stepping type ordered time interval dividing method | |
| CN114282443A (en) | Remaining service life prediction method based on MLP-LSTM supervised joint model | |
| CN113568774A (en) | A real-time anomaly detection method for multidimensional time series data using unsupervised deep neural networks | |
| CN116757898B (en) | Flood season pollution intensity accounting method and system based on predictive comparison | |
| CN118211082B (en) | Oil filter element residual life prediction method and system based on data analysis | |
| CN106649050A (en) | Multi-parameter running situation graphic representation method for time sequential system | |
| CN118900275A (en) | Industrial Internet cloud platform based on data intelligence | |
| CN110474862B (en) | A kind of network traffic abnormal detection method and device | |
| CN117370330A (en) | Prophet-SVM model-based data monitoring and early warning method of big data acquisition system | |
| CN107808209B (en) | Wind power plant abnormal data identification method based on weighted kNN distance | |
| CN119066369A (en) | A method, medium and system for estimating filter replacement time in ABS production process | |
| CN111625525A (en) | Environmental data repairing/filling method and system | |
| CN117828443A (en) | Water quality multi-index multi-step prediction method based on sequence decomposition and autocorrelation | |
| CN113673822A (en) | A flexible scheduling method and system | |
| CN108509990A (en) | A kind of sequential key assignments type industrial process data Parallel analytic method | |
| CN118427748A (en) | Autoformer and water quality index early warning method of double-branch attention structure | |
| CN114358190B (en) | Transfer learning method and device | |
| CN118313694B (en) | Intelligent expressway cooperative control method, system and device | |
| CN117931897B (en) | Temperature data operation monitoring method for heating device | |
| CN116776086B (en) | Signal fault discriminating method and device based on self-attention mechanism self-encoder | |
| CN118736512B (en) | Automatic identification method and system for mine transport vehicle | |
| CN119204315A (en) | Urban planning design method and system |
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 |