CN115016421A - Batch quality control method for flexible manufacturing of pressure powder in aerospace initiating explosive devices - Google Patents
Batch quality control method for flexible manufacturing of pressure powder in aerospace initiating explosive devices Download PDFInfo
- Publication number
- CN115016421A CN115016421A CN202210892457.2A CN202210892457A CN115016421A CN 115016421 A CN115016421 A CN 115016421A CN 202210892457 A CN202210892457 A CN 202210892457A CN 115016421 A CN115016421 A CN 115016421A
- Authority
- CN
- China
- Prior art keywords
- batch
- initiating explosive
- product
- pressure
- quality control
- 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.)
- Granted
Links
- 238000004519 manufacturing process Methods 0.000 title claims abstract description 99
- 239000002360 explosive Substances 0.000 title claims abstract description 55
- 230000000977 initiatory effect Effects 0.000 title claims abstract description 49
- 239000000843 powder Substances 0.000 title claims abstract description 48
- 238000003908 quality control method Methods 0.000 title claims abstract description 31
- 238000000034 method Methods 0.000 claims abstract description 122
- 230000008569 process Effects 0.000 claims abstract description 97
- 239000003814 drug Substances 0.000 claims abstract description 15
- 238000003825 pressing Methods 0.000 claims description 37
- 238000010923 batch production Methods 0.000 claims description 15
- 230000035882 stress Effects 0.000 claims description 13
- 238000004364 calculation method Methods 0.000 claims description 6
- 230000003111 delayed effect Effects 0.000 claims description 4
- 238000005259 measurement Methods 0.000 claims description 4
- 230000032683 aging Effects 0.000 claims description 3
- 238000004458 analytical method Methods 0.000 claims description 3
- 238000000605 extraction Methods 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- YHXISWVBGDMDLQ-UHFFFAOYSA-N moclobemide Chemical compound C1=CC(Cl)=CC=C1C(=O)NCCN1CCOCC1 YHXISWVBGDMDLQ-UHFFFAOYSA-N 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 3
- 230000007704 transition Effects 0.000 claims description 3
- 229940079593 drug Drugs 0.000 claims 1
- 239000000126 substance Substances 0.000 claims 1
- 238000003745 diagnosis Methods 0.000 abstract description 7
- 238000012797 qualification Methods 0.000 abstract 1
- 239000003795 chemical substances by application Substances 0.000 description 6
- 239000007787 solid Substances 0.000 description 5
- 230000008859 change Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 238000011038 discontinuous diafiltration by volume reduction Methods 0.000 description 4
- 239000002245 particle Substances 0.000 description 4
- 238000011160 research Methods 0.000 description 3
- 238000005056 compaction Methods 0.000 description 2
- 230000008878 coupling Effects 0.000 description 2
- 238000010168 coupling process Methods 0.000 description 2
- 238000005859 coupling reaction Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000009826 distribution Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 239000011148 porous material Substances 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 238000005482 strain hardening Methods 0.000 description 2
- 230000001131 transforming effect Effects 0.000 description 2
- 230000005483 Hooke's law Effects 0.000 description 1
- 238000007476 Maximum Likelihood Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000001364 causal effect Effects 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000001739 density measurement Methods 0.000 description 1
- 238000005474 detonation Methods 0.000 description 1
- 238000006073 displacement reaction Methods 0.000 description 1
- 229920001971 elastomer Polymers 0.000 description 1
- 239000000806 elastomer Substances 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000014759 maintenance of location Effects 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 238000002156 mixing Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000002787 reinforcement Effects 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B19/00—Programme-control systems
- G05B19/02—Programme-control systems electric
- G05B19/418—Total factory control, i.e. centrally controlling a plurality of machines, e.g. direct or distributed numerical control [DNC], flexible manufacturing systems [FMS], integrated manufacturing systems [IMS] or computer integrated manufacturing [CIM]
- G05B19/41875—Total factory control, i.e. centrally controlling a plurality of machines, e.g. direct or distributed numerical control [DNC], flexible manufacturing systems [FMS], integrated manufacturing systems [IMS] or computer integrated manufacturing [CIM] characterised by quality surveillance of production
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B2219/00—Program-control systems
- G05B2219/30—Nc systems
- G05B2219/32—Operator till task planning
- G05B2219/32252—Scheduling production, machining, job shop
-
- 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
- Y02P90/02—Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS]
Landscapes
- Engineering & Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- Manufacturing & Machinery (AREA)
- Quality & Reliability (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Feedback Control In General (AREA)
Abstract
The invention discloses a batch quality control method for flexibly manufacturing a pressure powder of an aerospace initiating explosive device, and mainly relates to the field of quality control of the aerospace initiating explosive device; the method comprises the following steps: s1, quantitatively describing the flexible manufacturing process of the pressure powder for the aerospace initiating explosive devices; s2, carrying out dynamic Bayesian modeling on the manufacturing process, and extracting key factors influencing the production quality; s3, defining the input and output processes of the pressurized medicine production process based on an EWMA algorithm; s4, performing batch quality control based on an EWMA algorithm; the method is used for performing batch quality control on flexible manufacturing of the pressure-loading medicine of the aerospace initiating explosive device based on the dynamic Bayesian network and the EWMA, realizes deviation diagnosis and quality control on the manufacturing process, and finally improves the qualification rate of products.
Description
Technical Field
The invention relates to the field of aerospace initiating explosive device quality control, in particular to a batch quality control method for flexible manufacturing of aerospace initiating explosive device loading powder.
Background
The manufacturing of the aerospace initiating explosive device is a typical batch production process, the manufacturing process has the characteristics of large data volume, multiple parameters and complex specification, and deviation sources such as part manufacturing errors, positioning errors, assembly errors and operation defects in the manufacturing process all affect the quality of finished products. The existing initiating explosive device production technology is still difficult to achieve full-process automatic production, the process links with strong manufacturability still depend on manual experience, and the quality control depends on the proficiency of operators to a certain extent. This results in fluctuations, transmissibility and coupling in the measured values of the quality control, which causes cumulative errors in the production process, and causes difficulties in locating and identifying the product deviation sources, which in turn leads to uncertainty in the control. Due to a plurality of deviation factors influencing the quality of finished products in the production process of initiating explosive devices, the quality control difficulty is higher.
Disclosure of Invention
The invention aims to solve the problems in the prior art and provides a batch quality control method for flexible manufacturing of a pressure-loading agent of an aerospace initiating explosive device.
In order to achieve the purpose, the invention is realized by the following technical scheme:
the batch quality control method for the flexible manufacturing of the pressure loading medicine of the aerospace initiating explosive device comprises the following steps:
s1, quantitatively describing the flexible manufacturing process of the pressure powder for the aerospace initiating explosive devices;
s2, carrying out dynamic Bayesian modeling on the flexible manufacturing process of the pressure-loading explosive of the aerospace initiating explosive device, and extracting key factors influencing the production quality;
s3, defining the input and output processes of the pressurized medicine production process based on an EWMA algorithm;
and S4, performing batch quality control based on the EWMA algorithm.
Preferably, the step S1 includes the steps of:
s11, introducing a Barersian equation to quantitatively describe the process of loading the pressing powder, and obtaining the relation between the pressure and the strain of the contact point of the pressing powder as follows:
lgP m ＝lgP+L(β-1)；
wherein P is the pressure of the pressing process, P m P, P for pressing the powder to a pressure corresponding to the compact state m The units of (A) are all MPa;
l is a pressing factor corresponding to the ratio between the elastic modulus of the powder and the pressing stress;
beta is equal to epsilon +1 and is the relative volume of a pressed compact;
s12, introducing a Chuanbei Hough pressing equation, and obtaining the relation between the pressure and the density in the pressing process as follows:
wherein M is the modulus of elasticity, ρ 0 Is the initial density of the powder, p m The density in the powder dense state is denoted by ρ as the current density.
Preferably, the step S2 includes the steps of:
s21, starting from the actual production flow of the pressure loading of the aerospace initiating explosive devices, constructing a mapping relation from a deviation source node in the process to an observation node for quality control based on a deviation source analysis and uncertainty reasoning method, and establishing a dynamic Bayesian network on the basis;
s22, completing dynamic Bayesian network modeling of the loading process of the aerospace initiating explosive device through deviation factor extraction, deviation source node and observation node definition, data processing, network structure determination and network parameter determination processes, and then simplifying the dynamic Bayesian network model according to the Markov property of a random process and the constraint conditions of time invariance of state transition probability and the like.
Preferably, the step S3 includes the steps of:
s31, the EWMA algorithm adjusts the system inputs by minimum variance control, assuming k is batch and a set of observations { x } k The output prediction is expressed as:
y k+1 ＝λx k +λ(1-λ)x k-1 +...+λ(1-λ) k-1 x 1 ；
s32, defining the process input u ═ lgP, the process output
The production process is described as follows: y is k ＝βu k-1 +α k ；
Wherein u k-1 For the input at the beginning of the batch production process, y k Alpha being the output of the batch production process k And beta is the linear model parameter of the real physical process, beta is the process gain, alpha k Is a disturbance in the process;
s33, introducing an ARIMA model, wherein the common disturbance form in the process of manufacturing the pressure-charged explosive of the aerospace initiating explosive device in discrete batches is represented as IMA (1,1) disturbance:
α k ＝α k-1 +ε k -θε k-1 ；
wherein epsilon k Is white noise, theta is a model parameter, and when theta is 1, alpha is k ＝ε k The disturbance term is degenerated into a normally distributed white noise sequence; when the value of theta is smaller, the correlation of the IMA (1,1) sequence is stronger; when the value of theta is equal to 0,
the disturbance item becomes the superposition of all historical white noises;
s34, using the estimated value a 0 And b parameter α k And beta, initializing to obtain a prediction model for controlling between batches:
Y k+1 ＝bu k +a k ；
and then continuously iterating in the EWMA algorithm to obtain:
a k ＝λ(Y k -bu k-1 )+(1-λ)a k-1 ,0≤λ≤1；
wherein, λ is a variable discount factor of the EWMA algorithm, and is related to a model parameter θ, a model mismatching degree β/b and a time delay item;
s35, after completing the iteration of each batch, calculating the input of the next batch by the EWMA controller:
wherein, T k Is the target value for the kth batch.
Preferably, the step S4 includes the steps of:
s41, the equipment parameters are reset and then appear as jump of the control target on the flexible manufacturing model, and in the next batch:
wherein T is 0 Hopping for a control target;
s42, control when measurement data is delayed, and when observation data is delayed, the model intercept term is updated as follows:
a t+1 ＝λ(Y t-d -bu t-d )+(1-λ)a t ,0≤λ≤1；
s43, aging in a specific direction, and carrying out a linear model of the pressure charge production process with specific direction drift:
y k ＝βu k-1 +α k +δk；
u k-1 for the input at the beginning of the batch production process, y k Alpha being the output of the batch production process k And beta is the linear model parameter of the real physical process, beta is the process gain, alpha k δ is the drift term for the disturbance in the process.
Preferably, a double exponential weighted moving average control algorithm is introduced in the step S43, and the estimated value a is used 0 And b parameter α k And beta, initializing to obtain a prediction model of batch-to-batch control containing a drift term:
Y k+1 ＝bu k +a k +D k ；
the input of the next batch is calculated by a double-exponential weighted moving average control algorithm:
continuously iterating in a double-exponential weighted moving average control algorithm to obtain:
a k ＝λ 1 (Y k -bu k-1 )+(1-λ 1 )a k-1 ；
D k ＝λ 2 (y k -bu k -a k-1 )+(1-λ 2 )D k-1 。
preferably, in step S4, the production process of hybrid manufacturing is represented as:
wherein n is the number of batches within a period, in each batch with 0 < n < j, the input of product 1 is used as the process input, in each batch with j < n < i, the input of product 2 is used as the process input, corresponding to different process outputs y ik+n ；β 1 、β 2 The process gains, α, of the process models of product 1 and product 2, respectively 1k 、α 2k Corresponding disturbance of the product 1 and the product 2 respectively accords with an IMA (1,1) disturbance model which has no memory;
α k ＝α k-1 +ε k -θε k-1 ；
wherein epsilon k Is normally distributed white noise, and theta is a model parameter;
using a 1 、a 2 For parameter alpha 1k 、α 2k To estimate, b 1 、b 2 For parameter beta 1 And beta 2 By estimation, a batch-to-batch control prediction model for mixed manufacturing of two products can be obtained:
preferably, a control method based on a machine is adopted for the mixed production;
assume that the control target of product 1 is T 1 Control target for product 2 is T 2 Then, the next batch of inputs based on the EWMA algorithm of the machine is:
and (3) obtaining a parameter estimation value through iterative calculation:
preferably, in step S4, a product-based control method is adopted for the mixed production;
control target for product 1 is T 1 The next batch of inputs based on the EWMA control algorithm of the product are:
and (3) iteratively calculating in an EWMA algorithm to obtain a parameter estimation value:
the production process and parameter calculation of product 2 is the same as that of product 1.
Compared with the prior art, the invention has the beneficial effects that:
1. the invention designs a control algorithm for the batch process of the production of the aerospace initiating explosive device to carry out batch control, can correspondingly change the parameters or the structure of the controller when the process characteristics of different scenes change, provides an optimal set value curve, and ensures the control quality of the batch control under various conditions.
2. The invention carries out dynamic Bayesian network modeling on the manufacturing process of the pressure-loading explosive of the aerospace initiating explosive device, extracts the deviation factors which have larger influence on the quality of the finished product, and optimizes the control quality.
3. The invention realizes a model with simple structure and strong robustness by utilizing an EWMA exponential weighting moving average algorithm, improves the control quality compared with the prior method, and improves the accumulative error of the production process and the difficulty in positioning and identifying the product deviation source.
Drawings
FIG. 1 is a block diagram of an EWMA controller;
FIG. 2 is a flow diagram of dynamic Bayesian network modeling;
fig. 3 is an EWMA controller work flow diagram.
Detailed Description
The invention will be further illustrated with reference to the following specific examples. It should be understood that these examples are for illustrative purposes only and are not intended to limit the scope of the present invention. Further, it should be understood that various changes or modifications of the present invention may be made by those skilled in the art after reading the teaching of the present invention, and these equivalents also fall within the scope of the present application.
Both deviation source diagnosis and batch-to-batch control are the prior art, but due to the particularity of aerospace initiating explosive device production and manufacturing, the existing deviation source diagnosis and batch-to-batch control cannot be directly used.
In the aspect of deviation source diagnosis, the current deviation source diagnosis methods commonly used at home and abroad mainly include a method based on an analytic mathematical model, a method based on pattern matching and a method based on knowledge expression. The production process of the aerospace initiating explosive device has the characteristics of multiple varieties and small batch, the historical data volume of each model of product is limited, the deviation source diagnosis depends on data support, and when a detected data sample is small or is missing, the coupling relation between variables is difficult to accurately describe, so that more researches are needed for the deviation source diagnosis of the aerospace initiating explosive device under a small data set. Meanwhile, the production process of the aerospace initiating explosive device contains more uncertainty factors, and the uncertainty in the process is measured, which is also a problem which is not solved by the existing method.
In the aspect of batch control, an iterative learning control algorithm is applied to batch control, historical data and a process model are fused, and an iterative learning control function is added to an outer ring on the basis of a closed ring of the process model. On the basis of an iterative learning algorithm, a two-dimensional system theory applied to batch control is developed. The two-dimensional control theory decomposes a system into two dimensions of time and batch, and describes dynamic characteristics in two independent directions by using a state space model. In the two-dimensional model system, the time-varying characteristics of the production process are reflected on the time dimension, and the repeatability of disturbance is reflected on the batch dimension, so that other control requirements can be met on the premise of keeping the system stable. The research on the batch-to-batch control is mainly divided into two aspects of a state estimation algorithm and a state estimator. Deep reinforcement learning is applied to training of batch controllers, a new strategy network is embedded in a learning model, and the network is divided into a linear part and a nonlinear part so as to improve the prediction performance of the batch controllers on process change and provide a special reward function to balance target tracking and production parameter fluctuation.
At present, the application research of batch control in the production and manufacturing of aerospace initiating explosive devices is less, how to design a control algorithm aiming at the batch process of the production of the aerospace initiating explosive devices accurately adjusts the parameters of a controller when the process characteristics change, gives an optimal set value curve, ensures the control quality of the controller under various conditions, and still is a problem to be solved.
The embodiment is as follows: the invention relates to a batch quality control method for flexible manufacturing of an aerospace initiating explosive device loading and pressing agent, which carries out dynamic Bayesian network modeling on an aerospace initiating explosive device loading and pressing agent manufacturing process, extracts key factors having large influence on the quality of a finished product, carries out modeling and quality control on the aerospace initiating explosive device loading and pressing agent manufacturing process by utilizing an EWMA (equal weighted moving average) exponential weighting algorithm, realizes a model with simple structure and strong robustness, and comprises the following steps:
s1, quantitatively describing the production process of the pressed powder, and specifically comprising the following steps:
and S11, introducing a Barersian equation to quantitatively describe the process of loading the pressed medicine. The Barershen equation is based on an ideal elastic stress-strain relation, and factors such as powder flowability, pressing time, friction force in the pressing process, work hardening and the like are not considered.
For the contact point of the pressing agent, the relationship between the pressure and the strain is as follows:
wherein, sigma is the pressing pressure, epsilon is the deformation of the grain, and A is the contact area between particles in the pressing process.
And calculating and transforming the equation to obtain:
lgP m ＝lgP+L(β-1)；
wherein P is the pressure of the pressing process, P m The unit is MPa for the pressure corresponding to pressing to a compact state, and L is a pressing factor corresponding to the ratio between the elastic modulus of the powder and the stress of the pressing. β ═ ε +1, relative volume of green compacts.
S12, introducing a northwest Chuanfu pressing equation, and introducing a concept of volume reduction rate by combining an actual powder pressing curve on the basis of the above equation. The northwest Hough pressure equation assumes that the powder particles can only withstand the inherent yield limit of the material, that the unit pressures at each point in the powder layer are equal, that the total sum of the external pressure on each section and the actual pressure on the section is balanced, and that the displacement probability of the powder particles is proportional to the size of the pores around the powder particles when the powder is compressed.
The volume reduction during powder compaction is described as follows:
wherein, V 0 Is the initial volume of the powder under no pressure, V is the volume of the powder under pressure P, and the units are cm 3 C is the volume reduction rate of the powder, and a and b are constant terms.
Due to the fact that the factor of the volume reduction rate is considered by the northeast China's equation, the accuracy is better under the conditions of small pressure and medium pressure, but the hardening phenomenon in the powder pressing process is not considered, and therefore the northeast China's equation is still a certain distance away from the actual powder pressing process.
For an ideal elastomer, the relationship between stress and strain is described using hooke's law:
where σ is stress, ε is strain, and M is elastic modulus.
Considering the viscosity of the powder, there is a stress relaxation during pressing, and the relationship between stress and strain is described using maxwell's equations:
wherein, tau 1 Representing the stress relaxation time.
When the elastic body has strain relaxation, the relationship between the stress and the strain is described by using a Karl Wen equation:
wherein, tau 2 Expressing strain relaxation time, wherein eta is viscosity coefficient and satisfies the condition that eta is M tau 2 。
When both stress relaxation and strain relaxation exist and the stress and strain are in a Linear relationship, the elastic hysteresis body at this time is a Standard Linear Solid (SLS) and shows the following relationship:
standard linear solids are more accurate than maxwell solids and calvin solids. However, in the actual pressing process, the properties of the powder body cannot be fully described using a standard linear solid, since after sufficient relaxation, the powder body is work hardened, at which time the stress and strain no longer exhibit a linear relationship.
The powder body at this time is described by the following nonlinear hysteresis body:
wherein n < 1 is a coefficient indicating work hardening, and the smaller n, the more pronounced the hardening tendency.
Stress retention sigma during pressing 0 When the time is not changed, the user can select the time,
after sufficient relaxation, tau 2 T, at this time
And considering the dimension, the above formula can be simplified as follows:
for powder bodies, during compaction, where the reduction in pore volume leads to a reduction in overall volume, the density increases with constant mass, with the strain defined as follows:
wherein, V 0 Is the initial volume of the powder, V m For a volume pressed to a dense state, V is the current volume, corresponding to p 0 Is the initial density of the powder, p m The density in the powder dense state is denoted as ρ.
Transforming to obtain:
as can be seen from the above formula, the processAfter transformation lgP and
the linear relation is formed, the equation considers the influence of press hardening, and the method is suitable for the hardness of the powder body and the pressing pressure.
S2, performing dynamic Bayesian network modeling on the flexible manufacturing process of the pressure loading medicine of the aerospace initiating explosive device, and extracting factors which have a large influence on the product quality in the production process as key variables in the quality control of subsequent batches.
And S21, analyzing deviation factors of the manufacturing process. The process of loading the pressure powder into the aerospace initiating explosive device is a step-by-step and repeated process. The part deviation, the positioning deviation, the measurement deviation and the operation deviation in each round of working procedure can influence the parameters of the next round of working procedure, the transmissibility is represented as probability transfer among time slices in a dynamic Bayesian network, and finally the detonation performance and the safety of finished products are related. Based on the actual production flow of loading and pressing the explosive in the aerospace initiating explosive devices, a mapping relation from a deviation source node in a process to an observation node for quality control is constructed based on a deviation source analysis and uncertainty reasoning method, and a dynamic Bayesian network is established on the basis.
And S22, modeling the dynamic Bayesian network. The dynamic Bayesian network modeling of the pressure medicine loading process of the aerospace initiating explosive device is completed through the processes of deviation factor extraction, deviation source node and observation node definition, data processing, network structure determination, network parameter determination and the like. And after modeling is finished, calculating posterior probability, if the posterior probability conforms to the historical data, conforming the model, and if the posterior probability does not conform to the historical data, adjusting the network structure and the parameters to re-model until the posterior probability conforms to the requirements.
The bayesian formula is defined as:
the Bayesian network represents the joint probability distribution in the form of a directed acyclic graph, wherein the variables are represented by nodes, the dependency relationship among the variables is represented by directed edges, and the dependency strength of the lower-layer nodes on the upper-layer nodes is measured by using a conditional probability table. For the root node, its conditional probability table represents the distribution of prior probabilities.
The dynamic Bayesian network expands the network according to time slices on the basis of the Bayesian network, and realizes dynamic reasoning and prediction according to the causal relationship between the front time slice and the rear time slice. Assuming that N nodes exist in a dynamic Bayesian network, the corresponding characteristic variable is X 1 ,X 2 ,...,X n Node X i Is used for the upper node set of (n) (X) i ) X representing time slice t i By X i [t]That, the following relationship exists:
according to the formula, the network state of the current time slice is influenced by all the preorder time slices, the complexity of the network is exponentially improved along with the increase of the time slices, and reasoning and prediction by using the network become very difficult. Therefore, finally, the dynamic Bayesian network model is simplified according to the constraint conditions of Markov property of the random process and time invariance of state transition probability. The dynamic bayesian network modeling flow chart is shown in fig. 2.
And S3, defining input and output of the pressurized medicine production process based on an EWMA algorithm. And quantitatively describing key parameters in the process of manufacturing the pressed powder, and defining input and output of the system.
S31, the EWMA algorithm adjusts the system input through minimum variance control. The EWMA controller has two modes of updating model gain and model intercept, analyzes historical data to obtain an offline estimation value of process gain or disturbance based on an offline model, iterates among product batch data through an EWMA filter, and estimates a future value by using moving average data, so that the purposes of adjusting system input and eliminating deviation are achieved. The EWMA algorithm has a weight parameter, also called discount factor, represented by lambda, and the value of the weight parameter is between 0 and 1. The larger the numerical value of the discount factor is, the higher the weight of the latest measured value is, and the stronger the real-time property of the output is; the smaller the value of the discount factor, theThe higher the weight of the historical value, the stronger the stationarity of the output. Assume k is batch and a set of observations is { x } k The output prediction can be expressed as:
y k+1 ＝λx k +(1-λ)y k ；
the above formula recursion can be obtained:
y k+1 ＝λx k +λ(1-λ)x k-1 +...+λ(1-λ) k-1 x 1 ；
the EWMA algorithm has different weights for all historical data in the sequence, and forms a memory for the data sequence as the batch moves forward and decreases exponentially. A block diagram of an EWMA controller acting on a control system is shown in fig. 1.
And S32, input and output definition. In the manufacture of powder compacts in batches, the trend of parameter variation is expressed as a process with constant gain, for the relationship between the compression pressure and the density measurement in the compacting process, the process input u ═ lgP and the process output are defined
The production process is described using the following relationship:
y k ＝βu k-1 +α k ；
wherein u is k-1 For the input at the beginning of the batch production process, y k Alpha being the output of the batch production process k And beta is the linear model parameter of the real physical process, beta is the process gain, alpha k Is a disturbance in the process.
An ARIMA model, i.e., an autoregressive differential moving average model, was introduced to make estimates of current values based on time series historical values and prediction errors. The method is composed of an autoregressive term, a difference term and a moving average term. The form of disturbance common in the discrete batch aerospace initiating explosive device loading manufacturing process is represented as IMA (1,1) disturbance:
α k ＝α k-1 +ε k -θε k-1 ；
wherein epsilon k Is white noise, theta is a model parameter, and when theta is 1, alpha is k ＝ε k The disturbance term is degenerated into a normally distributed white noise sequence; when the value of theta is smaller, the correlation of the IMA (1,1) sequence is stronger; when the value of theta is equal to 0,
the disturbance term becomes a superposition of all the historical white noise.
Using the estimated value a 0 And b parameter α k And beta, initializing to obtain a prediction model for controlling between batches:
Y k+1 ＝bu k +a k ；
and then continuously iterating in the EWMA algorithm to obtain:
a k ＝λ(Y k -bu k-1 )+(1-λ)a k-1 ,0≤λ≤1；
wherein λ is a variable discount factor of the EWMA algorithm, and is related to a model parameter θ, a model mismatch β/b and a time delay term.
After completing the iteration of each batch, the input for the next batch is calculated by the EWMA controller:
wherein, T k Is the target value for the kth batch.
The workflow of the EWMA controller is obtained as shown in fig. 3.
And S4, performing batch quality control based on the EWMA algorithm. The manufacturing process of the pressure-loading agent for the aerospace initiating explosive devices is a flexible manufacturing process, an EWMA algorithm is introduced, different disturbance conditions are analyzed according to various common scenes and production condition changes in the production process, and corresponding changes are made to a controller to ensure the control performance of the controller.
S41, the equipment parameters are reset and then appear as jump of the control target on the flexible manufacturing model, and in the next batch:
wherein, T 0 For controlling the jump of the target, the above equation describes an initial deviation with unknown magnitude and direction, and the system structure is not changed, which is equivalent to:
y k ＝βu k-1 +α k +α 0 ；
as the number of production batches increases, the controller can track the deviation and gradually restore the system output to the control target.
And S42, control when the measurement data lags. When the observation data lags, the model intercept term is updated as follows:
a t+1 ＝λ(Y t-d -bu t-d )+(1-λ)a t ,0≤λ≤1；
the hysteresis value is estimated by analyzing historical data and combining maximum likelihood estimation, and the performance of the controller is ensured.
S43, aging in a specific direction. A linear model of the pressure charge production process containing specific direction drift:
y k ＝βu k-1 +α k +δk；
u k-1 for the input at the beginning of the batch production process, y k Alpha, an output of the batch production process k And beta is the linear model parameter of the real physical process, beta is the process gain, alpha k Delta is the drift term for the disturbance in the process.
The conventional EWMA algorithm does not track the drift term, and thus, the drift error cannot be completely eliminated in the control process. A double exponential weighted moving average control algorithm D-EWMA is introduced. The D-EWMA algorithm has two EWMA filters in the feedback link, and can play a role in prediction and correction.
Using the estimated value a 0 And b parameter α k And beta, initializing to obtain a prediction model of batch-to-batch control containing a drift term:
Y k+1 ＝bu k +a k +D k ；
the input for the next batch is calculated by the D-EWMA controller:
continuously iterating in the D-EWMA algorithm to obtain:
a k ＝λ 1 (Y k -bu k-1 )+(1-λ 1 )a k-1 ；
D k ＝λ 2 (y k -bu k -a k-1 )+(1-λ 2 )D k-1 ；
and S44, mixing the products. The EWMA control method based on the mixed production of two products can be also suitable for the mixed production of a plurality of products. The production process of hybrid manufacturing can be expressed as:
wherein n is the number of batches within a period, in each batch with 0 < n < j, the input of the product 1 is used as the process input, in each batch with j < n < i, the input of the product 2 is used as the process input, and the process outputs y are different correspondingly ik+n 。β 1 、β 2 The process gains, α, of the process models of product 1 and product 2, respectively 1k 、α 2k The corresponding disturbance of the product 1 and the product 2 respectively conforms to an IMA (1,1) disturbance model, and the model has no memory.
α k ＝α k-1 +ε k -θε k-1 ；
Wherein epsilon k Is normally distributed white noise, and theta is a model parameter.
Using a 1 、a 2 For parameter alpha 1k 、α 2k Carry out an estimation, b 1 、b 2 For parameter beta 1 And beta 2 By estimation, a batch-to-batch control prediction model for mixed manufacturing of two products can be obtained:
there are two control methods for the case of hybrid production: a machine-based control method and a product-based control method.
1. EWMA algorithm based on machine
Assume that the control target of product 1 is T 1 Control target for product 2 is T 2 Then, the next batch of inputs based on the EWMA algorithm of the machine is:
and (3) obtaining a parameter estimation value through iterative calculation:
2. control target for product 1 is T 1 Because the process parameters of each product on the production line are independent of each other, the input and output of the product 1 are independent of other product parameters. The production process and the parameter calculation mode of the product 2 are the same. The next batch of inputs based on the EWMA control algorithm of the product are:
and (3) iteratively calculating in an EWMA algorithm to obtain a parameter estimation value:
Claims (9)
1. the batch quality control method for the flexible manufacturing of the pressure loading chemicals of the aerospace initiating explosive devices is characterized by comprising the following steps of:
s1, quantitatively describing the flexible manufacturing process of the pressure powder for the aerospace initiating explosive devices;
s2, carrying out dynamic Bayesian modeling on the flexible manufacturing process of the pressure-loading explosive of the aerospace initiating explosive device, and extracting key factors influencing the production quality;
s3, defining the input and output processes of the pressurized medicine production process based on an EWMA algorithm;
and S4, performing batch quality control based on the EWMA algorithm.
2. The batch quality control method for the flexible manufacturing of the pressurized medicine of the aerospace initiating explosive device according to claim 1, wherein the step S1 includes the steps of:
s11, introducing a Barersian equation to quantitatively describe the process of loading the pressing powder, and obtaining the relation between the pressure and the strain of the contact point of the pressing powder as follows:
lgP m ＝lgP+L(β-1)；
wherein P is the pressure of the pressing process, P m P, P for pressing the powder to a pressure corresponding to the compact state m The units of (A) are all MPa;
l is a pressing factor corresponding to the ratio between the elastic modulus of the powder and the pressing stress;
beta is equal to epsilon +1 and is the relative volume of a pressed compact;
s12, introducing a Chuanbei Hough pressing equation, and obtaining the relation between the pressure and the density in the pressing process as follows:
wherein M is the modulus of elasticity, ρ 0 Is the initial density of the powder, p m The density in the powder dense state is denoted by ρ as the current density.
3. The batch quality control method for the flexible manufacturing of the pressurized medicine of the aerospace initiating explosive device according to claim 1, wherein the step S2 includes the steps of:
s21, starting from the actual production flow of the pressure loading of the aerospace initiating explosive devices, constructing a mapping relation from a deviation source node in the process to an observation node for quality control based on a deviation source analysis and uncertainty reasoning method, and establishing a dynamic Bayesian network on the basis;
s22, completing dynamic Bayesian network modeling of the loading process of the aerospace initiating explosive device through deviation factor extraction, deviation source node and observation node definition, data processing, network structure determination and network parameter determination processes, and then simplifying the dynamic Bayesian network model according to the Markov property of a random process and the constraint conditions of time invariance of state transition probability and the like.
4. The batch quality control method for the flexible manufacturing of the pressure powder for the aerospace initiating explosive device according to claim 1, wherein the step S3 includes the steps of:
s31, the EWMA algorithm adjusts the system inputs by minimum variance control, assuming k is batch and a set of observations { x } k The output prediction is expressed as:
y k+1 ＝λx k +λ(1-λ)x k-1 +...+λ(1-λ) k-1 x 1 ；
s32, defining the process input u ═ lgP, and the process output
The production process is described as follows: y is k ＝βu k-1 +α k ；
Wherein u is k-1 For the input at the beginning of the batch production process, y k Alpha being the output of the batch production process k And beta is the linear model parameter of the real physical process, beta is the process gain, alpha k Is a disturbance in the process;
s33, introducing an ARIMA model, wherein the common disturbance form in the process of manufacturing the pressure-charged explosive of the aerospace initiating explosive device in discrete batches is represented as IMA (1,1) disturbance:
α k ＝α k-1 +ε k -θε k-1 ；
wherein epsilon k Is white noise, theta is a model parameter, and when theta is 1, alpha is k ＝ε k The disturbance term is degenerated into a normally distributed white noise sequence; when the temperature is higher than the set temperatureThe smaller the value of theta, the stronger the correlation of the IMA (1,1) sequence; when the value of theta is equal to 0,
the disturbance item becomes the superposition of all historical white noises;
s34, using the estimated value a 0 And b parameter α k And beta, initializing to obtain a prediction model for controlling between batches:
Y k+1 ＝bu k +a k ；
and then continuously iterating in the EWMA algorithm to obtain:
a k ＝λ(Y k -bu k-1 )+(1-λ)a k-1 ,0≤λ≤1；
wherein, lambda is a variable discount factor of the EWMA algorithm, and is related to a model parameter theta, a model mismatching degree beta/b and a time delay item;
s35, after completing the iteration of each batch, the EWMA controller calculates the input of the next batch:
wherein, T k Is the target value for the kth batch.
5. The batch quality control method for the flexible manufacturing of the pressurized medicine of the aerospace initiating explosive device according to claim 1, wherein the step S4 includes the steps of:
s41, the equipment parameters are reset and then are shown as the jump of the control target on the flexible manufacturing model, and in the next batch:
wherein T is 0 Hopping for a control target;
s42, control when measurement data is delayed, and when observation data is delayed, the model intercept term is updated as follows:
a t+1 ＝λ(Y t-d -bu t-d )+(1-λ)a t ,0≤λ≤1；
s43, aging in a specific direction, and carrying out a linear model of the pressure charge production process with specific direction drift:
y k ＝βu k-1 +α k +δk；
u k-1 for the input at the beginning of the batch production process, y k Alpha, an output of the batch production process k And beta is the linear model parameter of the true physical process, beta is the process gain, alpha k δ is the drift term for the disturbance in the process.
6. The method for controlling batch quality in flexible manufacturing of pressurized drugs for aerospace initiating explosive devices according to claim 5, wherein a double exponential weighted moving average control algorithm is introduced in step S43, and the estimated value a is used 0 And b parameter α k And beta, initializing to obtain a prediction model of batch-to-batch control containing a drift term:
Y k+1 ＝bu k +a k +D k ；
the input of the next batch is calculated by a double-exponential weighted moving average control algorithm:
continuously iterating in a double-exponential weighted moving average control algorithm to obtain:
a k ＝λ 1 (Y k -bu k-1 )+(1-λ 1 )a k-1 ；
D k ＝λ 2 (y k -bu k -a k-1 )+(1-λ 2 )D k-1 。
7. the batch quality control method for the flexible manufacturing of the pressurized medicine of the aerospace initiating explosive device according to claim 1 or 5, wherein in the step S4, the production process of the hybrid manufacturing is expressed as:
wherein n is the number of batches within a period, in each batch with 0 < n < j, the input of product 1 is used as the process input, in each batch with j < n < i, the input of product 2 is used as the process input, corresponding to different process outputs y ik+n ；β 1 、β 2 The process gains, α, of the process models of product 1 and product 2, respectively 1k 、α 2k Corresponding disturbance of the product 1 and the product 2 respectively accords with an IMA (1,1) disturbance model which has no memory;
α k ＝α k-1 +ε k -θε k-1 ；
wherein epsilon k Is normally distributed white noise, and theta is a model parameter;
using a 1 、a 2 For parameter alpha 1k 、α 2k Carry out an estimation, b 1 、b 2 For parameter beta 1 And beta 2 By estimation, a batch-to-batch control prediction model for mixed manufacturing of two products can be obtained:
8. the batch quality control method for the flexible manufacturing of the pressure powder for the aerospace initiating explosive devices according to claim 7, wherein in step S4, a machine-based control method is adopted for the mixed production;
assume that the control target of product 1 is T 1 Control target for product 2 is T 2 Then, the next batch of inputs based on the EWMA algorithm of the machine is:
and (3) obtaining a parameter estimation value through iterative calculation:
9. the batch quality control method for the flexible manufacturing of the pressurized medicine of the aerospace initiating explosive device according to claim 7, wherein in the step S4, a product-based control method is adopted for the mixed production situation;
control target for product 1 is T 1 The next batch of inputs based on the EWMA control algorithm of the product are:
and (3) iteratively calculating in an EWMA algorithm to obtain a parameter estimation value:
the production process and parameter calculation of product 2 is the same as that of product 1.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210892457.2A CN115016421B (en) | 2022-07-27 | Batch-to-batch quality control method for flexible manufacturing of loading and pressing of aerospace initiating explosive device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210892457.2A CN115016421B (en) | 2022-07-27 | Batch-to-batch quality control method for flexible manufacturing of loading and pressing of aerospace initiating explosive device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115016421A true CN115016421A (en) | 2022-09-06 |
| CN115016421B CN115016421B (en) | 2024-07-02 |
Family
ID=
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103176425A (en) * | 2013-03-13 | 2013-06-26 | 西北工业大学 | Design method of multivariate exponentially weighted moving average controller |
| CN107479509A (en) * | 2008-03-06 | 2017-12-15 | 应用材料公司 | Yield prediction feedback for control device engineering system |
| CN110375598A (en) * | 2019-06-11 | 2019-10-25 | 重庆大学 | One kind being directed to priming system product testing system analysis method for reliability |
| CN112884348A (en) * | 2021-03-12 | 2021-06-01 | 重庆大学 | Method for diagnosing production deviation source of aerospace initiator based on dynamic Bayesian network |
| CN113807014A (en) * | 2021-09-17 | 2021-12-17 | 沈阳工业大学 | Non-parametric self-adaptive dynamic EWMA control chart multi-objective optimization design method considering statistics and economy |
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107479509A (en) * | 2008-03-06 | 2017-12-15 | 应用材料公司 | Yield prediction feedback for control device engineering system |
| CN103176425A (en) * | 2013-03-13 | 2013-06-26 | 西北工业大学 | Design method of multivariate exponentially weighted moving average controller |
| CN110375598A (en) * | 2019-06-11 | 2019-10-25 | 重庆大学 | One kind being directed to priming system product testing system analysis method for reliability |
| CN112884348A (en) * | 2021-03-12 | 2021-06-01 | 重庆大学 | Method for diagnosing production deviation source of aerospace initiator based on dynamic Bayesian network |
| CN113807014A (en) * | 2021-09-17 | 2021-12-17 | 沈阳工业大学 | Non-parametric self-adaptive dynamic EWMA control chart multi-objective optimization design method considering statistics and economy |
Non-Patent Citations (5)
| Title |
|---|
| HUANG, DL: "Run-to-run control of batch production process in manufacturing systems based on online measurement", 《COMPUTERS & INDUSTRIAL ENGINEERING》, no. 141, 17 March 2020 (2020-03-17), pages 1 - 15 * |
| ZHOU, JX: "Using improved dynamic Bayesian networks in reliability evaluation for flexible test system of aerospace pyromechanical device products", 《RELIABILITY ENGINEERING & SYSTEM SAFETY》, vol. 210, 7 July 2021 (2021-07-07), pages 1 - 13 * |
| 侯明阳: "基于变折扣因子EWMA的混合产品批间控制方法", 《控制理论与应用》, 15 October 2009 (2009-10-15), pages 1137 - 1142 * |
| 尚敖男: "起爆器生产任务调度与监控系统研究", 《中国优秀硕士学位论文全文数据库》, 15 April 2022 (2022-04-15) * |
| 王辉阳: "航天火工品装压药质量控制与性能评估", 《万方数据知识服务平台》, 1 November 2023 (2023-11-01) * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Ahn et al. | Online tuning fuzzy PID controller using robust extended Kalman filter | |
| CN105760933A (en) | Method and apparatus for fixed-pointing layer-wise variable precision in convolutional neural network | |
| CN107688850A (en) | A kind of deep neural network compression method | |
| CN111539132B (en) | Dynamic load time domain identification method based on convolutional neural network | |
| CN109308063B (en) | Fermentation process stage division method based on score matrix | |
| CN111881601B (en) | Soil and rock dam material constitutive model construction method based on deep learning | |
| CN110609476B (en) | Multivariate nonlinear dynamic system model prediction control method based on Gaussian process model | |
| CN116504341B (en) | Sequential singular value filtering method for data-driven identification partial differential equation | |
| CN111144017A (en) | FF-RVM-based multi-period intermittent process soft measurement modeling method | |
| CN112001115A (en) | Soft measurement modeling method of semi-supervised dynamic soft measurement network | |
| CN110398942B (en) | Parameter identification method for industrial production process control | |
| CN113343559B (en) | Reliability analysis method for response surface of iterative reweighted least square method extreme learning machine | |
| CN114692507A (en) | Counting data soft measurement modeling method based on stacking Poisson self-encoder network | |
| CN115016421A (en) | Batch quality control method for flexible manufacturing of pressure powder in aerospace initiating explosive devices | |
| CN115016421B (en) | Batch-to-batch quality control method for flexible manufacturing of loading and pressing of aerospace initiating explosive device | |
| KR102457539B1 (en) | Structural Performance Based Inverse Material Design Method | |
| Havinga et al. | Estimating product-to-product variations in metal forming using force measurements | |
| CN113779724B (en) | Intelligent fault prediction method and system for filling packaging machine | |
| CN115062542A (en) | Polymerization reaction process quality prediction method based on two-dimensional steady LSTM | |
| CN114254416B (en) | Soil stress-strain relation determining method based on long-term and short-term memory deep learning | |
| CN111160464B (en) | Industrial high-order dynamic process soft measurement method based on multi-hidden-layer weighted dynamic model | |
| CN114720129A (en) | Rolling bearing residual life prediction method and system based on bidirectional GRU | |
| CN114529040A (en) | On-line prediction method for assembly error of electromechanical product | |
| CN110866643B (en) | Fermentation process quality variable prediction method based on maximum secondary mutual information criterion regression | |
| CN112818455A (en) | Bridge structure response monitoring 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 | ||
| GR01 | Patent grant |