CN112363462A - Static-dynamic cooperative sensing complex industrial process running state evaluation method - Google Patents
Static-dynamic cooperative sensing complex industrial process running state evaluation method Download PDFInfo
- Publication number
- CN112363462A CN112363462A CN202011050080.3A CN202011050080A CN112363462A CN 112363462 A CN112363462 A CN 112363462A CN 202011050080 A CN202011050080 A CN 202011050080A CN 112363462 A CN112363462 A CN 112363462A
- Authority
- CN
- China
- Prior art keywords
- state
- formula
- data
- static
- evaluation
- 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
- 238000011156 evaluation Methods 0.000 title claims abstract description 87
- 238000004519 manufacturing process Methods 0.000 title claims abstract description 38
- 238000000034 method Methods 0.000 claims abstract description 149
- 230000008569 process Effects 0.000 claims abstract description 122
- 239000013598 vector Substances 0.000 claims abstract description 53
- 230000003068 static effect Effects 0.000 claims abstract description 24
- 230000007704 transition Effects 0.000 claims abstract description 20
- 230000008859 change Effects 0.000 claims abstract description 19
- 238000013210 evaluation model Methods 0.000 claims abstract description 17
- 238000005516 engineering process Methods 0.000 claims abstract description 8
- 238000005065 mining Methods 0.000 claims abstract description 5
- 239000011159 matrix material Substances 0.000 claims description 36
- 230000008447 perception Effects 0.000 claims description 11
- 230000003416 augmentation Effects 0.000 claims description 6
- 230000003190 augmentative effect Effects 0.000 claims description 3
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 150000001875 compounds Chemical class 0.000 claims description 3
- 230000006870 function Effects 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 claims description 3
- 238000011112 process operation Methods 0.000 claims description 3
- 238000003745 diagnosis Methods 0.000 abstract description 2
- 239000003245 coal Substances 0.000 description 31
- 238000002360 preparation method Methods 0.000 description 12
- 239000007788 liquid Substances 0.000 description 6
- 238000004458 analytical method Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 238000000926 separation method Methods 0.000 description 4
- 230000002159 abnormal effect Effects 0.000 description 3
- 230000008901 benefit Effects 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000013459 approach 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
- 238000009776 industrial production Methods 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 238000004540 process dynamic Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000012887 quadratic function Methods 0.000 description 2
- 239000000126 substance Substances 0.000 description 2
- 239000000725 suspension Substances 0.000 description 2
- 230000005856 abnormality Effects 0.000 description 1
- 238000009825 accumulation Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 229910052500 inorganic mineral Inorganic materials 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000005272 metallurgy Methods 0.000 description 1
- 239000011707 mineral Substances 0.000 description 1
- 239000002994 raw material Substances 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 230000001052 transient effect 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
- G05B2219/32368—Quality control
-
- 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)
- General Factory Administration (AREA)
Abstract
A static-dynamic cooperative sensing complex industrial process running state evaluation method comprises the steps of firstly, carrying out static-dynamic characteristic cooperative sensing information mining by using a KPI-drive SFA algorithm, and establishing an off-line evaluation model of a running state; introducing a sliding window technology, calculating the similarity of variation information between the score vector and the first-order difference of the online data and the score vectors and the first-order differences of each state grade, and calculating static and dynamic evaluation indexes according to the similarity; formulating an effective evaluation rule, and realizing online identification of the process running state according to the size of the static evaluation index; realizing on-line identification of the process running state change trend according to the size of the dynamic evaluation index, and completing comprehensive evaluation of each state and the transition process; aiming at non-optimal variables, the influence of independent variables is reduced through a data-driven fault diagnosis method based on sparse learning, and the accurate positions of the non-optimal variables are traced according to the group contribution GWC. The method can effectively ensure the quality of industrial products.
Description
Technical Field
The invention belongs to the technical field of evaluation of operation states in industrial production processes, and particularly relates to a static-dynamic cooperative sensing evaluation method for operation states in a complex industrial process.
Background
Conventional process monitoring only focuses on the occurrence of abnormal conditions, and due to process disturbances and uncertainties, even in a normal operating state, the process may deviate from an optimal operating point, and a non-optimal or even poor operating state occurs. Particularly, in typical process industries such as mineral processing, metallurgy and the like in China, the raw materials change frequently, the operation environment is complex, the working condition fluctuation is severe, the equipment operation state is not good, the product quality and the process parameters cannot be comprehensively detected in real time, the non-optimal state of the production process is frequent, and the operation control effect is difficult to meet the actual production requirement.
Although some methods for evaluating the production process exist in the prior art, the process is easy to be abnormal due to dynamic change caused by feedback control in the actual production process, and the problems of serious false alarm and missing report phenomena exist, so that the static-dynamic characteristics of the process cannot be cooperatively sensed, an accurate and reliable running state evaluation model cannot be established, the generalization capability of the evaluation model is poor, and the precision is low. For this reason, many approaches to process dynamics emerge. For example: dynamic PLS, multi-scale PCA, state-space equations, etc., but none of these methods clearly distinguish dynamic information from process steady state, resulting in insensitivity to dynamic anomalies. In addition, the increasingly intense competition has led enterprises to place greater demands on product quality and production efficiency, greatly increasing the degree of automation and complexity of industrial processes.
Disclosure of Invention
Aiming at the problems in the prior art, the invention provides a static-dynamic cooperative sensing complex industrial process running state evaluation method, which effectively solves the problems that the conventional evaluation method has incomplete sensing on working condition information and has the phenomena of missing report and false report, and can effectively ensure the quality of industrial products.
In order to achieve the above object, the present invention provides a static-dynamic cooperative sensing complex industrial process operation state evaluation method, which specifically includes the following steps:
the method comprises the following steps: performing static-dynamic characteristic collaborative perception information mining by using a KPI-drive SFA algorithm, and establishing an off-line evaluation model of an operation state;
s11: collecting the production processProduction data obtaining input data matrix X is belonged to RN×mAnd the output data matrix Y ∈ RNWherein N is the number of samples, m is the number of process variables, R is the set of real numbers, RN×mA real number matrix representing dimensions N × m;
s12: the columns of the input data matrix X are processed with zero mean and unit variance, which are marked as XaAnd the difference between two input data points in each column time series is recorded as DeltaXa(ii) a The output data matrix Y is also normalized and denoted as Ya；
S13: for input matrix XaOutput matrix YaSum matrix Δ XaOperating a KPI-drive SFA algorithm, and specifically performing the following process:
a1: let YaAny one column in the sequence is equal to the initial ua；ΔXaIs equal to the initial ra；
A2: calculating a load vector w according to equation (1)a；
In the formula, alpha is a regularization parameter and is used for balancing two targets;
a3: load vector w is expressed according to equation (2)aStandardizing;
a4: x is determined according to the formula (3)aIs projected to waObtaining XaScore t ofa(ii) a Dividing Δ X according to equation (4)aIs projected to waTo obtain Δ XaScore r ofa；
ta＝Xawa (3)；
ra＝ΔXawa (4)；
A5: solving for vector c according to equation (5)a；
A6: c according to formula (6)aCarrying out standardization treatment;
a7: recalculating u according to equation (7)a；
A8: according to u in A7aU from step A1aWhether the precision is the same or whether the precision is converged or not is judged, if yes, A8 is executed, otherwise, A2 is executed;
a9: calculating the load vector p of the matrix X according to equation (8)a；
A10: calculating the component u according to equation (9)aWith respect to taRegression coefficient b ofa；
A11: the residual matrix X is obtained according to the formula (10)a+1(ii) a The residual matrix Y is obtained according to the formula (11)a+1；
A12: mixing Xa、YaIs replaced by Xa+1、Ya+1Iterating from steps a1 to a11 until the desired number of features are extracted;
a13: calculating an objective function S according to formula (12);
S＝RX (12)；
wherein R ═ W (P)TW)-1；
S14: calculating a comprehensive economic related index R (i) according to a formula (13);
s15: according to the index R (i), eliminating the characteristics irrelevant to the quality in R to obtain Rq；
S16: historical data (x, y) is assumed, and data (x) therein is knownc,yc) The corresponding status level c;
s17: calculating score vectors t for respective states according to equation (14)c；
tc＝Rqxc (14)；
Wherein c is the number of operating state levels;
s18: scoring the vector t for each statecTime sequence is augmented, and the first order difference of each state score vector is calculated to obtain
Step two: introducing a sliding window technology, calculating the similarity of the score vector and the first-order difference of the online data and the variation information between the score vectors and the first-order differences of the state grades according to an offline evaluation model, and calculating static and dynamic evaluation indexes according to the similarity; making an effective evaluation rule, and then completing comprehensive evaluation of each state and the transition process;
s20: constructing a sliding data window X for time K according to equation (15)k；
Xk＝[xk-H+1,Lxk]T (15)；
In the formula, H is the width of a data window;
s21: constructing a time sequence augmentation matrix, then carrying out standardization preprocessing, and recording standardized data as
S22: introducing online sample data into offline evaluation model
And calculating the online sample data according to the formula (16)
Middle h sample
Score vector of
Wherein H is k-H +1, L, k; c is 1,2, L, C;
s23: time sequence augmentation is carried out on the score vector of the online sample data, and the first-order difference of the score vector of the online data is calculated to obtain
S24: calculating the distance between the score vector and the center of each state grade of the quality-related set according to formula (17)
In the formula (I), the compound is shown in the specification,
a score vector for each steady state;
s25: calculating the state grade first-order difference center distance of the first-order difference and quality related set of the score vector according to the formula (18)
Wherein E is an offset;
s26: according to the formula (19) using
Defining static evaluation indexes of online data relative to various state grades
S27: according to the formula (20) using
Defining dynamic evaluation indexes of online data relative to various state grades
S28: the process running state is evaluated on line according to the evaluation index,
firstly, when
When the process variation information is consistent with the variation information in the state grade, the process operation state can be judged to be
(ii) when the condition (i) is not satisfied, but the condition
If so, indicating that the process running state is in the state grade conversion process, namely the current process is gradually changed;
③ when
In time, the change speed of the process variation information related to the quality in the online data is not obviously changed, and the process state can be judged to be unchanged;
fourthly when
When the process state is in a transition state, the process state is judged to be deteriorated;
fifthly, when
When the process state is in a transition state, the process state is determined to be optimized;
s29: integrating the results of S28 to give the operation state evaluation result of static-dynamic characteristic cooperative sensing;
(ii) when the conditions (i) and (iii) in S28 are satisfied, the state rank is derived from the static information
And the dynamic information indicates that the process state does not change, and the running state of the process can be judged to be
When the condition (c) in the S28 is satisfied but the condition (c) is not satisfied, the state level obtained from the static information is not changed and
the dynamic information indicates that the process state is not changed, and the running state of the process can be judged to be
When the conditions of the second time and the fourth time in the S28 are met, the state is obtained from the static information to change, the dynamic information shows that the process state is developed towards a worse direction, and the running state of the process can be judged to be developed from the previous time state to the worse process state;
when the conditions II and V in the S28 are met, the state is obtained from the static information to change, and the dynamic information shows that the process state is developed towards a better direction, so that the running state of the process can be judged to be developed from the previous state to the better process state;
if none of the above conditions is satisfied, maintaining the last evaluation result.
When the evaluation result is in a non-optimal or transitional state, tracing the non-optimal factors, and determining the accurate position of the non-optimal variables, wherein the specific steps are as follows:
b1: setting x to contain m variables, and dividing the process variables into B non-repeated groups according to the formula (21) according to time;
in the formula, xkIs m in the dimensionkK is 1, L, B, and
b2: selecting lambda according to formula (23);
in the formula, gammaMIs the corresponding control limit;
b3: rewriting formula (17) into a quadratic form according to formula (24);
b4: initialization f0And calculating M maximum eigenvalue gammak；
B5: calculating a gradient v (f) according to equation (25);
v(f)＝2M(f-x) (25)；
b6: updating f according to equation (26)k；
B7: calculate fkResidual errors are carried out until the precision requirement is met, and B8 is executed; otherwise, B5 is executed;
b8: f and A are calculated according to the formula (27);
A＝{k:||fk||＞0} (27)；
b9: sparse group contribution { fkAnd (4) forcing the group contribution of the independent variables to be zero, and calculating a group contribution rate GWC according to a formula (28)kDetermining the accurate position of the non-optimal variable by observing the GWC contributed by each group;
GWCk＝||fk|| (28)。
according to the invention, through a KPI-drive SFA algorithm, the relevant information of the comprehensive economic indicators of KPI (Key Performance indicators) is merged into the dynamic information extracted by SFA (Slow motion analysis), static-dynamic characteristic cooperative perception information in complex industrial process data is deeply mined, an off-line evaluation model of the operation state is established, then, a sliding window technology is introduced, the similarity of the score vector and the first-order difference of the on-line data and the variation information between the score vector and the first-order difference of each state grade are calculated, the evaluation of the process steady state and transition is completed by utilizing the similarity of the characteristic information and the similarity of the first-order difference information, and a unified framework of the operation state evaluation based on the static-dynamic characteristic cooperative perception is provided on the basis. In addition, when the running state is not optimal, accurate tracing and positioning of the time period when the non-optimal state changes and the non-optimal factor variable can be realized through a non-optimal factor tracing method based on sparse learning, and a theoretical basis can be provided for field operators to adjust the production strategy in real time. The off-line evaluation model of the running state established by the running state evaluation method based on static-dynamic characteristic cooperative sensing has practical significance for the coal dressing process of heavy media. By means of the method, the production process condition can be monitored in real time, and the condition that the production working condition deviates from the optimal state can be found in time. Therefore, a good foundation is laid for optimizing and adjusting the technological process in the coal dressing process, improving the raw coal dressing quality and finally improving the economic benefit of enterprises. The method can solve the problems of serious phenomena of missing report and false report caused by incomplete process perception in the complex industrial process.
Drawings
FIG. 1 is a flow chart of a dense medium coal separation process;
FIG. 2 is a flowchart of the present invention for tracking the operation status evaluation and non-optimal factors based on static-dynamic characteristics cooperative sensing;
FIG. 3 is a schematic diagram showing the evaluation results of the operational state of a PLS-based complex industrial process;
FIG. 4 is a schematic diagram of the evaluation result of the operation state of the complex industrial process based on static-dynamic characteristic cooperative sensing;
FIG. 5 is a schematic diagram of group contribution rate results based on sparse learning non-optimal factor tracing;
fig. 6 is a schematic diagram of non-optimal factor tracing results of non-optimal state levels and adjacent level transition stages.
Detailed Description
As shown in fig. 1 and fig. 2, the present invention provides a static-dynamic cooperative sensing method for evaluating an operation state of a complex industrial process, which specifically includes the following steps:
the method comprises the following steps: performing static-dynamic characteristic collaborative perception information mining by using a KPI-drive SFA algorithm, and establishing an off-line evaluation model of an operation state;
s11: acquiring production data of the production process to obtain an input data matrix X belonging to RN×mAnd the output data matrix Y ∈ RNWherein N is the number of samples, m is the number of process variables, R is the set of real numbers, RN×mA real number matrix representing dimensions N × m;
s12: the columns of the input data matrix X are processed with zero mean and unit variance, which are marked as XaAnd the difference between two input data points in each column time series is recorded as DeltaXa(ii) a The output data matrix Y is also normalized and denoted as Ya；
S13: for input matrix XaOutput matrix YaSum matrix Δ XaOperating a KPI-drive SFA algorithm, and specifically performing the following process:
a1: let YaAny one column in the sequence is equal to the initial ua；ΔXaIs equal to the initial ra；
A2: calculating a load vector w according to equation (1)a；
In the formula, alpha is a regularization parameter and is used for balancing two targets;
a3: load vector w is expressed according to equation (2)aStandardizing;
a4: x is determined according to the formula (3)aIs projected to waObtaining XaScore t ofa(ii) a Dividing Δ X according to equation (4)aIs projected to waTo obtain Δ XaScore r ofa；
ta＝Xawa (3)；
ra＝ΔXawa (4)；
A5: solving for vector c according to equation (5)a；
A6: c according to formula (6)aCarrying out standardization treatment;
a7: recalculating u according to equation (7)a；
A8: according to u in A7aU from step A1aWhether the precision is the same or whether the precision is converged or not is judged, if yes, A8 is executed, otherwise, A2 is executed;
a9: calculating the load vector p of the matrix X according to equation (8)a；
A10: calculating the component u according to equation (9)aWith respect to taRegression coefficient b ofa；
A11: the residual matrix X is obtained according to the formula (10)a+1(ii) a The residual matrix Y is obtained according to the formula (11)a+1；
A12: mixing Xa、YaIs replaced by Xa+1、Ya+1Iterating from steps a1 to a11 until the desired number of features are extracted;
a13: calculating an objective function S according to formula (12);
S＝RX (12)；
wherein R ═ W (P)TW)-1；
S14: calculating a comprehensive economic related index R (i) according to a formula (13);
s15: according to the index R (i), eliminating the characteristics irrelevant to the quality in R to obtain Rq；
S16: historical data (x, y) is assumed, and data (x) therein is knownc,yc) The corresponding status level c; the historical data (x, y) is from known historical data;
s17: calculating score vectors t for respective states according to equation (14)c；
tc＝Rqxc (14)；
Wherein c is the number of operating state levels;
s18: scoring the vector t for each statecTime sequence is augmented, and the first order difference of each state score vector is calculated to obtain
Step two: introducing a sliding window technology, calculating the similarity of the score vector and the first-order difference of the online data and the variation information between the score vectors and the first-order differences of the state grades according to an offline evaluation model, and calculating static and dynamic evaluation indexes according to the similarity;
s20: constructing a sliding data window X for time K according to equation (15)k；
Xk＝[xk-H+1,Lxk]T (15)；
In the formula, H is the width of a data window;
s21: constructing a time sequence augmentation matrix, then carrying out standardization preprocessing, and recording standardized data as
S22: introducing online sample data into offline evaluation model
And calculating the online sample data according to the formula (16)
Middle h sample
Score vector of
Wherein H is k-H +1, L, k; c is 1,2, L, C;
s23: time sequence augmentation is carried out on the score vector of the online sample data, and the first-order difference of the score vector of the online data is calculated to obtain
S24: calculating the distance between the score vector and the center of each state grade of the quality-related set according to formula (17)
In the formula (I), the compound is shown in the specification,
a score vector for each steady state;
s25: calculating the state grade first-order difference center distance of the first-order difference and quality related set of the score vector according to the formula (18)
Wherein E is an offset;
s26: according to the formula (19) using
Defining static evaluation indexes of online data relative to various state grades
S27: according to the formula (20) using
Defining dynamic evaluation indexes of online data relative to various state grades
S28: the process running state is evaluated on line according to the evaluation index,
firstly, when
When the process variation information is consistent with the variation information in the state grade, the process operation state can be judged to be
(ii) when the condition (i) is not satisfied, but the condition
If so, indicating that the process running state is in the state grade conversion process, namely the current process is gradually changed;
③ when
In time, the change speed of the process variation information related to the quality in the online data is not obviously changed, and the process state can be judged to be unchanged;
fourthly when
When the process state is in a transition state, the process state is judged to be deteriorated;
fifthly, when
When the process state is in a transition state, the process state is determined to be optimized;
s29: integrating the results of S28 to give the operation state evaluation result of static-dynamic characteristic cooperative sensing;
(ii) when the conditions (i) and (iii) in S28 are satisfied, the state rank is derived from the static information
And the dynamic information indicates that the process state does not change, and the running state of the process can be judged to be
When the condition (c) in the S28 is satisfied but the condition (c) is not satisfied, the state level obtained from the static information is not changed and
the dynamic information indicates that the process state is not changed, and the running state of the process can be judged to be
When the conditions of the second time and the fourth time in the S28 are met, the state is obtained from the static information to change, the dynamic information shows that the process state is developed towards a worse direction, and the running state of the process can be judged to be developed from the previous time state to the worse process state;
when the conditions II and V in the S28 are met, the state is obtained from the static information to change, and the dynamic information shows that the process state is developed towards a better direction, so that the running state of the process can be judged to be developed from the previous state to the better process state;
if none of the above conditions is satisfied, maintaining the last evaluation result.
The non-optimal operation state is all operation states except the optimal operation state, which includes a good state, a normal state, a bad state, and a transition stage between the stable states. Generally, an enterprise manager is in an optimal operating state for a production process. However, when the industrial production is inevitably in a non-optimal stage, a suitable method needs to be found to trace the reasons for the operation in the non-optimal state, so as to provide an effective basis for production adjustment.
The existing non-optimal factor tracing method only considers the variable which causes the non-optimal operation state in the current time window, the identification method cannot comprehensively consider the accumulation effect of the non-optimal factor on the non-optimal occurrence time period, when the non-optimal factor variable and other process variables have strong coupling, the identification result is unreliable, and other non-critical variables are easily identified by mistake. The application provides a non-optimal factor tracing method based on sparse learning.
On the basis of the Lasso, a Group Lasso regularization technology is adopted, non-optimal factor variables with Group sparsity are recovered from a quadratic function, the overall contribution of a certain variable in the past is punished through the Group Lasso, and the overall contribution rate of the irrelevant variables is forced to approach zero. Based on the above method, the present application rewrites formula (17) to xTQuadratic functional form of Mx. The total Contribution of a variable in a past period is punished through GroupLasso, and the accurate position of a non-optimal variable causing the process state to change is traced according to Group-Contribution GWC (Group-Wise Containment).
When the evaluation result is in a non-optimal or transitional state, tracing the non-optimal factors, and determining the accurate position of the non-optimal variables, wherein the specific steps are as follows:
b1: setting x to contain m variables, and dividing the process variables into B non-repeated groups according to the formula (21) according to time;
in the formula, xkIs of the dimension ofmkK is 1, L, B, and
b2: selecting lambda according to formula (23);
in the formula, gammaMIs the corresponding control limit;
b3: rewriting formula (17) into a quadratic form according to formula (24);
b4: initialization f0And calculating M maximum eigenvalue gammak；
B5: calculating a gradient v (f) according to equation (25);
v(f)＝2M(f-x) (25)；
b6: updating f according to equation (26)k；
B7: calculate fkResidual errors are carried out until the precision requirement is met, and B8 is executed; otherwise, B5 is executed;
b8: f and A are calculated according to the formula (27);
A＝{k:||fk||＞0} (27)；
b9: the group contribution rate of the independent variables tends to be zero, so the sparse group contribution { fkAnd (4) forcing the group contribution of the independent variables to be zero, and calculating a group contribution rate GWC according to a formula (28)kDetermining the accurate position of the non-optimal variable by observing the GWC contributed by each group;
GWCk＝||fk|| (28)。
firstly, mining static-dynamic characteristic cooperative perception information by using a KPI-DrivenSFA algorithm, and establishing an off-line evaluation model of an operation state; then, a sliding window technology is introduced, the similarity of variation information between the score vector and the first-order difference of the online data and the score vector and the first-order difference of each state grade is calculated, and static and dynamic evaluation indexes are calculated according to the similarity. The online identification of the process running state is realized according to the size of the static evaluation index by formulating an effective evaluation rule; realizing on-line identification of the process running state change trend according to the size of the dynamic evaluation index, and completing comprehensive evaluation of each state and the transition process; and then, recovering non-optimal factor variables with Group sparsity from a quadratic function by adopting a Group Lasso Group regularization technology on the basis of Lasso through a data-driven fault diagnosis method based on sparse learning, punishing the total contribution of a certain variable in the past period through Group Lasso, and tracing the accurate positions of the non-optimal variables according to Group contribution GWC (Group-WisetContribustion).
In the process of evaluating the running state, a data window with the width of H is introduced as a basic analysis unit in consideration that single sample data is not enough to represent the running state of the production process and is easy to be interfered by various factors; then, the similarity of the score vector and the first order difference of the k time data and the variation information between the score vectors of each state grade and the first order difference thereof is calculated, and the static and dynamic evaluation indexes are calculated according to the similarity
The online identification of the process running state is realized according to the size of the static evaluation index by formulating an effective evaluation rule; and realizing on-line identification of the process running state change trend according to the size of the dynamic evaluation index, and completing comprehensive evaluation of each state and the transition process.
Simulation analysis:
the dense medium coal preparation process is a typical flow industrial process running in a severe open environment, various uncertain factors and interference are frequent, so that the data quality is low, the current process monitoring and running state evaluation methods and the like are not comprehensive in sensing working condition information, and serious problems such as missing report, false report and the like occur.
By knowing the dense medium coal preparation process and analyzing the influence factors influencing the coal preparation quality, 5 parameter variables are finally determined as process variables for evaluation and research of the running state. Because the dense medium suspension density, the cyclone inlet pressure, the qualified medium barrel liquid level, the coal slime content and the magnetic substance concentration are process parameters which have great influence on the coal preparation quality in the dense medium coal preparation process, the 5 variables are selected as process variables. The determination of the KPI index of coal is generally performed by off-line testing of the ash content of coal, and therefore, the ash content of coal is used as an output variable.
1) Establishment of offline evaluation model
26000 groups of data are collected from field data of a centralized control center of a certain coal preparation plant, 8000 groups of data with relatively complete information are selected from the field data for offline modeling through screening processing, and nearly 1081 groups of data are selected for online evaluation.
The data of the coal preparation process are divided mainly according to mechanism knowledge and process data characteristics of the coal preparation process and combined with experience in the actual production process, and then the process data are divided according to ash content. In the present application, the coal preparation process is divided into three stable states, which are respectively indicated by numbers, i.e. c is 1,2 and 3. The state grades corresponding to the three states are excellent, good and medium respectively. The larger the overflow ash content, the poorer the quality of the coal, and the poorer the state grade thereof. The relationship between the coal overflow ash and the corresponding state rank is shown in table 1.
Table 1: overflow ash and corresponding status grade
2) Static-dynamic characteristic cooperative sensing based on-line evaluation of running state of heavy-medium coal separation process
1081 groups of data are randomly selected from actually acquired sample data, including data of three steady state levels and three transition states. Wherein, the state level has 240 groups of excellent data, 300 groups of excellent data in the state level and 228 groups of excellent data in the state level. In addition, there are a good-to-good transition data 40 set, a good-to-medium transition data 40 set, and a medium-to-good transition data 30 set. The parameters in the experiment are set as follows: the threshold of the evaluation index is selected to be epsilon 0.85, and the width of the online window data is set to be H3.
In order to better embody the advantages of the complex industrial process operation state evaluation method based on static-dynamic characteristic collaborative perception, the evaluation result of the complex industrial process operation state based on PLS is mentioned. As shown in fig. 3, three sections (1), (2), and (3) in the drawing represent online data and evaluation index values of three state levels, respectively. It can be found that between the online data 1 to 240, the evaluation index evaluation results of the online data and the excellent level are slightly higher than those of the other two state levels, but the threshold requirement of 0.85 cannot be met, and the evaluation result has large fluctuation range, although a basic trend can be seen, the error rate is high. The main reasons are that the disturbance of the coal preparation process is more, the data quality is low, and the influence on the traditional method is larger.
For this purpose, the online evaluation result of the running state of the complex industrial process based on the static-dynamic characteristic collaborative perception is given. As shown in fig. 4, three parts (1), (2), and (3) in the figure represent online data and evaluation index values of three state levels, respectively, and fig. 4 represents a process dynamics analysis result. As can be seen from FIG. 4, the evaluation indexes of the online data and the quality grade are almost all greater than 0.85 between the online data 1-240, and the evaluation index is equal to 0.5 in FIG. 4, which indicates that the process is in a steady state. The process at this stage is in a good state according to the judgment criterion. In the online data 240-280, the evaluation indexes of the online data and the three state levels are all less than 0.85, the evaluation index value in the graph (2) is in an increasing trend as a whole, the evaluation index value in the graph (1) is in a decreasing trend as a whole, and in the graph (4), the evaluation index of the online data between 240-280 is less than 0.5, which indicates that the operation state is developing to a poor state and is in a good state at the previous moment. And in combination with the evaluation criterion, the process in the interval can be judged to be in the condition of transition from the excellent state to the good state. Compared with the method shown in the figure 3, the method for evaluating the running state of the complex industrial process based on the static-dynamic characteristic cooperative sensing is more comprehensive in process sensing, can further judge whether the running state is in a stable state or a transition state, and is remarkably improved in accuracy. The KPI-drive SFA method can better and more stably extract the intrinsic change law in the process, so that the problem of data quality is made up to a certain extent, and the identification precision is improved. The same can analyze the operation status of the "good", "medium" status in fig. 4.
Since the threshold setting of the evaluation index is empirically selected, the setting of the threshold size affects the final evaluation result. The smaller the threshold, the more beneficial the identification of a steady state, but the less likely a transient state will be identified and classified into its neighboring steady state. Table 2 shows the state level misidentification cases when the threshold values are 0.85, 0.8, and 0.7, respectively. The false recognition rate in the present application indicates the ratio of the number of samples that are actually inconsistent with the result of online evaluation to the total number of online evaluation data. As can be seen from table 2, the misrecognition rate is not so large, but a tendency is also seen in which the lower the evaluation threshold value is, the greater the misrecognition rate is.
Table 2: on-line evaluation error recognition rate
In addition, since coal preparation data actually collected from a factory contains a large amount of noise interference and outliers, an offline evaluation model of the conventional operation state evaluation method loses generalization capability due to the influence of the noise interference and the outliers, so that the evaluation result loses accuracy in online evaluation, and a relatively accurate operation state online evaluation result can be obtained from the graph (4). It can therefore be concluded that a comparatively satisfactory evaluation of the operating state can be achieved without or with little interference, i.e. with comparatively good data quality.
3) Non-optimal factor tracing in heavy-medium coal separation process
The coal dressing process of the heavy medium is evaluated by using an evaluation method based on static-dynamic characteristic cooperative perception, and when the coal dressing process is operated in a non-optimal state, variables causing the non-optimal process need to be traced.
FIG. 5 is a non-optimal factor tracing result based on variable contribution rate, wherein the abscissa 1-5 represents five process variables of the coal preparation process, respectively, such as dense medium suspension density, medium barrel liquid level, cyclone inlet pressure, magnetic concentration, and coal slime content. FIG. 5 is a non-optimal factor tracing result based on variable contribution rates; the fact that the contribution rates of the liquid level of the medium barrel and the concentration of the magnetic substances in the 'good' stage are approximate can be clearly found, and non-optimal variables cannot be accurately indicated, is caused by the fact that slight abnormal variables are mistakenly identified as main contribution variables due to the fact that the non-optimal variables and other process variables have strong coupling, and accordingly non-optimal factor tracing based on the variable contribution rates cannot accurately indicate the state variables affecting the process. FIG. 6 is a non-optimal factor tracing result based on sparse learning, and a non-optimal factor variable can be found to be the liquid level of the medium barrel from a 'good' stage, so that the situation is met; also taking the "middle" stage as an example, analyzing GWC, it was found that the contribution rate of the medium tank level was high relative to other variables, thus the factor that caused the process to be in non-optimum was mainly the medium tank level, whereas the "middle" stage of fig. 5 could not clearly indicate the main non-excellent factor.
In summary, according to the conventional variable contribution rate method, when the process is not significantly degraded, since the operating states "good" and "medium" are both normal states and no abnormality occurs, no significant degradation occurs, the given recognition result is often not clear enough, and the contribution rates of all variables are not zero in general, which cannot highlight the variables that result in non-optimal. The method based on sparse learning can make the contribution of the independent variable sparse to zero, and cause the contribution indication of the non-optimal variable to be more prominent and definite, and the identification result is consistent with the result of process analysis and is the non-optimal condition caused by the fact that the liquid level of the medium barrel is not well controlled. Too low and too high liquid level of the medium barrel can cause fluctuation of the inlet pressure of the cyclone, and the coal dressing effect is poor.
In conclusion, the static-dynamic characteristic cooperative sensing and deep extraction of the operation state information in the industrial process are realized by the static-dynamic characteristic cooperative sensing-based operation state evaluation method. The evaluation on the steady state and the transition of the process is completed by utilizing the similarity of the characteristic information and the similarity of the first-order difference information, when the running state is not optimal, the non-optimal factor tracing method based on sparse learning is provided, the non-optimal factor variable is accurately traced based on low-quality process data, and the influence of irrelevant variables is reduced. And finally, based on the actual production data in the dense medium coal separation process, the effectiveness and the practicability of the method are comprehensively analyzed and verified. Therefore, a good foundation is laid for optimizing and adjusting the technological process in the coal dressing process, improving the raw coal dressing quality and finally improving the economic benefit of enterprises.
Claims (2)
1. A static-dynamic cooperative perception complex industrial process running state evaluation method is characterized by comprising the following steps:
the method comprises the following steps: performing static-dynamic characteristic collaborative perception information mining by using a KPI-drive SFA algorithm, and establishing an off-line evaluation model of an operation state;
s11: acquiring production data of the production process to obtain an input data matrix X belonging to RN×mAnd the output data matrix Y ∈ RNWherein N is the number of samples, m is the number of process variables, R is the set of real numbers, RN×mA real number matrix representing dimensions N × m;
s12: the columns of the input data matrix X are processed with zero mean and unit variance, which are marked as XaAnd the difference between two input data points in each column time series is recorded as DeltaXa(ii) a The output data matrix Y is also normalized and denoted as Ya；
S13: for input matrix XaOutput matrix YaSum matrix Δ XaOperating a KPI-drive SFA algorithm, and specifically performing the following process:
a1: let YaAny one column in the sequence is equal to the initial ua；ΔXaIs equal to the initial ra；
A2: according to the formula(1) Calculating a load vector wa；
In the formula, alpha is a regularization parameter and is used for balancing two targets;
a3: load vector w is expressed according to equation (2)aStandardizing;
a4: x is determined according to the formula (3)aIs projected to waObtaining XaScore t ofa(ii) a Dividing Δ X according to equation (4)aIs projected to waTo obtain Δ XaScore r ofa；
ta＝Xawa (3)；
ra＝ΔXawa (4)；
A5: solving for vector c according to equation (5)a；
A6: c according to formula (6)aCarrying out standardization treatment;
a7: recalculating u according to equation (7)a；
A8: according to u in A7aU from step A1aWhether it is the same or refinedJudging whether convergence is achieved or not, if convergence is achieved, executing A8, otherwise, executing A2;
a9: calculating the load vector p of the matrix X according to equation (8)a；
A10: calculating the component u according to equation (9)aWith respect to taRegression coefficient b ofa；
A11: the residual matrix X is obtained according to the formula (10)a+1(ii) a The residual matrix Y is obtained according to the formula (11)a+1；
A12: mixing Xa、YaIs replaced by Xa+1、Ya+1Iterating from steps a1 to a11 until the desired number of features are extracted;
a13: calculating an objective function S according to formula (12);
S＝RX (12)；
wherein R ═ W (P)TW)-1；
S14: calculating a comprehensive economic related index R (i) according to a formula (13);
s15: according to the index R (i) independent of the quality in RIs eliminated to obtain Rq；
S16: historical data (x, y) is assumed, and data (x) therein is knownc,yc) The corresponding status level c;
s17: calculating score vectors t for respective states according to equation (14)c；
tc＝Rqxc (14)；
S18: scoring the vector t for each statecTime sequence is augmented, and the first order difference of each state score vector is calculated to obtain
Step two: introducing a sliding window technology, calculating the similarity of the score vector and the first-order difference of the online data and the variation information between the score vectors and the first-order differences of the state grades according to an offline evaluation model, and calculating static and dynamic evaluation indexes according to the similarity;
s20: constructing a sliding data window X for time K according to equation (15)k；
Xk＝[xk-H+1,L xk]T (15)；
In the formula, H is the width of a data window;
s21: constructing a time sequence augmentation matrix, then carrying out standardization preprocessing, and recording standardized data as
S22: introducing online sample data into offline evaluation model
And calculating the online sample data according to the formula (16)
Middle h sample
Score vector of
Wherein H is k-H +1, L, k; c is 1,2, L, C;
s23: time sequence augmentation is carried out on the score vector of the online sample data, and the first-order difference of the score vector of the online data is calculated to obtain
S24: calculating the distance between the score vector and the center of each state grade of the quality-related set according to formula (17)
In the formula (I), the compound is shown in the specification,
a score vector for each steady state;
s25: calculating the state grade first-order difference center distance of the first-order difference and quality related set of the score vector according to the formula (18)
Wherein E is an offset;
s26: according to the formula (19) using
Defining static evaluation indexes of online data relative to various state grades
S27: according to the formula (20) using
Defining dynamic evaluation indexes of online data relative to various state grades
S28: the process running state is evaluated on line according to the evaluation index,
firstly, when
When the process variation information is consistent with the variation information in the state grade, the process operation state can be judged to be
(ii) when the condition (i) is not satisfied, but the condition
If so, indicating that the process running state is in the state grade conversion process, namely the current process is gradually changed;
③ when
In time, the change speed of the process variation information related to the quality in the online data is not obviously changed, and the process state can be judged to be unchanged;
fourthly when
When the process state is in a transition state, the process state is judged to be deteriorated;
fifthly, when
When the process state is in a transition state, the process state is determined to be optimized;
s29: integrating the results of S28 to give the operation state evaluation result of static-dynamic characteristic cooperative sensing;
(ii) when the conditions (i) and (iii) in S28 are satisfied, the state rank is derived from the static information
And the dynamic information indicates that the process state does not change, and the running state of the process can be judged to be
When the condition (c) in the S28 is satisfied but the condition (c) is not satisfied, the state level obtained from the static information is not changed and
the dynamic information indicates that the process state is not changed, and the running state of the process can be judged to be
When the conditions of the second time and the fourth time in the S28 are met, the state is obtained from the static information to change, the dynamic information shows that the process state is developed towards a worse direction, and the running state of the process can be judged to be developed from the previous time state to the worse process state;
when the conditions II and V in the S28 are met, the state is obtained from the static information to change, and the dynamic information shows that the process state is developed towards a better direction, so that the running state of the process can be judged to be developed from the previous state to the better process state;
if none of the above conditions is satisfied, maintaining the last evaluation result.
2. The method for evaluating the running state of the complex industrial process based on static-dynamic cooperative sensing as recited in claim 1, wherein when the evaluation result is in a non-optimal or transitional state, non-optimal factor tracing is performed, and the accurate position of a non-optimal variable is determined, and the method comprises the following specific steps:
b1: setting x to contain m variables, and dividing the process variables into B non-repeated groups according to the formula (21) according to time;
in the formula, xkIs m in the dimensionkK is 1, L, B, and
b2: selecting lambda according to formula (23);
in the formula, gammaMIs the corresponding control limit;
b3: rewriting formula (17) into a quadratic form according to formula (24);
b4: initialization f0And calculating M maximum eigenvalue gammak；
B5: calculating a gradient v (f) according to equation (25);
v(f)＝2M(f-x) (25)；
b6: updating f according to equation (26)k；
B7: calculate fkResidual errors are carried out until the precision requirement is met, and B8 is executed; otherwise, B5 is executed;
b8: f and A are calculated according to the formula (27);
A＝{k:||fk||＞0} (27)；
b9: sparse group contribution { fkAnd (4) forcing the group contribution of the independent variables to be zero, and calculating a group contribution rate GWC according to a formula (28)kDetermining the accurate position of the non-optimal variable by observing the GWC contributed by each group;
GWCk＝||fk|| (28)。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011050080.3A CN112363462B (en) | 2020-09-29 | 2020-09-29 | Static-dynamic cooperative sensing complex industrial process running state evaluation method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011050080.3A CN112363462B (en) | 2020-09-29 | 2020-09-29 | Static-dynamic cooperative sensing complex industrial process running state evaluation method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112363462A true CN112363462A (en) | 2021-02-12 |
| CN112363462B CN112363462B (en) | 2021-11-16 |
Family
ID=74507531
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011050080.3A Active CN112363462B (en) | 2020-09-29 | 2020-09-29 | Static-dynamic cooperative sensing complex industrial process running state evaluation method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112363462B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113848836A (en) * | 2021-09-30 | 2021-12-28 | 中国矿业大学 | ILSTM network-based complex industrial process running state evaluation method |
| CN114384870A (en) * | 2021-11-10 | 2022-04-22 | 中国矿业大学 | Complex industrial process running state evaluation method based on kernel local linear embedding PLS |
| CN114527731A (en) * | 2022-02-22 | 2022-05-24 | 中国矿业大学 | Industrial process running state evaluation method based on supervision probability slow characteristic analysis |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20020029115A1 (en) * | 2000-09-07 | 2002-03-07 | Mori Seiki Co., Ltd. | Machine tool maintenance system |
| JP2003223219A (en) * | 2002-01-31 | 2003-08-08 | Toshiba Corp | Apparatus and method of state evaluation for monitoring plant, and evaluation program for plant data |
| CN104062953A (en) * | 2014-06-12 | 2014-09-24 | 东北大学 | Gold hydrometallurgy process running state online evaluation method |
| CN105278520A (en) * | 2015-06-29 | 2016-01-27 | 中国矿业大学 | Complex industrial process running state evaluation method and application based on T-KPRM |
| CN108037747A (en) * | 2017-12-11 | 2018-05-15 | 东北大学 | Golden hydrometallurgy whole process hierarchical block operating status on-line evaluation method |
-
2020
- 2020-09-29 CN CN202011050080.3A patent/CN112363462B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20020029115A1 (en) * | 2000-09-07 | 2002-03-07 | Mori Seiki Co., Ltd. | Machine tool maintenance system |
| JP2003223219A (en) * | 2002-01-31 | 2003-08-08 | Toshiba Corp | Apparatus and method of state evaluation for monitoring plant, and evaluation program for plant data |
| CN104062953A (en) * | 2014-06-12 | 2014-09-24 | 东北大学 | Gold hydrometallurgy process running state online evaluation method |
| CN105278520A (en) * | 2015-06-29 | 2016-01-27 | 中国矿业大学 | Complex industrial process running state evaluation method and application based on T-KPRM |
| CN108037747A (en) * | 2017-12-11 | 2018-05-15 | 东北大学 | Golden hydrometallurgy whole process hierarchical block operating status on-line evaluation method |
Non-Patent Citations (3)
| Title |
|---|
| STYLIANI SARRI: "Performance Assessment of Linear State Estimators Using Synchrophasor Measurements", 《IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT》 * |
| 常玉清: "基于改进的动态因果图推理算法的复杂工业过程运行状态在线评价", 《控制理论与应用》 * |
| 邹筱瑜: "基于分层分块结构的流程工业过程运行状态评价及非优原因追溯", 《自动化学报》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113848836A (en) * | 2021-09-30 | 2021-12-28 | 中国矿业大学 | ILSTM network-based complex industrial process running state evaluation method |
| CN114384870A (en) * | 2021-11-10 | 2022-04-22 | 中国矿业大学 | Complex industrial process running state evaluation method based on kernel local linear embedding PLS |
| CN114527731A (en) * | 2022-02-22 | 2022-05-24 | 中国矿业大学 | Industrial process running state evaluation method based on supervision probability slow characteristic analysis |
| CN114527731B (en) * | 2022-02-22 | 2023-12-26 | 中国矿业大学 | Industrial process operation state evaluation method based on supervision probability slow feature analysis |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112363462B (en) | 2021-11-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112363462B (en) | Static-dynamic cooperative sensing complex industrial process running state evaluation method | |
| CN104932488B (en) | A kind of Model Predictive Control Performance Evaluation and diagnostic method | |
| CN109884892B (en) | Process industrial system prediction model based on cross correlation time-lag grey correlation analysis | |
| CN105278520B (en) | Based on T-KPRM complex industrial process evaluation of running status methods and application | |
| CN112001527A (en) | Industrial production process target data prediction method of multi-feature fusion deep neural network | |
| CN111752147B (en) | Multi-working-condition process monitoring method with continuous learning capability and improved PCA (principal component analysis) | |
| CN110879580A (en) | Analysis and monitoring method for large-range non-steady transient continuous process | |
| CN110942258B (en) | Performance-driven industrial process anomaly monitoring method | |
| CN111413949B (en) | Method for reducing fault early warning false alarm rate of industrial process | |
| CN112904810A (en) | Process industry nonlinear process monitoring method based on effective feature selection | |
| Zhang et al. | A novel plant-wide process monitoring framework based on distributed Gap-SVDD with adaptive radius | |
| CN111639304A (en) | CSTR fault positioning method based on Xgboost regression model | |
| Åsnes et al. | Predictive maintenance and life cycle estimation for hydro power plants with real-time analytics | |
| CN114625009A (en) | Fault detection method based on system identification and optimal filtering | |
| CN114139446A (en) | On-line detection soft measurement method for components in special rectification process | |
| CN107563656B (en) | Method for evaluating running state of gold hydrometallurgy cyaniding leaching process | |
| CN111338310B (en) | Industrial process steady-state working condition identification and classification method | |
| Li et al. | Data-driven modeling and operation optimization with inherent feature extraction for complex industrial processes | |
| CN115310561B (en) | Electromagnetic valve fault monitoring method based on integrated instant learning | |
| CN114527731A (en) | Industrial process running state evaluation method based on supervision probability slow characteristic analysis | |
| Peng et al. | Phase partition and fault diagnosis of batch process based on keca angular similarity | |
| CN114200914A (en) | MW-OCCA-based quality-related early fault detection method | |
| Chu et al. | Latent variable techniques based operating performance assessment for the dense medium coal preparation process | |
| Wang et al. | Maximizing Anomaly Detection Performance Using Latent Variable Models in Industrial Systems | |
| Fernández Berrocal et al. | Evaluation and Interpretation on Data Driven ROP Models From Engineering Perspectives |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |