CN115616919B - Electromechanical product sensor optimal configuration method - Google Patents
Electromechanical product sensor optimal configuration method Download PDFInfo
- Publication number
- CN115616919B CN115616919B CN202211421113.XA CN202211421113A CN115616919B CN 115616919 B CN115616919 B CN 115616919B CN 202211421113 A CN202211421113 A CN 202211421113A CN 115616919 B CN115616919 B CN 115616919B
- Authority
- CN
- China
- Prior art keywords
- sensor
- individual
- fault
- optimization
- electromechanical product
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
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
- G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
- G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
- G05B13/04—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators
- G05B13/042—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators in which a parameter or coefficient is automatically adjusted to optimise the performance
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Testing Or Calibration Of Command Recording Devices (AREA)
Abstract
The invention provides an electromechanical product sensor optimal configuration method, wherein an electromechanical product comprises one or more sensors, the sensors are used for monitoring the state of the electromechanical product, the method comprises the steps of obtaining the corresponding relation between different sensor sets and fault mode sets by analyzing the fault mode and the fault mechanism of the electromechanical product, constructing a fault mode and sensor relation matrix, constructing a sensor configuration optimization model comprising a sensor optimization target model and a sensor optimization constraint model, solving the sensor configuration optimization model based on a discrete multi-objective particle swarm algorithm, and obtaining an optimal sensor set as the sensor configuration of the electromechanical product. The method and the device automatically screen the configuration selection of the sensor, have better detection performance on the premise of ensuring higher reliability of the system, obtain the optimal configuration selection of the electromechanical product sensor, quickly and accurately select the sensor, and reduce the complexity of the electromechanical product/system.
Description
Technical Field
The invention relates to the field of product layout configuration, in particular to an electromechanical product sensor optimal configuration method.
Background
Comprehensive fault diagnosis, prediction and health management become an advanced technology gradually adopted by airplanes, accurate acquisition of information is the basis of function realization of a PHM system, and acquisition of signals can not leave a sensor, so that reasonable sensor selection and layout play an important role in accurate acquisition of information of electromechanical products.
The existing aviation electromechanical products are limited by installation positions and system complexity, sensors are less in arrangement, state perception of the products cannot be achieved, along with improvement of attention to visual maintenance of the aviation electromechanical products, and in order to master a fault evolution rule, the system gradually requires that the electromechanical products are additionally provided with the sensors to monitor states. The sensor type, function, characteristics are various, generally speaking, the designer will carry out failure mode and influence analysis to the product, provide the sensor set of monitoring all failure modes of monitored object. However, not all sensors are necessary, and for fault detection, the same type of fault can be characterized by different monitoring parameters, which have signal redundancy and different effects on fault diagnosis and prediction. In addition, if a large number of sensors are used in the system to monitor critical parameters in real time, due to the limitation of the structural characteristics of the equipment, the installation and layout modes of the sensors may affect the working state of the equipment. Therefore, under the condition of meeting the requirement of the testing index of the PHM system, the theory and the method for optimizing the configuration of the sensor are researched, how to find the balance point between the sensor installation requirement and the PHM testing requirement and configure the proper sensor have very important practical value. In the prior art, multiple targets are artificially integrated to form a single target which is comprehensively considered, and then a single-target optimization method is adopted, so that the influence of artificial subjective factors is large, the mode is single and solidified, a combination mode of multiple sensors cannot be obtained, and the practical application is not facilitated.
Disclosure of Invention
In order to solve the defects in the prior art, the invention provides an electromechanical product sensor optimal configuration method. According to the method, a sensor optimized configuration multi-constraint multi-target evaluation index is set by constructing a relation matrix of the faults of the electromechanical products and the sensors; because several targets of the multi-target optimization problem may be contradictory, a single solution cannot make each target achieve the optimal condition; therefore, the configuration selection of the sensor is automatically screened by designing a discrete multi-target intelligent optimization method, the detection performance is better on the premise of ensuring higher reliability of the system, and the optimal configuration selection of the electromechanical product sensor is obtained.
According to a first aspect of the present invention, there is provided an electromechanical product sensor optimal configuration method, wherein the electromechanical product includes one or more sensors, and the sensors are used for monitoring the state of the electromechanical product, the method includes:
step 10: by analyzing the failure mode and failure mechanism of the electromechanical product, the corresponding relation between different sensor sets and the failure mode set is obtained, and a failure mode and sensor relation matrix is constructed.
Step 20: and constructing a sensor configuration optimization model, wherein the sensor configuration optimization model comprises a sensor optimization target model and a sensor optimization constraint model.
Step 30: and solving the sensor configuration optimization model based on the discrete multi-target particle swarm algorithm to obtain an optimal sensor set as the sensor configuration of the electromechanical product.
Further, the method for optimally configuring the sensor of the electromechanical product provided by the present invention is characterized in that the step 10 further comprises: the relation matrix of the fault modes and the sensors is a matrix with m rows and n columnsThe rows of matrix D represent m failure modes of the systemThe columns of the matrix D represent the n sensors of the electromechanical product,The value is 0 or 1, and the like,to representCan detect,To representIs unable to detect。
Further, the sensor optimization configuration method for the electromechanical product provided by the present invention is characterized in that the sensor optimization target model requires the minimum number of sensors in the sensor set, the minimum sensor cost and the minimum sensor failure probability, and the sensor optimization target model includes: the number of sensors is minimum:in whichThe number of the jth sensor; the cost of the sensor is lowest:whereinThe cost of the jth sensor; the probability of sensor failure is minimal:whereinThe failure probability of the jth sensor.
Further, the method for optimizing configuration of the sensor of the electromechanical product is characterized in that the sensor optimization constraint model requires that the sensor set meets preset conditions of fault coverage capability, fault detection rate, fault isolation rate and fault false alarm rate, and the sensor optimization constraint model comprises the following steps: fault coverage capability: for any kind of faultAt least one sensor is ensured to detect it, i.e.(ii) a The fault detection rate is as follows: ratio of the total probability of a correctly detected failure mode to the total probability of the failure mode, i.e.Wherein, in the process,is a failure modeThe probability of the occurrence of the event is,for the purpose of the corrected detection capability of the sensor,(ii) a Fault isolation rate: the ratio of the probability of correct isolation to component level failure to the total probability of detected failure mode, i.e.Wherein, I is a component fault mode set capable of fault isolation; fault false alarm rate: the ratio of the number of false alarm faults generated by the diagnosis to the total number of faults detected, i.e.。
Further, the method for optimally configuring the sensor of the electromechanical product provided by the present invention is characterized in that step 30 comprises:
step 301: initializing calculation parameters in the discrete multi-target particle swarm algorithm, wherein the calculation parameters comprise a population scale N, a sensor variety number N and a learning factor、Inertial weight parameters、The size O of an external archiving space, the degree of variation u, the maximum iteration number M and constraint target parameters are set.
Step 302: generating a group by using a chaotic initialization strategy according to the group size N and the sensor type number NDimension initial matrixDiscretizing to initialize individual position and speed to obtain initial populationWherein, the individual speed represents the probability that the individual position takes 1 or 0, the individual position is 1 to represent that the sensor is installed at the position, the individual position is 0 to represent that the sensor is not installed at the position, and the initial iteration number is set。
Step 303: optimizing target model calculations from sensorsSelecting the population according to the individual optimal strategy according to the fitness value of each individualAccording to the global optimum strategy, selectingThe population extremum of (1).
Step 304: and adjusting the inertia weight of the discrete multi-target particle swarm algorithm according to a self-adaptive strategy, and then updating the individual speed and position in the swarm.
Step 305: according to a mutation strategy, forThe middle individual performs mutation operation with a certain probability to change the position of the particles, and calculates the mutation according to an optimization objective functionThe fitness value of each individual.
Step 306: and modifying the fitness value of the population according to the sensor optimization constraint model.
Step 307: and updating the external archive, and adjusting the scale of the external archive by adopting a congestion degree strategy.
Step 308: updating populations according to individual optimal strategiesUpdating the population according to the global optimal selection strategyThe population extremum of (1).
Step 309: if the iteration stop condition is satisfiedOutputting an optimal sensor configuration set, namely a Pareto optimal leading edge, and selecting a proper set x from an optimal solution set as optimal configuration according to application requirements; if the iteration stop condition is not satisfied, the counter is updatedAnd returns to step 304.
Further, the method for optimally configuring the electromechanical product sensor provided by the present invention is characterized in that step 302 further comprises: according to the initial valueComputing chaotic sequences through logistic mappingTo obtain an initial matrixThe logistic mapping is:wherein, in the process,control parameters of(ii) a According toFor the initial matrixDiscretizing to an initial discrete location, whereinIs composed ofA random number in between.
Further, the electromechanical product sensor optimal configuration method provided by the invention is characterized in that the individual optimal strategy comprises the following steps: if the individual present position dominates the individual extremum, updating the individual extremum into the individual present position; if the individual extremum dominates the individual present position, the individual extremum remains unchanged; if the individual present position and the individual extremum do not dominate each other, the individual extremum is randomly selected to be updated to the individual present position or remain unchanged.
Further, the optimal configuration method for the electromechanical product sensor provided by the invention is characterized in that the global optimal strategy comprises the following steps: and selecting the individuals as the group extremum of the current individuals from an external archive in a roulette mode or a tournament selection mode, wherein the external archive comprises a group of non-inferior solution sets, the group of non-inferior solution sets are the individuals with better quality in the group, and each individual is not mutually dominant.
Further, the method for optimally configuring the electromechanical product sensor provided by the present invention is characterized in that step 304 further comprises: adjusting the inertial weights according to an adaptive strategy(ii) a Updating the individual speed in the population as follows:(ii) a Updating the individual positions in the population as follows:(ii) a Wherein the content of the first and second substances,、is a random number, subscriptRepresenting the ith individual type d sensor,the individual velocities of the t +1 th iteration and the t-th iteration respectively,individual positions, functions, of the t +1 th and t-th iterations, respectively,Is composed ofA random number in between.
Further, the method for optimally configuring the sensor of the electromechanical product provided by the present invention is characterized in that the step 305 further includes: as the number of iterations t increases, the proportion of individuals involved in the variation decreases non-linearly.
Further, the method for optimally configuring the sensor of the electromechanical product provided by the present invention is characterized in that step 306 further includes: and obtaining the individual constraint value according to the sensor optimization constraint model, and when the individual constraint value does not meet the constraint target parameter requirement, setting the individual fitness value as a larger value or an upper limit value so that the individual cannot enter the group extremum and the individual extremum.
Further, the method for optimally configuring the sensor of the electromechanical product provided by the present invention is characterized in that step 307 further comprises: for a newly generated alternative individual, if the alternative individual is dominated by any individual in an external archive, the alternative individual is rejected; if any individual of the external archive can not dominate the alternative individual, adding the alternative individual into the external archive; if the storage individual in the external archive is dominated by the alternative individual, rejecting the storage individual in the external archive; when the external archive size reaches a maximum, a crowdedness policy is employed to limit the external archive size.
According to a second aspect of the present invention, there is provided a computer apparatus comprising: a memory to store instructions; and a processor for invoking the instructions stored by the memory to perform the electromechanical product sensor optimization configuration method of the first aspect.
According to a third aspect of the present invention, there is provided a computer readable storage medium, characterized by instructions stored thereon, which, when executed by a processor, perform the method for optimized configuration of electromechanical product sensors of the first aspect.
Compared with the prior art, the technical scheme of the invention at least has the following beneficial effects:
1. blindness and redundancy of manual sensor selection are avoided, cost is high, efficiency is low, time is consumed, and detection capability of electromechanical products/systems is improved;
2. comprehensive screening is carried out on Failure modes and Failure Mechanism Analysis (Failure Mode, mechanism and Effect Analysis, FMMEA) of electromechanical products, and the system has good detection performance on the premise of ensuring high reliability based on sensor optimization configuration indexes;
3. aiming at the problem that the optimal configuration of the sensor is a set coverage and multi-target combination optimization, in order to take the influence of each optimization target into consideration, the improved discrete multi-target intelligent optimization algorithm is adopted, the sensor is selected quickly and accurately, and the complexity of an electromechanical product/system is reduced.
It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention, as claimed.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and together with the description, serve to explain the principles of the invention.
FIG. 1 is a flow diagram illustrating a method for optimized configuration of an electromechanical product sensor in accordance with an exemplary embodiment.
FIG. 2 is a schematic diagram illustrating a process for solving a sensor configuration optimization model based on a discrete multi-objective particle swarm optimization algorithm according to an exemplary embodiment.
FIG. 3 is a Pareto front for sensor layout optimization according to an exemplary embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. In addition, the technical features involved in the embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.
The embodiment of the invention provides an optimized configuration method of an electromechanical product sensor, which is used for automatically analyzing a balance point between the installation requirement of the electromechanical product sensor and the PHM test requirement, has better detection performance on the premise of ensuring higher reliability of a system, and searches for proper sensor configuration. The optimal configuration method for the electromechanical product sensor, as shown in fig. 1, includes steps 10 to 30:
the electromechanical product comprises one or more sensors, and the sensors are used for monitoring the state of the electromechanical product.
Step 10: by analyzing the fault mode and the fault mechanism of the electromechanical product, the corresponding relation between different sensor sets and the fault mode set is obtained, and a fault mode and sensor relation matrix is constructed.
And analyzing an electromechanical product FMMEA (frequency modulated Membrane electrode assembly) to obtain a sensor set for monitoring all fault modes of the monitored object. Based on this, a logical matrix between the set of failure modes and the set of sensors is established, describing the correlation of the failure with the sensors.
The traditional Failure Mode impact and hazard Analysis method (Failure Mode impact Criticality Analysis, FMECA) mainly obtains three types of information: a specific section/component failure mode list; the effects of each failure mode, including local effects and final effects; the criticality of each failure mode is classified into I-IV, with the I-th being most critical. Conventional FMECAs are beneficial for identifying product reliability and redundancy alternatives, but are deficient in meeting equipment health monitoring design requirements. For example, conventional FMECAs do not provide prognostic or symptom information regarding failure modes; the requirements of the sensor and the installation position thereof required for observing the symptoms of the failure mode and influencing are not involved; the health management techniques required for fault diagnosis and prognosis are not fully embodied; it is common to see that each part of the product is independent. In consideration of the defects, the method is expanded on the basis of the traditional FMECA, the research on the Failure Mechanism, the Failure detection and identification technology is strengthened and integrated into the traditional FMECA while the Failure Mode Analysis is carried out, and a sensor set for monitoring all Failure modes of the monitored object is provided through a sensor configuration obtained through Failure Mode and Failure Mechanism Analysis (FMMEA) of electromechanical products. Based on this, a logical matrix between the set of failure modes and the set of sensors is established, describing the correlation of the failure with the sensors.
Step 20: and constructing a sensor configuration optimization model, wherein the sensor configuration optimization model comprises a sensor optimization target model and a sensor optimization constraint model.
And formulating a corresponding sensor configuration optimization model according to the actual application requirements, wherein the sensor configuration optimization model comprises a sensor optimization target model and a sensor optimization constraint model and is used for optimizing and guiding the selection of the sensor.
In the optimal configuration process of the sensor system, it is considered that the number of sensors needs to be reduced as much as possible to minimize the cost. In addition, in practical engineering application, the system performance is of vital importance, and the minimum fault probability of the sensor system is taken as another target in consideration of the influence caused by the reliability of the sensor and external factors, so that the optimization problem becomes a multi-target optimization problem.
The realization of the optimization design target needs to be carried out under a plurality of constraint conditions, and for a detection system, the problems of detection, isolation, false alarm and the like of a fault mode are important aspects capable of reflecting the effectiveness of the detection system, so that the establishment of a fault detection rate constraint model, a fault isolation rate constraint model and a false alarm rate constraint model of the system is an effective way for obtaining an optimization result, and in addition, constraint indexes can be expanded and deleted according to different detection targets.
The purpose of the optimal configuration of the sensors is to find a sensor set which meets the requirements of fault detection rate, fault isolation rate, false alarm rate and the like, simultaneously minimize the number of sensors in the set, cost and fault probability, and simultaneously consider the installability of the aeronautical and electromechanical products.
And step 30: and solving the sensor configuration optimization model based on a discrete multi-target particle swarm algorithm to obtain an optimal sensor set serving as the sensor configuration of the electromechanical product.
The sensor configuration problem simultaneously comprises two problems of set coverage and multi-target combination optimization, and the sensor configuration optimization model is solved by utilizing an improved discrete multi-target particle swarm algorithm aiming at the problems of sensor parameter integer selection and multi-target optimization. The improved discrete multi-target particle swarm optimization method is used, multiple constraint targets do not need to be subjected to prior synthesis, the relation among the optimization targets is considered in a balanced mode, various strategies are adopted to obtain a plurality of sensor configuration combinations meeting the evaluation indexes of the detection system, and one combination mode is selected according to the actual engineering to perform configuration.
For the basic multi-objective particle swarm optimization (MOPSO), when the multi-objective optimization problem is solved, the solutions obtained by the objective functions may be contradictory, and it is difficult to find a solution under the condition that the optimal solution can be obtained by each objective function. In the process of solving the multi-objective optimization problem, some compromise solutions should be found as much as possible, and the compromise solutions can meet the condition that each objective function achieves the better condition as much as possible. This set of solution sets may be called the non-dominated solution set or Pareto optimal solution set.
A multi-objective optimization problem consisting of n optimization objectives can be described as:
in the formula (I), the compound is shown in the specification,the m decision variables are represented by a number m of decision variables,is the target solution vector space. The goal of the optimization is to find the targetSo thatThe optimization is achieved on the premise that equality and inequality constraints are met.
In the embodiment, the sensor optimization configuration is a set coverage and multi-objective combination optimization problem, the sensors are selected as integer programming problems, the sensor optimization model is solved by using the improved discrete form, namely a binary particle swarm algorithm, and the obtained non-dominated solution set or Pareto optimal solution set is solved to serve as the sensor configuration of the electromechanical product.
In some embodiments, step 10 further comprises: the fault mode and sensor relation matrix is a matrix with m rows and n columnsThe rows of the matrix D represent the m failure modes of the systemThe columns of the matrix D represent the n sensors of the electromechanical product,The value is 0 or 1, and the like,to representCan detect,To representIs unable to detect。
The function of the PHM system of the electromechanical product can not be separated from the information acquired by the sensor, if the configuration of the sensor is in a problem, partial faults can not be identified or misjudgment can occur, and unpredictable results can be easily caused, so that the selective configuration of the sensor needs to meet the requirements of covering and effectively detecting key faults of the electromechanical product. The failure mode and sensor type of the electromechanical product can be obtained by performing FMMEA on the electromechanical product, and a relationship matrix between a failure mode set and a sensor set needs to be established firstly.
If the system has m fault modes and the optional sensor measurement signals have n types, the relation matrix of the fault and the sensor of the system can be recorded as. The rows of the matrix represent failure modes and the columns represent sensor types, as shown in table 1 below.
TABLE 1 Fault vs. sensor matrix
The fault and sensor relation matrix describes the correlation between the fault set and the sensor set, however, in the actual system, the sensor is influenced by the detection reliability of the sensor and external factorsFailure F may not necessarily be detected 100% i Occurs. Therefore, it is necessary to analyze the fault detection capability of each type of sensor in consideration of the reliability of the sensor and correct the internal value of the relationship matrix according to the actual detection capability of the sensor.
In some embodiments, the sensor optimization objective model requires a minimum number of sensors in the sensor set, a sensor costThe sensor optimization target model comprises the following steps of: the number of sensors is minimum:whereinThe number of the jth sensor; the cost of the sensor is lowest:whereinCost for the jth sensor; the probability of sensor failure is minimal:whereinIs the failure probability of the jth sensor.
In the process of optimal configuration of the sensor system, it is considered that the number of sensors needs to be reduced as much as possible so as to minimize the cost. In addition, in practical engineering application, the system performance is of vital importance, and the minimum fault probability of the sensor system is taken as another target in consideration of the influence caused by the reliability of the sensor and external factors, so that the optimization problem becomes a multi-target optimization problem.
In some embodiments, the sensor optimization constraint model requires the sensor set to satisfy preset conditions of fault coverage capability, fault detection rate, fault isolation rate and fault false alarm rate, and the sensor optimization constraint model includes:
The fault detection rate is as follows: ratio of the total probability of a correctly detected failure mode to the total probability of the failure mode, i.e.
Wherein, the first and the second end of the pipe are connected with each other,as a failure modeThe probability of occurrence of the event is determined,for the purpose of the corrected detection capability of the sensor,;
fault isolation rate: the ratio of the probability of correct isolation to component level failure to the total probability of detected failure mode, i.e.
Wherein, I is a component fault mode set capable of fault isolation; fault false alarm rate: the ratio of the number of false alarm faults generated by the diagnosis to the total number of faults detected, i.e.
The realization of the optimization design target needs to be carried out under a plurality of constraint conditions, and for a detection system, the problems of detection, isolation, false alarm and the like of a fault mode are important aspects capable of reflecting the effectiveness of the detection system, so that the establishment of a fault detection rate constraint model, a fault isolation rate constraint model and a false alarm rate constraint model of the system is an effective way for obtaining an optimization result, and in addition, constraint indexes can be expanded and deleted according to different detection targets.
In some embodiments, as shown in fig. 2, step 30 specifically includes steps 301 to 309:
step 301: initializing calculation parameters in the discrete multi-target particle swarm algorithm, wherein the calculation parameters comprise a population size N, a sensor variety number N and a learning factor、Inertial weight parameter、The size O of an external archiving space, the degree of variation u, the maximum iteration number M and constraint target parameters are set.
Step 302: generating a group by using a chaotic initialization strategy according to the group size N and the sensor type number NDimension initial matrixDiscretizing to initialize individual position and speed to obtain initial populationWherein, the individual speed represents the probability that the individual position takes 1 or 0, the individual position is 1 to represent that the sensor is installed at the position, the individual position is 0 to represent that the sensor is not installed at the position, and the initial iteration number is set。
Step 303: optimizing target model calculations from sensorsSelecting the population according to the individual optimal strategy according to the fitness value of each individualAccording to the global optimum strategy, selectingThe population extremum of (1).
Step 304: and adjusting the inertia weight of the discrete multi-target particle swarm algorithm according to a self-adaptive strategy, and then updating the individual speed and position in the population.
Step 305: according to a mutation strategy, forThe middle individual performs mutation operation with a certain probability to change the position of the particles, and calculates the mutation according to an optimization objective functionOf each individual.
Step 306: and modifying the fitness value of the population according to the sensor optimization constraint model.
Step 307: and updating the external archive, and adjusting the scale of the external archive by adopting a congestion degree strategy.
Step 308: updating populations according to individual optimal strategiesUpdating the population according to the global optimal selection strategyThe group extremum of (1).
Step 309: such asIf the iteration stop condition is satisfiedOutputting an optimal sensor configuration set, namely a Pareto optimal leading edge, and selecting a proper set x from the optimal solution set as optimal configuration according to application requirements; if the iteration stop condition is not satisfied, the counter is updatedAnd returns to step 304.
Specifically, in some embodiments, step 302 further comprises: according to the initial valueComputing chaotic sequences through logistic mappingTo obtain an initial matrixThe logistic map is:wherein, in the step (A),control parameters(ii) a According toFor the initial matrixDiscretizing to an initial discrete location, whereinIs composed ofA random number in between.
In the embodiment, the velocity vector of the discrete multi-target particle swarm algorithm is no longer a position change, but is taken as the probability that the position of an individual is 1 or 0, and the individual is selected to be 1 or 0 at the corresponding position according to the magnitude of the velocity, so as to indicate whether a sensor is installed at the position.
The population initial distribution has important influence and effect on the iterative process after the algorithm, the reasonable initial distribution has positive effect on all performances of the algorithm, and particularly the reasonable initial distribution can help the algorithm to improve the early-maturing condition. The diversity of the population requires that the initial distribution of the individuals must be as even as possible. A good optimal solution requires that the individual initial distributions should occupy as much solution space as possible.
The motion state with randomness generally obtained in a deterministic equation is called chaos. The chaos optimization is a novel optimization method, which utilizes the characteristic ergodic characteristic of a chaos system to carry out optimization search and does not require that a target function has continuous and differentiable properties.
Generating a chaotic variable by applying Logistic mapping in the form ofWhen it comes toWhen the value is 4, the system is completely in a chaotic state.
Obtaining an initial discrete position by discretizing the chaotic sequence, wherein the discretization mode is as follows:
to indicate whether a sensor is installed at that location.
For single-target optimization problems, the magnitude of the objective function value can directly reflect the quality of the obtained solution. However, for the multi-objective optimization problem, due to the mutual non-dominance relationship, the individuals are not comparable, and the number of the non-dominance individuals is increased along with the execution of the algorithm, due to the memory limitation, a part of the non-dominance individuals are discarded, which greatly destroys the diversity of the understanding. Therefore, selecting a better optimal individual strategy has also become a key factor affecting the performance of the algorithm.
In some embodiments, the individual optimization strategies include: if the individual present position dominates the individual extremum, updating the individual extremum into the individual present position; if the individual extremum dominates the individual present position, the individual extremum remains unchanged; if the individual present position and the individual extremum do not dominate each other, the individual extremum is randomly selected to be updated to the individual present position or remain unchanged.
In some embodiments, the global optimal policy comprises: and selecting the individuals as the group extremum of the current individuals from an external archive in a roulette mode or a tournament selection mode, wherein the external archive comprises a group of non-inferior solution sets, the group of non-inferior solution sets are the individuals with better quality in the group, and each individual is not mutually dominant.
A group of non-inferior solution sets is stored in the external archive, the group of solution sets represents individuals with better quality in the group, the individuals are selected from the external archive in a roulette mode or a tournament selection mode to serve as the global optimal solution of the current individuals, excellent information shared by an information sharing mechanism can be ensured, and the diversity of the group is ensured by randomly selecting the nature of the individuals.
In some embodiments, step 304 further comprises: adjusting inertial weight according to adaptive strategy:
the individual speed in the updated population is:
updating the individual positions in the population as follows:
wherein, the first and the second end of the pipe are connected with each other,、is a random number, subscriptRepresenting the ith individual type d sensor,the individual velocities for the t +1 th iteration and the t-th iteration respectively,individual positions, functions, of the t +1 th and t-th iterations, respectively,Is composed ofA random number in between.
In the embodiment, although the particle swarm optimization algorithm has the advantages of simple algorithm structure, few parameters, high convergence speed and the like, the particle swarm optimization algorithm is easy to fall into a local extreme point, so that a global optimal solution cannot be obtained. There are two reasons for this phenomenon: firstly, the property of the function to be optimized and secondly, the diversity of individuals in the calculation process disappears rapidly due to the reasons of improper parameter design, individual scale selection and the like of the algorithm in the operation process, so that the algorithm is premature.
Therefore, the constant initial inertial weight is improved to improve the global optimization capability of the algorithm. Adopting an inertia weight adaptive adjustment strategy:
initial stage of iterationThe size is large, the individual is in the development process, the global convergence capability of the individual is strong, and the local convergence capability is weak; end of iterationSmaller, individual in the mining process, weak overall convergence ability of the individual, strong local convergence ability.
In this embodiment, the velocity vector in the discrete particle swarm algorithm is no longer a position change, but is a probability of 1 or 0 being taken as an individual position. Selecting 1 or 0 of the individual at the corresponding position according to the speed, indicating whether a sensor is installed at the position, wherein the probability value can be represented by a sigmoid functionTo describe it.
To prevent saturation of the sigmoid function, the speed of the individual is set within a certain range, modified by:
in some embodiments, step 305 further comprises: as the number of iterations t increases, the proportion of individuals involved in the variation decreases non-linearly.
Generally, the particle swarm optimization has a fast convergence speed, but the fast convergence can cause the population to fall into a locally optimal solution. Therefore, in order to prevent the population from falling into a locally optimal solution (pseudo Pareto frontier), mutation operations are performed on individuals with a certain probability to change the particle positions. One of the variation methods is that the variation proportion is reduced nonlinearly along with the iteration process, and the variability is large at the initial stage of the iteration process so as to enhance the global exploration capability of the algorithm; in the later stage of the iterative process, the variation influence is weakened, and the stability of the algorithm is ensured.
In some embodiments, step 306 further comprises: and obtaining the individual constraint value according to the sensor optimization constraint model, and when the individual constraint value does not meet the constraint target parameter requirement, setting the individual fitness value as a larger value or an upper limit value so that the individual cannot enter the group extremum and the individual extremum.
In this embodiment, the fitness value is a value of each function in the optimization target model, and under the requirement of the constraint target parameter of the optimization constraint model, when the constraint value calculated by the individual at this moment does not satisfy the set constraint target parameter, the fitness value (i.e., the optimization target) calculated by the individual at this moment can be given a larger value or an upper limit value. The constraint processing strategy is adopted because in the individual updating process, some individuals may not meet the constraint condition, but the optimization targets calculated by the individuals are inversely small, which affects the judgment. Because the optimization targets are minimum values, through the constraint processing strategy, when the optimization target function of the individual not meeting the constraint condition is endowed with a larger value, the extreme value of the selected group and the extreme value of the individual can be avoided, and the competitiveness of the individual and other individuals meeting the constraint condition is reduced.
In some embodiments, step 307 further comprises: for a newly generated alternative individual, if the alternative individual is dominated by any individual in the external archive, the alternative individual is rejected; if any individual of the external archive can not dominate the alternative individual, adding the alternative individual into the external archive; if the storage individual in the external archive is dominated by the alternative individual, rejecting the storage individual in the external archive; when the external archive size reaches a maximum, a crowdedness policy is employed to limit the external archive size.
Multi-objective evolutionary algorithms typically employ an external population to hold all non-dominated solutions that have been discovered, this external population is often referred to as external archiving. The loss of the found non-dominant solution can be prevented through external archiving, and therefore the convergence performance of the algorithm is guaranteed. Each iteration of the multi-objective particle swarm algorithm produces a set of non-inferior solutions, and therefore, during the operation of the algorithm, an external archive is applied to store the non-inferior solutions produced each generation. Initially the external archive is empty and as the iteration progresses, the non-inferior solutions produced by each generation are used to update the external archive. The adopted update strategy is as follows:
a) If the newly generated alternative individual is dominated by any individual in the external archive, the alternative individual is rejected;
b) If any individual of the external archive cannot dominate the alternate individual, it is added to the external archive.
c) If the stored individuals in the external archive are dominated by the alternative individuals, the stored individuals are rejected in the external archive.
d) When the external archive scale reaches the maximum value, a constraint policy is adopted to limit the external archive scale, the constraint policy includes a clustering policy and a congestion degree policy, and the congestion degree policy is adopted in this embodiment.
To verify the effectiveness of the method of the present invention, the embodiment of the present invention provides an example of performing optimal configuration on the sensor layout of a certain electromechanical product:
the system has 7 failure modes and can provide 10 sensor measuring points. The requirements of the fault detection rate and the fault isolation rate of the product are not lower than 95% and 90%, respectively, and the requirement of the false alarm rate is not higher than 10%. The fault-sensor relationship matrix of the product is shown in table 2, and it is assumed that the fault prior probability and the cost of each sensor are respectively shown in table 3 and table 4.
TABLE 2 product failure mode and sensor relationship matrix
TABLE 3 Prior probability of failure
TABLE 4 sensor cost and failure Rate
Considering the actual detection performance of the sensor as follows:whereinFor sensor failure rate, the post-failure-sensor relationship matrix characterizes the probability that a failure can be effectively detected by the sensor. The modified relationship matrix is shown in table 5:
TABLE 5 corrected product failure mode and sensor relationship matrix
The algorithm sets the parameters as follows: the population size N =60; the number of sensor types n =10; inertial weight parameter=1.2; minimum value=0.4; maximum number of iterations M =100; learning factor(ii) a Meshing n =20; the unified mutation operator percentage u =0.5; external archivingSize O =100. The adopted fitness value evaluation functions are two optimization targets of sensor failure rate and sensor cost respectively.
The total of 4 sets of external archives (black circle parts, overlapping) obtained through 100 iterations are calculated, pareto fronts of the external archives are shown in fig. 3, and the Pareto fronts are mutually independent, and can be used as sensor optimization layout schemes, and it needs to be further explained that two different types of sensor sets, namely, an optimization result 3 and an optimization result 4, have the same cost and total failure rate. The red circle represents the solution set that meets the constraints of fault detection rate and the like but is dominated by Pareto. Therefore, in the actual selection process, the expert can further comprehensively consider the solution set in the external archive to preferentially select the target of interest.
Further, each group of sensor selection sets is analyzed in sequence according to the sensor cost:
(1) And an optimization result 1: <xnotran> x = [0,0,0,1,1,1,0,0,1,1]; </xnotran>
And calculating to obtain: the number of sensors: 5; total cost of the sensor: 22000; total failure rate of sensors: 0.121; fd =99.91%; fi =92.89%; fa =7.43%.
(2) And an optimization result 2: x = [0,0,0,1,1,0,0,1,1 ];
and calculating to obtain: the number of sensors: 5; total cost of the sensor: 23000; total failure rate of sensors: 0.081; fd =99.93%; fi =94.57%; fa =6.29%.
(3) And an optimization result 3: <xnotran> x = [0,1,1,1,0,0,0,1,1,1]; </xnotran>
And calculating to obtain: the number of sensors: 6; total cost of the sensor: 37000; total failure rate of sensors: 0.062; fd =99.93%; fi =97.08%; fa =6.18%.
(4) And an optimization result 4: x = [1,0,1, 0, 1];
and calculating to obtain: the number of sensors: 6; total cost of the sensor: 37000; total failure rate of sensors: 0.062; fd =99.93%; fi =97.08%; fa =6.18%.
The optimized layout was compared to the unoptimized full sensor configuration, with the results shown in table 6:
TABLE 6 comparison of post-optimization results with pre-optimization experimental results
The comparison shows that the unoptimized configuration fault isolation rate does not meet the requirements, and the adopted improved discrete multi-target particle swarm algorithm can self-adaptively screen out a group of Pareto optimal solution sets under the condition that the requirements of the fault isolation rate, the detection rate and the fault false alarm rate are met, and the selection combinations in the solution sets are not dominant, and can be used as the basis of sensor layout. The number and the cost of the optimized sensors are greatly reduced, the lower total failure rate of the sensors is guaranteed, and the reliability of the monitoring system is remarkably improved. If the structural difficulty of the sensor mounting part is further considered, deep analysis can be additionally carried out on each scheme, and a configuration scheme with more practical applicability is selected.
Other embodiments of the invention will be apparent to those skilled in the art from consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the invention following, in general, the principles of the invention and including such departures from the present disclosure as come within known or customary practice within the art to which the invention pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the invention being indicated by the following claims.
It will be understood that the invention is not limited to the precise arrangements described above and shown in the drawings and that various modifications and changes may be made without departing from the scope thereof. The scope of the invention is limited only by the appended claims.
Claims (8)
1. An electromechanical product sensor optimal configuration method, wherein the electromechanical product comprises one or more sensors, and the sensors are used for monitoring the state of the electromechanical product, and the method comprises the following steps:
step 10: acquiring corresponding relations between different sensor sets and a fault mode set by analyzing fault modes and fault mechanisms of electromechanical products, and constructing a fault mode and sensor relation matrix;
step 20: constructing a sensor configuration optimization model, wherein the sensor configuration optimization model comprises a sensor optimization target model and a sensor optimization constraint model;
step 30: solving a sensor configuration optimization model based on a discrete multi-target particle swarm algorithm to obtain an optimal sensor set as the sensor configuration of the electromechanical product;
the step 10 further comprises:
the fault mode and sensor relation matrix is a matrix with m rows and n columnsThe rows of the matrix D represent the m failure modes of the systemThe columns of the matrix D represent the n sensors of the electromechanical product,The value of the compound is 0 or 1,to representCan detect,To representIs unable to detect;
The sensor optimization objective model requires a minimum number of sensors, a minimum sensor cost, and a minimum probability of sensor failure in a sensor set, and includes:
the sensor optimization constraint model requires a sensor set to meet preset conditions of fault coverage capacity, fault detection rate, fault isolation rate and fault false alarm rate; the sensor optimization constraint model comprises:
The fault detection rate is as follows: the ratio of the total probability of a correctly detected failure mode to the total probability of the failure mode, i.e.
Wherein the content of the first and second substances,as a failure modeThe probability of the occurrence of the event is,for the purpose of the corrected detection capability of the sensor,;
fault isolation rate: the ratio of the probability of correct isolation to component level failure to the total probability of detected failure mode, i.e.
Wherein, I is a component fault mode set capable of fault isolation;
fault false alarm rate: the ratio of the number of false alarm faults generated by the diagnosis to the total number of faults detected, i.e.
The step 30 comprises:
step 301: initializing calculation parameters in the discrete multi-target particle swarm algorithm, wherein the calculation parameters comprise a population scale N, a sensor variety number N and a learning factor、Inertial weight parameter、Setting a constraint target parameter according to the external archiving space size O, the variation degree u and the maximum iteration number M;
step 302: generating a group by using a chaotic initialization strategy according to the population size N and the sensor type number NDimension initial matrixDiscretizing the population to initialize individual positions in the population and randomly initializing individual speed to obtain an initial populationWherein, the individual speed represents the probability that the individual position takes 1 or 0, the individual position is 1 to represent that the sensor is installed at the position, the individual position is 0 to represent that the sensor is not installed at the position, and the initial iteration number is set;
Step 303: optimizing target model calculations from sensorsSelecting the population according to the individual optimal strategy according to the fitness value of each individualAccording to the global optimum strategy, selecting the individual extreme valueThe group extremum of (1);
step 304: adjusting the inertia weight of the discrete multi-target particle swarm algorithm according to a self-adaptive strategy, and then updating the individual speed and position in the swarm;
step 305: according to a mutation strategy, forThe middle individual performs mutation operation with a certain probability to change the position of the particles, and calculates the mutation according to an optimization objective functionA fitness value for each individual of (a);
step 306: modifying the fitness value of the population according to the sensor optimization constraint model;
step 307: updating the external archive, and adjusting the scale of the external archive by adopting a congestion degree strategy;
step 308: selecting populations according to individual optimal strategiesAccording to the global optimum strategy, selecting the individual extreme valueThe group extremum of (1);
step 309: if the iteration stop condition is satisfiedOutputting an optimal sensor configuration set, namely a Pareto optimal leading edge, and selecting a proper set x from the optimal solution set as optimal configuration according to application requirements; if the iteration stop condition is not satisfied, updating the counterAnd returning to step 304;
the individual optimal strategy comprises:
if the current position of the individual dominates the extreme value of the individual, updating the extreme value of the individual into the current position of the individual;
if the individual extremum dominates the individual present position, the individual extremum remains unchanged;
if the current position of the individual and the extreme value of the individual are not mutually dominant, the extreme value of the individual is randomly selected to be updated to the current position of the individual or kept unchanged;
the global optimal strategy comprises:
and selecting the individuals as the group extremum of the current individuals from an external archive by roulette or a tournament selection mode, wherein the external archive comprises a group of non-inferior solution sets, the group of non-inferior solution sets are individuals with better quality in the group, and each individual is independent of each other.
2. The method for optimized configuration of electromechanical product sensors according to claim 1, wherein said step 302 further comprises:
according to the initial valueComputing chaotic sequences through logistic mappingTo obtain an initial matrixThe logistic map is:wherein, in the step (A),control parameters;
3. The method for optimized configuration of electromechanical product sensors according to claim 1, wherein said step 304 further comprises:
wherein, the first and the second end of the pipe are connected with each other,、is a random number, subscriptRepresenting the ith individual type d sensor,the individual velocities of the t +1 th iteration and the t-th iteration respectively,individual positions, functions, of the t +1 th and t-th iterations, respectively,Is composed ofA random number in between.
4. The method for optimized configuration of electromechanical product sensors according to claim 1, wherein said step 305 further comprises:
as the number of iterations t increases, the proportion of individuals involved in the variation decreases non-linearly.
5. The method for optimized configuration of electromechanical product sensors according to claim 1, wherein said step 306 further comprises:
and obtaining the individual constraint value according to the sensor optimization constraint model, and when the individual constraint value does not meet the constraint target parameter requirement, setting the individual fitness value as a larger value or an upper limit value so that the individual cannot enter the group extremum and the individual extremum.
6. The method for optimized configuration of electromechanical product sensors according to claim 1, wherein said step 307 further comprises:
for a newly generated candidate individual to be selected,
if the alternative individual is dominated by any individual in the external archive, the alternative individual is rejected;
if any individual of the external archive can not dominate the alternative individual, adding the alternative individual into the external archive;
if the storage individual in the external archive is dominated by the alternative individual, rejecting the storage individual in the external archive;
when the external archive size reaches a maximum, a crowdedness policy is employed to limit the external archive size.
7. A computer device, comprising:
a memory to store instructions; and
a processor for invoking the memory-stored instructions to perform the electromechanical product sensor optimization configuration method of any of claims 1-6.
8. A computer-readable storage medium storing instructions that, when executed by a processor, perform the method of optimally configuring electromechanical product sensors according to any one of claims 1 to 6.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211421113.XA CN115616919B (en) | 2022-11-15 | 2022-11-15 | Electromechanical product sensor optimal configuration method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211421113.XA CN115616919B (en) | 2022-11-15 | 2022-11-15 | Electromechanical product sensor optimal configuration method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN115616919A CN115616919A (en) | 2023-01-17 |
CN115616919B true CN115616919B (en) | 2023-03-17 |
Family
ID=84879513
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202211421113.XA Active CN115616919B (en) | 2022-11-15 | 2022-11-15 | Electromechanical product sensor optimal configuration method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115616919B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117713598A (en) * | 2023-08-07 | 2024-03-15 | 荣耀终端有限公司 | Motor control strategy generation method, driving method and electronic equipment |
Family Cites Families (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104571088B (en) * | 2014-12-26 | 2018-01-05 | 北京控制工程研究所 | Satellite control system Multipurpose Optimal Method based on fault diagnosability constraint |
CN105678423B (en) * | 2016-01-28 | 2019-09-24 | 西北工业大学 | Fault diagnosis system Optimum sensor placement method based on D-M (Determiner-Measure) construction model |
CN112036540B (en) * | 2020-09-07 | 2023-11-28 | 哈尔滨工程大学 | Sensor number optimization method based on double-population hybrid artificial bee colony algorithm |
EP4320520A1 (en) * | 2021-04-07 | 2024-02-14 | Ohio State Innovation Foundation | Propagation-based fault detection and discrimination method and the optimization of sensor deployment |
CN113204915B (en) * | 2021-04-14 | 2023-05-12 | 中国人民解放军空军工程大学 | PHM design method based on CPS |
-
2022
- 2022-11-15 CN CN202211421113.XA patent/CN115616919B/en active Active
Also Published As
Publication number | Publication date |
---|---|
CN115616919A (en) | 2023-01-17 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109657797B (en) | Fault diagnosis capability analysis method based on hybrid diagnosis Bayesian network | |
Svärd et al. | Automated design of an FDI-system for the wind turbine benchmark | |
CN106845526B (en) | A kind of relevant parameter Fault Classification based on the analysis of big data Fusion of Clustering | |
KR102320706B1 (en) | Method for setting model threshold of facility monitoring system | |
CN111597760B (en) | Method for obtaining gas path parameter deviation value under small sample condition | |
CN115616919B (en) | Electromechanical product sensor optimal configuration method | |
CN112231980B (en) | Engine life prediction method, storage medium and computing device | |
CN107909194B (en) | System-level testability design multi-objective optimization method | |
CN112000015A (en) | Intelligent BIT design method for heavy-duty gas turbine control system controller module based on LSTM and bio-excitation neural network | |
JP2013196698A (en) | System monitoring | |
CN104539601A (en) | Reliability analysis method and system for dynamic network attack process | |
CN109240276B (en) | Multi-block PCA fault monitoring method based on fault sensitive principal component selection | |
CN114925536A (en) | Airborne system PHM testability modeling and diagnosis strategy optimization method and device | |
US20050144537A1 (en) | Method to use a receiver operator characteristics curve for model comparison in machine condition monitoring | |
CN110941902A (en) | Lightning stroke fault early warning method and system for power transmission line | |
Tsai et al. | Sensor abnormal detection and recovery using machine learning for IoT sensing systems | |
CN112883651A (en) | System-level testability design multi-objective optimization method based on improved PBI method | |
CN111079348B (en) | Method and device for detecting slowly-varying signal | |
CN106886620B (en) | Spacecraft test resource optimal configuration method | |
KR102320707B1 (en) | Method for classifiying facility fault of facility monitoring system | |
CN113884807A (en) | Power distribution network fault prediction method based on random forest and multi-layer architecture clustering | |
Li et al. | Remaining life prediction of cores based on data-driven and physical modeling methods | |
Yin et al. | A network security situation assessment model based on BP neural network optimized by DS evidence theory | |
CN111240881A (en) | Web service combination abnormity diagnosis method based on Bayesian network | |
Valdes et al. | Towards conservative helicopter loads prediction using computational intelligence techniques |
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 |