CN-121480994-B - Method and system for evolution of alarm rules in production process of operation context awareness

CN121480994BCN 121480994 BCN121480994 BCN 121480994BCN-121480994-B

Abstract

The invention relates to the technical field of industrial automation and intelligent manufacturing, and particularly discloses an evolution method and system of an alarm rule in a production process with operation context sensing. By fusing multi-source real-time and historical data, a dynamic deduction model is constructed, the abnormal triggering probability of each production unit is predicted in a prospective mode, and a vulnerability unit candidate set is generated. And screening out a core intervention unit based on the virtual simulation and the context weighted comprehensive production loss evaluation. And determining the minimum effective intervention combination capable of controlling the residual risk and a matched optimization strategy thereof through enumeration combination, three-dimensional priority ordering and nested process scheduling optimization verification cycle. And (3) putting the combination into an online environment for verification, curing the combination into a formal alarm rule if the combination is successful, analyzing the situation deviation and feeding back an updated model if the combination is failed, and forming a self-evolution mechanism of perception-evaluation-optimization-verification-evolution, thereby realizing continuous self-adaption of the alarm rule to a dynamic production environment.

Inventors

XIE YIXIANG
ZHANG HAIQIANG
ZHANG LIJUAN

Assignees

浙江常青树信息技术有限责任公司
青岛凯兴电子技术有限公司
浙江蔚蓝芯片科技有限公司

Dates

Publication Date: 20260508
Application Date: 20260107

Claims (10)

1. A method of operating context aware production process alarm rule evolution comprising: Based on the current production running context data, a dynamic deduction model is constructed, abnormal triggering probability and time sequence distribution of the equipment in a future time window are predicted, and a vulnerability unit candidate set is generated; Performing simulation analysis on units in the vulnerability unit candidate set, evaluating comprehensive production loss caused by failure of each unit, and screening out a core intervention unit according to a predefined economic impact factor and a loss threshold; Wherein, a comprehensive production loss index of the comprehensive production loss is defined For quantizing a set of current cells Expected losses at the overall production system level in the event of a failure or disconnection, the integrated production loss index is a contextually weighted multi-objective loss aggregation: And wherein: Representing a set of causes The expected delivery delay loss due to the failure, Indicating that an additional energy consumption loss is expected, Indicating an expected degradation loss of product quality, , , Respectively in the current context The loss weight coefficient dynamically determined reflects the relative importance of different production targets in the current scene; Through nested optimization and verification circulation, the control strategy is reset for each priority combination and the comprehensive loss of the system is evaluated until the minimum effective intervention combination which can enable the comprehensive loss of the system to meet the preset requirement is screened out; wherein, for a selected combination, in a nested optimization and verification loop The goal of the process schedule optimizer is to minimize the remaining set of units The optimization variables include control parameter sets Standby path enable flag Buffer policy parameters The optimization problem is formalized as: ; Wherein Represents the first A process, equipment or safety constraint condition, To obtain new system configuration after optimization for constraint total number and re-evaluate ; The method comprises the steps of carrying out online operation verification on a minimum effective intervention combination, packaging a current context, a unit combination and an optimization strategy into a formal alarm rule and storing the formal alarm rule if verification is successful, analyzing the difference between an actual operation context and a dynamic deduction model assumption if verification is failed, feeding back a difference factor to update the dynamic deduction model and a unit set, and re-entering a screening and optimizing process until the verification is successful and the formal alarm rule is generated.
2. The method of claim 1, wherein the operation context data comprises a human configuration matrix, an order queue, an order priority weight, a real-time energy cost coefficient acquired by a manufacturing execution system, an environment parameter vector acquired by an internet of things sensor network, an equipment maintenance status and health status code acquired by an equipment management platform, and a real-time equipment sensing information stream acquired by an industrial bus or an internet of things gateway.
3. The method for evolution of production process alarm rules based on operation context awareness according to claim 1, wherein the dynamic deduction model adopts a hybrid architecture based on fusion of physical mechanism and data driving, and comprises the following steps: Simulating a process disturbance path based on material balance and energy balance; predicting the probability of the health state of the device based on survival analysis or depth time sequence prediction; And fusing the process disturbance simulation result, the equipment health probability prediction and the context constraint weight, and calculating the comprehensive trigger probability and time distribution of each unit for triggering the system-level alarm in a future time window.
4. The method of claim 1, wherein the integrated production loss is calculated by context weighted multi-objective loss aggregation, the multi-objective loss including expected delivery delay loss due to failure, expected additional energy consumption loss, and expected product quality degradation loss; Wherein each loss term dynamically determines a weight coefficient according to the current running context.
5. The method of claim 1, wherein the prioritizing is based on a three-dimensional scoring function, comprising: scale inverse index And (3) calculating: the scale reverse index encourages the selection of a combination with a small number of intervention units, and accords with the minimum intervention principle; Risk index And (3) calculating: Wherein Is a combination of All units in the prediction time window The joint probability is kept normal at the same time; Context fitness index And (3) calculating: Wherein Is an evaluation combination With the current running context Is a matching function of (a); Based on scale inverse index Risk index And context fitness index Calculate the composite priority score for each combination : Wherein, the method comprises the steps of, 、、 Respectively scale inverse index Risk index And context fitness index An assigned weight; and ordering all the combinations according to the comprehensive priority score from high to low to form a priority queue.
6. The run context aware production process alarm rule evolution method of claim 1, wherein in the nested optimization and validation loop, for each candidate combination, the control parameter set is adjusted by the process schedule optimizer, the alternate path flag and the buffer strategy parameters are enabled to minimize the overall production loss of the remaining set of units in the simulation environment, and re-evaluate if the loss is below the dynamic tolerable loss threshold; And (3) performing virtual island disconnection simulation on the system configuration after the nested optimization verification cycle in a digital twin environment, recalculating the comprehensive production loss index caused by the residual unit set, and comparing the comprehensive production loss index with a tolerable loss threshold dynamically updated based on the current context to determine whether verification is passed.
7. The method for generating alarm rules for running context aware production process according to claim 1, wherein the stage of on-line running verification deploys a minimum effective intervention combination and its matched optimization strategy as temporary alarm rules, monitors actual production indexes during a preset verification period, encapsulates current context, unit combination and optimization strategy as formal alarm rules if the indexes meet expectations, stores the formal alarm rules into a rule base, and triggers a situation deviation analysis flow if verification fails.
8. The method of claim 7, wherein the analysis of the context deviation after verification failure comprises: comparing the actual running context data sequence with the dynamic deduction model assumption context; Identifying a key difference factor as a model misalignment cause; and feeding the difference factors back to the dynamic deduction model for updating, and re-triggering the complete flow from model construction to online verification.
9. The method for evolution of an alarm rule in a production process based on operation context awareness according to claim 1, further comprising recording operation context characteristics, unit combination states, optimization strategy parameters and verification period performance data corresponding to a formal alarm rule when the rule is generated, forming a reusable rule knowledge item, and supporting subsequent rule matching and scenario self-adaptation calling.
10. An early warning system for a method of operating a context aware production process alarm rule evolution according to any one of claims 1 to 9, the system comprising: The system comprises an operation context sensing and model construction module, a dynamic deduction model, a first run vulnerability unit candidate set, a data processing module and a data processing module, wherein the operation context sensing and model construction module is configured to acquire multidimensional operation context data of a current production system; The simulation evaluation and core unit screening module is configured to perform virtual island disconnection simulation on each unit in the vulnerability unit candidate set, dynamically weight-calculate a comprehensive production loss index caused by unit failure based on the current running context in the simulation process, screen out core intervention units with the comprehensive production loss exceeding the threshold value and needing to be intervened according to the predefined tolerable loss threshold value, and form a core intervention unit set; the combined optimization and verification decision module is configured to perform combined space exploration on the core intervention unit set, enumerate all possible non-empty unit subsets, and perform multidimensional priority ranking according to the combined scale, the combined risk probability and the adaptation degree with the current context; performing nested optimization and verification cycles on each candidate combination according to the priority order, wherein the system control strategy is reset and optimized for the current candidate combination through a process scheduling optimizer for the rest unit set; the comprehensive production loss of the optimized system facing the risk of the residual units is evaluated in a digital twin environment, and the process is iteratively executed until the minimum effective intervention combination and the matched optimization strategy which can lead the comprehensive loss of the system to meet the preset requirement are screened out; The on-line verification and rule evolution module is configured to deploy the minimum effective intervention combination and matched optimization strategy to an actual production system for on-line operation verification, package the current operation context characteristics, unit combination and optimization strategy into a formal alarm rule and store the formal alarm rule into a rule base if verification is successful, analyze the difference between the actual operation context and the context used in a model construction stage if verification fails, and feed back the identified key difference factors to the operation context sensing and model construction module so as to trigger the update of a dynamic deduction model and the re-execution of a subsequent module until the formal alarm rule with successful verification is generated.

Description

Method and system for evolution of alarm rules in production process of operation context awareness Technical Field The invention relates to the technical field of industrial automation and intelligent manufacturing, in particular to a method and a system for evolution of an alarm rule in a production process with operation context sensing. Background In the modern industrial production process, the alarm system is an important tool for guaranteeing production safety and stable operation. Conventional alarm rules are typically based on fixed threshold or static logic settings, such as employing statistical process control methods or empirical rules based on historical fault data. Although the method can identify the abnormality to a certain extent, the limitation of the method is increasingly prominent, firstly, the fixed threshold cannot adapt to dynamic fluctuation of multidimensional operation contexts such as raw material change, equipment aging, order priority adjustment and the like in the production process, false alarm or missing alarm is easy to cause, and secondly, the existing alarm rules lack quantitative evaluation on the whole influence of the production system, alarm storm or submerged critical alarm is often caused, the load of operators is increased, and the real risk is delayed. In recent years, with the development of industrial big data and artificial intelligence technology, part of research is beginning to introduce a data driving method for alarm optimization, such as an abnormality detection model based on machine learning. However, the method still has obvious defects that on one hand, most models only pay attention to single equipment or technological parameters, the cooperative influence of coupling effect among equipment in a production system and a global production target is not fully considered, and on the other hand, the existing method generally lacks an evolution mechanism, rules are difficult to self-adaptively adjust along with the change of a production environment once deployed, and the alarm effect is attenuated along with time. In addition, the traditional method is always in a simple mode of alarm-manual treatment in an alarm response strategy, is not optimized with a process scheduling and control system, cannot provide a system-level self-adaptive adjustment scheme before or at the same time of alarm triggering, and is difficult to realize real intelligent early warning and self-healing control. In summary, the existing alarm technology has significant defects in the aspects of context awareness capability, system-level influence evaluation, optimization, self-adaptive evolution and the like, and restricts the effective application of the alarm technology in complex and dynamic production environments. Therefore, it is needed to propose an intelligent alarm rule generation method capable of perceiving an operation context in real time, quantitatively evaluating an alarm influence, and having self-optimization and continuous evolution capabilities. Disclosure of Invention In order to overcome the problems of context awareness, system-level influence evaluation, optimization and self-adaptive evolution of the existing alarm system in the background art, the invention provides a production process alarm rule evolution method and system for running context awareness. The technical scheme of the application is as follows: According to an aspect of the present application, there is provided a method of operating context-aware production process alarm rule evolution, comprising: Based on the current production running context data, a dynamic deduction model is constructed, abnormal triggering probability and time sequence distribution of the equipment in a future time window are predicted, and a vulnerability unit candidate set is generated; Performing simulation analysis on units in the vulnerability unit candidate set, evaluating comprehensive production loss caused by failure of each unit, and screening out a core intervention unit according to a predefined economic impact factor and a loss threshold; Through nested optimization and verification circulation, the control strategy is reset for each priority combination and the comprehensive loss of the system is evaluated until the minimum effective intervention combination which can enable the comprehensive loss of the system to meet the preset requirement is screened out; The method comprises the steps of carrying out online operation verification on a minimum effective intervention combination, packaging a current context, a unit combination and an optimization strategy into a formal alarm rule and storing the formal alarm rule if verification is successful, analyzing the difference between an actual operation context and a dynamic deduction model assumption if verification is failed, feeding back a difference factor to update the dynamic deduction model and a unit set, and re-entering a screening and optimizing proce