Embodiment
Figure 1 shows that a kind of cloud service platform of the safety in production towards industrial and mining enterprises, be integrated with the functions such as the real time data processing of support safety in production service cloud, production safety management business cooperation and system management, realize the Real-Time Monitoring to enterprise's production safety and early warning.
Specifically, the described safety production cloud service platform system towards industrial and mining enterprises, comprise secure cloud service platform portal sub-system, subsystem developed by system management and related tool collection, application service layer subsystem, virtual resource layer subsystem, platform service supporting layer subsystem, access and adaptation layer subsystem, security service resource subsystem, infrastructure services supporting layer subsystem.
Described secure cloud service platform portal sub-system is responsible for realizing information inquiry and management to outside based on web2.0 technology, comprise government's supervision platform, enterprise's application platform, trade supervision platform, device data and the property tax data that government supervises enterprise operation are in real time responsible for by described government supervision platform, described enterprise application system is responsible for externally providing value added service and application query, and described trade supervision platform is used for monitoring and the feedback of product quality.
Described system management and related tool collection are developed subsystem and are comprised the exploitation of cloud service instrument, deployment, monitoring, safety management, log management and configuration.
Described application service layer, comprise mass data processing system and business cooperation operating system, wherein said mass data processing system is responsible for, to data acquisition, Data Integration, data management and provide standardized service interface to provide data flow for secure cloud service platform door, and described business cooperation operating system is responsible for building Mathematical Modeling to collaborative task and cross-domain task, be specially and business datum is gathered, the service content that cloud platform is provided and the business datum of collection are integrated, the development of prediction follow-up business, described cross-domain task specifically comprises the management and supervision in different business field, work between the different departments of enterprise is realized task coordinate.
Described virtual resource layer, mainly for Cloud Server provide virtual after digital resource, specifically comprise knowledge services resource pool, produce Service Source pond, data message resource pool, described virtual resource layer includes these original provided knowledge services, produces the data message resource of service, pass through Intel Virtualization Technology, in the middle of this partial data content integration to application service layer, at any time can by described mass data processing unit and business cooperation cell call.
Described access and adaptation layer subsystem, mainly from relevant criterion and verifying test system, obtain data resource, comprise security service resource in addition, specifically comprise production equipment data, standardization fundamental norms, laws and regulations, safety management system, accident source historical data, educational training knowledge, safety in production drop into, organization accesses from third party's service adapter with responsible, hidden danger inspection message.
Described platform service supporting layer is to above-mentioned application service layer, virtual resource layer, access Adaptation Layer Service, there is provided cloud service to manage and support engine, transaction collaborative logic$$$$ engine, knowledge accumulating and classification engine, described cloud service manages with support engine as described application service layer provides cloud service to register, issues, nullifies, cloud service search, scheduling, combination, cloud service performs and monitoring, described transaction collaborative logic$$$$ engine provides process management for business cooperation, cost accounting, credit evaluation, described knowledge accumulating and classification engine provide industry multiple resource to disperse knowledge acquisition for business cooperation, domain knowledge modeling, domain knowledge assembles classification, and for other responsible operation management in described platform service supporting layer, operation management, terminal software is developed, the module of platform development instrument is described virtual resource layer and access and Adaptation Layer Service, many tenants service is responsible in wherein said operation management, order management, delivery management, payment management, user management, integration managing, safety management is responsible in described operation management, performance management and optimization, system configuration, mass data is fault-tolerant to be managed with confidence level, described terminal software exploitation, comprise the fusion management of heat transfer agent, Service Source graphic interface and pervasive man-machine interaction instrument, and platform development instrument, described platform service supporting layer all passes through cloud computing by infrastructure services supporting layer, cloud network, cloud stores unified support.
Based on industrial and mining enterprises' real-time data processing system of cloud computing, as shown in Figure 2, be made up of reliable Networks of Fiber Communications five major part of data acquisition equipment, Data Integration equipment, data management apparatus, standardized service interface equipment and high speed, form the industrial and mining enterprises' real time data integrated processing system based on cloud computing
The network configuration of this data integrated system shown in data acquisition equipment: Fig. 2, wireless senser cluster is responsible for the data gathering industrial and mineral scene, and transfer data to industrial and mineral Data Integration equipment, on described Data Integration equipment, standard configuration HTTP Proxy module proxy server and cloud management server, both all uses ubuntu server blade plate, as the node accepting management, every platform node server has all installed node manager, the major function of this manager realizes the management to KVM virtual machine, comprise: the control command of (1) reception cloud manager carries out the deployment of KVM virtual machine, the operations such as start and stop, (2) monitor the availability of local KVM virtual machine, when local KVM virtual machine is made mistakes unavailable, attempt starting the identical KVM virtual machine instance of another one, (3) behaviour in service of each KVM resources of virtual machine local of monitoring in real time, and the real time resources service condition of KVM virtual machine is sent to cloud manager, (4) extended operation of KVM virtual machine is automatically performed according to the resource service condition that virtual machine is real-time, (5) receive the virtual machine (vm) migration instruction from cloud manager, meanwhile, every platform node server all also will dispose multiple KVM virtual machine, and dispose node server in each virtual machine.All above-mentioned node servers all share same network data store.And this locality of each node server stores and is only used for buffer memory each virtual machine self-operating data, and the image file of virtual machine and embodiment all can be buffered in shared network data storage, the migration operation of high availability and virtual machine more easily can be supported like this.In addition, the data of industrial and mineral data acquisition, also can be randomly distributed in network memory by the control of cloud Control Server and proxy server, and are worked in coordination with and cross-domain task cooperation by the application software finishing service on each node server.Shown in Fig. 3 is be provided with wireless sensor data to gather cluster, described wireless sensor data harvester is according to application component function needs in system, data or analog interface is passed through with by industrial and mineral data equipment, wired or wireless interface connects, and frequency acquisition is set to function need highest frequency and precision carry out data sampling, the data buffer storage gathered is in the memory of self, and data compression is transmitted into remote monitoring center through wireless transmit/receive units after becoming consolidation form.Wherein also can carry out synchronous or asynchronous communication between each wireless sensor devices, composition data gathers cluster.The fundamental diagram of wireless data collection device shown in Fig. 4, comprises; Core processing unit, touch-screen, camera, microphone, clock, memory, power management, interface circuit, sensor circuit form; Described core processing unit comprises embedded controller, audio interface circuit, video interface circuit, touch screen interface circuit, multi-path serial interface, hi-speed USB interface, wireless communication module, debugging interface and sensor interface circuitry; Described touch screen interface circuit, memory, audio interface circuit, multi-path serial interface, wireless communication module, debugging interface and hi-speed USB interface are bi-directionally connected with embedded controller respectively; Described touch-screen and touch screen interface circuit are bi-directionally connected; Described microphone and camera connect the input of audio interface circuit and video interface circuit respectively; Described multi-path serial interface and hi-speed USB interface connect industrial and mineral data equipment respectively; The respective input of the output termination embedded controller of described video interface circuit; Embedded controller is connected with internet; Power supply management device is used for providing electric energy, and power acquisition photovoltaic solar panel and the combined form of batteries, can ensure the long-term work that equipment can be effectively stable, also can avoid laying power line; Wherein use the charging-discharging controller that MC3063 chip is formed, to provide stable voltage, acceleration, pressure, temperature, optical sensor and load signal transducer is provided with in sensor unit, realize the calculating to displacement by acceleration signal and load signal, monitored by pressure and temperature sensors towards ambient.Core processing unit is used for collection computing and the storage of sensor signal, and wireless communication module realizes reception and the transmission of wireless signal; Embedded controller can adopt CC2530ZigBee chip, to realize transmission and the reception of the 2.4GHzISM band signal based on ZigBee easily, such as when being used on trepan, trepan comprises motor, quadric chain, walking beam, support, pumping rod; Motor drives walking beam motion by quadric chain, and walking beam drives and moves up and down with pumping rod, and the acceleration signal transducer on data acquisition unit and load signal transducer just can be arranged in pumping rod, to gather the kinematic parameter of pumping rod.In addition, shown data acquisition unit can also be arranged on industrial and mineral electrical appliance equipment, for real-time video, audio frequency, digital monitoring running parameter, obtain operating state, and the data of collection are uploaded onto the server by multi-protocol gateway, server is analyzed, to draw its running status according to the working condition of data to relevant device gathered.
Data storage layer: be provided with cloud data storage pool, for in industrial and mining enterprises from the mass data in various different pieces of information source, it is processed to the subject data of the rear generation unified interface corresponding with different business, and these subject datas are stored in distributed file system, when there being task requests, according to different task requests, multinode, the parallel computation of multitask and analysis are carried out to the data stored in described distributed file system, analysis result is represented accordingly according to different application.As shown in Figure 5, the cloud storage pool of data processing provided by the invention, draw together at least one cloud storage administration node, at least one cloud memory space and at least one virtual unit, cloud storage administration node, cloud memory space and virtual unit form privately owned cloud.In embodiment as shown in Figure 4, this cloud storage architecture comprises two cloud memory nodes, three cloud memory spaces and multiple virtual unit, and three cloud memory spaces and multiple virtual unit form a privately owned cloud.Cloud storage administration node a and cloud storage administration node b is interconnected, mutually take between two cloud storage administration nodes, between two cloud storage administration nodes can equally loaded and wherein a cloud storage administration node take over each other when breaking down, to ensure the unfailing performance during system cloud gray model based on this cloud storage architecture.Wherein, two cloud storage administration nodes can manage three cloud memory spaces and multiple virtual unit respectively, its management comprise to three cloud memory spaces and multiple virtual unit carry out newly-built, delete and configure, to carry out system backup, recovery and dilatation.Data directory is had in each cloud storage administration node, data directory is for recording the relevant information of cloud memory space and virtual unit, and cloud storage administration node finds corresponding cloud memory space and virtual unit by the related data in the data directory of its inside.In addition, each cloud storage administration node is provided with unified application program access entrance, and this application program access entrance is application programming interfaces, application program/serve by calling this application program access entrance access cloud memory space.Each virtual unit in Fig. 4 all can be mapped as raw device, memory field, file system or a memory file system etc. in operating system, and the disk array of virtual management physical memory, built-in disk and various interface, agreement.In addition, because multiple virtual unit c1, virtual unit c2 ..., virtual unit cn are the identical virtual unit of characteristic, so, in the present embodiment, virtual unit c1, virtual unit c2 ..., virtual unit cn can form a virtual unit group b, simplify the management to multiple virtual unit by this virtual unit group b.Multiple virtual unit can distinguish a memory space in a corresponding physical storage device or a physical storage device.In the present embodiment, virtual unit a1, virtual unit a2 ..., virtual unit an are connected with physical storage device a1, physical storage device a2 ..., physical storage device an respectively, thus by the data information memory from industrial and mineral collection in worksite in above-mentioned physical storage device.Outside application program/service a, b is connected with cloud storage administration node a, b in cloud storage architecture respectively.Owing to being provided with unified application program access entrance on cloud storage administration node, therefore, application program/serve the interface function by invokes application access entrance thus access corresponding cloud memory space.Application program/service can indicate access or the corresponding information accessed in any one cloud memory space by application program access entrance or manage cloud memory space and virtual unit.Application program/serve by the upper cloud storage administration node of invokes application access entrance connection, then application program/service can indicate the cloud memory space, filename, side-play amount, accessing operation etc. of access by API, the information connecting inner data dictionary that cloud storage administration node imports into according to these API completes accessing operation by finally operating to be assigned on one or more concrete physical storage device, returns access results finally by cloud storage administration node.
Data computation layer: corresponding to cloud data computing platform, this cloud data computing platform for call the Real-time Collection that stores in cloud data storage pool to data calculate its characteristic according to industrial and mineral system business public relation respectively, set up general data computational analysis model, as numerical value such as computing equipment performance, load capacity, operating efficiency, safe coefficient, environmental parameters, the data computation module of the virtual system on Node Controller can realize adding flexibly and setting, upgrade data analysis computing module, obtain corresponding calculated value.The structure of data center as shown in Figure 6.Comprise: it is interconnected by network adapter and proxy server and data network, internet that cloud controls the period of service, monitoring and control multiple memory node and on virtual machine, the task control module of master controller is responsible for carrying out unified allocation of resources management to downstream node controller, comprise interpolation, delete and the operation such as the phisical drive of any amount of readable data of Migration Control System and storage medium, administrative model is responsible for carrying out analyzing and processing to the load of collecting and resource using information, then transfers to controller to control.Computing node comprises the virtual machine of any amount, the resources of virtual machine that each intra-node comprises a Node Controller responsible node inside controls, the distribution management by synchronization of resource is responsible for by coordinated management engine, run application in virtual machine resource, if resource monitor, Network Performance Monitor, early warning monitoring device are at interior multiple Network Performance Monitors.For data statistics flow process under Network Performance Monitor introduction, as shown in Figure 7, real time load and monitoring resource model, according to sampled data type, select to preset device type, and then call the service behaviour data that the statistical model analyzing and processing for this equipment samples, generation equipment ideal Distribution data, model calculating is carried out in conjunction with actual samples data again by controller, export the data of actual allocated, judgment device resource service condition and load level thereof, and then instruct the distribution of main controlled node control system equipment resource, and for load monitoring module, as shown in Figure 8, in system cloud gray model, according to collected data message, the curve that real-time monitoring generates, observe the exceptional value appearance whether curve has catastrophe point or do not meet matched curve.If exist, then illustrate that Rush Hour appears in this application of current data center, proof system resource requirement needs to carry out larger variation.Now need to analyze system resource behaviour in service fast, whether can meet resource requirement when system resource reaches heap(ed) capacity.Because be outlier, current obtained parameter can not represent the relational model between overall load, performance and resource, if but not in time process can have a great impact the performance of system, so need to carry out analyzing and processing to outlier in time.If system resource Pooled resources meets current demand, then directly give cloud controller to process, the current outlier of quick solution, if resource does not meet current demand, just start alternate device resource, period uses simple linear regression model (LRM) to predict the operating load of next 5 minutes, and simple linear regression model (LRM) effectively can catch operating load Changing Pattern in time, even more complicated historical data also can conclude its load of prediction easily.The operating load of prediction assesses as the input of model the performance that device resource requirements needed for existing workload and system can reach.The factor of many complexity all can affect the performance of application program, the such as change etc. of environmental parameter, operating personnel's quantity, at this moment KCCA algorithm and remote related algorithm can be adopted to realize multi-variate statistical analysis modeling, analyze the impact that multiple influencing factor is brought systematic function simultaneously, real-time adjustment model parameter, generates desirable device resource allocation data.The distribute data calculated can by the standardized query interface on main controlled node or communication interface, initiatively carries out query manipulation or be passively pushed to user by user.Wherein need the data message of real-time collecting for those, need the data that upgrade in time, the service quality of guarantee data center.
Data access layer: cloud service access modules is provided with cloud service access interface, for the triggering according to application component, finds Cloud Server by fiber optic network, can obtain corresponding calculated value from cloud data computing platform in real time fast.The interface service that data access layer provides comprises the cloud service access interface that Various types of data server provides.
Above four layers are connected successively by light switched communication network that is reliable and high speed, and this communication looped network and power circuit tight association, the fibre circuit of high-speed communication is applied to all intelligent cells along power circuit, for enterprise provides powerful high speed information communication port.Compare the data acquisition in traditional industrial and mineral system, the data that data handling system of the present invention can expand high in the clouds store and computing function, and the concrete logic judging function based on data model, complete the unified monitoring of different functions by configuring different hardware devices or system, the equipment of these difference in functionalitys or system have separately independently data processing unit.And traditional real-time data acquisition and processing unit are made up of the multiple systems isolated, Data duplication collection and also data incomplete, observing and controlling, metering, protection, automatic safety device are realized by different hardware units locality.Its defect is that data acquisition repeats, and data are difficult to share, and system application function is difficult to effectively work in coordination with.The application function of the observing and controlling in the present invention, protection, metering, automatic safety device is all realized by the clustering form of functional unit; by high-speed network communication technology; really realize the unified process of data, calculate, share, be conducive to the reliability of raising system, the installation cost reducing system and maintenance cost.
The data message of safety in production enterprise practical productive accumulation is carried out unified management and storage by above-mentioned magnanimity information processing cloud platform, in addition, in order to provide the function such as prediction, abnormality detection based on data in production process, realize the safety in production of enterprise, managers also wish the abnormal conditions in the production of energy Timeliness coverage, find out reason, and propose counter-measure in time, keep normally carrying out of production.Normal and abnormal two kinds of situations are divided into the judgement of the condition of production, so it is classified as classification problem, adopt the good SVMs of classifying quality to carry out the abnormal judgement of the condition of production so long.By the achievement data of keeping the safety in production in analytical database, product quality, composition, practical productivity is selected to pass judgment on the whether abnormal supported data of the condition of production as index, box traction substation analysis and correlation analysis are carried out to support initial data, obtains influencing each other between data.Select data sample, because magnitude gap between variable is comparatively large, first need to carry out needing to carry out standardization to variable data,
i=1,2,3,......,n
Wherein, Vi is former variate-value, and μ is the mean value of former variate-value, and σ is former variable standard deviation.
After standardization, training and the test of SVMs decision model can be carried out, as shown in Figure 9, training sample set data are used to carry out model training, its training algorithm adopts SMO algorithm (SequentialMinimalOptimization, the minimum optimization of sequence), draw the support vector of its disaggregated model, calculate the parameter of decision function f (x) according to support vector.After training, the modified model SVMs that obtains is not optimum, and this is because some optimum configurations in initial reference template and algorithm can affect the result of training.By selecting the design parameter in layering core and selecting software metrics, the model that can be more optimized.The tree form data structure that the software module that input needs carry out predicting in the modified model SVMs trained is corresponding, obtain the output between [-1 ,+1], being greater than 0 if exported, there is not exception in the product quality condition of production; Otherwise output is less than 0 and means that the condition of production of product quality is abnormal.
Virtual resource layer described in Fig. 1 and also need to extract and produce relevant standard, knowledge, data etc. between access with adaptation layer and help to produce and decision-making, this just requires that cloud platform can search useful knowledge in a database.For object and the requirement of the intension of knowledge and characteristic, Knowledge Conversion, Immunity Operator (Immuneerator) is also incorporated in standard Evolutionary Programming Algorithm by we.Knowledge, Knowledge Conversion are combined with theory of immunity, the immune programming algorithm of knowledge based transfer realizes knowledge-based reasoning and finds.In the operating process of reality, immune algorithm is a kind of global optimization approach grown up on genetic algorithm basis, the problem that most genetic algorithm can solve, immune algorithm can both effectively solve and efficiency is better than genetic algorithm, the optimizing ability utilizing immune algorithm good can be accessed all nodes at virtual resource layer and wireless sensor network and be completed Data Collection and the minimum selecting paths scheme of total energy consumption, finally achieves the energy ezpenditure reducing wireless sensor network, the problem reducing network system time delay, improve virtual resource layer interactive efficiency.As described in Figure 10, first, according to optimized target and condition, concrete analysis Sum decomposition is carried out to solved problem, extracts the most basic characteristic information or feature set; Secondly, this characteristic information is processed, to be translated into a kind of scheme of Solve problems under local environment or optimal constraint conditions; Finally, this scheme changed into Immunity Operator in a suitable form and be used for producing new individuality.The process of knowledge based and Knowledge Conversion, on the basis of reasonable drawing immune vaccine, realizes immunoevolution by vaccine inoculation and Immune Selection, effectively to treat the priori asking problem, improves individual fitness.
For immune collaborative fault diagnosis software workflow, in immunological investigation, mutually promoting and suppressing phenomenon to can be understood as a kind of distinctive coevolution form-immune coevolution between various immunocyte, immune coevolution mechanism can be used for reference, feature and collaborative diagnosis pattern is solved for excitation system troubleshooting issue, propose a kind of many diagnostic models coevolution Diagnostic Strategy, the main composition key element that immunity collaborative diagnosis calculates is each immunodiagnosis cell colony, the diagnosis evolution algorithm of immunodiagnosis cell and cell population regulation mechanism etc., evolve and adopt basic immune algorithm, immunity collaborative diagnosis strategy can formalized description as follows: ICED=(CPD, CPDN, CEDA, CPCM), wherein: CPD: immunocyte diagnosis population, CPDN: immunocyte diagnosis population number, the immune collaborative diagnosis algorithm of diagnosis population,
Wherein CED
ij={ PCM
ij, CEDP
ij, EDV
ij, i, j=1,2,, CPDN: the i-th diagnosis cell population is with the Cooperative Mode of a jth diagnosis cell population, and PCMij represents synchronous task assessment, determine that this task has cooperated the need of with other intelligent bodies, CEDPij represents synchronous evolutionary operator, and EDVij representative synchronously evaluates operator; DA
i={ s
i, g
i, p
i, f
i, d
i; i=1; 2 ..., the Immune Evolutionary Algorithm of CPN: the i-th diagnosis cell population; si represents the selection strategy of i-th diagnosis cell population; gi represents the evolutional operation operator of i-th diagnosis cell population, comprises cell clone, clone inhibition etc., and pi represents the execution probability of the evolutional operation operator of i-th diagnosis cell population; fi represents the close fitness function of i-th diagnosis cell population, and di represents the concentration function of i-th diagnosis cell population; The quantity of the contained individuality of CNi: the i-th diagnosis cell population; CPCM={CPO, CPD, CPA, CPE}: immunocyte Advanced group species control model, CPO represents population scale operator, and CPD represents population concentration adjustment, and CPA represents Evolution of Population target function; CPE represents population evaluation method.After employing foregoing description mode, if: CPD (k)={ CPD
1 k, CPD
2 k..., CPD
cPN krepresent that kth is for immune cell population, then kth is expressed as immune cell population coevolution: CEDA{CPD (k) }=CEDA{{CPD
1 k, CPD
2 k..., CPD
cPN k, then in immune collaborative diagnosis from kth-1 generation evolve to kth generation can be expressed as: CPCM (k): CEDA{CPD (k-1) } → CEDA{CPD (k), k=1,2 ...Abnormality diagnosis process is roughly divided into six stages: 1. genI=1, and failure diagnosis territory is set to equipment fault set F={F1 ..., Fi ..., FM}, diagnostic model set is FDM={FDM1 ..., FDMi, ..., FDMM}, M are equipment fault pattern number, and FDMi is i-th fault diagnosis model; 2. for certain diagnostic task FDM_TASK, initialization fault diagnosis model colony FDM, distributes each diagnostic model weight w={w1 ..., wi ..., wM}, wi represent the diagnosis importance degree of diagnostic model FDMi in tracing trouble task FDM_TASK; 3. each fault model FDMi is to diagnostic task FDM_TASK, provides the sub-conclusion FDR={FDR1 of each self diagnosis ..., FDRi, ..., FDRM}, calculates the individual FDMi affinity degree of each diagnostic model and concentration, evaluate FDM_TASK failure diagnosis effect, if meet failure diagnosis termination condition, then turn to 6.; 4. genI=genI+1, to fault diagnosis model colony FDM, chooses colony of new generation based on affinity degree and concentration value from previous generation colony; 5. Immunity Operator (clone, sudden change, suppression) is applied in the individuality of colony, obtains new fault diagnosis model colony FDM, and distribute new diagnostic model weight w={w1 ..., wi ..., 3. wM}, turn to; 6. failure diagnosis terminates.The failure diagnostic process in a certain moment, immune collaborative fault diagnosis software module 5 will complete above-mentioned whole Diagnosis behavior, after the diagnostic result be comparatively satisfied with, just terminate this diagnostic process.
In addition, very to the knowledge accumulating in platform service supporting layer and classification engine, along with the production and construction of safety in production enterprise, the corresponding increase of value volume and range of product of data information, between data, relation day is also beneficial complicated, traditional qualitative analysis is carried out to security incident and has been not suitable with the needs of complex data in a large number, propose to utilize the analysis tool of correlation rule to excavate the characteristics and rules of security incident data, find the Strong association rule between accident occurrence type and " people-machine-environment-management " each factor, early warning is carried out to accident.Collect and excavate relevant historical disaster data and design prospecting data, obtain the data that construction accident is relevant.Data scrubbing is carried out to construction accident data, ensures the exact specification of data.Imperfect in construction accident data record, inconsistent, also vicious information etc., therefore, for ensureing effective with data in post analysis, this one-phase needs to clear up these type of data, mainly solves the human error in data file foundation, and in data file, some set up Multidimensional Data Model to the larger particular values of statistic analysis result impact.Figure 11 gives a kind of data analysing method based on Association Rule Analysis flow process, its reason occurred according to the feature accident of construction accident, setting data attribute, from database, extract casualty data and reconnoitre data formation accident sample data, its data cube is set up from accident pattern, traffic injury time, equipment parameters three dimensions, and utilize the Aggregation Query of sql like language and enable statement to operate this data cube, complete the search procedure of frequent predicate set and strong rule.Produce frequent predicate set.Concrete demand fulfillment minimum support and min confidence, minimum support is set and is counted as 3, thus determine the predicate collection not having Concept Hierarchies dimension, min confidence can collect minimum support with item and count the conditional probability represented and express, thus utilizes the apriori traditional improved to produce frequent predicate set.Produce Strong association rule.Further can by describing the real implication meeting the correlation rule of minimum support threshold value and minimal confidence threshold, the such as time in month, equipment parameters carries out interaction prediction to contingent accident rank, the early warning of industrial and mineral accident will be contributed to, auxiliary enterprises science decision, in concrete mining process, each threshold value can by staff and domain expert's common setups, that is whether rule is correct, applicable, depend on the setting of boundary condition.
The feature of, close in cooperation mutual frequent for industrial and mining enterprises' production safety management, the type of service of combing enterprise, the safety management realizing supporting business cooperation is worked in coordination with, and also specifically comprises following technology:
According to external appreciative standard, combine again the production technology level of China and the difference of each site surrounding simultaneously, the method for step analysis and fuzzy mathematics can be adopted to carry out overall merit, to judge that whether it is for major hazard source.Dangerous reason to various dangerous matter sources is comprised to the assay of major hazard source, accident occurrence probability, the coverage etc. of consequence.The evaluation of major hazard source is one of key measure controlling substantial industrial accident.Major hazard source evaluation should be evaluated two aspects from natrual danger evaluation and real danger and carry out.Natrual danger evaluation mainly reflects material inherent characteristic, and the feature of dangerous substance production process and dangerous unit are inner, external ambient conditions.Real danger evaluation is the counteracting factor considering various danger on the former basis, reflects people and occurs in control accident and control the subjective initiative in damage sequence expansion.Add the external environment condition counteracting factor and propose following evaluation model:
(B111) i in formula---the physical hazard evaluation of estimate of i-th kind of material; (B112) j---the process dangerous evaluation of estimate of jth kind material; Wij---the coefficient correlation of jth item technique and i-th kind of physical hazard; B12---the serious evaluation of estimate of accident; The factor is offset in B21---process equipment container building structure; B22---peopleware offsets the factor; The factor is offset in B23---safety management.
Method during usual employing dangerous matter sources classification, determines a border circular areas by dead radius, and this border circular areas is exactly the scope of this dangerous matter sources impact.The scope of this border circular areas is applicable for the impact of exploding, but just inapplicable for some other situation, such as relates to Leakage Gas, just can not determine according to dead radius completely.Judge accurately and classification dangerous matter sources, the result then evaluated by it takes corresponding prevention, emergency treatment.The object of major hazard source Fast Evaluation classification, is on the data message basis that substantial risk source data is included, carries out Fast Evaluation and classification, carry out macroscopical hierarchical monitoring and management with convenient responsible departments of the government to major hazard source to major hazard source.Dangerous matter sources fast appraisement method is mainly evaluated the damage sequence that major hazard source may cause, to predict that dead radius that accident occurs is for primary evaluation index, carries out the classification of major hazard source with the size of dead radius.First select the numbering of dangerous matter sources, then search the related data of database according to this numbering, analyze after obtaining data, judge the species number of the material comprised in these dangerous matter sources data, whether toxic, thus a point several approach processes this dangerous matter sources.Wherein most important part is exactly calculate noxious substance, inflammable, the dead radius of the injury model of explosive material, comes to dangerous matter sources classification finally by maximum dead radius.
In addition, enterprise can also strengthen the control of production scene safety management and production process by cloud platform.To the hidden danger that production process and material, installations and facilities, equipment, passage, operating environment etc. exist, analysis and control should be carried out.Work permit management is implemented to the dangerous higher activity such as operation in fire operation, restricted clearance, temporary electricity operation, high-rise working, strictly fulfils examination and approval procedures.As shown in figure 12, based on the process model mining algorithm of Petri, the process model of non-Petri network modeling is converted to Petri network, solves to excavate from daily record and hide this open problem of task, make to excavate the task node comprising not tape label in the Petri network obtained.Safety in production process logs is analyzed, the data of production process is assessed, stops hidden danger in time.May be used for based on this Petr algorithm the workflow engine supporting a kind of composite service, comprising: the database of interface layer, key-course, physical layer, accumulation layer and the flow instance for storage service; After described workflow engine is disposed, received the business information of operation system or the transmission of other interface systems by described interface layer, other interface systems described comprise resource management system, service release management system, billing and accounting system; Described interface layer provides the interface of three kinds of modes, comprises api interface, Corba interface, web Service interface, is convenient to being connected of workflow engine and operation system; Described business information comprises the flow instance of business, the current link performance of flow instance; After described key-course receives the business information of described interface layer transmission, according to the flow process route control method supporting composite service, come the generation of the flow instance of service control, scheduling, decomposition, merging, end; And determine that the flow instance of business is that automatic flow arrives next link, still need original place to wait for; Simultaneously described key-course calls the circulation result that the performance of the current link of the method record flow instance that described physical layer provides and described flow process route control method are determined, result of flow is returned by described interface layer; The method that described physical layer provides for: to newly-increased, the amendment of the management object described by workflow engine inside, delete and the operation of inquiry, described management object comprises: the flow instance object of business, flow process routing object, the current link object of flow instance, flow tasks object; Wherein, the specific tasks of the link execution of each flow instance of described flow tasks object factory; Described accumulation layer is by the flow instance information of the described business of the persistent preservation of described database.Can also merge a kind of verification method of Optimized Operation in described flow operations process, concrete steps are:
(1) scheduling scheme generates: according to product marketing plan, purchasing of raw materials plan, equipment maintenance plan, product (intermediate, raw material) inventory information and equipment capacity, the resources of production take, consume, production cost etc. produces constraint information, the data file generating object module sets up the Mathematical Modeling of Optimized Operation, and according to the optimization aim (maximum productivity, maximum profit or meet sales order) of user's setting, and model solver, calculate Optimized Operation scheme.
(2) abnormal by calculating the production that precognition may occur: to gather product marketing plan, purchasing of raw materials plan (comprising arrival situation), inventory information, Plant maintenance plan, energy supply plan etc. current in enterprise resource planning, temporally carry out tracking by promiscuous device constraint (comprising Batch Process equipment and continuous producing apparatus), the constraint of capacity-constrained, logistical balancing and energy constraint to calculate, predict the production that may occur abnormal.
(3) Optimized Operation scheme (data that step 1 obtains) is combined with the production that may occur abnormal (data that step 2 obtains), utilize the visual ESCPetri-Nets network technology virtual production process of figure.Take scheduling time as axis, dynamic graph display workshop equipment working condition, material balance (buying-stock-sale), simplation verification production scheduling situation.Centered by material, with scheduling time change, graphically the variation tendency of certain material of Dynamic Announce consumes relevant equipment operation condition with certain moment to certain material production.