Structural redundancy as robustness assurance of complex geoengineering systems

Page created by Cody Perez
 
CONTINUE READING
E3S Web of Conferences 166, 11003 (2020)                                                            https://doi.org/10.1051/e3sconf/202016611003
ICSF 2020

         Structural redundancy as robustness assurance of complex
         geoengineering systems
         Alina Dychko1,*, Igor Yeremeyev2, Natalya Remez1, Serhii Kraychuk3, and Natalia Ostapchuk3
         1NationalTechnical University of Ukraine “Igor Sikorsky Kyiv Polytechnic Institute”, Institute of Energy Saving and Energy
         Management, 37 Peremohy Ave., Kyiv, 03056, Ukraine
         2Taurida National V. I. Vernadsky University, 33 Ivana Kudri Str., Kyiv, 04000, Ukraine
         3Rivne State University of Humanities, 12 Stepana Bandery Str., Rivne, 33000, Ukraine

                       Abstract. The present paper provides that robustness is the facility of complex computer-aided
                       geoengineering systems of keeping its feature invariable along the certain period of time. The detecting of
                       indications which testify that system is not responding adequately to input disturbances may be realized by:
                       comparing real data on assembly output with number of reference artefacts which correspond to specific states
                       of disturbances; comparing real data on assembly output with a normal reaction to the actual set of input
                       signals. To increase the resilience of monitoring data of geoengineering system the majority principle and the
                       Byzantine agreement method are used.

      1 Robustness problem                                                   failure of the currently operating structure is detected.
                                                                                 Every of mentioned ways of duplicating has intrinsic
      Robustness [1–3] characterizes the complex computer-                   preferences and limitations.
      aided geoengineering systems (CCAS) facility of keeping                    Redundancy is considering as an additional type of
      its feature invariable along the specified time. After this            structural performance indicator that is defined as a
      time such characteristics may gradually deteriorate but                measure of warning available prior to system catastrophe
      with a decrease in quality of operation within predefined              [4]. The redundancy can be evaluated and its
      limits (by reducing dynamic and static accuracy,                       quantification can be formulated with the help of
      increasing response time, increasing transient intervals,              characteristics of the geoengineering systems functions.
      reducing possible additional functions, performance/cost               For optimization of their operation it’s important to
      metrics, etc.).                                                        control the effects of maintenance on lifetime functions
          Robustness study helps to understand better the                    and redundancy [4].
      characteristics of the complex systems, geoengineering,                    The existence of redundancy assists in enhancing the
      energy efficiency and building energy sustainability in                safety and reliability of a system in its intact state; and
      particular [2]. It’s used when there is a need for a                   mitigating the sensitivity or vulnerability of the structure
      functional layer to provide built-in protection to ensure              to localised damage under an accidental situation. To
      that it is robust with respect to requests that are issued at          assure such characteristics of the systems it’s possible to
      instances that are incompatible with its current state and             take into account various stages where redundant
      could therefore cause catastrophic system failure [3].                 structures are involved in and to develop essential means
          There are the following ways of robustness assurance:              of designing for redundancy which can readily be
      structural      redundancy;     procedural      redundancy;            integrated into safety decision-making [5]. Such approach
      informational redundancy; combination of the all or some               should be considered in underground and on-ground
      of above mentioned redundancies.                                       construction, in mining and the extraction of minerals, in
          Structural redundancy presumes the duplicating                     the assessment of subsidence and subsidence of soil bases
      structures utilization while each of them is capable to                etc. Generalizing the problem of structural redundancy, it
      realize all the necessary procedures and control actions               should be emphasized that it implies the introduction into
      inherent to CCAS. There are the following ways of using                the system of additional equipment, structured in such
      duplicating structures:                                                way that even in case of failure of a certain part of the
      - Simultaneous functioning of the two (or three and more)              equipment, the system will continue to function
      structures (dynamic or “hot” backup) with their activity               successfully [6, 8, 11]. All these need the reliable control
      results analysis and the structure of fault occurred                   of information, which is impossible without system
      deactivating if the specifics of wrong operation is                    reliability engineering and its predicting with reliability
      detected;                                                              monitoring scheme under changing environmental
      - One of structures is operating when the other one is in              conditions [7, 9-11].
      standby reserve and shall be activated only after the
         *
             Corresponding author: aodi@ukr.net
© The Authors, published by EDP Sciences. This is an open access article distributed under the terms of the Creative Commons Attribution License 4.0
(http://creativecommons.org/licenses/by/4.0/).
E3S Web of Conferences 166, 11003 (2020)                                                        https://doi.org/10.1051/e3sconf/202016611003
ICSF 2020

      2 Structural redundancy                                              comparing real data on assembly output with a normal
                                                                          reaction to the actual set of input signals (to detect the fact
      The simplest example of structural redundancy is the                of stability loss).
      hardware duplicating and use in standby reserve either in               Incidentally the distribution of points of set Yi f, which
      dynamic or “hot” backup. In the first variant the one               characterizes the real output signal is compared with
      hardware is operating but the second one is idle or on              distribution of points of reference signal Yіe and the
      maintenance prevention. In the second variant the both              Euclidean metrics
      hardware operate simultaneously and supplementary
      monitoring hardware analyzes the operation of both
                                                                                                ,     =   ∑         −                (3)
      complexes and makes a decision about what complex
      generates the more reliable information [12]. Each mode
      has the positive and negative distinguishing features. For          is estimated.
      example, the duplicating structure with standby reserve is              If D[Yi f, Yіe] < |δ|, where δ – maximum acceptable
      developed in such way that after predetermined time of              deviation of the real pattern from reference, the assembly
      operation trsb when the probability of trouble-free                 is considered as correct one or corresponding to certain
      operation                                                           reference artefact.

                        =               ,                      (1)        3 Procedure of imbalance identification
                                                                          in systems
      (where m – trouble-free life) exceeds a certain set
      threshold, it is realized the transition to operating reserve       The algorithm that implement the above operations is
      set and the operated hardware is put on prevention for a            presented in Fig.1-A –Fig.1-D and the conceptual layout
      time tmp, during which the troubleshooting operations are           of subsystem for some devices mismatching detecting and
      carry out. It is the essential fulfilment of the condition          adjusting their features – in Fig.2. The following
      tstb > tmp while the more Δt = tstb – tmp then the system’s         abbreviation is used here: UID – unit for information
      fault-tolerance is better. If the operating complex fails the       distribution, which depending on identical duplicating
      standby reserve complex should be activated (if the last            units’ у1 and у2 state and subsystem’s mode of operation,
      one isn’t on prevention). This duplication approach is              feeds the information on both units’ inputs (if the loss of
      easy-to-work and requires the minimum of redundant                  stability is not detected), locks out the inputs of both units
      hardware. But in case of failure and the need to switch to          (if the loss of stability is detected and faulty unit is
      a backup set it is a loss of time connected with procedures         identified) or locks out input of faulty unit (while
      for switching units, which leads to loss share information.         correction procedure of the wrong data has place); UAM
      If the standby reserve complex is on maintenance                    – unit of alternative models of assembly for the events of
      prevention the information drop may even result in overall          routine situation (proper operation, distortions due to
      fault. Therefore in such a case when loss of information            specific common causes); URS – unit (source) of
      and temporary stoppage of CCAS functioning are                      reference signals for testing the nodes and performing
      intolerable the dynamic or “hot” backup is used. In this            procedures for correcting the characteristics of defective
      case probability of failure-free operation along the time thb       nodes; UIE – unit of information’s evaluation (the fact of
      is estimated as:                                                    loss of stability detection, fault unit identification, the
                                                                          optimal model selection); MU – memory unit for
                                    ℎ        ℎ
                              =         2        .             (2)        intermediate results of evaluations and corrections saving;
                                                                          UFC – unit of feature correction which implements the
           In the case of “hot” reservation the complementary             correction calculations in nodal points and choice of
      hardware is demanded for state analysis of every                    interpolation polynomials for calculations correction in all
      complexes and decision making of which of the blocks                the residuary points of unit features.
      provides more reliable information. This hardware should                In addition besides of majorization for fault tolerance
      also be duplicated in order to ensure its trouble-free              the Byzantine agreement approach is used, which differs
      operation. Therefore this way of structure redundancy is            in that degree of redundancy need not to be as uneven
      characterized by surplus of the complementary hardware              number.
      as compared with standby reserve. Fault tolerance
      property in this case is guaranteed along the shorter time          4 Principles of the resilience increase
      interval that is thb< tsbr. But along that interval no losses
      of information in its unexpected fault from operating               As mentioned above, the further development of the
      hardware occurs.                                                    method of structural redundancy consists in the use of the
           The detecting of indications which testify that                majority principle, or the principle of voting, which
      assembly (subsystem, geoengineering system) is not                  requires the introduction of multiple redundancy of data
      responding adequately to input signals or disturbances,             processing facilities, where multiplicity is an odd number
      i.e. that it is a some loss of stability, may be realized by:       equal to “3” or higher than it. When using majorization,
       comparing real data on assembly output with number                the probability of a system failure can be determined from
      of reference artefacts which correspond to specific states          the expression
      of disturbances (for detecting the cause of stability loss);

                                                                      2
E3S Web of Conferences 166, 11003 (2020)                                                           https://doi.org/10.1051/e3sconf/202016611003
ICSF 2020

                                    Start
                                                                                               і: → 1

                                   i: → 1
                                                                                   To disable the DTIі input

                   E

                                                1                                       To bring test уе
                                    Ті=1                                                    to DTIі

                                    0
                                                                                     To compare уі and уе
                             Data entry in DTI і

                                                                                                                 1
                                                                                               Δуе>2ξ
                                                1
                                       І=2
                                                                                               0                To inform about
                                   0
                                                                                                                anomaly in DTIі

                                  і+1: → і

                       E                                                                                         0
                                                                                                         і2ξ

                                                                                                   To save уj

                                  y= 0,5(у1+у2)

                                                                                           1
                                                                                                     j=N
                                                                             A
                           To put y (or у1) on output
                                     of DTI                                                                0

                                                                                                   j+1: → j

                                        Stop

      Fig. 1. The block-diagram of algorithm of imbalance identification in homogeneous subsystems which duplicates one another (A).

                                                                     3
E3S Web of Conferences 166, 11003 (2020)                                                       https://doi.org/10.1051/e3sconf/202016611003
ICSF 2020

                                               А

                          To build according to yj the smoothed
                                      curve yз = f(x)
                                                                                    To determine the corrections
                                                                                          Δyjk= yint j - yAKj

                                          K:→ l

                                                                                      1
                                                                                                 j=N
                           To determine the metrics μК, which
                           describes the grade of membership                                         0
                                   уз=f(x) to yAK=f(x)

                                                                                               j+1: → j

                                      To record μК

                                                                                      To determine the interpolating
                                                                                 polynomial dependence by data Δyj-1, Δyj
                                                                                                and Δyj+1
                                                           1
                                           k=K

                                          0
                                                                                                                     D

                                        k+1: → k
                                                                                                 ТiA: → 1

                                   μopt = min(μ1, …, μk)
                                                                                                     E

                     v                                             B

                                           j: → 1

                                                               0
                                           μopt>2ξ

                                           1

                                               C

      Fig. 1. The block-diagram of algorithm of imbalance identification in homogeneous subsystems which duplicates one another (B).

                                                                       4
E3S Web of Conferences 166, 11003 (2020)                                                           https://doi.org/10.1051/e3sconf/202016611003
ICSF 2020

                                          C

                                 Δyjk = yфj - yAkj

                                                     1
                                     y jk >2ξ

                                      0
                                                                To save the
                                                                         y jk

                                                     1
                                         j=N

                                     0

                                          C               To range saved             :

                                                            y mk    max max
                                                                                    δ1
                                                          ……………………………

                                                             y nk   min min
                                                                                    δ2

                                                                                          1
                                                               δ2 = δy(m-1)k

                                                                     0

                                                                                          1
                                                               δ2 = δy(m-1)k                  m – catastrophic point

                                                                     0

                                              G                   ξ:→ ξ1

                                                                         B                              F

      Fig. 1. The block-diagram of algorithm of imbalance identification in homogeneous subsystems which duplicates one another (C).

                                                                          5
E3S Web of Conferences 166, 11003 (2020)                                                           https://doi.org/10.1051/e3sconf/202016611003
ICSF 2020

                                        F
                                                                                      μq opt = min(μq1,… μqQ)

                                   q: → 1
                                                                                                                    1
                                                                                              μq opt > 2ξ

                     Smooth substitution of variable in                                                0
                        points (m-1), m, (m+1) by                                                                       G
                       assistance of catastrophe q
                                ycat q =f(x)
                                                                                       To define corrections
                                                                                       Δyjkcat= y int j - yAkcatj

                                                      1
                                    q= Q                                                                            1
                                                                                                   j= N
                                    0
                                                                                               0

                                  q+1: → q
                                                                                              j+1: → j

                                   q: → 1
                                                                           To record interpolating polynomials for the
                                                                           points (m-1), m, (m+1) [disaster struck], and
                                                                               interpolating polynomials for points
                             To define the μq,
                                                                                      Δуj-1, Δyj,, Δyj+1 if j≠m
                         characterized by grade of
                          membership y cat q to yAk

                                                     1
                                    q=Q                                                            D

                                    0

                                  q+1: → q

      Fig. 1. The block-diagram of algorithm of imbalance identification in homogeneous subsystems which duplicates one another (D).

                                                                           systems using majorization principle with different
                        = ∑                 (1 − )              (4)        redundancy values are given in Table 1.
                                                                               In addition to the majority principle, the so-called
      where n – number of sequentially connected blocks in the             “Byzantine agreement” is used to increase the resilience,
      channel; l – number of channels, i.e. number of                      characterized in that the multiplicity of redundancy is not
      redundancies; j – number of properly operating channels              necessarily characterized by an odd number. The
      (j≥ k , where k = { l / 2 } – the closest integer in excess of       formalism of the “Byzantine agreement” is
      l/2; = -t pM / – probability of a single unit failing for            effective when the total number of redundant (duplicate)
      time tpM under failure conditions M.                                 blocks (duplication rate) is n ≥ 2 k + m + 1, where the
          Comparative characteristics of geoengineering                    blocks with failures are no more than k and m is the

                                                                       6
E3S Web of Conferences 166, 11003 (2020)                                                         https://doi.org/10.1051/e3sconf/202016611003
ICSF 2020

      number of blocks that did not fail.                                       If the answer is different, a third repetition is initiated
          When using Byzantine agreement the denied blocks                 and if its result is equal to one of the previously obtained
      may remain connected with the system in any way and                  during the previous two steps of the decision, the answer
      continue to function. It is only necessary to add that in            is recognized as valid and the procedure ends. If, however,
      such systems the source of the initial data is assumed to            the decision obtained in the third step does not coincide
      be quite reliable.                                                   with any obtained in the previous steps, a steady rejection
          In a simpler case, such an approach should be that the           signal may be issued. Sometimes a steady rejection signal
      result of the re-decision is compared with the first                 can be provided after most 4…256 series solutions differ
      decision and if they are the same, the decision is                   from each other by a value exceeding a predetermined
      recognized as valid and the control procedure is                     threshold or double standard deviation.
      completed.

                                                                    Inputs

                                                                     UID

                                                                   UAM

                                                                                                        У2
                                     У1

                                                                       URS

                                                                       UIE

                                                         MU                          UFC

                                                                    Output Y

      Fig. 2. Conceptual layout of subsystem for some devices mismatching detecting and adjusting their features

                                                                      7
E3S Web of Conferences 166, 11003 (2020)                                                       https://doi.org/10.1051/e3sconf/202016611003
ICSF 2020

           Table 1. Comparative characteristics of majorized systems            2012. doi: 10.1109/EDCC.2012.16
                  with different meanings of redundancy rate.
                                                                            4.  N.M. Okasha, D.M. Frangopol, Reliability
                                                    Probability of              Engineering and System Safety 95, 5 (2010)
                     Probability of no-             system output           5. Z.X. Fang, H.T. Fan, Procedia Engineering 14,
      Majorization
                   failure operation RM          emergency (1–RM)               (2011)
                                                   R=0.9     R=0,99
           2 from 3        RM = 3R2– 2R3          2.8x10-2   3x10-4
                                                                            6. B.S. Dhillon, Engineering maintenance: a modern
           3 from 5     RM = 10R3 -15R4 +6R5     8.56x10-3 1x10-5               approach (CRC Press, 2002)
                          RM = 35R4 – 84R5                                  7. M. Ram, Modeling and simulation based analysis in
           4 from 7                              2.73x10-3   3x10-7
                            +70R6 – 20R7                                        reliability engineering (CRC Press, 2018)
                             RM=125R5-                                      8. I.H. Witten, E. Frank, M. Hall, Data Mining.
           5 from 9        420R6+540R7-          8.91x10-4   1.3x10-8
                            315R8+70R9
                                                                                Practical Machine Learning Tools and Techniques,
                             RM=462R6-                                          3rd edn (Morgan Kaufmann, 2011)
                         1980R7+3465RR8-                                    9. R.W. Scholz, Environmental literacy in science and
           6 from 11                             2.96x10-4
You can also read