abductive and deductive approach in learning from examples method for technological decisions making

Selection and peer-review under responsibility of the Vilnius Gediminas Technical University doi: 10.1016/j.proeng.2013.04.074 11th International Conference on Modern Building Materials,

Trang 1

Procedia Engineering 57 ( 2013 ) 583 – 588

Selection and peer-review under responsibility of the Vilnius Gediminas Technical University

doi: 10.1016/j.proeng.2013.04.074

11th International Conference on Modern Building Materials, Structures and Techniques,

MBMST 2013

Abductive and Deductive Approach in Learning from Examples Method

for Technological Decisions Making Konczak Anetaa ∗, Paslawski Jerzyb

a, b Division of Construction Engineering and Management, Institute of Structural Engineernig, Faculty of Civil and Environmental Engineering, Poznan University of Technology,

5, Piotrowo Str., PL-60-965 Poznan, Poland

Abstract

One of the fundamental problems in building engineering is to ensure conformance between the planned and the actual course of construction works What may help solve this problem is a database with examples of performance of similar processes in analogous environmental conditions that enables estimating the basic process parameters (time, cost, quality, etc.) Case-based reasoning in a hybrid advisory system may play an important role between the rule-based reasoning and the machine learning when the data that has been gathered enables inference based on analogies with the completed processes, but their number is still inadequate for application of machine learning Introduction of the abductive approach is intended to enable a detailed analysis of causes of nonconformances – a limited number of examples facilitate this This is an example of cooperation between a human (search for causes based on intuition) and

a computer (systematic gathering of examples) The authors also plan to use simulations to verify the abductive hypotheses In the final part of the article, an example of application of case-based reasoning in the process of delivery of ready-mix concrete is presented

Keywords: Hybrid advisory system, case-based reasoning, simulation, abductive approach

1 Introduction

During performance of a construction project there are usually many disruptions and difficulties that force the persons involved in the project to change their decisions and to look for new solutions Consequently, the typical approach to management of construction projects, based on a single technological and organizational scenario, is very risky Given the changing conditions, a contractor is required to stop the works or to continue to risk financial losses or failure to meet the quality requirements Thanks to accumulation and use of experiences, it is possible to gain an easy access to alternative solutions They provide an opportunity to continue the processes but most importantly enable achieving the assumed results related to schedule, costs, quality, etc

In order to achieve the rational results during performance of a project, it is thus necessary to:

• use a knowledge base containing examples of prior performances of processes in a broad spectrum of environmental conditions;

• monitor the environment in order to evaluate and record data for future projects

Based on previous projects, one may estimate the process parameters and use them to select the optimum solution to a given decision-making problem

In order to estimate the basic parameters of process performance options, one may use different methods [1]:

• multiple regression analysis (MRA) [2], [3];

* Corresponding author

E-mail address: aaneta.konczak@put.poznan.pl; bjerzy.paslawski@put.poznan.pl

Trang 2

• artificial neural networks (ANN) [4];

• simulation methods [5]; and

• case-based reasoning methods [1, 6–8, 20]

An important success driver is to combine various methods in order to achieve synergy (hybrid systems): an analogy resulting from similarity between recurring construction processes performed in similar conditions, an abductive approach aimed at analyzing the causes of irregularities in the performance of processes, and simulation modeling aimed to verify the hypotheses generated in the process of analyzing the cases

2 Advisory systems supporting technological decision making

Advisory systems based on knowledge gathering provide an opportunity to reduce the problems associated with difficulties in planning of construction processes At the early stage of development of such systems, it is reasonable to use rule-based reasoning similar to the concept of traditional expert systems, based on knowledge contained in standards, instructions, procedures, and rules elaborated by experts At the next stage, as the knowledge from cases is gathered, one may introduce case-based reasoning, based on a relatively small number of examples of performance of a specific construction process At the next stage, one may use machine learning based on the use of artificial neural networks and simulation These methods may be implemented only if a sufficiently large database with cases is available

Use of advisory systems is justified only when the possibility to index the cases is used skillfully and critical evaluation

of the process performance results is performed Experience and intuition of the system user is thus necessary in order to achieve synergy between the capacities of a human and a computer (considered to be one of the key objectives of creation and implementation of hybrid advisory systems)

The functioning of the hybrid advisory system assumed using various tools/methods depending on the unique characteristics of the problem and the progress achieved in the development of the system (simulation [14], flexibility, ANN [4], MCDM [15, 16], evolution algorithms [17], material modification [18] and case analysis [19]

3 Case-Based Reasoning

The use of knowledge from previous cases, which is the topic of this article, is based on based reasoning The case-based reasoning (CBR) consists in solving problems case-based on previous analogous cases [9] The CBR speeds up the acquisition of the ability to handle difficult situations It is based on:

• use of previous cases in order to explain new situations [8];

• use of previous cases in order to critically and rationally evaluate new situations;

• reasoning based on previous cases in order to interpret new cases;

• development of new solutions in new situations based on previous experiences [10]

Fig 1 Algorithm of case-based knowledge use

Trang 3

Monitoring of another performance of a recurring construction process is the source of valuable information that can be

analyzed and used to supplement the knowledge base When planning new cases, decision makers use the knowledge based

on the previous performances of the process in similar conditions Parameters of similar cases are used to select a new

solution The applied solution must be monitored and controlled in order to evaluate it later and describe it by analyzing its

results at later stages

When solving problems using the CBR method, it is also important to see the failures that cause the decision maker's

expectations to be unfulfilled Regardless of the results of the decisions made, each action enables drawing conclusions and

contributes to the enlargement of the knowledge base Cases with positive outcomes can be used to solve problems in new

cases Failures, on the other hand, serve the purpose of warning against potential new problems Consequently, each

analyzed (and described) case should be included in the database to be used in the future (Fig 1)

4 Abductive and deductive approach in CBR

The key to proper use of knowledge is using both deductive thinking and abductive thinking This is because it is not

enough to supplement the knowledge base with parameters from previous experiences only It is also important to describe

those cases that will enable supplementing the knowledge base with information on the causes of possible failures Thus,

each monitored case must be analyzed from the point of view of the causes of possible disruptions

Abduction and deduction are integral problem-solving methods Deduction consists in drawing conclusions from what is

known, e.g the best medicine for flue is aspirin Abduction explains what is known (or likely) to us, e.g fever is caused by

flue

As far as the algorithm of knowledge use in construction process management is concerned, abduction is used for

analyzing and searching for causes of deviations The basic model of abduction is the following [21]:

D

C →

D

The C → description pertains to the following relation: cause of disturbance →disturbance One must keep in mind D

that abduction is a reasoning system based on hypotheses The cause that one is looking for is only a suspicion regarding the

occurring deviations Consequently, explanation of the reasons behind the deviations does not guarantee that a solution will

be found when another similar disruption occurs The logic of abduction is the following [11]:

1 !P

) , (

2¬AK P

)

*, (

3¬AK P

( )

4 ,Apres K H P

1

5 H m eets aditional criteria S , … , Sn

) ( ,

6ThusCr H

Cr

H Thus,

In formula (2), P is the objective, i.e learning the cause of the disruptions K and K* (as an expansion of K) are the

knowledge bases of the subject, and A(K,P) is the achievement of the objective based on the K base, which is impossible

based on current resources The objective of abduction is to elaborate hypothesis H which will enable satisfying the demand

of “purported cognitive achievability” (when some additional criteria S1, , Sn, have been met) It is reasonable to consider

the hypothesis (Cr(H)) and possibly to accept it (Hcr)

Deduction reasoning, on the other hand, consists in using the gathered information and logical selection of a solution

that, in specific performance conditions at the building site, appears to be the most advantageous

Trang 4

5 Proposed approach

The key element of proposed approach is based on abductive approach, which is included for eliminate causes of

disruptions and create a base of net cases

In order to improve the ability to make proper decisions in abductive and deductive thinking, it is understandable that

knowledge from experience should be used as a way to find the most similar case (CBR abduction and deduction) Such a

case, in turn, enables finding the best solution to the occurring problems

The abduction model adapted to CBR method must be expanded to include an analogous case Q’ and analogous

disturbance D’ which has analogous causes C’ The basic model of CBR abduction is the following [21]:

D

C → '

~ D D

D'

The proposed method of learning from examples is based on the systematic collection of data, in which the deduction is

to be used in finding the best solutions based on previous and similar cases The abductive approach is designed to find the

causes of disruptions and eliminating anomalies in stored cases (Fig 2) The result is pure/net case

Fig 2 Learning cycle based on abduction and deduction approach

6 Case study

An example of application of knowledge from experiences described herein is concreting of reinforced concrete

structures The consecutive stages of the concreting cycle are loading of the concrete truck, transport to the building site,

testing of the concrete mix (if it does not meet the requirements, it is sent back to the plant), unloading (if the pump is busy,

the concrete truck must wait until it is unloaded), washout (and possible waiting in a line), and return to the concrete plant

(Fig 3) The random nature of the loading, delivery, and pouring of the concrete mix leads to situations where the concrete

trucks are unused (usually the concrete plant and the pumps are regarded as the leading machines)

However, in order to find an analogous case, one must first determine the preset parameters of the process that are

important from the point of view of the analogy being searched for Graham and Smith [12] given five key factors of this

type: the type of the concreted element, the month, the weather, the concreting volume, and the number of concrete trucks

delivering the concrete mix

According to Dunlop and Smith [13], other factors that have significant impact on the efficiency of concreting are the

distance between the concrete plant and the building site, the capacity of the concrete truck, the capacity of the pump, and

the type and age of the pump

The example discussed in the article is delivery of concrete mix to make a 50 cm thick substructure layer at the toll

collection point on a motorway In the case in question, there was no need to use a pump The gathering of cases needed to

determine the efficiency was started by identifying the criteria describing the cases gathered in the base:

1 Element type (slab = SLA, wall = WAL, foundation = BAS, other = OTH) – this refers to a change in the efficiency

depending on the shape and dimensions of structural elements;

Trang 5

Fig 3 Delivery of ready-mix concrete

2 Temperature (to –15 °C- EL, from –14 °C to –5 °C- L, from –4 °C to –0 °C- HL, from 1 °C to 5 °C- LM, from 6 °C to

15 °C- M, from 16 °C to 25 °C- MH, above 26 °C – EH) – this refers to the air temperature during the transport and the pouring of the concrete mix;

3 Weather (sunny, overcast, rainy, snowy, sunny spells) – this refers to the weather conditions at the time of delivery and pouring of the concrete mix; Subjective criterion based on the user’s opinion;

4 Concreted volume (0-9, 10-19, 20-29, , 340-349, above 350 [m3]) – the volume of the structural element;

5 Capacity of the concrete trucks (6-8 – L, 9-10 – M, 11-12 – H [m3]) – the nominal capacity of the concrete trucks;

6 Number of concrete trucks (1-5 – L, 6-10 – M, 11-15 – H, above 16 – EH [ea]) – the number of concrete trucks used to transport the concrete mix;

7 Working hours (6:01-7:00 AM, 7:01-9:00 AM, 9:01-12:00 AM, 12:01-2:00 PM, 2:01-4:00 PM, 4:01-6:00 PM, 6:01 PM-6:00 AM) – the times at which the concrete mix is being transported

A part of the data table for several cases is shown below (Table 1)

Table 1 Data on concrete mix deliveries

No Pour type Temperature Weather Vol of truck No of

trucks

Working hours Time of the ride Abduction

1 SLA LM overcast 6 9 6:01PM-6:00AM 0:42

4 SLA LM overcast 6 9 6:01PM-6:00AM 0:58 accident on the route

5 SLA LM overcast 7,5 9 6:01PM-6:00AM 0:38

Based on the cases above, it can be concluded that the main reason for delays and reduced efficiency of concreting is the changing conditions of traffic

In such conditions, two basic solutions are available: 1) to adjust the number of concrete trucks to match the variable traffic problems (e.g at 6.01 PM/7.00 AM – 9 trucks; 9.01 AM/2.00 PM and 4.01PM/6.00 PM – 11 trucks, and 7.01 AM/9.00 AM and 2.01 PM/4.00 PM – 13 trucks); 2) to change the location of the concrete mix plant (maybe install a field mix plant to support such a large building project)

Trang 6

7 Conclusions

The approach to supporting decision making processes related to selection of construction process performance variants based on the abductive and deductive approach in learning from examples method described here, with practical examples enable drawing the following conclusions:

1 The diversity of sources of disruptions in construction management justifies using the abduction approach

2 Gathering knowledge from cases is an intermediate stage of development of a hybrid advisory system that fits between rule-based reasoning and machine learning, e.g artificial neural networks

3 The proposed concept can be implemented during performance of cyclic construction processes by a specialized contractor

4 It is recommended to use an irregularity elimination mechanism based on abduction reasoning

5 Use of simulations and case-based reasoning will enable achieving synergy, thus increasing the effectiveness and efficiency of the proposed method

Acknowledgements

The authors would like to thank the Institute of Structural Engineering of the Poznan University of Technology for the support from its statutory activities fund

References

[1] Koo, C., Hong, T., Hyun, C., & Koo, K., 2010 A CBR-based hybrid model for predicting a construction duration and cost based on project characteristics in multi-family housing projects, Canadian Journal of Civil Engineering 37(5), pp 739-752 http://dx.doi.org/10.1139/L10-007 [2] Le-Hoai, L., Young Dai Lee, Y D., Nguyen, A T., 2013 Estimating time performance for building construction projects in Vietnam, KSCE Journal of Civil Engineering 17(1), pp 1-8 http://dx.doi.org/10.1007/s12205-013-0862-3

[3] Kaplinski, O., Janusz, L., 2006 Three phases of multifactor modelling of construction processes, Journal of Civil Engineering and Management 12(2),

pp 127-134

[4] Gajzler, M., 2011 Neural networks in the advisory system for repairs of industrial concrete floors, Computer Assisted Mechanics and Engineering Sciences 18(4), pp 255-263

[5] Liu, X., Wang, Z., Jin, D., 2011 Risk evaluation of cost for hydropower construction under risk fixed schedule probability using monte carlo simulation method, Technics Technologies Education Management 6(2), pp 287-299

[6] Arditi, D., Tokdemir, O B., 1999 Using case-based reasoning to predict the outcome of construction litigation, Computer-Aided Civil and Infrastructure Engineering 14(6), pp 385-393

[7] Dogan, S., Arditi, D., Murat Gunaydin, H., 2006 Determining attribute weights in a CBR model for early cost prediction of structural systems, Journal

of Construction Engineering and Management 132(10), pp 1092-1098

[8] Ryu, H., Lee, H., Park, M., 2007 Construction planning method using case-based reasoning (CONPLA-CBR), Journal of Computing in Civil Engineering 21(6), pp 410-22 http://dx.doi.org/10.1061/(ASCE)0887-3801(2007)21:6(410)

[9] Aamodt, A., Plaza, E., 1994 Case-based reasoning: foundational issues, methodological variations, and System Approaches, AI Communications 7(1),

pp 39-59

[10] Kolodner, J L., 1992 An introduction to case-based reasoning, Artificial Intelligence Review 6(1), pp 3-34 http://dx.doi.org/10.1007/BF00155578 [11] Gabbay, D M., Woods, J., 2006 Advice in Abductive Logic, Logic Journal of the IGPL 14(2), pp 189-219

[12] Graham, D., Smith, S., 2004 Estimating the productivity of cyclic construction operations using case-based reasoning, Advanced Engineering Informatics 18(1), pp 17–28 http://dx.doi.org/10.1016/j.aei.2004.03.001

[13] Dunlop, P., Smith, S., 2003 Estimating key characteristics of the concrete delivery and placement process using linear regression analysis, Civil Engineering and Enviromental Systems 20(4), pp 273-29 http://dx.doi.org/10.1080/1028660031000091599

[14] Jaskowski, P., Czarnigowska, A., Tokarski, Z., Sobotka, A., 2013 Simulation for selection road works equipment, Baltic Journal of Road and Bridge Engineering [in Press]

[15] Zavadskas, E K., Turskis, Z., 2011 Multiple criteria decision making (MCDM) methods in economics: an overview, Technological and Economic Development of Economy 17(2), pp 397-427 http://dx.doi.org/10.3846/20294913.2011.593291

[16] Zavadskas, E K., Turskis, Z., Tamosaitiene, J., 2011 Selection of construction enterprises management strategy based on the SWOT and multi-criteria analysis, Archives of Civil and Mechanical Engineering 11(4), pp 1063-1082 http://dx.doi.org/10.1016/S1644-9665(12)60096-X

[17] Jaskowski, P., Sobotka, A., 2006 Scheduling construction projects using evolutionary algorithm, Journal of Construction Engineering and Management 132(8), pp 861-870 http://dx.doi.org/10.1061/(ASCE)0733-9364(2006)132:8(861)

[18] Slowik, M., 2012 Modelling of the inverse creep of road bitumen modified with SBS copolymer, Baltic Journal of Road and Bridge Engineering 7(1),

pp 68-75 http://dx.doi.org/10.3846/bjrbe.2012.10

[19] Ustinovichius, L., Rasiulis, R., Ignatavičius, C., Vilutien, T 2012 Analysis of waterproofing defects and technology development for car parking roofs: Lithuanian case, Journal of Civil Engineering and Management 18(4), pp 519-529 http://dx.doi.org/10.3846/13923730.2012.701231 [20] Yau, N., Yang, J., 1998 Case-based reasoning in construction management, Computer-Aided Civil and Infrastructure Engineering 13(2), pp 143-150 http://dx.doi.org/10.1111/0885-9507.00094

[21] Sun, Z., Finnie, G., Weber, K., 2005 Abductive case based reasoning, International Journal of Intelligent Systems, 20(9), pp 957-983.http://dx.doi.org/10.1002/int.20101

Định dạng
Số trang	6
Dung lượng	337,5 KB