Professional Documents
Culture Documents
Computational Aspects and Applications in Large Scale Networks NET 2017 Nizhny Novgorod Russia June 2017 Valery A. Kalyagin
Computational Aspects and Applications in Large Scale Networks NET 2017 Nizhny Novgorod Russia June 2017 Valery A. Kalyagin
https://textbookfull.com/product/large-scale-scientific-
computing-11th-international-conference-lssc-2017-sozopol-
bulgaria-june-5-9-2017-revised-selected-papers-1st-edition-ivan-
lirkov/
https://textbookfull.com/product/algorithms-for-computational-
biology-4th-international-conference-alcob-2017-aveiro-portugal-
june-5-6-2017-proceedings-1st-edition-daniel-figueiredo/
https://textbookfull.com/product/information-science-and-
applications-2017-icisa-2017-1st-edition-kuinam-kim/
https://textbookfull.com/product/estimating-functional-
connectivity-and-topology-in-large-scale-neuronal-assemblies-
statistical-and-computational-methods-vito-paolo-pastore/
Computational
Aspects and
Applications
in Large-Scale
Networks
NET 2017, Nizhny Novgorod, Russia,
June 2017
Springer Proceedings in Mathematics & Statistics
Volume 247
Springer Proceedings in Mathematics & Statistics
This book series features volumes composed of selected contributions from
workshops and conferences in all areas of current research in mathematics and
statistics, including operation research and optimization. In addition to an overall
evaluation of the interest, scientific quality, and timeliness of each proposal at the
hands of the publisher, individual contributions are all refereed to the high quality
standards of leading journals in the field. Thus, this series provides the research
community with well-edited, authoritative reports on developments in the most
exciting areas of mathematical and statistical research today.
Editors
Computational Aspects
and Applications
in Large-Scale Networks
NET 2017, Nizhny Novgorod, Russia,
June 2017
123
Editors
Valery A. Kalyagin Oleg Prokopyev
Higher School of Economics Department of Industrial Engineering
National Research University University of Pittsburgh
Nizhny Novgorod, Russia Pittsburgh, PA, USA
This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface
This volume is based on the papers presented at the 7th International Conference on
Network Analysis held in Nizhny Novgorod, Russia, June 22–24, 2017. The main
focus of the conference and this volume is centered on the development of new
computationally efficient algorithms as well as underlying analysis and optimization
in large-scale networks. Various applications in the context of social networks,
power transmission grids, stock market networks, and human brain networks are
also considered. The previous books based on the papers presented at the 1st–6th
Conferences International Conference on Network Analysis can be found in [1–6].
The current volume consists of three major parts, namely, network computational
algorithms, network models, and network applications.
The first part of the book is focused on algorithmic methods in large-scale
networks. In the Chapter “Tabu Search for Fleet Size and Mix Vehicle Routing
Problem with Hard and Soft Time Windows”, the authors present an efficient
algorithm for solving a computationally hard problem in a real-life network setting.
In the Chapter “FPT Algorithms for the Shortest Lattice Vector and Integer Linear
Programming Problems”, fixed-parameter tractable (FTP) algorithms are discussed.
The main parameter is the maximal absolute value of rank minors of matrices in the
problem formulation. The Chapter “The Video-Based Age and Gender Recognition
with Convolution Neural Networks” reviews the problem of age and gender
recognition methods for video data using modern deep convolutional neural net-
works. A new video-based recognition system is implemented with several
aggregation methods that improve the method identification accuracy. The Chapter
“On Forbidden Induced Subgraphs for the Class of Triangle-König Graphs” con-
siders a hereditary class of graphs with the following property: the maximum
cardinality of a triangle packing for each graph of the class is equal to the minimum
cardinality of a triangle vertex cover. Then, some minimal forbidden induced
subgraphs for this class are described. The Chapter “The Global Search Theory
Approach to the Bilevel Pricing Problem in Telecommunication Networks” deals
with the hierarchical problem of optimal pricing in telecommunication networks.
v
vi Preface
New methods of local and global search for finding optimistic solution are devel-
oped. In the Chapter “Graph Dichotomy Algorithm and Its Applications to Analysis
of Stocks Market”, a new approach to graph complexity measure is presented. This
new measure is used to study short-term predictions of crises at the stock market. In
the Chapter “Cluster Analysis of Facial Video Data in Video Surveillance Systems
Using Deep Learning”, a new approach is proposed for structuring information in
video surveillance systems. The first stage of the new approach consists of grouping
the videos, which contain identical faces. Based on this idea, a new efficient and fast
algorithm of video surveillance is developed. The Chapter “Using Modular
Decomposition Technique to Solve the Maximum Clique Problem” applies the
modular decomposition technique to solve the weighted maximum clique problem.
Developed technique is compared against the state-of-the-art algorithms in the area.
The second part of the book is focused on network models. In the Chapter
“Robust Statistical Procedures for Testing Dynamics in Market Network”, a sign
similarity market network is considered. The problem of testing dynamics in market
network is formulated as the problem of homogeneity hypotheses testing. Multiple
testing techniques to solve this problem are developed and applied for different stock
markets. The Chapter “Application of Market Models to Network Equilibrium
Problems” describes an extension of the network flow equilibrium problem with
elastic demands and develops a new equilibrium type model for resource allocation
problems in wireless communication networks. In the Chapter “Selective
Bi-coordinate Variations for Network Equilibrium Problems with Mixed Demand”,
a modification of the method of bi-coordinate variations for network equilibrium
problems with mixed demand is proposed. Some numerical results that confirm
efficiency of the method are presented. The Chapter “Developing a Model of
Topological Structure Formation for Power Transmission Grids Based on the
Analysis of the UNEG” studies the nodes degrees distribution in the United National
Electricity Grid (UNEG) of Russia. This study allows to develop a new Random
Growth Model (RGM) to simulate UNEG, and to identify the key principles for the
network formation. In the Chapter “Methods of Criteria Importance Theory and
Their Software Implementation”, a general approach for solution of the multicriteria
choice problem is developed on the basis of the criteria importance theory. These
methods are implemented in the computer system DASS. The Chapter “A Model of
Optimal Network Structure for Decentralized Nearest Neighbor Search” discusses
the problem of the best structure of network for the decentralized nearest neighbor
search algorithm. Mathematical programming model for the problem is proposed.
Optimal structures of small size networks are computed. A generalization for
large-scale networks is also discussed. In the Chapter “Computational Study of
Activation Dynamics on Networks of Arbitrary Structure”, new results on modeling
dynamical properties of collective systems are presented. A general technique for
reducing the problems to SAT problem is developed. This approach is then applied
to computer security networks. The Chapter “Rejection Graph for Multiple Testing
of Elliptical Model for Market Network” uses the symmetry condition of tail
Preface vii
We would like to thank all the authors and referees for their efforts. This work is
supported by the Laboratory of Algorithms and Technologies for Network Analysis
(LATNA) of the National Research University Higher School of Economics and by
RSF grant 14-41-00039.
References
1. Goldengorin, B.I., Kalyagin, V.A., Pardalos, P.M. (eds.): Models, algorithms and technologies
for network analysis. In: Proceedings of the First International Conference on Network
Analysis. Springer Proceedings in Mathematics and Statistics, vol. 32. Springer, Cham (2013)
2. Goldengorin, B.I., Kalyagin, V.A., Pardalos, P.M. (eds.): Models, algorithms and technologies
for network analysis. In: Proceedings of the Second International Conference on Network
Analysis. Springer Proceedings in Mathematics and Statistics, vol. 59. Springer, Cham (2013)
3. Batsyn, M.V., Kalyagin, V.A., Pardalos, P.M. (eds.): Models, algorithms and technologies for
network analysis. In: Proceedings of Third International Conference on Network Analysis.
Springer Proceedings in Mathematics and Statistics, vol. 104. Springer, Cham (2014)
4. Kalyagin, V.A., Pardalos, P.M., Rassias, T.M. (eds.): Network models in economics and
finance. In: Springer Optimization and Its Applications, vol. 100. Springer, Cham (2014)
5. Kalyagin, V.A., Koldanov, P. A., Pardalos, P.M. (eds.): Models, algorithms and technologies
for network analysis. In: NET 2014, Nizhny Novgorod, Russia, May 2014. Springer
Proceedings in Mathematics and Statistics, vol. 156. Springer, Cham (2016)
6. Kalyagin, V.A., Nikolaev, A.I., Pardalos, P.M., Prokopyev, O.A. (eds.): Models, algorithms
and technologies for network analysis. In: NET 2016, Nizhny Novgorod, Russia, May 2016.
Springer Proceedings in Mathematics and Statistics, vol. 197. Springer, Cham (2017)
Contents
ix
x Contents
xiii
xiv Contributors
Igor Konnov Kazan Federal University, Kazan, Russia; Kazan Federal University,
Institute of Computational Mathematics and Information Technologies, Kazan,
Russia
Nadezhda Kostyakova Higher School of Economics, National Research
University, Moscow, Russia
Egor Krivov Skolkovo Institute of Science and Technology, Moscow, Russia;
Kharkevich Institute for Information Transmission Problems, Moscow, Russia;
Moscow Institute of Physics and Technology, Moscow, Russia
Anvar Kurmukov Higher School of Economics, National Research University,
Moscow, Russia
Rodion Laptsuev Higher School of Economics, National Research University,
Moscow, Russia
Alexander Lyashuk Higher School of Economics, National Research University,
Moscow, Russia
Ilya Makarov Higher School of Economics, National Research University,
Moscow, Russia
Sergey Makrushin Financial University under the Government of the Russian
Federation, Moscow, Russia
Nataliya Matveeva Higher School of Economics, National Research University,
Moscow, Russia
Dmitry Meinster Higher School of Economics, National Research University,
Moscow, Russia
Dmitry B. Mokeev Department of Algebra, Geometry and Discrete Mathematics,
Lobachevsky State University of Nizhny Novgorod, N.Novgorod, Russia;
Laboratory of Algorithms and Technologies for Networks Analysis, Higher School
of Economics in Nizhny Novgorod, N.Novgorod, Russia
Ayagoz Mussabayeva Higher School of Economics, National Research
University, Moscow, Russia
Andrey Pavlovich Nelyubin Mechanical Engineering Research, Institute of the
RAS, Moscow, Russia
Alexey Nikolaev Laboratory of Algorithms and Technologies for Network
Analysis, National Research University Higher School of Economics, Nizhny
Novgorod, Russia
Andrei V. Orlov Matrosov Institute for System Dynamics and Control Theory of
SB RAS, Irkutsk, Russia
xvi Contributors
Abstract The paper presents a tabu search heuristic for the Fleet Size and Mix
Vehicle Routing Problem (FSMVRP) with hard and soft time windows. The objective
function minimizes the sum of travel costs, fixed vehicle costs, and penalties for soft
time window violations. The algorithm is based on the tabu search with several
neighborhoods. The main contribution of the paper is the efficient algorithm for a
real-life vehicle routing problem. To the best of our knowledge, there are no papers
devoted to the FSMVRP problem with soft time windows, while in real-life problems,
this is a usual case. We investigate the performance of the proposed heuristic on
the classical Solomon instances with additional constraints. We also compare our
approach without soft time windows and heterogeneous fleet of vehicles with the
recently published results on the VRP problem with hard time windows.
Keywords Soft time windows · Fleet size and mix · Vehicle routing · Variable
neighborhood search · Tabu search
1 Introduction
The practical interest of this problem consists in its application to various spheres of
the real life: delivery of goods to supermarkets [26], waste collection [3], school bus
routing [20], mail delivery [25], transportation of people with disabilities [8], and
many others.
However, real-life vehicle routing problems differ considerably from classical
ones having various constraints such as hard and soft time windows, different vehi-
cles, capacity constraints, split deliveries, and others. All these additional features
make the process of finding the solution extremely difficult and time-consuming.
VRP problems with a heterogeneous fleet of vehicles are frequently encountered in
logistics. In real-world applications, fleets of vehicles with different capacities and
costs are more common than homogeneous ones [1]. There are two types of het-
erogeneous fleet VRP [16, 17]. The first one is Fleet Size and Mix Vehicle Routing
Problems (FSMVRP). In FSMVRP, there are no limits on the number of available
vehicles of each type. The second category is Heterogeneous Fleet Vehicle Routing
Problems (HFVRP), where the number of vehicles of each type is limited. In this
paper, the FSMVRP is considered [4, 23, 29, 30].
According to Laporte [18], time windows are also one of the most vital constraints
used in real-world applications. There are two types of them: hard and soft time
windows. In real life, it is rarely possible that a customer exposes only one type
of time windows. The most common situation is when there are certain working
hours of a customer—hard time window, and a desirable period of time for receiving
an order—soft time window. In the academic world, much attention is paid to the
Vehicle Routing Problem with hard Time Windows (VRPTW) [2, 5, 6, 10, 12, 13,
21, 25, 27, 28].
Fewer papers deal with the vehicle routing problem with soft time windows [9, 14,
22, 24]. And the papers considering both the heterogeneous fleet of vehicles and soft
time windows are not known to the authors. Note that the existing VRP models with
Soft Time Windows (VRPSTW) also have hard time window constraints. So hereafter
VRPSTW will denote a VRP problem with both soft and hard time windows.
In this paper, we suggest an efficient approach for the Fleet Size and Mix Vehicle
Routing Problem with Soft Time Windows (FSMVRPSTW). Our algorithm is based
on tabu search with variable neighborhoods. We use swap and relocate neighbor-
hoods for inter-route optimization and swap neighborhood for intra-route. A special
procedure is developed for soft time windows optimization.
The paper is organized as follows. In the next section, we provide an integer
linear programming model for the considered problem. In Sect. 3, the pseudocode
and description of our algorithm are presented. Section 4 contains computational
experiments and results, and Sect. 5 concludes the paper.
2 Mathematical Model
The fleet size and mix vehicle routing problem with soft time windows can be
described as follows. Let G (V0 , A) be a complete directed graph, where
Tabu Search for Fleet Size and Mix Vehicle Routing Problem ... 5
The objective of the FSMVRPSTW problem is to choose vehicles and their routes
to serve all the customers with minimum total cost including travel costs, fixed vehicle
costs, and penalties for soft time window violations. The following mixed integer
linear programming model for the fleet size and mix vehicle routing problem with
hard and soft time windows is introduced.
Decision variables
xiklj ∈ {0, 1}, xiklj 1, if vehicle l of type k travels by the arc (i, j) ∈ A in the
solution.
ykl ∈ {0, 1}, ykl 1, if vehicle l of type k is used in the solution for serving
customers.
z ikl ∈ {0, 1}, z ikl 1, if vehicle l of type k serves customer i in the solution.
bikl ≥ 0, beginning of service at customer i by vehicle l of type k.
vi−kl
≥ 0, vehicle l of type k violation of the earliest preferred time at customer i.
vi+kl
≥ 0, vehicle l of type k violation of the latest preferred time at customer i.
Objective function
n
n
n
kl
fk ykl + ci j xiklj + α kl
vi− + vi+ → min (1)
k∈K l1 k∈K l1 i∈V0 j∈V0 k∈K l1 i∈V
6 M. Batsyn et al.
Constraints:
n
z ikl 1, ∀i ∈ V (2)
k∈K l1
ykl ≥ z ikl , ∀k ∈
K , ∀l 1, n, ∀i ∈ V (3)
x0ikl kl
xi0 ykl , ∀k ∈ K , ∀l 1, n (4)
i∈V i∈V
xiklj x klji z ikl , ∀k ∈ K , ∀l 1, n, ∀i ∈ V (5)
j∈V0 j∈V0
qi z ikl ≤ Q k , ∀k ∈ K , ∀l 1, n (6)
i∈V
bklj ≥ bikl + si + ti j − M 1 − xiklj , ∀k ∈ K , ∀l ∈ 1, n, ∀i ∈ V0 , j ∈ V (7)
ei ≤ bikl ≤ li , ∀k ∈ K , ∀l 1, n, ∀i ∈ V (8)
kl
vi− ≥ ẽi − bikl ,∀k ∈ K , ∀l 1, n, ∀i ∈ V (9)
kl
vi+ ≥ bikl ˜
− li , ∀k ∈ K , ∀l 1, n, ∀i ∈ V (10)
The objective function (1) minimizes the total cost of all routes including travel
costs, fixed vehicle costs, and penalties for soft time window violations. Constraints
(2) require that each customer should be serviced exactly once. Inequalities (3)
guarantee that a vehicle is used in the solution and its fixed cost is taken into account
only if it serves at least one customer. Equalities (4, 5) specify that every used vehicle
has to leave the depot, arrive to all the appointed customers, leave them, and at the end
return to the depot. Constraints (6) require the total demand of all customers in one
route to satisfy the capacity of the vehicle serving it. Inequalities (7) guarantee that
the service of the customer is not started before the vehicle completes service at the
previous customer and travels to this one. Here, we use big-M notation, which makes
the inequality to be automatically satisfied for any values of variables, when xiklj 0.
Finally, constraints (8) require beginning times to lie inside hard time windows and
constraints (9, 10) determine how much the soft time windows are violated.
3 Algorithm Description
The developed algorithm is based on the tabu search and variable neighborhood
search metaheuristics. The main function in our algorithm is TABUSEARCH()
(Algorithm 1). The idea of the developed algorithm for the fleet size and mix vehicle
routing problem with soft time windows consists in application of tabu search with
several neighborhoods to an initial greedy solution. One of the advantages of the
Tabu Search for Fleet Size and Mix Vehicle Routing Problem ... 7
tabu search is in tabu list which forbids some edges and thus diversifies the search.
This approach allows avoiding loops and going further from a local optimum out
from the neighborhood of the current best solution. New solutions are generated by
means of 2-opt, swap and relocate operators taking into consideration the tabu edges.
If a better value of the objective function is found and the new solution is feasible in
terms of hard time windows and vehicle loads, then it replaces the current optimal
solution. In such a case, the algorithm tries to optimize further the found solution by
means of intra-route improvement operators. If the solution is infeasible, then the
function value is penalized. All these steps are carried out until the current solution
remains infeasible during SC (Stopping Criterion) iterations.
The algorithm of the initial solution construction for the stated problem is based on
Solomon [28] insertion heuristic. First, the route consisting of only one customer,
which is the farthest from the depot, is created. Afterward, new customers are grad-
ually added to the current route taking into account hard time windows and capacity
constraints. To evaluate the insertion cost for the new customer u between two adja-
cent customers i and j criterion c1 (i, u, j) is computed. And to choose the best
customer for insertion we compute criterion c2 (u).
Let (r1 , r2 , . . . , rm ) be the current route where the first and the last points are the
depots: r1 0, rm n + 1 (vertex n + 1 is a copy of vertex 0). For each unserved
customer, we determine the best insertion place taking into account the change in
travel cost, beginning times, and fixed vehicle cost. The evaluation of these factors
is taken into account in the c1 (i, u, j) criterion calculation. After application of this
procedure, the best insertion place with the least insertion cost is chosen. Next,
for every customer u the c2 (u) criterion is calculated. It takes into account how
expensive is to reach this customer from the depot—direct travel cost c0u , what is
the cost of the smallest vehicle to serve the route with this customer, and what is
the cheapest insertion cost c1∗ (u). The calculation of these criteria is implemented in
SolomonInsertion().
Tabu Search for Fleet Size and Mix Vehicle Routing Problem ... 9
To better determine the best insertion place for the customer in terms of vehicle
costs, Dullaert et al.’s [7] approach AROS (Adapted Realistic Opportunity Savings)
10 M. Batsyn et al.
is used. The idea of Realistic Opportunity Savings (ROS) was suggested by Golden
et al. [11]. The principle of the method consists in taking into account the opportunity
to fill the unused capacity of a bigger vehicle needed after a new customer u insertion.
The new vehicle cost is taken into account only if the new total demand q q + qu
in the route is greater than the capacity Q of the current vehicle and requires at least
the vehicle of capacity Q . The difference F(q ) − F(q) shows how greater is the
fixed cost F(q ) of the smallest vehicle needed to serve the new total demand q than
the fixed cost F(q) of the current vehicle. The additional item F (Q − q ) is called
the opportunity savings. Here, Q − q is the unused capacity of the bigger vehicle
which gives an opportunity to add more customers with total demand not greater than
Q − q . So we can avoid using for such customers any vehicle having the capacity
not greater than Q − q . The fixed cost of the largest such vehicle is F (Q − q ).
As the tabu search implies, when the initial solution is known we should find the
neighbor solutions of it and choose one of them to be the current. Swap and relocate
operators are chosen as the neighborhood construction methods in the developed
algorithm. A notice should be made that while generating neighbor solutions it is
allowed to break restrictions regarding hard time windows and capacity constraints.
This is done to diversify the search because often the way to the best solution lies
through infeasible ones. After neighbor solutions are constructed, to avoid the sit-
uation of repeating solutions, the random one from the 3% of the best generated
solutions with the minimal function value is chosen as the current.
When a good feasible solution is found sometimes possible changes within each
route can be made which can lead to the decrease in the objective function value. Hard
time windows constraints set a certain sequence of customers in a route therefore
not all solution improvement methods are suitable for this operation. We considered
only neighbors’ swap. For details, see Braysy and Gendreau [2].
Besides taking into account soft time window violation penalty in the objective
function, the soft time window optimization procedure is developed. The principle
of it is in the following: for each route starting from its last customer, the method
tries to distribute soft time windows intervals with their smallest violations among
all the nodes in the route trying to use the latest preferred time where possible.
Tabu Search for Fleet Size and Mix Vehicle Routing Problem ... 11
4 Computational Experiments
The experiments are carried out on Intel Xeon X5690 3,47 GHz with the following
parameter values determined empirically: Stopping Criterion SC 1500, Tabu List
size |TL| 24, diversification parameter σ 0.833, the penalty for one time unit of
soft time windows violation α = 4. As the main goal of the article is to develop the
algorithm for the fleet size and mix vehicle routing problem with soft time win-
dows—the problem regarding which the published papers are not found—therefore
the effectiveness of the proposed heuristic is tested on the VRPTW problems. Our
results are compared with the latest VRPTW results from the article of Jawarneh and
Abdullah [15].
As we compare the performance of our algorithm with the one from the article
Jawarneh and Abdullah [15], the computation times for every instance are set nearly
12 M. Batsyn et al.
the same: from 1 to 10 min. This is done by setting an appropriate Number of Shakes
(NS parameter) in the main algorithm.
Table 1 shows that in 34 cases of the 56 our approach is better than the approach
proposed by Jawarneh and Abdullah [15]. In eight cases, our algorithm shows the
same performance and for the remaining 14 cases, we have a slightly lower per-
formance. The best results we have on random instances (R101–R211), where we
obtain better solutions on 18 instances of the 23, and our improvement reaches 10%.
We have good results on random clustered instances (RC101–RC208), where our
algorithm gives lower cost on 10 instances of the 16, and the maximal improvement
is 6.4%. Finally, on clustered instances (C101–C208), our best improvement is 4.4%;
we are better on six instances of the 17 and have the same result on eight instances.
As the fleet size and mix vehicle routing problem with hard and soft time windows
is not considered in literature we have generated instances for the computational
experiments. As a base, we take benchmarks of Liu and Shen [19] (Table 2) and add
soft time windows with different sizes of 10, 20, 30, and 40% of the hard time window
size. The position of the determined soft time window inside the hard one is chosen
randomly with uniform distribution. In each instance, there are several vehicle types
with different capacities and fixed costs denoted with letters A, B, C, D, E, and F. For
each VRPTW instance, we generate three instances with large, medium, and small
fixed vehicle costs marked with letters a, b, c correspondingly. In Tables 3, 4, 5 and 6,
the average results for each class of instances are presented. The generated instances
are available in the following link: https://nnov.hse.ru/en/latna/benchmarks.
The computational experiments show that the size of soft time windows influences
the value of the objective function: the larger is the soft time window interval the
better is the solution. However, the travel cost of routes in a solution can be even
smaller in case of narrow soft time windows, because only short routes can satisfy
such time windows. At the same time, such a solution will have more routes and a
greater fixed cost of used vehicles.
Tabu Search for Fleet Size and Mix Vehicle Routing Problem ... 13
Table 1 Comparison of the results for the VRPTW with Jawarneh and Abdullah [15]
Instances Jawarneh and Proposed heuristic Improvement, %
Abdullah
C101 828,94 828,94 0,00
C102 828,94 828,94 0,00
C103 835,71 832,07 0,44
C104 885,06 847,45 4,44
C105 828,94 828,94 0,00
C106 828,94 828,94 0,00
C107 828,94 828,94 0,00
C108 831,73 828,94 0,34
C109 840,66 828,94 1,41
C201 591,56 591,56 0,00
C202 591,56 591,56 0,00
C203 593,21 636,08 −6,74
C204 606,9 648,07 −6,35
C205 588,88 588,88 0,00
C206 588,88 588,49 0,07
C207 590,59 602,99 −2,06
C208 593,15 588,49 0,79
R101 1643,18 1626,42 1,03
R102 1476,11 1452,20 1,65
R103 1245,86 1232,13 1,11
R104 1026,91 979,63 4,83
R105 1361,39 1362,88 −0,11
R106 1264,5 1246,08 1,48
R107 1108,11 1075,48 3,03
R108 994,68 955,35 4,12
R109 1168,91 1178,57 −0,82
R110 1108,22 1114,86 −0,60
R111 1080,84 1081,08 −0,02
R112 992,22 991,46 0,08
R201 1197,09 1152,34 3,88
R202 1092,22 1077,94 1,32
R203 983,06 893,26 10,05
R204 845,3 775,19 9,04
R205 999,54 983,99 1,58
R206 955,94 929,16 2,88
R207 903,59 858,53 5,25
R208 769,96 749,70 2,70
(continued)
Another random document with
no related content on Scribd:
The back is
made with one full
row of squares,
and two rows of
buttons at the
bottom, besides
the finishing
squares (see Fig.
50), then the swell
of the back is
carried up to
within 4 inches of Fig. 50.
the upper edge,
1⅛ inches being allowed for the swell. Then the top of the back is
finished with a large roll, of about 5 inches swell or girth, so that the
back has only one row of buttons in the upper sweep.
The arm-pieces, Fig. 51, are made in a
peculiar way, and the modus operandi is
rather difficult of explanation. In the place of
the usual arm-piece block, a piece of plank,
2¾ inches wide by ½ inch thick, is fitted in
with the usual sweep to it. Now fit four pieces
of single fly buckram, to form as it were a
Fig. 51.
funnel, the shape of the arm-piece desired;
then sew seaming lace to the two edges of
the funnel, which will show inside of the body; to the lower edge sew
a piece of cloth in smooth, so that it will cover the bottom of the
funnel or cylinder. To the same lace edge, blind sew in a piece of
cloth, for the purpose of forming a wrinkled roll on the inner face of
the funnel, wrinkled 1 inch in fulness for every 3 inches in length, and
as full the other way as desired, for it ought to be full enough to
come out with the bottom side quarter. Next, blind sew the outer
edge of this roll to the other or top seaming lace, and stuff lightly with
hair, thus forming one roll on the inner face, and having the lower
face covered with the smooth cloth.
Now blind sew another piece of cloth to the top lace, as in the
other case, to form another wrinkled roll on top of the funnel or
cylinder, but the outer edge of this roll is to be finished by nailing to
the outside of the piece of plank just mentioned. All this sewing is of
course to be done on the bench, one side of the funnel to be left
open for this purpose. Next, nail in the bottom side quarter made up
in squares, and then nail the side of the funnel, which is fitted against
the arm-board, to the board and over the quarter, thus finishing the
lower part. Then nail the fourth on top side to the top edge of arm-
board; next, stuff from the front the funnel, pretty solid, and finish the
top roll, which up to now has been left open, into the outside of the
arm-board, thus completing the arm-piece, which shows two
wrinkled rolls divided by two rows of seaming lace.
The door fall, Fig. 52, is made on three
fly buckram pasted together, but one fly is
cut off about an inch from the top to allow
the fall to hinge. The fall is made about 12
inches deep, the lower edge being
circular. The broad lace is bent to the
required shape, and the corners sewed
Fig. 52. and put on to the buckram, which is cut to
the shape intended for the fall. Mark
where the inside edge of the lace comes
all round the buckram, then mark 1¼ inches from the mark; paste a
piece of carpet into the buckram to come within ½ inch from this
mark, which will make the edge of the carpet 1⅜ inches from the
edge of the lace; cover this carpet with a piece of cloth, pasting on
the buckram; take a piece of seaming lace, long enough to reach
round the fall, and sew a piece of cloth to it, for the purpose of
forming a wrinkled roll round three sides of the fall, inside the broad
lace and outside the sewed carpet, between both, with 1 inch fulness
to each 3 inches of length; sew this seaming lace and roll to the
buckram at the mark, 1¼ inches from the broad lace, gather the
other edge with a running string, and sew down and stuff lightly,
finishing in such a manner that the broad lace shall cover this sewed
edge; next, paste on the broad lace and cover the wrong side with
silk or muslin. When dry, stitch both edges.
Fig. 53. Fig. 54.
Figs. 53 and 54 show two styles of trimming for a door. Fig. 53 is
made as follows:—Paste out three flies of buckram, and lay off for
block or biscuit pattern, leaving space enough all round for a broad
lace border, and at the top leave double the space. The top space is
formed into a plain cloth roll of the same goods as the job is trimmed
with.
In this case the trimming is brown cloth; the broad lace is silk and
worsted of a shade much lighter than the cloth. The diamond-shaped
and connecting figures are worsted and are raised.
The card-pocket is made of tin and covered with Turkey morocco
the colour of the trimming.
In Fig. 54 it will be noticed that the style is somewhat different from
the other; the surface of the door is trimmed plain, the fall alone
being stuffed. The fall is stuffed in diamond form and enclosed with a
lace border. This pocket runs up under the fall to the top, and is there
nailed.
The following remarks on lining and trimming are taken from
“Cassell’s Technical Educator:”—
“We may with advantage say a word to our carriage lace-makers,
who seem to have made but scant use of the various Schools of
Design for the improvement of their taste in producing new and
suitable patterns in the manufacture of their goods. For a long period
we had nothing but the old scroll or flower pattern, which was
handed down from father to son as if by a fixed law. At length, when
it was felt that some change was required, the absence of all taste in
design was shown in the production of entirely plain worked laces,
which deprived carriage linings of their chief element of lightness and
beauty. Thanks, however, to the taste and discernment of Messrs.
Whittingham & Walker, who, perhaps, have devoted more attention
to this branch of industry than any other house in London, the trade
was relieved from the necessity of either adhering to the old pattern
or of adopting the opposite scheme. They introduced small neat
designs in laces eminently adapted to the purpose, and in 1857 they
registered a pattern, now extensively known as the double diamond
pattern, which has not only become general in England, but is largely
patronised throughout Europe and America. This, and kindred
patterns, exactly fulfil modern requirements, and give us the
necessary relief without extreme.
“But with materials well and tastefully selected the trimmer has still
his work to do. The lining of a carriage is divided into many different
parts, all of which have to be designed. Canvas or paper patterns
have to be cut to these, and properly fitted before the material is
touched with which the carriage is to be lined. Wherever superiority
of workmanship is to be shown in this department, the French
method of trimming is adopted as being more elegant than the
English. We shall therefore confine our observations to this method.
“In adopting the French method silk is mostly used in the place of
morocco, and its peculiarities consist in the manner of quilting. The
different squabs are made up in horizontal pipes or flutes, which are
tufted in different ways. To proceed, cut a pattern in strong paper the
size and shape of the space to be trimmed, and draw on it with a
pencil the pipes, also mark the position of the tufts. On large pieces
only mark one-half, the other half being the same. The pipes of the
back are usually 12 inches high and from 3 to 5 inches wide.
“The position of the tufts is considerably varied. Next stretch a
piece of strong muslin in the stretching frame, lay the paper pattern
on it, and mark the position of the tufts with an awl. Mark the lines of
the pipes on the muslin with red chalk or pencil. In the same manner
the pattern must be transferred upon the inside of the material used
for covering, making of course due allowance for the depth of the
pipes; about 3 inches is a fair average for fulness at the top, 1 inch
for the height of the pipes, and 1½ inches for the width. For the last
pipe an extra allowance is made in a narrow strip sewed on to it.
“Next lay a quantity of hair on the frame and form it the swell
desired. Keep the hair in position with a few long stitches, and lay
the silk over it. Commence tufting in the middle of the lower row of
pipes, and continue equally to both sides. Silk cords stretched into
the channels between the pipes were at one time considered
elegant, but their main merit was that they aided materially in
preserving the original shape of the pipes. Backs are usually made
couch-shaped, with a roll all round on the top, which at the same
time form the elbow-pieces on the sides. In elegant carriages this roll
is often elaborately executed in a helical or screw-like shape, and
continued from the door-pillar down to the seat-frame, being made
by winding silk cords around the roll. These silk cords appear as a
single thread, but in reality there are three different cords which are
wound at even distances. A style of trimming much used of late both
in France and Germany for low backs, is a row of pipes at the lower
end, which are pinched to points at the top, and above these are
three rows of regular squares. Squares are preferred to diamonds as
they are softer.
“Usually the back is laid on spiral springs, which are fastened as
follows:—The back of the body is covered with coarse muslin, after
being slightly stuffed, and on this muslin four rows of seven small
springs each are set. For the lowest row, springs a little stronger may
be used than for the other rows. The highest row is set about 1½
inches below the edge of the back-board, and the lowest row at 6
inches above the seat-frame. The springs having been sewn on with
a bent needle, are tied first from right to left and then from top to
bottom. A thin cord will answer for this purpose.
“The cord is first cut in lengths, and when the tying begins about 6
inches are allowed at the ends. The cord is wound about the third
ring of the first and last spring in each row, and afterwards the first
ring is brought into the right position with the piece of cord allowed
over. This will make the spring stand upright, and it can be raised or
lowered on one side. The springs being thus all placed in position,
they are finally tied crossways.
“The squab, in this instance, is worked in coarse muslin or canvas,
stiffened with a little thin paste. It is set in the frame and marked as
we have described above. When the cushions press against the
back and side pieces, frequently no stuffing is made, but simply a
piece of fine linen is sewed reverse to the main piece, and this is
called the ‘false finish.’ In fine work the stuffing extends clear to the
seat-frame.
“Of course each of these variations requires a different calculation
for the muslin at the back as well as for the cover. For the latter an
allowance of 1½ inches is made for the pipes from the lower to the
upper end, and also for the points an addition of ¾ inch. For each
square in the height 1½ inch has to be calculated. The folds of
squares when laid over springs being diagonal, easily draw apart
when stretched out, while the folds of diamonds running up and
down may be drawn tighter to a certain degree of stretching.
“For the upper row of squares we have to allow for the backs at
least double what we have to allow for the other rows—namely, 3
inches. For the width of every pipe an addition of 1½ inches is
calculated.
“Both cover and muslin being thus marked, we commence to draw
in the tufts. Every point marked on the cover has to lie exactly on the
corresponding one of the muslin. The lowest tufts are first drawn in;
then turn the frame and commence on one side at top, every point of
the pipes being singly stuffed and the folds adjusted. This being
done, every fold of the squares can be tufted right through, stuffed,
and folded. Squares are easier to be worked than diamonds, but
pointed pipes give more trouble than the ordinary straight ones.
Fig. 57.
It is desirable to direct attention to the proper horsing of carriages,
that the owners of carriages and horses may so adapt their plans as
to get the most satisfactory result from their arrangements. Not
unfrequently a carriage is ordered for one horse only; when it is
partly made, or perhaps finished, fittings are ordered for two horses;
and it sometimes happens that the two horses put to the light one-
horse carriage are coach-horses, between sixteen and seventeen
hands in height. Such horses, though well adapted to a family
carriage, are quite out of their place attached to a light one. Although
they can draw it at a good pace, and over almost any obstacle in the
road, and do their journey without fatigue, the carriage suffers
sooner or later. The lounging of such horses against a light pole, the
strain thrown on the pole in case of a horse tripping, the certain
breakage that must occur in case of a fall, and the risk of overturning
the carriage, should all be considered before putting a very light
carriage behind very large horses. It also sometimes happens that
miniature broughams and other very small carriages, built as light
and as slight as safety will allow, are afterwards used with a pair of
horses. In such cases, if accidents do not occur through the great
strain of a long pole acting as a lever on very light mechanism, the
parts become strained, do not work as they were intended to do, and
necessitate constant repair from not being adapted to the work put
upon them. Carriage owners should, in their own interest, have their
carriages and horses suited to what they ought and can undergo,
bearing in mind that there are advantages and disadvantages both
with heavy and light carriages. The former are easier and more
comfortable to ride in; they are safer for horses, drivers, and riders;
and the necessary repairs are less frequently required. The lighter
carriages follow the horses more easily, and can therefore do a
longer day’s journey; and, although the necessary repairs may come
more frequently, the saving of the horses may be an advantage that
many persons will consider of the utmost importance. Such light
carriages should, however, be made of the choicest materials and
workmanship, that they may do the work required of them.
A feature in the financial department of coach-making must not be
overlooked, as it has much influence on an important trade. In former
times a large proportion of the carriages were built to order for the
owners; the reverse is now the case; most persons select a finished
carriage which pleases their taste, or an advanced one, and get it
completed their favourite colour. This, of course, necessitates the
employment of a larger capital to meet the altered state of trade,
which now requires so large a stock of carriages to be kept ready for
use.
The excessive competition of recent years has so reduced the
profit on each carriage, that in order to carry on his business without
loss, the builder has to require a prompt payment from his customer
instead of giving a long credit.
The modern system enables the coach-builder to make his
purchases for ready money, and so buy not only better in quality but
at a less cost than for extended credit, in order that he might in his
turn give long credit to his customers, so that he is now obliged to
depend on small profits and quick returns by turning over his capital
more rapidly. He is not now, as much as in former times, the agent of
the persons who supply the materials that he and his workmen
convert into a carriage, but rather the designer, capitalist, and
director of those who seek his service or custom, whether to supply
labour or materials.
From the Government returns we find that carriages of all sorts
have increased from 60,000, in 1814, to 432,600 in 1874—a benefit
to the general population, it is clear, as well as to the workmen. In
1874, 125,000 carriages paid the Government duty.
The valuable library and fine series of photographs of state and
other carriages of the Coach and Coach Harness Makers’ Company
are open to coach artisans every Saturday afternoon. Tickets of
admission may be obtained at the principal coach-builders in
London.
In Calcutta there are several coach-builders of good reputation,
and who employ large numbers of native workmen. Messrs. Dyke
employ 600 hands; Messrs. Stewart and Co., 400; and Messrs.
Eastman, 300. The men are chiefly Hindoos, and are clever and
industrious, but have a singular habit of sitting down to their work.
Owing to the prejudices of the people in regard to the use of animal
fat, the labourers who have to use grease are chiefly
Mahommedans. The wages in the trade vary from sixpence to two
shillings per day.
“In Hindostan” (says Mr. Thrupp) “there are a large number of
vehicles of native build. It has been frequently remarked that there is
little change in Eastern fashions, that tools and workmen are
precisely as they were a thousand years ago, and the work they
produce is precisely the same. In examining, therefore, what is now
done by Indian coach-builders, we are probably noticing carriages of
a similar, if not identical, sort with those in use three thousand years
ago. The commonest cart in Hindostan is called ‘hackery’ by
Europeans; it is on two wheels, with a high axletree bed and a long
platform, frequently made of two bamboos which join in front and
form the pole, to which two oxen are yoked; the whole length is
united by smaller pieces of bamboo, tied together, not nailed. In
France two hundred years ago there was a similar cart, but the main