Capacity Expansion Problem for Large Urban Transportation Networks
Capacity Expansion Problem for Large Urban
Transportation Networks
Tom V.Mathew1and Sushant Sharma2
Abstract:A traf?c network design problem attempts to?nd optimal network expansion policies under budget constraints.This can be formulated as a bilevel optimization problem:the upper level determines the optimal link capacity expansion vector and the lower level determines the link?ows subject to user equilibrium conditions.The upper level is a capacity expansion problem which minimizes the total system cost and can be solved using any optimization algorithm.In the present study,genetic algorithm?GA?is used in the upper level because of its modeling simplicity and ability to handle large problems.The proposed model is?rst applied to a small sized network and then to a medium sized test network and the results are compared with other existing solution approaches.The sensitivity analysis of the model is performed by designing the networks at different demand levels.The resilience of the solution when demand increases the design demand is also carried out.Finally,the network design for the city of Pune,India was taken as a case study.This is a large sized network having1,131links and370nodes.The capacity expansion is carried out under various budget scenarios and the results are discussed.This study shows the potential of GA to obtain a high quality solution for large network design problems.
DOI:10.1061/?ASCE?0733-947X?2009?135:7?406?
CE Database subject headings:Network design;Optimization;Computation;Traf?c assignment;Urban areas;Traf?c capacity.
Introduction
One of the options that strikes transportation engineers and plan-ners,alike,while considering the growth in traf?c demand is to expand the capacity of existing congested links or build new links.Traf?c planners experience a number of constraints,while designing for urban areas.They have to overcome some of the invaluable socio-economic,environmental,budget,and space im-pediments to further development.
The planner has to make a decision while considering,on the one hand,maximization of bene?ts derived and minimization of total system cost under limited budget and on the other hand, behavior of the road users.In such cases,selecting new links and adding capacity to existing links of a network becomes an impor-tant problem.This problem is stated as a network design problem ?NDP?.
The objective of NDP is to achieve a system optimal solution by choosing optimal decision variables in terms of capacity ex-pansion values.This decision,taken by the planner,affects the route choice behavior of road users and needs to be considered while assigning traf?c to the network.In general,NDP can be formulated as a bilevel problem which has an upper level repre-senting a system optimal design and a lower level representing travelers route choice behavior.We may classify the lower level problem as the Nash game and whole NDP as the Stackelberg game.In game theory,the Nash equilibrium?named after John Nash,who proposed it?is a kind of optimal collective strategy in a game involving two or more players,where no player has any-thing to gain by changing only his or her own strategy?Fisk 1984?.On the other hand,a Stackelberg game,named after the German economist Heinrich von Stackelberg,is a leader-follower game where the leader takes a move and the follower responds to it?Lim et al.2005?.The key assumption is that the leader must have perfect information of the responses of the follower.Al-though the leader cannot intervene in the follower’s decision,he can modify the follower’s decision for his own advantage.In network design problems,the Nash game refers to the traf?c as-signment based on Wardrop’s user equilibrium principle.The Stackelberg game,then,refers to the link expansion decisions where the system designer anticipates the responses of the road users?Lim et al.2005?.
Network design models that are concerned with adding indi-visible facilities?for example,a lane addition?are said to be dis-crete NDP,whereas those dealing with divisible capacity enhancements?for example road widening?are said to be continu-ous NDP.It should be noted that discrete models can easily allow the investment to signi?cantly affect the mean free speed of pro-posed links;this seems to be dif?cult in the case of continuous models.Continuous models,on the other hand,have the advan-tage that the optimal levels of improvement?with the correspond-ing investment?for each link is determined by the model. Continuous network design models with convex investment costs usually result in minor increases in practical capacity of many links proposed for the improvement.This may be desirable if the purpose of the model is to improve or maintain the existing trans-portation network rather than to construct new roads?Abdulaal and LeBlanc1979?.In practice,networks used in transportation planning are quite large,and so the continuous network design
1Associate Professor,Transportation Systems Engineering,Dept. of Civil Engineering,Indian Institute of Technology Bombay,Powai, Mumbai400076,India?corresponding author?.E-mail:vmtom@civil. iitb.ac.in
2Research Scholar,Transportation Systems Engineering,Dept. of Civil Engineering,Indian Institute of Technology Bombay,Powai, Mumbai400076,India.E-mail:sushantsharma@iitb.ac.in Note.This manuscript was submitted on June15,2006;approved on December22,2008;published online on June15,2009.Discussion pe-riod open until December1,2009;separate discussions must be submit-ted for individual papers.This paper is part of the Journal of Transportation Engineering,V ol.135,No.7,July1,2009.?ASCE, ISSN0733-947X/2009/7-406–415/$25.00.
model appears to be a good compromise between network accu-racy and model sophistication.
Due to the intrinsic complexity of model formulation the net-work design problem has been recognized as one of the most dif?cult and challenging problem in transportation.In spite of substantial research in this area there is still a scope for improve-ment.In particular,an ef?cient convergent algorithm for large scale bilevel traf?c modeling and optimization is yet to be devel-oped?Yang and Bell2001?.Therefore,this study is an attempt to ?nd optimal capacity expansion for a large city network. Literature Review
The?rst discrete network design formulation was proposed by LeBlanc?1975?and later Abdulaal and LeBlanc?1979?extended it to a continuous version.This network design problem with continuous investment variables subject to equilibrium assign-ment was formulated as a nonlinear unconstrained optimization problem.Hook–Jeeves pattern search algorithm?H-J algorithm?and Powell’s algorithm were used for solving these models on a medium sized realistic network?Abdulaal and LeBlanc1979?. Since then,several variations of the network design problem were studied extensively.Optimization of road tolls under condi-tion of queuing and congestion?Yan and Lam1996;Yin2000?, optimization of reserve capacity of a whole signal controlled net-work?Wong1997;Yin2000?,estimation of trip matrix?Maher et al.2001?,and optimization of traf?c signal?Maher et al.2001?are some variations of the network design problems.All these problems are normally formulated as a bilevel programming problem in which the lower lever problems are either determinis-tic or stochastic user equilibrium.The upper level problems are variants of system optimum design with decision variables spe-ci?c to the problem at hand.There are some attempts to formulate them as an equivalent single level problem.For example,Meng et al.?2001?proposed an equivalent single level continuously dif-ferentiable optimization model for the conventional bilevel con-tinuous network design problem and solved it using Lagrangian algorithm.
Bilevel formulations of the network design problem are nonconvex and nondifferentiable and therefore getting global optimum solution is not easy?Yan and Lam1996?.These prob-lems are dif?cult to solve and designing ef?cient algorithms is still considered to be one of the most challenging tasks in trans-portation?Meng et al.2001?.Therefore,several solution ap-proaches have evolved over the past few decades.Initial approaches used heuristic algorithms,which may give near opti-mal solution or local optimum solutions?Steenbrink1974;Allsop 1974?,or methods like equilibrium decomposed optimization ?EDO??Suwansirikul et al.1987?,which are computationally ef-?cient but result in suboptimal solutions.Sensitivity analysis of user equilibrium assignment was proposed by Tobin and Friesz
?1988?.Yang?1995,1997?developed ef?cient sensitivity analysis based models for the traf?c control problem.A similar approach is adopted to?nd reserve capacity of signal controlled networks ?Wong and Yang1997?.However,the sensitivity analysis based algorithm lacks theoretical guarantees?Meng et al.2004?.Chiou ?1999?explored a mixed search procedure to solve an area traf?c control optimization problem con?ned to equilibrium?ows where local optimum solution can be effectively found using the gradi-ent search method.Chiou?2005?exploited a descent approach by implementing gradient based methods.Lim et al.?2005?followed a bilevel formulation similar to a Stackelberg game approach.The relation between link?ows and design parameters were formed as a closed form function by using a logit route choice model.
The heuristic solution approaches also have their own limita-tions as they are not suitable for large real networks problems since they require the approximate derivatives to have the same sign as original derivatives,which is dif?cult to verify in many cases?Meng et al.2001?.In a bilevel formulation,user equilib-rium problems at a lower level can be formulated as variational inequality problems?Dafermos1980;Smith1979?and this ap-proach was attempted by Marcotte?1983?.However,it suffers from the limitation of using path?ows in the formulation,which makes it dif?cult to use in large network problems?Meng et al. 2001?.With the advent of computing power,computationally de-manding tools with potential for easy implementation and global optimum solution are becoming increasingly popular.For in-stance,the simulated annealing?SA?algorithm is applied by Friesz et al.?1992?for network design problems.Although simu-lated annealing has the potential to obtain good solutions in a short time,it is not able to improve these solutions even if more time is given.In spite of various intriguing attempts to solve the bilevel programming problem,these algorithms are unfortunately either incapable of?nding the global optimum or require very complex gradient computation or unable to handle networks of realistic size?Yin2000?.Therefore,designing an ef?cient solu-tion algorithm for a large scale continuous NDP remains a chal-lenging task?Meng et al.2001?.
Genetic algorithm?GA?is yet another tool that has emerged as an ef?cient and simple implementation of several nonsmooth op-timization problems?Yin2000?.GA was successfully utilized for optimal road pricing and reserve capacity of a signal controlled road network?Yin2000?.The motivation of using GAs is due to its globality,parallelism,and robustness.In addition,GAs are simple and powerful in their search for improvement and not fundamentally limited by restrictive assumptions about the search space?assumptions concerning continuity,existence of deriva-tives,and other matters??Yin2000?.Unlike simulated annealing, GA is a slow starter and is able to improve the solution consis-tently given suf?cient time.Ceylan and Bell?2004?used a GA-based approach for traf?c signal control problems.The GA based model is also used for very large transit route network design and frequency setting problems?Aggarwal and Mathew2004?.All the bilevel algorithms have been tested only on very small test net-works and few have attempted medium size networks.However, there appears to be no study on the application of the model on large real networks.Therefore,this paper is an attempt to use genetic algorithm for capacity expansion problems,an important extension to bilevel network design problems focusing on its ap-plication to a real large scale network.
Model Formulation
Determining the optimal link capacity expansions is formulated as a bilevel continuous network design problem?CNDP?with the upper level problem minimizing the system travel cost subject to user’s travel behavior.This behavior is represented in the lower level using the Wardropian user equilibrium principles.The upper level problem is an example of system optimum assignment and can be solved using any ef?cient algorithm.
The following notation has been used for CNDP formulation: A=set of links in the network,q=vector of?xed origin destina-tion?OD?pair demands,q rs?q=demand from node r to node s; K=set of paths or routes between OD pair r and s;f=vector of
path?ows,f=?f k rs?;x=vector of equilibrium link?ows;x=?x a?, y=vector of link capacity expansions;y=?y a?,?a,k rs=1if route k between OD pair rs uses link a;and0otherwise;and g a?y a?=improvement cost function for arc a.The upper level problem is to minimize the total system travel time under a budget constraint and is formulated as
Upper level
Minimize Z y=?a?x a t a?x a,y a???1?subject to
??a g a?y a??B?2?
y a?0:?a?A?3?It should be noted that the term?g a?y a?represents the total im-provement expenditure and the budget constraints can be written
as??
a g a?y a??B.The upper level will give a trial capacity expan-
sion vector y a and will be translated into new link capacities.
Based on the new link capacity values,the link?ows can be computed by solving the following formulation:
Lower level
Minimize Z x=?a?0x a t a?x a,y a?dx?4?subject to
??k f k rs=q rs:k?K;r,s?q;?5?x a=?r?s?k?a,k rs f k rs:r,s?q;a?A;k?K?6?
f k rs?0:r,s?q;k?K?7?
x a?0:a?A?8?Note that the solution to the above formulation will give Ward-rop’s user equilibrium link?ows.The link cost function that this study uses is the popular Bureau of Public Roads?BPR?equation which is given by
t a?x a,y a?=t a o?1+?a?x a c a+y a??a??9?
where t a o=free?ow time on link a;x a=link?ow;and?a and ?a=link speci?c constants.The sum of the capacity expansion y a and the base capacity c a represents the improved link capacity. Since BPR Eq.?9?,is a monotonically increasing convex function and the travel time on the link a depends on the?ows on that link alone,the lower level formulation given by Eqs.?4?–?8?is convex.Therefore,there is a unique global solution and can be computed by any ef?cient convex combination method,like the Frank–Wolfe algorithm.One should note that the key model as-sumptions that have been considered are as follows:
1.The assignment is static,i.e.,deterministic user equilibrium
assignment;
2.The OD matrix is?xed and for peak hour traf?c;and
3.The road geometries other than width and length have not
been considered while calculating the improvement scheme. Solution Approach
A genetic algorithm based iterative procedure is adopted to get the solution.The?owchart of the solution approach is given in Fig.1.The algorithm starts in the upper level by reading all the inputs,like network details,demand matrix,budget,link expan-sion cost functions,and travel time function.The upper level algorithm will give a trial capacity expansion vector y?and will be translated into new network capacities.The upper level then invokes the lower level with these new link capacities.At the lower level,the demand matrix is assigned to the network based on the user equilibrium principles.This is accomplished by solv-ing the lower level problem given by Eqs.?4?–?8?,using the Frank–Wolfe algorithm.The application of this algorithm pro-vides user equilibrium link?ows x?.These link?ows are passed to the upper level.From the link?ow,the objective function, which is the sum of system travel time and the total link expan-sion cost,is computed.This objective function value is passed to the genetic algorithm operators,which then computes a new trial capacity expansion vector y?based on the objective function val-ues.This new trial capacity expansion vector is passed to the lower level.This procedure is repeated until convergence.
An important step in this method is the transformation of the problem into an equivalent formulation on which GA can apply its operators.Suppose the capacity expansion vector y=?y1,y2,y3ˉy n?:R n→R is mapped to the objective function value Z y??tness function in GA parlance?.Suppose further that
each decision variable y a can take the value from a domain
bounded by?y
a
min,y
a
max??R.Then,the decision variable y a requires?binary bites to represent it in a binary form.This string length?is the minimum value that satis?es the following relation:
?y a max?y a min?
?
??2??1??10?
where?=required precision?Goldberg1989;Deb1998?.To con-vert this binary string back to a decimal,the following linear mapping can be used:
y a=y a min+
?y a max?y a min?
?2??1???j=0??1b j2j
?,b j??0,1??11?
where b j represents the binary digit at j th position?Goldberg 1989?.Similarly,each variable of the y can be represented as a binary string.The binary string of each variable is concatenated to get the solution string or the chromosome which forms an in-stance of the solution?Fig.2?.The above coding is necessary as the GA operators can work only on the coded variable to generate
Fig.1.Flowchart representing solution approach
a new trial solution.The decoded variables are used in evaluating the trial solution by computing the corresponding objective func-tion value.The most important GA operators are reproduction,crossover,and mutation ?Goldberg 1989?.
The second aspect is to handle the budget constraint ???a g a ?y a ??B ?in the upper level formulation ?Eqs.?1?–?3??.GA is naturally suited for handling the unconstrained functions.Here,the constrained formulation is transformed into an equiva-lent unconstrained function ?k by an exterior penalty function method ?Rao 1996?as follows:?k =??x ,r ?=
?a ?x a t a ?x a ,y a ??+r ?Max ?0,??a
g a ?y a ??B ??
?12?
where,?k =unconstrained objective;r =penalty parameter;and
Max ?0,??a g a ?y a ??B ?=constraint violation term.The function value ?k is passed to GA to compute ?tness function.Example Network 1
To investigate whether the solution of the bilevel problem using genetic algorithm gives the global optimum value,an example network having four nodes and ?ve links ?Fig.3?is considered.
The network details,link parameters,and demand data are adopted from the study by Suwansirikul et al.?1987?.For the purpose of comparison,the upper level formulation given by Eqs.?1?–?3?is rewritten as follows:
Minimize Z y =?a ?x a t a ?x a ,y a ?+?g a ?y a ??
?13?
subject to
y a ?0:??A
?14?It should be noted that the term ?g a ?y a ?represents the total im-provement expenditure.The parameter ?=relative weight of im-provement costs and travel costs,or dual variable for the budget
constraints ?Suwansirikul et al.1987?.The lower level formula-tion is the same as reported by Eqs.?4?–?8?.In order to ascertain the effect of the genetic algorithm input values,a parameter study was conducted on this network.It was found from the study that the model gave the best performance at a crossover probability of 0.4,mutation probability of 0.05,and a population size of 10and 100generations.First,the genetic algorithm model is applied to this network.Then,a complete enumeration ?or exhaustive search ?method is adopted to ?nd the global optimal solution.Finally,these results are compared with solutions from other ex-isting major algorithms like modular in-core nonlinear optimiza-tion system ?MINOS ?,Hook-Jeeves ?H-J ?,and equilibrium decomposition optimization ?EDO ??Suwansirikul et al.1987?.The results are tabulated in Table 1.
The ?rst column of Table 1shows the optimum solution ob-tained by solving the problem by complete enumeration.Note that the complete enumeration gives the global optimum solution,and the objective function value obtained is 1,200.58units.The objective function values obtained by MINOS and genetic algo-rithm is the same as the solution obtained by complete enumera-tion.This shows the ability of GA to obtain a global optimum solution of NDP.Although MINOS is able to obtain global opti-mal solution it is not suitable for large networks ?Suwansirikul et al.1987?.Similarly,complete enumeration is prohibitively computational expensive even for small networks.Further,all three solutions gave different link capacities for the same objec-tive function values,suggesting multiple optimum solutions.In order to check the sensitivity of the demand,a sensitivity analysis was performed by varying the demands and compared with the solutions from existing algorithms ?Suwansirikul et al.1987?,and the results are reported in Table 2.Since the compu-tational effort needed to get the complete enumeration solu-
Table https://www.wendangku.net/doc/4215596906.html,parison of Results from Solving Example Network 1with Complete Enumeration and GA and Other Existing Algorithms Case demand 100Complete enumeration MINOS Hook-Jeeves
?H-J ?Equilibrium decomposition optimization ?EDO ?Genetic algorithm ?GA ?y 1 1.35 1.34 1.25 1.31 1.33y 2 1.20 1.21 1.20 1.19 1.22y 30.000.000.000.060.02y 40.950.970.950.940.96y 5 1.10 1.10 1.10 1.06 1.10Z y
1,200.58
1,200.58
1,200.61
1,200.64
1,200.58
Chromosome y (coded y)
1101110011 (11001)
y 1y 2..................y n
Fig.2.Representation of decision variables
4
Fig.3.Example 1:network having four nodes ?ve links
tion for all the demand levels of150,200,and300is quite high,
the values reported in Table2do not include solutions by
complete enumeration.The computational effort is high because
the precision required is0.01.The result shows the solutions ob-
tained by the GA models consistently outperforms the Hook-
Jeeves and EDO solutions and reaches very close to that of
MINOS.This analysis also con?rms the possibility of multiple
optimal solutions.
Example Network2:Sioux Falls Network
The second test network taken for the study is the Sioux Falls
network,which is probably the most extensively used test net-
work for the continuous network design problem.The Sioux falls
network,as shown in Fig.4,has24nodes,76links,and528
nonzero OD pairs.Among the76links,ten links?Links16,17,
19,20,25,26,29,39,48and74?are the candidate links for
capacity expansion.The network details and input parameters are
adopted from the study by Suwansirikul et al.?1987?.
The optimal capacity expansion vector and the corresponding
system cost is found using the proposed genetic algorithm model
and compared with other existing algorithms such as Hook-Jeeves ?H-J?,EDO,SA,sensitivity analysis based?SAB?,gradient pro-jection?GP?,conjugate gradient projection?CG?,quasi-Newton
projection?Qnew?,and Paratan version of gradient projection ?PT??Chiou2005?.The link expansion values and the objective function values from all these models and GA solution are tabu-lated in Table3.GA being a probabilistic model,the optimal solution is selected as the best outcome of several trials by vary-ing the random seed.GA*in Table3is the optimum solution thus obtained.It can be seen clearly from these results that the SA and GA are able to produce relatively good results;and among all the models GA is able to produce the best solution.It should be noted that in spite of relative closeness of objective function values the links expansion values are not consistent.This again con?rms that the problem has multiple optimal solutions.
In addition to the optimum solution,a suboptimal solution from one of the trial random seed reported in Table3as GA+. These two solutions have very close objective function values, but different expansion values.The purpose of reporting GA+is to study the sensitivity of the solution when demand exceeds the design value.The solution from SAB is noteworthy because it has very high expansion values for almost all the links.
In order to study the performance of the GA model with re-spect to other models,two types of sensitivity studies are con-ducted.First,network design is done at several demand levels. The base demand is multiplied by a factor?0.8,1.2,1.4,and1.6?and the model is applied.The total system cost and the number of Frank–Wolfe iterations performed under these demand levels are tabulated in Table4.
Table2.Sensitivity Analysis of Solution for Example Network1with GA and Various Other Existing Algorithms
Case MINOS Hook-Jeeves
?H-J?
Equilibrium
decomposition
optimization
?EDO?
Genetic
algorithm
?GA?
Demand=100
y1 1.34 1.25 1.31 1.33 y2 1.21 1.20 1.19 1.22 y30.000.000.060.02 y40.970.950.940.96 y5 1.10 1.10 1.06 1.10 Z y1,200.581,200.611,200.641,200.58 Demand=150
y1 6.05 5.95 5.98 6.08 y2 5.47 5.65 5.52 5.51 y30.000.000.020.00 y4 4.64 4.60 4.60 4.65 y5 5.27 5.20 5.20 5.27 Z y3,156.213,156.383,156.243,156.23 Demand=200
y112.9813.0012.8613.04 y211.7311.7512.0211.73 y30.000.000.020.01 y4 4.6410.3410.3310.33 y5 5.27117511.7711.78 Z y7,086.127,086.217,086.457,086.16 Demand=300
y128.4528.4428.1128.48 y225.7328.7526.0325.82 y30.000.000.010.08 y423.4023.4423.3923.39 y526.5726.5626.5826.48 Z y21,209.9021,209.9121,210.5421,210.06
The computational performance of these models is compared by observing the number of Frank–Wolfe ?FW ?evaluations,which eliminates the bias of the computing platforms.In addition,the literature supports such comparison in the form of a number of FW evaluations for design ?Chiou 2005?.The values written in parenthesis in Table 4are the number of FW iterations.Although the number of FW evaluations by GA is much higher than the
other algorithms,its solution obtained gives the lowest objective function value.The time taken by the GA model at a single de-mand level for network design of Sioux Falls was about 122s.The second sensitivity analysis is carried out to study the re-silience of the solution.This is done by loading a scalar product of the base demand matrix in the already expanded network ?that is the y a values are the same as that shown in Table 3?and com-pute the system cost under user equilibrium conditions.This will show the sensitivity of the solution when the demand changes after the construction.The system cost under various demand levels ?0.8–1.6?is shown in Table 5.A lower value indicates lower system travel time,and is an indication of higher model resilience when the demand varies.In this respect,the SAB model is least affected by the demand variation.However,this is due to ?as noted earlier ?the overestimation of expansion values at the expense of construction cost.To verify this,we have used the alternate GA solution ?GA +?,which incidentally gave higher val-ues of expansion with slightly higher system cost.As expected,the solution obtained in this case was much closer to the one obtained by SAB solution ?Tables 3and 5?.Thus,the difference between objective function values obtained by GA ?alternate so-lution ?and SAB is quite marginal.Overall,one can observe that the GA model is very resilient to the demand function.
Case Study
The objective of the case study is to demonstrate the ef?cacy of a GA based model for large capacity expansion problems.The net-work of Pune city in Maharastra state,India,is considered for the case study.This city has an area of 138km 2.The city is divided into 97zones out of which 85are internal zones and 12are external zones.There are 273road nodes and 97zone centroids with 1,131road links.Fig.5shows the network of the city.
The demand data includes 4,083nonzero origin-destination pairs.The peak hour trips within the network are 118,428passen-ger car units ?PCU ?.Various traf?c ?ow parameters like ?a ,?a ,free ?ow speed,and running speed for all the links were found after surveying.The parameters for the link travel time functions were developed for various classes of roads based on extensive
Table https://www.wendangku.net/doc/4215596906.html,parison of Results from Solving Sioux Falls Network with GA and Solution of Various Existing Algorithms Case H-J H-J EDO SA SAB GP CG QNew PT GA *GA +Init.y a 2.0 1.012.5 6.2512.512.512.512.512.5——y 16 4.8 3.8 4.59 5.38 5.74 4.87 4.77 5.30 5.02 5.17 5.24y 17 1.2 3.6 1.52 2.26 5.72 4.89 4.86 5.05 5.22 2.94 3.54y 19 4.8 3.8 5.45 5.50 4.96 1.87 3.07 2.44 1.83 4.72 4.85y 200.8 2.4 2.33 2.01 4.96 1.53 2.68 2.54 1.57 1.76 2.89y 25 2.0 2.8 1.27 2.64 5.51 2.72 2.84 3.93 2.79 2.39 1.75y 26 2.6 1.4 2.33 2.47 5.52 2.71 2.98 4.09 2.66 2.91 2.04y 29 4.8 3.20.41 4.54 5.80 6.25 5.68 4.35 6.19 2.92 6.03y 39 4.4 4.0 4.59 4.45 5.59 5.03 4.27 5.24 4.96 5.99 5.47y 48 4.8 4.0 2.71 4.21 5.84 3.76 4.40 4.77 4.07 3.63 5.77y 74 4.4 4.0 2.71 4.67 5.87 3.57 5.52 4.02 3.92 4.43 5.68Z y
82.50
82.61
84.50
81.89
84.38
84.15
84.86
83.19
84.19
81.74
82.88
Note:GA *=optimal solution obtained by GA,GA +=alternate solution generated to compare with SAB algorithm.In order for a uniform comparison,the objective function values computed from the capacity expansion values are by using the Frank–Wolfe algorithm implemented by the writers for this study.Therefore,the results may slightly differ from the one’s reported by Chiou ?2005?,but are on the same scale.Hook-Jeeves ?H-J ?,equilibrium decompo-sition optimization ?EDO ?,simulated annealing ?SA ?,sensitivity analysis based ?SAB ?,gradient projection ?GP ?,conjugate gradient ?CG ?,quasi-Newton projection ?QNew ?,Partan ?PT ?,and genetic algorithm ?GA ?
.
Fig.4.Example 2:Sioux Falls network having 24nodes,76links
?eld data.Link characteristics were collected and a trip table was generated for the whole network after validating the OD counts. Sample link parameters are shown in Table6.
In order to study the performance of the CNDP on this net-work,different scenarios,with budgets of25,50,100,and250 crores rupees?1crore rupee=$0.22million?have been consid-ered.The maximum capacity expansion of each link is assumed to be100%.The model is applied with GA parameters like popu-lation size of100,two point crossover with a probability of0.8, and a mutation probability of0.01.These parameters of the ge-netic algorithm run are chosen based on the experience gained from the study of the example networks and the values suggested in the literature for similar problems?Yin2000?.The convergence criteria adopted is the completion of10,000generations.This large value is suggested to ensure the convergence of the prob-lem.The computation was carried out in a Intel Xeon processor with four clusters and the clock speed of3.2GHz,1GB RAM and LINUX Fedora Core4operating system.The computation time taken by a single run for?nding optimal capacity expansion was about72h.
The optimal capacity expansion values under each budget sce-nario was obtained by the proposed CNDP algorithm The abso-lute network performance for all the budget scenarios is shown in Table7and the relative improvement with the base case is shown in parenthesis.From these results,it can be inferred that the av-erage travel time is reduced by26%with a spending of rupees25 crore?$5.5million?and it continues to reduce until it reaches 39%at higher budget.However,with a budget beyond rupees100 crore,there is only a marginal increase in the average travel time improvement.The corresponding improvements in the average link speeds are about5–9%and the maximum link speeds are only about2–3%.On the other hand,in the case of minimum link speeds,the improvement is in the range of16–96%.This is ex-pected since the algorithm selected the heavily congested links ?rst and maximum expansions were made.Therefore,the mini-mum travel time improved signi?cantly when the speed is im-proved under higher capacity.This in turn might result in diversion of traf?c to such links.Furthermore,the parameters of the BPR function were calibrated at the base capacity values.In reality,these values may change at higher capacities.However, this aspect is not considered,which may result in relatively low improvement of travel speed.In addition,this analysis also gives an insight into the optimal budget allocation level.
The performance of the network improved signi?cantly at lower budget as evident from Table7.However,the increase in budget beyond about100crore does not make any signi?cant difference to the network.The actual spending for each scenario is also shown in Table7.The actual spending is the product of the unit cost of construction and the proposed capacity expansion. Ideally,one would expect the budget and actual spending to match even at a very low and very high budget,but it is dif?cult to ensure constraint satisfaction at a very low and very high bud-get?see Table7for25crore and rupees250crore scenario?.At25 crore,the constraints are clearly violated.This is not unexpected, since we have adopted an exterior penalty function method to handle the constraints where the intermediate solution may lie in the infeasible region.It only suggests that the solution has not converged within speci?ed iterations.This is understandable since the budget of rupees25crore is extremely low for such a big and congested network.It is quite clear from the table that even at a
Table4.Sensitivity Analysis of Different Solutions Obtained for Sioux Falls Network with GA and Various Other Existing Algorithms at Various Values of Demand
Scalar/algorithm SAB GP CG QNew PT EDO IOA GA
0.8051.7648.3848.7848.8448.8149.5153.5848.92
?14??10??3??4??9??7??28??59?
1.0084.218
2.7182.538
3.0782.5383.5787.3481.74
?11??9??6??4??7??12??31??77?
1.20144.86141.53141.04141.6214
2.27149.39150.99137.92
?9??11??10??7??9??12??31??67?
1.40247.8246.04246.0424
2.74241.0825
3.39279.39232.76
?15??9??6??5??7??17??16??78?
1.6045
2.0143
3.64408.45409.04431.11427.56475.08390.54
?14??9??9??9??11??19??40??83?Note:Values written in bracket are number of Frank–Wolfe iterations performed.The different objective function values are for optimal expansion in each case,i.e.,every time the expansion values generated will be different for different demands.Sensitivity analysis based?SAB?;gradient projection?GP?; conjugate gradient?CG?;Partan?PT?,equilibrium decomposition optimization?EDO?,iterative optimization assignment?IOA?,and genetic algorithm ?GA?.
Table5.Sensitivity Analysis of Solutions Obtained by Varying Demand
Scalar/case H-J H-J EDO SA SAB GP CG QNew PT GA*GA+
Init.y a 2.0 1.012.5 6.2512.512.512.512.512.5——
0.849.9250.2750.1150.5854.9451.6851.7951.8851.8550.6252.43
1.08
2.6182.5084.5081.8984.8384.758
3.3783.848
4.7781.7482.88 1.2144.49141.75150.40140.48139.84147.95143.56143.58146.95143.09140.49 1.4260.64253.32273.30248.99240.8264.89253.48258.19263.0925
5.1024
6.05 1.6463.02448.84485.83439.05416.69465.2144
7.2145
8.5646
9.98446.05427.83 Note:GA*=final solution obtained by GA;GA+=alternate solution generated to compare with SAB algorithm.Hook-Jeeves?H-J?;equilibrium decom-position optimization?EDO?;simulated annealing?SA?;sensitivity analysis based?SAB?;gradient projection?GP?;conjugate gradient?CG?;quasi-Newton projection?QNew?,Partan?PT?,and genetic algorithm?GA?.
higher budget the actual spending are much less than the amount allotted.This may be attributed to the fact that only congested links are expanded and spending beyond this will result only in the escalation of construction cost.Further,this analysis also give an insight into the desirable budget for network improvement.The convergence history for each budget scenario is shown in Fig.6.From the graph,one can observe a very good conver-gence,especially at higher budget levels.This may be because demand pairs have very few alternate paths,whereas at a lower budget the shortest path will change even when there is only a slight change in the capacity,which results in several more itera-tions converging.
Conclusion
This study is an attempt to solve large network capacity expan-sion problem using genetic algorithm.This network design prob-lem has been formulated as a bilevel problem where the upper level determines the optimal link capacity expansion vector and the lower level determines the link ?ows subject to user equilib-rium conditions.The upper level is solved by genetic algorithm and the lower is solved using the Frank–Wolfe algorithm.The upper level module will give a trial capacity expansion vector and will be translated into new network capacities.This then in-vokes the lower level problem which gives user equilibrium
Table 6.Sample Link Input Parameters for Pune Network
From To Link length ?km ?Number of lanes Alpha ?a Beta ?a Initial capacity ?PCU/h ?Free ?ow speed ?km/h ?128340 1.0010.70 2.03,20040114258 2.2410.68 2.02,60035222254 1.7010.65 2.21,10035211253 2.0210.68 2.56004098255 1.6810.70 1.52,6005012367 1.0010.65 1.51,00035178
138
1.10
1
0.68
2.5
4,350
40
Table https://www.wendangku.net/doc/4215596906.html,parison of Improvement in Traf?c Flow Characteristics with Different Budgets Budget in rupees.?crores ?a Average travel time ?min ?
Average link speed ?km/h ?Minimum link speed ?km/h ?Maximum link speed ?km/h ?Total system
cost ?veh./h ?Actual spendings ?crores ?Base case 25.2325.79 6.6352.9149,799—25
18.6327.097.7953.8436,78148.15?26%?
?5%??17%??2%??26%?—5017.14
27.359.1654.3833,84650.01?32%?
?6%??38%??3%??32%?—10015.50
28.0513.0353.4530,59799.81?39%?
?9%??97%??1%??39%?—25015.34
28.2112.5753.9230.298122.63?39%?
?9%?
?89%?
?2%?
?39%?
—
a
Crore rupees=$0.22million.
N
1Km 01234
5678
91011121314151617181920212223242526272829303132333435363738394041424344454647484950
51
525354
5556575859606162636465666768697071727374
7576777879808182838485
86
87
88
8990
91
92
93
94
9596
97PUNE CITY
Fig.5.Pune city network with links and zone centroids
2e+06
4e+066e+068e+061e+071.2e+071.4e+071.6e+071.8e+072e+07
2000
4000
6000800010000
T o t a l S y s t e m T r a v e l C o s t
Number of Iterations
Fig.6.Convergence history for typical budgets
?ows.The link?ows are then passed to upper level.The upper
level then computes the objective function value,which is the
total system travel cost and the investment cost.This objective
function value is given to the genetic operators,which supplies a
new capacity expansion vector,and the process is repeated until
convergence.
The working of the proposed model is?rst tested with the help
of a hypothetical network reported in the literature.This study
demonstrates the ability of genetic algorithm to arrive at global
optimal solution.Next,a medium sized network,Sioux Falls,is
tested with the present model.The solution from this is compared
to all major existing solution approaches reported in the literature.
This study showed that the present model is able to give a solu-
tion much better than the rest of the approaches.Only the solution
using the simulated annealing approach is coming closer to the
present model.Although simulated annealing has the potential to
obtain good solutions in a short time,it is not able to improve
these solutions even if more time is given,while GA is a slow
starter that is able to improve the solution consistently when
given suf?cient time.In addition it is possible that SA may
discard potentially“good”solutions because it retains only a
single solution unlike GA which retains a population of solutions ?Thompson and Bilbro2000?.The sensitivity analysis of the model is performed by designing the networks at different de-
mand levels.The resilience of the solution when demand in-
creases the design demand is also carried out.This advocates the
use of GA as it offers a resilient solution.Finally,to achieve the objective of developing a network design procedure for the large transportation network,a case study is considered.Accordingly, the network of Pune city,India having1,131road links and370 nodes is solved under different budget levels.The results shows a large reduction in total system cost and average travel time with respect to the base case.The signi?cant increase in minimum speed of the links shows the ef?cacy of algorithm to select the most congested links for improvement.
The proposed model will be a very useful tool for planners for allocation of budgets and prioritization of links for improvements. Although GA may be computationally intensive and may take more time in?nding the optimal solution for a realistic large network design the computation time can be signi?cantly re-duced.Simple or canonical genetic algorithm is used in the present study.There are several ways of reducing the computa-tional time of GA like dynamic population size,distributed computing,hybrid systems,etc.Further,one should consider the fact that capacity expansion designs are not done frequently. Therefore,the limitation of high computational time may be offset by the bene?t from a better solution.Finally,GAs are identically suited for various extensions of the network de-sign problem like considering emission cost,demand uncertainty, etc.One more important aspect of network design is the timing and prioritization of improvements.Modeling the timing and prioritization of improvements will de?nitely improve bene?ts. For example,the projected demand will not be realized imme-diately.This can be accommodated by considering the demand realization over several horizon years.This approach will opti-mally decide,for instance,the links to be extended and so during various time periods within the study duration.However,this requires much more complex modeling.The complexity is at-tributed to factors like:?1?incorporation of multiple OD matrices in a single optimization problem;?2?resulting computation time and variable explosion;and?3?consideration of time value of money typically the net present value?NPV?.Some other possible extensions of the problem are to consider the variability in the demand?Ukkusuri et al.2007?and variability in travel time?Chen et al.2002;Clark and Watling2005?.Finally,im-provement in computational time by ef?cient GA implementation, the hybrid GA model,distributed computing,etc.,need to be explored.
Acknowledgments
The writers are grateful to Mr.Mallikarjun Setty and Dr.Tagore of the CES organization for providing Pune city data.They are also grateful to reviewers for their useful comments and suggestions.
References
Abdulaal,M.,and LeBlanc,L.J.?1979?.“Continuous equilibrium net-work design problem.”Transp.Res.,Part B:Methodol.,13,19–32. Aggarwal,J.,and Mathew,T.V.?2004?.“Transit route network design using genetic algorithms.”https://www.wendangku.net/doc/4215596906.html,put.Civ.Eng.,18?3?,248–256. Allsop,R.E.?1974?.“Some possibilities for using traf?c control to in-?uence trip distribution and route choice.”Proc.,6th Int.Symp.on Transportation and Traf?c Theory,D.J.Buckley,ed.,Elsevier,New York,345–375.
Ceylan,H.,and Bell,M.G.H.?2004?.“Traf?c signal timing optimization based on genetic algorithm approach,including driver’s routing.”
Transp.Res.,Part B:Methodol.,38,329–342.
Chen,A.,Ji,Z.,and Recker,W.?2002?.“Travel time reliability with risk sensitive travelers.”Transportation Research Record.1783,Transpor-tation Research Board,Washington,D.C.,27–33.
Chiou,S.W.?1999?.“Optimization of area traf?c control for equilibrium network?ows.”Transp.Sci.,33?3?,279–289.
Chiou,S.W.?2005?.“Bilevel programming for continuous network de-sign problem.”Transp.Res.,Part B:Methodol.,39,361–383. Clark,S.,and Watling,D.?2005?.“Modeling network travel time reli-ability under stochastic demand.”Transp.Res.,Part B:Methodol., 39,119–140.
Dafermos,S.?1980?.“Traf?c equilibrium and variational inequalities.”
Transp.Sci.,14?1?,42–54.
Deb,K.?1998?.Optimization for engineering design:Algorithms and examples,Prentice-Hall,New Delhi,India.
Fisk,C.S.?1984?.“Game theory and transportation systems modelling.”
Transp.Res.,Part B:Methodol.,18,301–313.
Friesz,T.L.,Cho,H.J.,Mehta,N.J.,Tobin,R.L.,and Anandalingam,
G.?1992?.“A simulated annealing approach to the network design
problem with variational inequality constraints.”Transp.Sci.,26?1?, 18–26.
Goldberg,D.E.?1989?.Genetic algorithms in search,optimization and machine learning,Addison-Wesley,Reading,Mass.
LeBlanc,L.J.?1975?.“An algorithm for discrete network design prob-lem.”Transp.Sci.,9,183–199.
Lim,Y.,Heydecker,B.G.,and Lee,S.?2005?.“A continuous network design model in stochastic user equilibrium based on sensitivity analysis.”J.Adv.Transp.,39?1?,63–79.
Maher,M.J.,Zhang,X.,and Vleit,D.V.?2001?.“A bilevel programming approach for trip matrix estimation and traf?c control problems with stochastic user equilibrium link?ows.”Transp.Res.,Part B:Meth-odol.,35,23–40.
Marcotte,P.?1983?.“Network optimization with continuous control pa-rameters.”Transp.Sci.,17,181–197.
Meng,Q.,Lee,D.-H.,Yang,H.,and Huang,H.-J.?2004?.“Transportation network optimization problems with stochastic user equilibrium con-straints.”Transportation Research Record.1882,Transportation Re-search Board,Washington,D.C.,113–119.
Meng,Q.,Yang,H.,and Bell,M.G.H.?2001?.“An equivalent con-
tinuously differentiable model and a locally convergent algorithm for the continuous network design problem.”Transp.Res.,Part B: Methodol.,35,83–105.
Rao,S.S.?1996?.Engineering optimization:Theory and practice,Wiley, New York.
Smith,M.J.?1979?.“Existence,uniqueness,and stability of traf?c equi-libria.”Transp.Res.,Part B:Methodol.,13,295–304. Steenbrink,P.A.?1974?.Optimization of transport network,Wiley,West Sussex,U.K.
Suwansirikul,C.,Friesz,T.L.,and Tobin,R.L.?1987?.“Equilibrium decomposed optimization:A continuous network design problem.”
Transp.Sci.,21?4?,254–263.
Thompson,D.R.,and Bilbro,G.L.?2000?.“Comparison of genetic algorithm with simulated annealing algorithm for design of an ATM network.”IEEE Commun.Lett.,4?8?,267–269.
Tobin,R.L.,and Friesz,T.L.?1988?.“Sensitivity analysis for equilib-rium network?ows.”Transp.Sci.,22?4?,242–250.
Ukkusuri,S.V.,Mathew,T.V.,and Waller,S.T.?2007?.“Robust net-work design problem under demand uncertainty.”Comput.Aided Civ.
Infrastruct.Eng.,22,6–18.Wong,S.C.?1997?.“Group-based optimisation of signal timings using parallel computing.”Transp.Res.,Part C:Emerg.Technol.,5?2?, 123–129.
Wong,S.C.,and Yang,H.?1997?.“Reserve capacity of a signal con-trolled road network.”Transp.Res.,Part B:Methodol.,31?5?,397–402.
Yan,H.,and Lam,W.H.K.?1996?.“Optimal road tolls under conditions of queuing and congestion.”Transp.Res.,Part A:Policy Pract., 30?5?,319–332.
Yang,H.?1995?.“Sensitivity analysis for queuing equilibrium and network?ows and its application to traf?c control.”https://www.wendangku.net/doc/4215596906.html,put.
Modell.,22,247–258.
Yang,H.?1997?.“Sensitivity analysis for elastic demand network equi-librium problem with applications.”Transp.Res.,Part B:Methodol., 31,55–70.
Yang,H.,and Bell,M.G.H.?2001?.“Transport bilevel programming problems:Recent methodological advances.”Transp.Res.,Part B: Methodol.,35,1–4.
Yin,Y.?2000?.“Genetic algorithms based approach for bi-level program-ming models.”J.Transp.Eng.,126?2?,115–120.
联想新员工培训分析教学内容
联想新员工培训分析 （一）培训方案 联想对新员工的指导从其进入公司的第一天开始。上班第一天，有新员工的“首天培训”，主要是熟悉工作环境和要求；紧接着是所在公司的新员工入职培训、集团的“入模子”培训及后续的轮岗培训，之后，新员工就可以返回各自的部门和岗位。前两个培训各需—周，轮岗通常需1—2周，经过这三个规范过程，新员工的培训内存就告一段落。 这当中的重点是集团统一的“入模子”培训。培训班为全脱产封闭式，共5天，其中一天是军训。课程包括联想简介、发展历史、奋斗目标、经营理念、行为规范和团队训练等，以及一系列团队活动，包括卡拉OK比赛、拔河比赛、篮球比赛等。培训班将一天分成几个时间段，每一段时间都作了统一安排。培训班的管理制度是依照公司对员工的要求来做，有严密的纪律，每天早晨出操点名，白天学院上课时，还要组织检查内务。从早晨起床到晚上熄灯都要对每个小组进行考察、打分。如上课迟到就要罚站，并对所在小组扣分，还会在最后的培训鉴定上标明。 培训班中，团队的协作作为—项重要内容贯穿始终。新学员刚刚进入培训班就被分为若干个小组，每组不超过20人，每个小组选拔3—4位骨干。之后的每一天，每一项活动都将按组进行打分，学习效果和最后成绩与这些分数有直接关系。
学习过程始终重视学员的参与，从互动式教学到各种团队活功，讲求人人参与，协调配合，积极竞赛。根据活动内容不同，各组组长会将组内人员进行不同组合和分工，每个人都有机会承担某一项活动的主力，一方面发挥新员工所长和能动性，另—方面注意挖掘每个人的潜能，有机会锻炼自己的弱项。因此，不论这个新员工是怎样的一个人，都有机会展示、发现并提升自己。 在拓展训练中，设有盲人行动、建寺庙和过硫酸河游戏、背摔等，这些训练主要为了提高学员的心理承受能力。如背摔，学员要站在一个高高的台子上闭着眼睛脸朝上似自由落体地摔向下面围起圈子的人们。为提高新员工的表达与胆量，培训班设有一个即兴主题演讲内容，每个新员工在拿到演讲题目后有两分钟的准备时间，然后他要站到一个很高的台子上进行两分钟的演讲，不管你是情急下的语无伦次，还是紧张的心跳、腿软甚至不知所措，你都必须坚持下来，完成演讲。类似这样的训练在子公司各自的培训课程完成之后，还要在集团统一的培训班上再度进行过关训练。几次下来，新员工能了解所必须具备的各种能力和自信心、他们会在今后的岗位上得到进一步的锻炼和提高。 联想入职培训的另一个特点是新员工指导人。公司在每一位新员工到岗之前都要指定一对一的指导人。指导人负责带新人并考察新员工在试用期间的表现能力等，其作用一是代行了人力资源部的考察职责，二是通过帮带行使部门职责。指导人已经在联想形成—种制度，对指导人的选择，公司也有一系列规范，包括要求和资格认定及指导
联想新员工培训内容
联想新员工培训内容 联想新员工培训后，在入职之初，都会发一份联想员工行为操守准则。这份准则共六条，规范了作为一名联想人应该具备的基本原则，包括做人诚实一致、做事公司分明、业务活动适度、对业务伙伴公正廉洁等。 在遵守公司制度履行岗位职责的基础上，认同联想文化。核心文化价值观是：“服务客户，精准求实，诚信共享，创业创新”。 联想认为，企业是有血型的，符合这个血型的人，就能成为联想的员工；不符合这个血型的人，联想与之无缘。联想管理学院就是要培养出具有联想血型的人。 联想需要三种血型的人： 第一种就是能独立做事的人； 第二种是能够带领一帮人做事的人； 第三种能审时度势，把握全局的领军人物。 如果你还不知道自己被贴上什么标签，我们奉劝你要赶紧行动起来，最好是用最快的速度找到这个标签，并赶紧到自己的大脑上。只有这样，你才能自信地告诉大家：我属于这个企业的一分子。 否则，你就很有可能面临“走人”的危险——因为，绝大多数的企业是不能容忍一个“异类”的存在的。 联想新员工培训包括企业文化培训、专业技能培训和联想员工时间管理培训。联想十分重视员工的时间管理，认为员工的工作不是从自己的任务开始，而是从掌握时间开始的。只有有效管理自己的时间，
减少非生产性工作所占用的时间，才是当务之急。 附：联想新员工时间管理培训 一、认识自己的时间 1. 记录自己的时间； 2. 管理自己的时间； 3. 集中自己的时间。 二、如何诊断自己的时间 第一步：要记录自己时间耗用的实际情形。 必须在“当时”立即加以记录，长期积累就可以保持一份时间记录簿，以便于自己检讨；你会发现自己的时间花费往往很乱，常常浪费在无谓的小事上，但是经过练习并且持之以恒，时间的利用是会有进步的。 第二步：先将“非生产性”的和“浪费时间”的活动找出来，消除这类活动；要做到着一点，有赖于自己发问下面的几个问题： 浪费时间，无助于成果。将时间记录本拿出来，逐项地问：“这件事如果不做，有何后果？”假使认为：“不会有任何影响”，那么要即刻停止,对于那些出于礼节性的事情,只要审度一下婉绝即可； 事实上，人总有一种倾向，高估自己的地位的重要性，认为许多事是非要恭亲不可，然而，实际上并非这样；在时间记录所载活动中,如果哪一项另请他人做更好,就要学会“授权给别人”，这样才能真正做好自己该做的工作。 有一项时间浪费的因素，我们是必须消除的：不要浪费别人的时间！
新员工培训存在的问题与对策
【摘要】及时、规范、全面的新员工入职培训是企业人力资源管理中不可忽视的一个重要环节。但有一部分企业在新员工入职培训实践中，存在培训缺乏系统性、规范性、培训内容简单及培训效果缺乏反馈和评估等误区。现代企业可通过树立以战略为导向的培训理念、加强新员工入职培训过程管理、确定针对性的新员工入职培训内容、选择灵活多样的新员工培训形式和建立完善的新员工入职培训效果反馈及评估体系等措施来提高培训的效果。 【关键词】新员工入职培训误区策略 新员工入职培训是企业为新进员工所专门设计并实施的培训，它在为企业塑造合格员工、传承企业文化、建设成功团队、赢得竞争优势等方面有着十分重要和独特的作用。新员工入职培训是一个企业录用的员工从局外人转变为企业人的过程，是员工从一个团体融入到另一个团体的过程，是员工逐渐熟悉、适应企业环境并开始初步规划职业生涯、定位自己的角色、开始发挥才能的过程。及时、规范、全面的新员工入职培训是企业人力资源管理中不可忽视的一个重要环节。日本松下电器公司的“入社教育”，联想集团的“入模子培训”，无不体现了对新员工入职培训的重视。研究新员工入职培训的误区及改进策略，有着重要的实践意义。 一、新员工入职培训中存在的误区 德隆商务咨询公司一项针对企业培训状况的调查显示在我国12个行业的百家企业中，有32%的企业根本不提供任何员工培训，有15%的企业只为员工提供最简单的入职培训。而在员工培训的满意度调查中，员工对入职培训的不满意高达67%。另据有统计表明，国内有近78%的企业没有对新进员工进行有效的培训，往往把新员工入职培训当作一个简单“行政步骤”。很多企业在新员工入职培训中存在一些误区。 1.新员工入职培训缺乏系统性和规范性 主要表现为：（1）很多企业在培训的时间安排上，随意性很大。如由于企业营业的需要，有些企业以工作忙，人手不足为借口，马上分配新招到的员工到相关部门开始工作，而不顾及对新员工的培训，只等有时间了再派新员工参加培训。这种无序的培训给培训部门带来了不必要的协调工作量，增加了培训的次数与时间成本，并且新员工没参加培训既增加了企业用人的风险性，又不利于新员工角色的迅速转换；（2）企业缺乏规范性的培训文本或讲义。为了维护企业培训水平，企业都应有自己的培训资料，对新员工入职培训必须掌握的知识、技能和态度都必须设定目标进行考核，只有这样才能使培训工作有持续性，并且能避免企业的后续培训的内容重叠和资源的浪费把新员工培训变成新员工欢迎典礼。许多企业在培训形式上，主要是领导发言、代表致辞、体检聚餐。 2.新员工入职培训内容简单 很多企业把新员工入职培训当作一个简单“行政步骤”，认为新员工培训只要了解一些企业基本情况则可，所谓的培训也就是安排一天或两天时间，在内容安排上主要是参观企业、讲解员工手册与企业的一些基本规章制度等培
联想员工轮岗制度
前言 2 第一章关于岗前培训与指导 3 第二章服务工程师指导12 第三章备件专员指导15 第四章非业务岗位新员工指导17 第五章新任站长指导18 附录1 岗位规范培训四步骤20 附录2 表格汇编22 表1 技术类新员工站内培训指导时间推进表22 表2 备件新员工站内培训指导时间推进表23 表3 新任站长站内实习指导时间推进表24 表4 岗位责任书学习和交流表25 表5 岗位工作规范交流总结表26 表6 试用期重点工作目标计划书27 表7 周工作目标计划表（一）28 表8 周工作目标计划表（二）29 表9 新员工轮岗实习考核表30 表10 新员工试用期考核表31 表11 新员工试用期工作总结表32 表12 联想电脑公司试用（见习）人员转正表33 表13 新员工提前转正申请表34 表14 延迟参加新员工入职培训申请表35
企业应对员工提供必要的培训，以保证他们具备完成工作所必备的知识和技能，这是企业在员工已经进入企业后所从事的提高这些员工价值的人力资源管理活动。 员工培训分为岗前培训与在职培训，岗前培训的主要目的是让员工尽快适应企业的工作环境；而在职培训涉及培养员工应该具有何种技能，以及如何提高培训人员的执行能力，更好地实施培训。员工培训是一个系统的过程，它通过提高员工的技能水平，增强员工对企业的规则和理念的理解以及改进员工的工作态度，旨在提高员工特征和工作要求之间的配合程度。 本手册主要讲述作为联想服务站如何进行站内新员工的岗前培训指导，供站内使用、执行。
第一章关于岗前培训与指导 1.1 岗前培训与指导的意义与作用 1.2 岗前培训与指导的步骤及方法 1.3 有关联想服务中心站的站内新员工培训与指导 1.1 岗前培训与指导的意义与作用 岗前培训与指导的意义 企业新员工岗前培训是使新员工熟悉企业、适应环境和形势的过程。新员工进入企业会面临“文化冲击”，有效的岗前培训可以减少这种冲击的负面影响。新员工在企业中最初阶段的经历对其职业生活具有极其重要的影响。新员工处于企业的边界上，他们不再是局外人，但是也没有有机地融入企业，因此会感到很大的心理压力。他们希望尽快地被企业接纳，结果，这一时期员工比以后的任何时期都容易接受来自企业环境的各种暗示。这些暗示的来源包括企业的正式文件、上级所作的示范、上级的正式指示、同事所作的示范、自己的努力所带来的奖惩、自己的问题所得到的回答和任务的挑战程度等。员工在组织的第一年是一个关键的时期。在这一阶段，员工尽力使自己与企业的要求相适应，这可以产生积极的工作态度和高的工作标准。 新员工刚刚进入一个企业时，他最关心的是学会如何去做自己的工作，以及与自己的角色相应的行为方式。一般来说，他们更愿意通过自己的观察和亲自的尝试来适应新的环境，而不大愿意通过询问上级和同事，或者阅读公司的政策手册来了解所需要的信息。因此，有效的岗前教育是一项技巧性很强的工作。 岗前培训与指导的作用 岗前培训是员工在企业中发展自己职业生涯的起点。岗前培训意味着员工必须放弃某些理念、价值观念和行为方式，要适应新企业的要求和目标，学习新的工作准则和有效的工作行为。公司在这一阶段的工作要帮助新员工建立与同事和工作团队的关系，建立符合实际的期望和积极的态度。员工岗前培训的目的是消除员工新进公司产生的焦虑，而新员工辅导有助于消除这些焦虑。具体而言，岗前培训的作用有以下几个方面： 第一，岗前培训是新员工进入团队过程的需要。新的环境给员工一种不确定感。一个
联想电脑公司新员工培训制度(DOC7页)
联想电脑公司 人力资源部文件 文件编号：QB/LX·RS· A0148-98发文日期： 1998/9/23第 1页，共8页 收文部门：公司各部 收文者：全体员工抄送：／／ 拟制：赵松林审核：王建中审批：牛红 关键字：新员工培训制度是否受控 :是□否□ 文件类别：管理规定、价格政策、请示报告、会议通知、备忘录、总结报告、会议纪要、问题处理、通报、技术报告、协议、文件转发、一般通知、其他 是否保密：是□ 1. 绝密 2.机密 3.秘密否□保密期限：/ 保密文件的销毁方式： 1.到期后由发文部门收回统一销毁 2.到期后由收文部门自行销毁 新员工培训制度（修订） 1．目的：为使电脑公司的新员工尽早了解联想的历史和文化、树立联想人的自豪感和自信心，并尽快熟悉工作环境、掌握必要的工作手段和技能，明确工作的规范和要求,以便 顺利地投入到工作之中去，根据《联想电脑公司培训工作管理章程》的有关条款， 特制定本制度。 2．适用范围：所有试用期的新员工，包括社会在职人员、应届毕业生以 及由集团其它部门转入的员工。 3．所有试用人员必须参加“新员工培训”（从集团其他部门转入的人员可免予参加集团培训）。 凡不参加“新员工培训”的，从社会招聘的在职人员不予转正，应届毕业生和从集团其他部门转入的人员不予定级。 4．新员工培训组成部分 4．1 电脑公司入职培训 4．2 集团公司新员工培训 4．3 电脑公司内部轮岗实习 5．培训内容 5．1电脑公司入职培训 1
5．1．1电脑公司宗旨、成长历程、经营业绩与未来发展目标 5．1．2电脑公司组织结构及各部宗旨与职责 5．1．3电脑公司主要规章制度（考勤、考核、工薪、财务、岗位 责任体系等） 5．1．4电脑公司的研发、制造现场参观 5．1．5客户意识与合作精神、团队建设 5．1．6通用岗位技能（基本礼仪、内部信息网、讲演等）5．2 集团新员工培训 5．2．1联想简介与历史 5．2．2联想企业文化 5．2．3规章制度等 5．2．4演讲技巧与团队精神 5．3 电脑公司轮岗实习 主要是针对本岗位工作的实际需要，选择与本岗位有业务关系的两 个岗位或两个处（分属两个部门），具体内容包括： 5．3．1实习部门工作和业务的基本情况和有关政策、制度 5．3．2实习岗位的具体工作流程和规范，明确相关接口人 5．3．3具体参与实习岗位工作过程，体会联想做事的方式和氛围6．培训时间安排 6．1 新员工培训每月安排一期，具体安排如下： 6．1．1每月整周第一周的周二至周五为电脑公司新员工入职培训； 6．1．2第一周的周日至第二周的周四为集团公司新员工培训，周 五新员工返回本部门接受其岗位指导人对其轮岗实习安排； 6．1．3第三周至第四周为新员工轮岗实习。 6．2 新员工试用一周（外地新员工须试用三周）后即参加最近一期的培训。 7．培训的组织、教学与指导 7．1 电脑公司入职培训 7．1．1公司人力资源部培训处负责整体策划、组织实施。 7．1．2每期培训须有一位总经理室成员给学员讲话，每位总经理 2
联想电脑公司新员工培训制度
联想电脑公司人力资源部文件 文件编号：QB/LX·RS·A0148-98 发文日期：1998/9/23 第 1 页，共8 页 收文部门：公司各部 收文者：全体员工抄送：／／ 拟制：赵松林审核：王建中审批：牛红 关键字：新员工培训制度是否受控:是□否□ 文件类别：管理规定、价格政策、请示报告、会议通知、备忘录、总结报告、会议纪要、问题处理、通报、技术报告、协议、文件转发、一般通知、其他 是否保密：是□ 1.绝密 2.机密 3.秘密否□保密期限：/ 保密文件的销毁方式： 1.到期后由发文部门收回统一销毁 2.到期后由收文部门自行销毁 新员工培训制度（修订） 1．目的：为使电脑公司的新员工尽早了解联想的历史和文化、树立联想人的自豪感和自信心，并尽快熟悉工作环境、掌握必要的工作手段 和技能，明确工作的规范和要求,以便顺利地投入到工作之中去， 根据《联想电脑公司培训工作管理章程》的有关条款，特制定本 制度。 2．适用范围：所有试用期的新员工，包括社会在职人员、应届毕业生以 及由集团其它部门转入的员工。 3．所有试用人员必须参加“新员工培训”（从集团其他部门转入的人员可免予参加集团培训）。凡不参加“新员工培训”的，从社会招聘的在职人员不予转正，应届毕业生和从集团其他部门转入的人员不予定级。 4．新员工培训组成部分 4．1电脑公司入职培训
4．2集团公司新员工培训 4．3电脑公司内部轮岗实习 5．培训内容 5．1电脑公司入职培训 5．1．1电脑公司宗旨、成长历程、经营业绩与未来发展目标 5．1．2电脑公司组织结构及各部宗旨与职责 5．1．3电脑公司主要规章制度（考勤、考核、工薪、财务、岗位 责任体系等） 5．1．4电脑公司的研发、制造现场参观 5．1．5客户意识与合作精神、团队建设 5．1．6通用岗位技能（基本礼仪、内部信息网、讲演等）5．2集团新员工培训 5．2．1联想简介与历史 5．2．2联想企业文化 5．2．3规章制度等 5．2．4演讲技巧与团队精神 5．3电脑公司轮岗实习 主要是针对本岗位工作的实际需要，选择与本岗位有业务关系的两 个岗位或两个处（分属两个部门），具体内容包括： 5．3．1实习部门工作和业务的基本情况和有关政策、制度 5．3．2实习岗位的具体工作流程和规范，明确相关接口人 5．3．3具体参与实习岗位工作过程，体会联想做事的方式和氛围
联想入职培训资料
目录 前言 (2) 第一课LTL网络培训手册 (3) 第二课关于计算机运用的其他问题 (10) 第三课商务流程 (13) 第四课LTL 财务知识 (18) 第五课时间管理概览 (19) 附：《联想之歌》
前言 为了完善集团公司的新员工“入模子”培训工作，巩固培训成果；也为了迅速使新员工转变观念、统一意志，认同LTL的经营理念和价值观以及掌握LTL的基本知识、方法和技能等内容，提升自己的角色转变意识和实际岗位工作能力，为今后的进一步提高创造条件，我们开展了这项新员工培训工作。 为便于培训工作的开展，经过我们的努力，编写了这套教材，以利于我们共同学习和提高。在此，我们对给予我们工作大力支持的调研部、秘书室、经营计划部、综合部以及部分事业部的同仁表示感谢，并对集团人力资源部和联想电脑公司人力资源部同仁的大力支持，致以诚挚的谢意。 由于我们水平有限和时间仓促，教材中的不妥之处在所难免，还望各位同仁不吝赐教，以便于把我们的工作做得更好。
他山之石…… 本章我们来学习一些较实用的技能方法，希望能给你带来一些便利，包括：LTL 网络的基本知识和操作、财务知识、时间管理等；另外，我们将不断丰富此篇内容，让更 多的联想同仁来分享。 第一课 LTL 网络培训手册 适用于LTL 网络用户 经营计划部 主编 一． 前 言 1. 此次培训的内容 讲解网络的基本知识。 讲解内部主页的内容。 收发Email 的方法和使用技巧。 掌握浏览Internet 的方法。 出差时怎样上网。 2．联想集团网络和LTL 网络的概况 二． 登录NT 域 目前我们LTL 的网络供大家登录的有三个域。分别为LTL 域、LTL_SW 域和RAS 域。 域，用通俗的话来说，就是由一台服务器（域控制器）管理的用户群的集合。如果想登 局域网 终端 终端 终端 终端 终端 终端 服 务 器 服 务 器 网 络 传 输 设 备 集团IT 部 POP3、 SMTP 服务器 代理 服务器 Proxy Server 防 火 墙 Internet 学习 提示
新员工的万能培训方案(完整版)
方案编号：YT-FS-4092-78 新员工的万能培训方案 (完整版) Develop Detailed Rules Based On Expected Needs And Issues. And Make A Written Plan For The Links To Be Carried Out T o Ensure The Smooth Implementation Of The Scheme. 深思远虑目营心匠 Think Far And See, Work Hard At Heart
新员工的万能培训方案(完整版) 备注：该方案书文本主要根据预期的需求和问题为中心，制定具体实施细则，步骤。并对将要进行的环节进行书面的计划，以对每个步骤详细分析，确保方案的顺利执行。文档可根据实际情况进行修改和使用。 在新员工到单位报到后，必须进行入厂、入校或 入公司教育，称这种教育为"引导"，即对新员工的工 作和组织情况作正式的介绍，让他们了解熟悉单位的 历史、现状、未来发展计划，他们的工作、工作单位 以及整个组织的环境、单位的规章制度、工作的岗位 职责、工作操作程序、单位的组织文化、绩效评估制 度和奖惩制度，并让他们认识将一起工作的同事等等。 此外。培训还要建立传帮带的师徒制度，使新职工更 快地熟悉环境，了解工作操作过程和技术，让他们知 道，如果碰到困难和问题，应该通过什么渠道来解决。 没有任何社会经历，就像一张白纸一样，踏上人 生旅途的新员工，对他们影响最大的是刚参加工作时 遇到的领导及老员工。对新员工的培养方式决定着新
员工的人生态度和生活目标，疏忽不得。 一、踏上工作岗位前的集中训练新来的员工，一般要经过领导人讲话、学习有关本公司知识、实习、集训等集中教育，然后再正式分配到各个部门。这种集中式的训练一般由人事部门组织，短则一个月，长则半年，也有将近一年的。这种训练的目的是要解决一些共同的问题，让新员工尽快熟悉和了解公司的基本情况。可采用通信教育或内部刊物发行的方式来实现。 训练内容应包括三大板块： 1.帮助新员工了解公司，培养员工的认同感。其作法可考虑讲述公司发展史。 主要负责人介绍、有关产品的生产销售状况、经营方针与发展目标等等。 2.在上述活动的基础上，要求新员工确定自己的工作态度和人生目标，同时为之提供有关交际、员工常识之类的小册子，帮助其尽快完成角色转换。 3.请新员工讲述对公司的感想，了解新员工在想
新大学生入职培训方案
新大学生入职培训方案 篇1 一、施行新员工“Ｃ。Ａ。Ｎ。”计划的背景及目的 20XX年，中信银行拥有30家一级分行、20家二级分行、493家营业网点、15，070名员工，并于07年4月在上海、香港两地同步上市，经营和管理步入一个新的开展时期。随着业务的快速开展，机构和人员规模不断增加，每年全行都有几千名新员工加入。其中，既有大学毕业生也有社会招聘人员，个人的经历、背景等存在很大差异。业务的快速增长和现有人员紧张的矛盾日益突出，要求新员工更快地适应岗位工作要求。 20XX年以前，全行新员工培训缺乏统一标准的要求，各分行在新员工培训内容、形式、组织管理模式等方面存在很大差异，培训效果也参差不齐。内容方面，有的分行仅对新员工进行点钞、辨伪、计算器等业务技能的培训，有的已经系统安排了企业文化、主要业务开展战略、礼仪、沟通、公文写作等方面的培训；形式方面，大部分分行停留在单一的集中面授培训阶段，做得比较好的分行已开始综合采用拓展训练、模拟银行上机操作、师傅带徒弟、管理培训生管理办法等多种形式；组织管理模式方面，普遍缺乏有效的激励约束机制，新员工主动学习的积极性和热情没有充分调动起来。一方面，新员工上岗后，对中信银行的业务和管理规章管理办法不够熟悉，难以很快独立
承担岗位工作，业务压力大而人员相对紧张的矛盾没有得到缓解，更重要的是存在潜在的操作风险；另一方面，由于缺乏合理的引导，新员工难以尽快融入新的团队和集体，原本对工作的满腔热情和神往转变成满心失望，造成很大的心理落差，出现新员工在试用期内离职的现象。 如何让全行的新员工都接受统一、规范化的培训？如何让他们具备基本的职业技能和素养，掌握岗位必备的知识和技能？如何让这些来自不同背景的新员工尽快接受中信文化，融入所在团队？如何缩短他们的适应期，更快地成为一名合格的员工？这些已成为全行培训管理者必须尽快解决的问习题。 二、新员工“Ｃ。Ａ。Ｎ。”计划的主要做法 从培训对象角度看，新员工是一个特殊群体，他们从事的岗位工作覆盖银行经营管理的各个方面，要做到统一，首先必须深入了解我们的客户，也就是全行各用人单位对新员工培训的期望，把握新员工群体共有特点，明确各个岗位对新员工的整体要求。为此，20XX年3至4月，中信银行总行培训中心启动了新员工培训调研活动，发放了近百份问卷，全面覆盖总行各部门、各分行人力资源部、业务部门及重点支行，了解了对新员工的培训期望、各分行现行的新员工培训做法。同时，对ＩＢＭ、微软、英特尔、索尼、惠普、海尔、联想、宜家等国内外企业新员工培训进行了调研。 在调研问卷分析中，我们结合逻辑思维方法和数量关系分析方法，对全部调研问卷分别从经营/管理、所在地域、分行规模等多个维度
物流企业新员工培训,范文
德 邦 物 流 企 业 新 员 工 培 训 的 策 划 方 案 姓名：杨兴团 学号：4 专业：10级市场营销教育
前言： 德邦物流公司有限公司经过几年的发展，已经形成了全国31个省级行政区的完善物流网络。所以为了更好的发展和逐步与国际物流水平接近，公司急切需要一批业务熟练、富有管理能力及其有物业业务经验的人士来公司为公司献计献策为公司未来的发展做充分准备，还需要原来的员工一如既往的努力，这个公司能走到现在的地步离不开他们的支持，如今物流企业也是多如牛毛，挑战更大了，但正如很多人都知道的那样，挑战与机遇并存。一个好的人才培训方案会给公司培养一群精英来为公司服务同时自己也可以达到自己人生目标。如今制定这个培训方案，目的就是给在岗员工一个对企业更深的认识，以及对自己的业务的一个熟知、熟练，用最短的时间给在岗员工补充各种知识，以便使员工有一个更好的发展空间和舞台，同时也为公司的发展，建立更好人才培训方案，以满足社会的发展需要和面临的各种挑战。 公司简介： 德邦是国家“AAAAA”级物流企业，主营国内公路零担运输业务，创始于1996年。截止2012年10月，德邦已在全国31个省级行政区开设直营网点2600多家，服务网络遍及国内550多个城市和地区，自有营运车辆5100余台，全国转运中心总面积超过80万平方米。 公司始终以客户为中心随时候命、持续创新，始终坚持自建营业网点、自购进口车辆、搭建最优线路，优化运力成本，为客户提供快速高效、便捷及时、安全可靠的服务体验，助力客户创造最大的价值。 公司秉承“承载信任、助力成功”的服务理念，保持锐意进取、注重品质的态度，强化人才战略，通过不断的技术创新和信息化系统的搭建，提升运输网络和标准化体系，创造最优化的运载模式，为广大客户提供安全、快速、专业、满意的物流服务。一直以来，公司都致力于与员工共同发展和成长，打造人企双赢。在推动经济发展，提升行业水平的同时，努力创造更多的社会效益，为国民经济的持
联想新员工培训
目录 前言 (2) 第一课 LTL网络培训手册 (3) 第二课关于计算机运用的其他问题 (10) 第三课商务流程 (13) 第四课 LTL 财务知识 (18) 第五课时间管理概览 (19) 附：《联想之歌》
前言 为了完善集团公司的新员工“入模子”培训工作，巩固培训成果；也为了迅速使新员工转变观念、统一意志，认同LTL的经营理念和价值观以及掌握LTL 的基本知识、方法和技能等内容，提升自己的角色转变意识和实际岗位工作能力，为今后的进一步提高创造条件，我们开展了这项新员工培训工作。 为便于培训工作的开展，经过我们的努力，编写了这套教材，以利于我们共同学习和提高。在此，我们对给予我们工作大力支持的调研部、秘书室、经营计划部、综合部以及部分事业部的同仁表示感谢，并对集团人力资源部和联想电脑公司人力资源部同仁的大力支持，致以诚挚的谢意。 由于我们水平有限和时间仓促，教材中的不妥之处在所难免，还望各位同仁不吝赐教，以便于把我们的工作做得更好。
他山之石…… 本章我们来学习一些较实用的技能方法，希望能给你带来一些便利，包括：LTL 网络的基本知识和操作、财务知识、时间管理等；另外，我们将不断丰富此篇内容，让更 多的联想同仁来分享。 第一课 LTL 网络培训手册 适用于LTL 网络用户 经营计划部 主编 一． 前 言 1. 此次培训的内容 讲解网络的基本知识。 讲解内部主页的内容。 收发Email 的方法和使用技巧。 掌握浏览Internet 的方法。 出差时怎样上网。 2．联想集团网络和LTL 网络的概况 二． 登录NT 域 目前我们LTL 的网络供大家登录的有三个域。分别为LTL 域、LTL_SW 域和RAS 域。 域，用通俗的话来说，就是由一台服务器（域控制器）管理的用户群的集合。如果想 局域网 终端 终端 终端 终端 终端 终端 服 务 器 服 务 器 网 络 传 输 设 备 集团IT 部 POP3、 SMTP 服务器 代理 服务器 Proxy Server 防 火 墙 Internet 学习 提示
【最新推荐】联想新员工入职培训-word范文 (3页)
【最新推荐】联想新员工入职培训-word范文 本文部分内容来自网络，本司不为其真实性负责，如有异议或侵权请及时联系，本司将予以删除！ == 本文为word格式，下载后可随意编辑修改！ == 联想新员工入职培训 联想的新员工入职培训 联想对新员工的指导从其进入公司的第一天开始。上班第一天，有新员工的“首天培训” ，主要是熟悉工作环境和要求;紧接着是所在公司的新员工入职培训、集团的“入模子”培训及后续的轮岗培训，之后，新员工就可以返回各自的部门和岗位。 前两个培训各需—周，轮岗通常需 1—2 周，经过这三个规范过程，新员工的培训内存就告一段落。这当中的重点是集团统一的“入模子”培训。培训班为全脱产封闭式，共 5 天，其中一天是军训。课程包括联想简介、发展历史、奋斗目标、经营理念、行为规范和团队训练等，以及一系列团队活动，包括卡拉 ok 比赛、拔河比赛、篮球比赛等。培训班将一天分成几个时间段，每一段时间都作了统一安排。培训班的管理制度是依照公司对员工的要求来做，有严密的纪律，每天早晨出操点名，白天学院上课时，还要组织检查内务。从早晨起床到晚上熄灯都要对每个小组进行考察、打分。如上课迟到、bp 机乱响就要罚站，并对所在小组扣分，还会在最后的培训鉴定上标明。 培训班中，团队的协作作为—项重要内容贯穿始终。新学员刚刚进入培训班就被分为若干个小组，每组不超过 20 人，每个小组选拔 3—4 位骨干。之后的每一天，每一项活动都将按组进行打分，学习效果和最后成绩与这些分数有直接关系。学习过程始终重视学员的参与，从互动式教学到各种团队活功，讲求人人参与，协调配合，积极竞赛。根据活动内容不同，各组组长会将组内人员进行不同组合和分工，每个人都有机会承担某一项活动的主力，一方面发挥新员工所长和能动性，另—方面注意挖掘每个人的潜能，有机会锻炼自己的弱项。因此，不论这个新员工是怎样的一个人，都有机会展示、发现并提升自己。 在拓展训练中，设有盲人行动、建寺庙和过硫酸河游戏、背摔等，这些训练主要为了提高学员的心理承受能力。如背摔，学员要站在一个高高的台子上闭着眼睛脸朝上似自由落体地摔向下面围起圈子的人们。为提高新员工的表达与胆量，培训班设有一个即兴主题演讲内容，每个新员工在拿到演讲题目后有两分钟的准备时间，然后他要站到一个很高的台子上进行两分钟的演讲，不管你是情急下的语无伦次，还是紧张的心跳、腿软甚至不知所措，你都必须坚持下来，完成演讲。类似这样的训练在子公司各自的培训课程完成之后，还要在集团统一的培训班上再度进行过关训练。几次下来，新员工能了
评价联想的新员工培训
联想新员工培训的分析 联想集团通过“入模子”培训，使公司的企业文化的建设得到持续发展，企业文化的传承的一致性得到了保持，同时使公司的管理制度得到了认可与执行。 一、企业文化的建设的持续性 企业文化的发展过程是一个长期积累的循序渐进的过程，成熟的企业文化具有丰富的底蕴，这种底蕴就是一种积累，一种延续，一种坚持；没有这种积累，这种延续，这种坚持，就不可能有丰富的、成熟的企业文化底蕴。不管你的企业如何发展，不管企业内部结构如何变化，其根源都是可以追朔久远的。通过入模子培训，联想集团的企业文化就可以一年一年、一代一代的传承下去。 二、企业文化传承的一致性 传承企业文化需要建立完善的企业文化传承机制，搭建文化理念的贯输通道，新员工企业文化培训是传承机制的一部分。在每位新人入司时，企业都有相关技能和素质培训。在这些培训过程中，不能忽视对企业文化的培训。只要每一个新来的人员都能感受到企业内在的这种文化，并加以理解和深入感知，企业文化就能够透过时间传递给企业的每一个人，不断被放大。 联想控股从小型企业到现在的龙头企业，从以前单一的IT领域到现在的多领域，这发展过程中保持企业文化的一致性是很重要的。任何企业都有自己的文化基因，之后进入的任何员工，可以用自己的思想去优化这一文化，却不能颠覆它，新员工必须如老员工一般，从心底认同企业的核心价值观。 三、企业的管理制度得到认可 在很多的公司，公司的管理制度很完善，但是却一直执行不下去，不按流程办事。联想能有今天的辉煌与柳传志重视按流程制度办事不无关系，他曾在公司强调：“员工做事，有流程制度的一定要按流程制度办事；流程制度有问题，那就先按流程制度办事，然后提出改进建议；没有流程制度，先按公司文化要求办，然后提出建设流程制度的建议。”要把制度执行下去，就要对执行的部门与个人进行培训，而联想的“入模子”培训，就把联想集团的企业文化以及管理制度等很好的贯彻了下去。 个人认为，企业最重要的的就是文化的传承，联想集团的“入模子”培训，