History of structural analysis & dynamics of Wärtsilä medium speed engines

Summary. This paper opens up the history of structural analysis and dynamics simulations of Wärtsilä engines. It cites already published articles and theses with some backgrounds information. It also discusses some of the backgrounds of the in-house tool development. Additionally, this paper presents the development of the computers and investment of the simulation capacity in order to understand how it has been the enabler of ever more complicated models. It lists the work done during ﬁfty decades. The authors sincerely attempt to make this article as reader-friendly as possible, even though there are over 220 references, which of course demonstrates how dedicated Wärtsilä has been in supporting numerical simulations research in the past ﬁve decades.


Introduction
Based on the feedback the authors have received it seems that the audience is very interested in finding out more about the history of the Finnish Structural Mechanics in the industry.Liljenfeldt's article Computers in diesel engine development [1] from the year 1992 served as a good starting point for this paper.The authors recommend it for the history enthusiast of computer-aided engineering.Hagelsberg's article Sinapinsiemen kylvetään [2] is also innovative, and despite the name it is describes the history of engine design.Hagelsberg was the head of the simulation team up until 1976, when Liljenfeldt took the position.
This paper outlines the history of simulation methods developed in the Wärtsilä fourstroke engine Research and Development, see Table 1 for some milestones.This paper can be used in the new employee induction process to familiarize them with the history of the used simulation methodologies.In addition, professors can use this paper in university education to teach next generation engine builders and research engineers about the research and method development of the modern R&D organization.The authors wanted to take a scientific approach and write this paper based on the published results.This way it will work as a future index for looking at previous research work about structural simulations of Wärtsilä engines.This paper is mainly written in a chronological order and divided in sections representing different decades.Due to a high number of references, all of them are collected under bigger umbrellas, to keep the storyline more readable and less fractioned.Some of the fascinating details or exceptional achievements will be highlighted to make the story more interesting to read.This paper is a scientific review of simulation method development work from the past five decades.
Writing this paper was unquestionably an exciting experience and can be recommended to all our colleagues in the industry.It is worthy to go through the history before "blindly" heading towards the future.As a reader, please take this as a challenge and write the history now, when it is still fresh in our memories.The future generation of engineering mechanics experts will surely appreciate the effort.These history articles will be used both internally in the company as an index of research, and externally in the universities as a helpful introduction for specifying the simulation needs of industry.

Starting from the seventies and continuing in the eighties
In the early seventies, Structural Analysts performed strength calculations with a pen, paper and slide rules, and towards the end of the decade, slide rules were little by little replaced by pocket calculators.Interestingly, the same engine components as today were the target of the simulations [5][6][7][8] including connecting rod, fuel pump, piston, crankshaft and cylinder liner.
The first numerical simulation in Wärsilä was Gösta Liljenfeldt's master thesis about fuel pump cavitation [6].Pentti Rajamäki developed the first internal Finite Element Method (FEM) solver in the year 1970 [2], see Figure 1, by using the first edition of the Zienkiewich's book The finite element method, which was published 1967, and remained the only book about FEM until 1971 [9].The FEM was taken into the production calculation usage quite quickly, as one can see from several documents.For example, Figures 2 and 3 show the usage of the solver.Triangular plane stress and plane strain elements enabled the analyst to study "complex" plane structures exposed to nodal forces.The means of varying the element thicknesses were considered the third dimension.Structural analysts had to walk to the Hartman building next to the marketplace in Vaasa, where they ran the analyses on Tietobotnia's IBM computer.Bookkeeping used the same IBM computer for salary counting, but then the common business-oriented language (COBOL) was used.They used a fairly small amount of nodes and elements for practical reasons.No computerized pre-and post-processing software or hardware was available.The structural analysts had to plot meshes and nodal displacements by hand, which was a very time-consuming task.Stresses could not be presented graphically with any reasonable amount of effort.Thus, they were read from numerical lists element by element, see Figure 3.
Wärtsilä Vasa Factory acquired its first computer in the late 1970s.The intention was to use this HP1000 minicomputer for laboratory and production test bench measurements, as well as monitoring purposes on the test bench.The core size was 32 kilowords, or 64 kb.Discs were not available; the structural analysts had to use the only mass storage device, which was a tape reader and punch device, used for both input and output.A FORTRAN compiler was available, and when a copy of a commercial FE-software SAP "was found on the market", it was installed and used.Its routines enabled us to carry out an FE analysis with more advanced element types such as quadratic plane stress and strain domains.Mesh size was still limited to a few hundred degrees of freedom.When Wärtsilä bought and installed a disc package with a capacity of a few 100 MBs in approximately 1980, the mesh size grew to at least 1000 degrees of freedom.The chicken and egg race started: more storage capacity and more powerful computers allowed us to use more accurate models, and in turn more accurate models asked for more and more storage capacity and power.See Figure 6.[10] The very first commercial FEM usage was Harry Holmgren's Stardyne FEM course project work, see Figure 5, where he calculated the Vaasa 32 main bearing cap [11].This was also the first published Finite Element Analysis (FEA) [12], see the Figures 7(a The second published FEA was Sjölind's master's thesis [13] on cylinder liner thermoelastic deformations 3D elements.However, the story behind the cooperation is a little bit bouncy.According to [14], Stig waited for boundary conditions for over one year in Vaasa.For one reason or another, he never got the needed boundary conditions from component experts, and left for Oulu with a never look back, and burn the bridges attitude.Nevertheless, he finalized the thesis, and it remains the second published FEA of Wärtsilä products.On the topic of master's theses and numerical computation development, the cylinder head has probably been the component that has benefited the most from the development of FEM.The earliest cylinder head FEA in Wärtsilä was performed already in 1983 [4], see Figure 11.Additionally, the piston FEA was performed three years later in 1986 [15].
At a very early stage, Wärtsilä focused on the engine dynamics [16][17][18][19] and their impact on the fundamentals.One of the early tasks of Adina FEM program, actually a test case, was to test its shell elements.The simulation manager selected the common base frame as the test case.An interesting detail was that there were some problems with bad element quality, and as a result, the Adina simulation crashed.Therefore, the analyst decided to make a queue of different versions of the mesh to debug the problem faster.However, this lead to the situation where all the server resources were "stolen" to the Research and Simulation team, which made the design team very unhappy because they used the same server for CAD work at the time.[10] What is most interesting about this is the fact that humans have not changed, even today with the "unlimited" server resources compared to those in the late eighties.However, some individual analysts can still steal all the resources to themselves, even with the same basic trick of simply generating a lot of simulations tasks to the queue.
Usually, behind all simulation tasks there is a real need for the results.Another example of the common base frame dynamics simulations is Rabb's master thesis [17], which was initiated by a field vibrations problem of twelve delivered 18-cylinder vee engines of Vasa 32 type.These aggregates had enormous vibrations issues, and the customer was already at a point where the only outcome is the cancellation of the order, in other words, he was willing to return the nonworking aggregates to Vaasa.Rabb, together with his colleague Martikainen did considerable work calculating different field fix options.Finally, it was discovered that the solutions was to add a bottom plate to the aggregates, which of course at this point was not a very attractive solution from the welder point of view.[10] The work was only halfway when the technically feasible solution had been found to fix the problem.Next, someone needed to go and visit the angry customer and explain the somewhat technical problem.Liljenfeldt was assigned the job to go around the world and meet the customer.The customer was not convinced, and he wanted to bring his expert Ph.D. for the meeting the following day.The expert was not convinced with the idea of finite element computations and stood his ground about how it is possible to solve these sizes matrices by hand.Liljenfeldt tried his best to convince the expert about the fact that this is the solution, we trust our simulation model and this change will fix the issue.At dinner, the customer gave partly in after revealing that their expert was a Ph.D. in history, and promised Wärtsilä to try their best with one aggregate.However, he demanded that Liljenfeldt stay and wait for these modifications to happen.Of course, this was not possible, it was Liljenfeldt's summer vacation, and he left back home to Finland.After the summer vacation, a letter was waiting from a genuinely angry customer who had already contacted Wärtsilä CEO about the poor customer assistance he had gotten from Liljenfeldt earlier.However, the story had a happy ending, as after the modification, a comparison of a simulated and measured model showed that the correlation was excellent.This meant that the simulated model was quite exact compared to the measurements.Of course, this led to the situation where customer satisfaction was up again.The customer even ordered five more aggregates after the problem was solved.[10] Another way that Wärtsilä has been building customer trust in numerical computations is the active contribution in CIMAC Congresses.For example, in the year 1985 Ahlqvist et al. [20] presented a CAD-enabled design process, and in the year 1987 Paro et al. presented the effect of compression ratio to the durability of the crankshaft [21], which is the key design question in engine concepts.Another interesting CIMAC contribution is the paper about the piston fatigue analysis from 1987 by Silvonen et al. [22].On the topic of pistons reliability, Wärtsilä's supplier also published relevant results, see [23].In a way, talking about tools is irrelevant.The crankshaft is the most loaded, and thus most calculated engine component, sometimes even called the heart of the engine.It is also the only component for which we have dedicated calculations rules, a standard for classification [24].In the nineties, crankshaft calculation methodology was developed and published in these theses [25][26][27][28][29].
Furthermore, the most calculated components are the cylinder head and cylinder liner.There are thermo-mechanical harsh loading conditions [30][31][32][33] present in these component Cylinder head simulations have always utilized the full available computing power, and even until recently it seems that the accuracy of the models compared with the wall clock time of the simulation running times is not satisfactory.However, previous as well as current methodologies have provided word class products of their time, and engineers have kept pushing the limits by increasing both the efficiency and power output of the engines.From very early on, Wärtsilä has recognized the need for engine dynamics simulations and the dynamic design of Gensets, a power packet of engine and generator combined with a common base frame.The common base frame was selected as the tunable component, which meant that its design depended on the combinations of engine and generator.The theses [34][35][36][37][38][39][40][41][42] present the methodology development.
In the early nineties MS-DOS-based in-house software for flexible mounting calculation was developed in collaboration with the Tampere University of Technology.It was based on the rigid body natural frequency calculation of a system where an engine or a generating set is standing on soft rubber mounts.This S.c.REMO software, combined with experimental modal analysis, which also started in the early nineties, was a powerful tool for designing engine installations.
Adina calculated the natural frequencies of engines and generating sets.It is commercial FEM software and the results was compared with experimental modal analysis results.Many studies were done to find suitable modeling techniques, and to reasonably simplify models without sacrificing too much accuracy.
In the nineties, a major development took place in engine dynamics simulation, see details in [43][44][45][46][47][48][49][50][51][52].The methodology of calculating only natural frequencies was finally considered to be insufficient, because of the increased excitation forces due to higher output with fewer emissions.Thus, a harmonic response calculation methodology was required and developed.This development started with studying engine dynamic excitation forces in detail and building computer software for their calculation.An important part of understanding the excitations, and the way affect the engine structure, is the visualization of the forces and internal moments.The need for quick calculation and visualization leads the DYNEX software development.Further development in the late nineties consisted of adding the export function of the dynamic forces to Finite Element software, enabling forced response simulation.It is possible to simulate the engine vibration response at a constant or varying speed.In the beginning, DYNEX made the export to Ideas "program" format, but when this format became obsolete, the export function was rewritten to support "Universal File Format," and the conversion to Abaqus input format was later added.DYNEX has played a major role in engine dynamics simulations for more than fifteen years, until the multi-body simulation tools' introduction to production usage.
Eventually, all loading in engines is fatigue loading by nature.Thus, fatigue calculation methodology needed to develop to guarantee human safety in the increasing competition of increasing the power outputs.These publications describe the development [53][54][55][56][57].
Gear train dynamics, namely camshaft gear train dynamics, have been under investigation since the nineties [58][59][60].Naturally, in the four-stroke engine, the camshaft torsional vibrations are especially important to handle, in order to design a smoothly running engine.The peculiarity of the medium speed engines are the multiple fuel pumps, even with each cylinder having one pump.These are significant loading, for example, in full power the pumps excitations to the camshaft are three times higher than excitations from valve lifting.
This takes us nicely to the next topic, elastohydrodynamic bearing calculations.They also started very early on [61] in 1994.The engine is full of highly loaded slider bearings, and their reliability has been and still is one of the cornerstones of the engine development.Pistons [62] have slider bearings and other demanding loads.
This paper focuses on structural analysis and dynamics, however the right boundary and loading conditions are needed, and can often be provided by CFD simulations.These theses [63][64][65][66][67][68][69] present some of the early CFD simulations.
Another, fatigue-related, phenomenon, namely fretting research need, was born.The start of the long-lasting method development is described in detail in these publications [85][86][87][88][89][90][91][92][93][94][95][96][97][98][99].Fretting typically occurs when the design scales upward.Thus, manufacturers with physically large products are usually aware of the fretting phenomena.Due to this fact, the fretting research is not as mature as the fatigue research, and therefore it provides a virgin ground for new development.See one example of the crankshaft counterweight fretting simulation model in Figure 8.
Something old was improved as well.Each decade attempted to improve the cylinder head calculation methodology.This time [100], the geometric model complexity had already reached the status where the full cylinder head assembly, together with the engine block, liner and all cylinder head equipment solve together.Of course, a lot of room for future improvements remains, and more of those are presented in the next section.
Continuing with the dynamics, camshaft dynamic calculation method needed to take into account the change in gear contacts, which changes the dynamics behavior and forced response of the camshaft.Theses [101][102][103] present the details.
References [104,105] show a new attempt towards the multi-body system modeling of the crankshaft and connecting rod.However, the organization was not yet ready for this new methodology, thus more about the virtual engine modeling is presented in the next section.Slider bearing calculations are as important as always, and one thesis was made on the topic [106] in 2006.CFD [107][108][109] also continued its development.
In the year 2007, the world economics were overheated and especially in the marine sector it seemed that there would be an unlimited amount of business growth before it all collapsed in 2008.Nevertheless, a new employee recruitment program called Wärtilä Young Professional was established to bait skilled resources.[10] The idea of the recruitment training program was to give a speedy start for new employees to catch up with the challenges of competencies of different types of engine expertise.Fundamentally, the idea of the program was to hire experts from different fields and give them all an introduction to all different competence areas.The analysis area with the team of Structural Analysis and Dynamics was one of the themes.Leaping to the modern computing age from 2010 until now A good start of the new decade was the Wärtsilä virtual engine simulation methodology development, see Figure 10 for an example of the model complexity.Papers [111][112][113][114][115][116][117][118][119][120][121][122][123][124][125][126][127][128][129] present the details.The idea of the virtual engine was to have as realistic an engine dynamics model as possible.Ideally, the same model is used for engine acoustics, then all the engine covers and similar are included in the simulation model.The same virtual engine models are typically used for modern crankshaft [130][131][132][133][134][135][136][137] simulations, including elastohydrodynamics [138,139] bearing models.In addition, in connecting rods calculations, a virtual engine modeling approach can be used [140,141].
Figure 10.Virtual engine model of powertrain of medium speed diesel engine [128] This decade was the time to get into the Simulation Process and Data Management (SPDM) [142][143][144][145][146][147][148][149][150][151][152][153][154].Some fundamental basic research was needed and performed.Eventually, the implemented system, Wärtsilä Digital Design Platform [152], contained all the product detail requirements and combined them into the actual simulation processes, which provided the values to the requirements.It also included task management tools, giving both line and project managers a real-time view of the status of project resources and tasks.
The cylinder head simulation methodology finally took a giant leap into the modern days, see details in [155][156][157][158][159][160][161][162].The state-of-the-art methodology of the whole cylinder head calculation process started from very detailed cylinder gas exchange modeling, continuing in the conjugate heat transfer [163] simulation model, which was run parallel to the gas exchange model to find the stable boundary conditions for the structural FEM calculations.The following FEA will take inputs from the conjugate heat transfer analysis, casting simulations and gas exchange analysis.The material model is microstructure as well as temperature dependent.After the FEA follows a combined low-high cycle fatigue analysis, see Figure 11 for example results.Those critical points where the desired lifetime is lower than expected are taken into detail crack propagation analysis.If the crack initiates, but does not propagate too far, it is called damage tolerant design and considered good enough.Next, it is just logical to continue with the fatigue methodology development review.The following references [160, give all the details.The major aspects to point out are micromechanics modeling and Bayesian statistics.Micromechanics modeling aims towards virtual fatigue testing, where the idea is to tune the micromechanics model to the exact interesting fatigue test points and then, when the micromechanics model can predict already measured test specimens, it will be used to interpolate and extrapolate the test matrix.This is true especially in the regions where the traditional fatigue test is either difficult or time consuming, namely testing the compress side of the Haigh-diagram and multi-axial fatigue loading.
What is more, fretting research has taken a huge leap as well; please see the details in the following references .The most significant achievement in fretting method development has been the operational calculations development.It does not matter how advanced the simulation methodology is unless one will get it to the production usage, where one will see fretting-safe designs in products.Currently, the authors are satisfied with how fast the Wärtsilä Structural Analysis and Dynamics department has been able to implement newly developed research results into the products.However, fretting remains in a way the untamed beast, and it will guarantee a plenty of future challenges for researchers.
The last paragraph is dedicated to loosely related research topics.Acoustics methodology went through some development [217,218].Big data measurements analysis [219,220] was something that needed to be tackled to get an understanding of the needed loading cases, so that they would represent real-world extreme loading conditions.All types of optimizations [221] are the cornerstone of the modern product development.Surprisingly, we have published an insufficient amount of information in the field.Some new studies on how to model brittle material were also performed [222], with the aim of human safety and reliability of our products.Finite element open source solver development [223][224][225] is something that was intentionally started to create a flexible and easy-to-use research platform.JuliaFEM is an attempt to create a common research platform for engineering mechanics numeric research in Finland.Currently, we are living an early stage, but academia has shown some interests.See also [226] for the start of the material modeling.

Conclusion
This paper attempted to open the structural analysis and dynamics method developments in the Wärtsilä engine division in the last five decades.As mentioned throughout the text, the authors consider Wärtsilä an excellent example of the industry, one that is dedicated to its product development so sternly that it is willing to do basic research to guarantee the needed competencies and skills in the company in order to make the most reliable products.We went through the early days of Finite Element Method (FEM) developed on the borrowed computer, and described the chicken and egg race of more significant computing results leading to more complex models that demand more computing resources.We added stories to the internal software development as well as the real driver, field problems, examples.In the final section, we leaped to a whole new level of simulation methodologies.
We have mentioned a few times when we believe that Wärtsilä has been an early adopter of the technology, or even the fundamental developer of it.Eventually, all the hard work paid off, and the Guinness World Records of the most efficient 4-stroke diesel engine was received in 2015, see Figure 9.
Figure 12 shows a part of the overwhelming task of going through all the references.Although it took a considerable amount of effort, the authors believe it was worthy for gaining a better understanding of what we have done and where we should be heading.It also revealed the somewhat surprising fact that Wärtsilä has not published much about optimization development as such.Structural analysts have mentioned optimization in some articles as a side note, but we have indeed not published as much as we have used it internally.

Figure 3 .
Figure 3. Vasa 32 connecting rod 1975: Numerical stress and displacement results on paper, drawn on a picture by hand.[1]

Figure 8 .
Figure 8. Finite element models for the fretting analysis[85] The program was somewhat successful and was later modified on into an internal training program called R&D Professionals.In the Structural Analysis and Dynamics team, we have one from the original training program and four from the follow-up programs.All of those who have participated have given much credit to the training programs.They have felt that their employer has invested in them.

Figure 11 .
Figure 11.Results of modern cylinder head simulation methodology, fatigue lifetime and stress-strain history of one material point[160]

Figure 12 .
Figure 12.Researcher's desk when writing this paper

Table 1 .
Wärtsilä Structural Analysis and Dynamics history milestones