Why data matters for hardware engineering companies

As a hardware engineering company, you might be excused for not paying too much attention to the latest trends in data science and artificial intelligence. After all, a lot of your daily work consists of exactly the kind of creative thinking and developing that is making these robots possible. You sell products made from aluminium, silicon and high-grade plastics, not click-per-view advertisements. You might not have a huge interest in the typical artificial intelligence stories about predictive maintenance or managing huge warehouses with robots.

On the other hand, what does matter is how good and how efficient you are at finding out what to create, at designing, manufacturing and selling. And that is where good data management is more and more essential. From evidence-based technology roadmaps to customer relationship management, enterprise resource management and many more. Automating all kinds of manual operations through scripts and macros helps to increase efficiency drastically, and reduces the possibility for errors.

So while as a hardware engineering company, you are maybe not focussed on ‘big data’ and artificial intelligence in itself, still being able to work with sizeable amounts of data and extracting knowledge out of them is quickly becoming an important competitive advantage.

If you went to bed last night as an industrial company, you’re going to wake up today as a software and analytics company.
– Jeffrey Immelt, former CEO of General Electric

Engineering data is typically not of the huge-quantity / low-quality type that gets all the attention. However, there are other reasons why it can still be a big challenge to work with design and manufacturing data. It is spread across many software tools, each performing its calculations, design or analysis with inputs and outputs that are manually copied and pasted. It is contained in documents, spreadsheets, presentations, emails, in different versions and local copies. There is no uniform way of indicating what number represents which parameter, let alone a clear structure in what numbers belong to what actual parts of the design.

At Valispace, we think of data storage, design automation and optimization as parts of a ‘hierarchy of needs’, inspired by ‘The AI Hierarchy of Needs’ as proposed by Monica Rogati from Data Collective. Just as in Maslow’s pyramid of needs, food and shelter are needed before self-actualization, in the engineering data pyramid a basic unified data storage and a certain data structure are needed before it is possible to perform advanced data analytics, apply optimization algorithms and machine learning.

Valispace engineering data pyramid, inspired by ‘The AI Hierarchy of Needs’.

At the bottom of the pyramid is data collection into a single database, available to all engineers collaborating on a certain product or project. Data is exchanged with as many other tools as possible, feeding in and reading it out again. This is the basic set-up to create a single source of ‘truth’ with the latest, up to date values.

Adding structure to that data is one step above in the pyramid. To allow for clear communication and analysis, values are structured using a simple data model. The relations to other values are also stored, for example through formulas. A difference can be made between a physical property (e.g. the mass of an engine blade of 0.3 kg) and other numbers describing such as a design margin or a maximum value. Keeping the data structure simple allows to put in many different kinds of data easily, from back-of-the-envelope spreadsheet calculations to the most elaborate fine-meshed analysis results, from complex state machine logic to large batches of manufacturing, calibration and testing data.

This structure allows for advanced ways of exploring engineering data and its connections, improving the efficiency of daily work and creating new insights. It becomes possible to filter and track data and the sensitivity of one parameter to other values. Interactions with parameters are based on their meaning rather than where they happen to be in a spreadsheet. This opens the door to new user interfaces such as voice, chat or augmented/virtual reality. You could imagine asking Alexa (or J.A.R.V.I.S.) for a breakdown of the latest design changes inside an important subsystem of your product and to detail their impact on the power budget. In fact, we have already set up a prototype Alexa skill for Valispace to do exactly that, get in touch if you would like to be a beta tester.

Building on these layers of data collection, structure and advanced exploration, powerful data analytics can be automated, providing oversight in charts, budgets, timelines or powerful custom scripts.

Finally, having engineering data stored in one place, but also connected, structured and analyzed, many features of a product or project can now be expressed as values inside Valispace. This provides a perfect platform to apply optimization algorithms and automation tools on a real engineering project, the kind of advanced algorithms that have until now been restricted to academic use or small subsets of data. For example, at IAC 2017 in Adelaide, Johannes Norheim from MIT presented a paper on optimization of a satellite design using geometric programming using Valispace and our MATLAB toolbox.

Summarizing:

  • Working with engineering data in an efficient way is a key competitive advantage, breaking up data silos and integrating knowledge across your company.
  • Advanced data analytics is based on structured data storage that forms a single source of truth and has automatic data exchanges with other engineering tools.
  • Exciting new user interfaces and methods of advanced optimization and machine learning become feasible with this data structure in place.

Interested? Take the first step today: centralize engineering data from fragmented spreadsheets and emails into a single place where they create value. To learn more about Valispace, have a look at our website and our documentation page, or try out the tutorials in the online demo.

Why you need more than a spreadsheet for your engineering data

When explaining the basic concepts of Valispace – collaboration, technical data exchange, connecting data with formulas – a question that often comes up is how Valispace compares with spreadsheets, and especially with cloud-based spreadsheet such as Excel Online and Google Sheets.

It is not a coincidence that spreadsheets are popular: storing data in tables is such a generic approach that it allows for thousands of different applications. Formulas that calculate the value in any particular cell depending on the other cells form a powerful calculation tool that is still relatively easy to learn (although hard to master). Charts can be created easily. Advanced functions like pivot tables and macros allow to set up more elaborate data analysis and automation scenarios.

A spreadsheet is the digital equivalent of a swiss army knife, a flexible tool to ‘get stuff done’ in our digital lives. Working with spreadsheets has become an important skill to survive and thrive in a digital world, so much so that it forms a mandatory part of secondary school education in many countries.

Swiss army knife
Swiss army knife

It is not a big stretch to imagine using spreadsheets to set up an engineering data exchange, and in fact this is exactly what is in place in many engineering companies: gather all data in one or multiple big ‘master’ tables, including the formulas to recalculate cells if other data changes. Put these on a local shared drive, a document management system or a cloud service such as Excel Online or Google Sheets, to allow for simultaneous editing by different users.

However, large and complex spreadsheets that are in use by multiple people are not a new phenomenon, and the challenges with them have been significant enough to merit quite some academic research and an annual conference. These challenges can include:

  • To keep the spreadsheet from becoming a data jungle and allow to retrieve the essential bits of information, typically some structure needs to be set up initially and carefully maintained through rules, active cleaning and review.
  • The data contains just numbers. For these numbers to be useful in calculations and formulas correction factors are often needed, for example to convert to the correct unit. Data manipulations like this need to be clearly documented to ensure other users work with the data in the correct way.
  • Changes are automatically propagated through the formulas, but their effect remains ‘hidden’ until the correct person happens to looks at the relevant data.

As with any human process, manually maintaining spreadsheets inevitably introduces mistakes. Some of these spreadsheet mistakes have become infamous, but it is increasingly clear that almost every spreadsheet has some bigger or smaller errors. The different types of mistakes in spreadsheets are included in the “Panko-Halverson Taxonomy of Spreadsheet Errors”. For example, ‘execution errors’ can include typos and wrong copies, and ‘planning errors’ can include logic errors inside formulas.

Spreadsheet error taxonomy
Taxonomy of spreadsheet errors, from Revising the Panko-Halverson Taxonomy of Spreadsheet Errors, R. R. Panko and S. Aurigemma, Decision Support Systems 2010.

While the rate of errors in spreadsheets is in itself not much different from human errors in other forms of data storage, reviewing a spreadsheet is notoriously difficult. Have a look at the picture below for a (still relatively straightforward) example. Difficult reviewing is part of the reason why many errors remain hidden and are even copied to other spreadsheets. It is not surprising that these spreadsheet mistakes are often the root cause of expensive last-minute fixes, leading to schedule slips and budget overruns.

Spreadsheet error example
Spreadsheet error example. How much time did it take you to find and fix the mistake?

Adding structure manually, performing data manipulations and taking care of changes all introduce errors in spreadsheets that are difficult to find and fix. How can this be solved? It is essential to set up a clear data structure that is shared between all collaborators, as well as automating as many actions as possible. It is exactly these two solutions that Valispace provides:

  1. Data structure: any data point inside Valispace is more than just a number. It has a name, a unit, a history, margins and many more. It typically is part of a data ‘product tree’ (see picture below). Formulas are constructed based on the name of the input values, rather than referring to an abstract cell number, which makes these formulas much easier to review. In every description or calculation where this value is used, it is always referring to the ‘single source of truth’, ensuring that calculations are consistent among each other.At the same time, very rigid data structures can easily make it complex to set up the data storage system and require significant training from all users. That is why a lot of flexibility remains within the basic Valispace ‘tree’ structure of components, Valis and their properties.

    Component tree
    Example component tree from Saturn V advanced tutorial.
  2. Automation: avoiding manual data manipulations greatly reduces the possibility of human errors. Valispace provides a range of different automation tools:
    • A lot of important but tedious calculations are performed by the Valispace algorithm behind the scenes, for example: automatic unit conversion, automatic propagation of design margins, notifications to the relevant users when something has changed that they need to be aware of.
    • Comparison charts, budget tables, relation charts between values and other data summary visuals are automatically to up to date.
    • Custom automation or scripting can be implemented through the Valispace Python and REST APIs.
    • Values included in other tools can easily be kept up to date with the Valispace add-ons for Microsoft Word, Excel, MATLAB,…

Through data structure and automation, Valispace solves the key issues of engineering data processing. While a ‘swiss army knife’ spreadsheet provides the ultimate flexibility in one tool, more complex data construction work requires dedicated tools to limit the amount of manual labor. Valispace is such a tool, a power drill for engineering data that provides efficient data storage and analysis.

Power drill
Power drill

How engineering students at KTH keep track of their satellite design data

You might have heard that satellites are massive objects that weigh several tons and cost millions of dollars to launch into space. Now, imagine scaling down both size and cost until you could hold a satellite in your own hands – a miniature satellite! The so called cubesats are already revolutionizing the satellite industry by making access to space more affordable. These mini-satellites are made of one or multiple standardized 10x10x10 cm cubic units which only weigh a little more than 1 kg each. They make it affordable for small companies, universities and research centres all over the world to build their own satellites carrying custom-made experiments. In 2017, there will be over 350 cubesats launched to space and the prediction is over 400 for next year.

Artist’s rendering of a cubesat in space.

The idea of MIST is to allow students to play an active part in designing, testing and launching a cubesat satellite.

The Royal Institute of Technology (KTH) in Stockholm is building MIST or MIniature STudent Satellite – a satellite designed only by students. “The idea of MIST is to allow students to play an active part in designing, testing and launching a cubesat satellite”, says Daniel Bogado, student in Aerospace Engineering and responsible for System Budgets in the MIST project. MIST was proposed to the KTH Space Center in 2014 and the work started in early 2015. Currently the 6th team is working on the development of the satellite. Each team stays with the project for half a year.

The experiments on the MIST are from different institutes and companies with many different objectives. Two of the experiments have their focus on radiation analysis, there is one biological experiment, one experiment is testing their self-healing/fault-tolerant computer system in space, another experiment is testing a semiconductor of silicon carbide. Furthermore, there is a piezoelectric motor being tested in space, a new propulsion technology and finally a camera. All these experiments pose different requirements on the satellite making the combination of these an interesting task for the students.

3D-printed model of the MIST satellite.

It’s very important that each new team member quickly gets up to speed in all areas necessary for their work.

Although the satellite is tiny, building a cubesat is a big challenge. The biggest hurdle in the MIST project is the change in personnel involved in the project each semester, many students stay only for half a year. It’s very important that each new team member quickly gets up to speed in all areas necessary for their work. With 8 experiments the MIST satellite has a high number of custom parts for a 10x10x30 cm (3 Unit) cubesat. Keeping a clear overview about all aspects of each experiment is a major challenge, as well as keeping every team member up to date with the most recent values and what the impact is if something changes.

Valispace allows everyone on the team to have the latest results available from everyone else’s work.

The MIST team has found a solution to this challenge by using Valispace to store their engineering data. They use Valispace as the central platform for saving all data regarding power consumption, mass budget and other important characteristics of each experiment and subsystem of the satellite. This allows everyone on the team to have the latest results available from everyone else’s work and therefore make the right assumptions in their calculations and analysis.

With small satellites, mass and power constraints are very strict. You want to maximize the number of experiments and their operation time, but there is only so much power that the tiny solar panels can generate. On top of that, part of the satellite orbit is in eclipse, meaning that the Earth blocks the sunlight, and the satellite must rely on the power stored in batteries. A small change in the power consumption of an experiment can completely change the overall performance of the mission. The MIST team uses Valispace for the creation of a mass and power budget for the satellite. The stored data is used to create a power schedule for the experiments and allow for an easier overview of the system budgets, to ensure that the system is performing at its best capacity.

MIST mass budget created in Valispace.

Right now, the mechanical design of the MIST satellite is being finalized and after that the required thermal analysis of the whole system will be done. Even a small cubesat must be carefully tested before being launched into space to be sure that all experiments work as they should. Once the satellite is in space, no repairs can be done. MIST will be launched sometime after 2019. You can read more about the MIST satellite and find the newest updates on their blog.

The 6th MIST team together with the Swedish astronaut Christer Fuglesang.

Evidence-based Technology Roadmapping

Big manufacturing companies for complex hardware usually have a product portfolio of thousands of products and product families. They range from basic technologies, integrated into components or subsystems up until the final products.

Imagine that you are the responsible for the company’s technical strategy at a major drone company. When creating a product strategy or an investment strategy for your R&D efforts, the main question you would want to answer usually is:

How can I ensure that my R&D investments today, lead to competitive products tomorrow?

The status quo

A Classical way of approaching this problem is to rely exclusively on the experience of managers and chief-engineers. A typical simple powerpoint analysis for your drone manufacturing business could look like this:

A gut-feeling-based Technology Roadmapping slide
The problem: Good Questions hardly get answered

This information is a good start, but it should not be the final information that you base your company’s R&D investment decisions on. Any company that wants to survive in a fast-paced market, needs to perform a much deeper technical analysis, to be able to answer questions like these:

> How much further would each of our 8 portfolio drones fly, if we hit the target for all of our R&D investments?

> How will our products compare to drones announced from competitors for critical performances, such as range, cost per flighthour and max-operating windspeed if we only achieve one of the goals?

> Can we perform quick concurrent engineering studies for the long-term potentials or do we need an entire team for months to work on it?

Evidence-Based Technology Roadmapping

To gain insights from your current portfolio and more importantly to plan its future, your engineers need to model the technological interdependencies between technologies and products:

> How does the flight-time correlate to the battery capacity and mass of a specific drone?
> How does the rotor size affect the lift?
> What is the impact of the landing gear to the center of gravity?

These models can also contain financial and process information:

> How is the cost-per-flighthour calculated?
> How much time in the production process is spent on assembly?

Models of technical interdependencies are the key for R&D investment strategies.

This information should be made available as simple formulas or as complex simulations but the key to success is to link them and not analyse them separately. Any change to a technology or product must directly affect all linked entities.
Once the information has been modeled, across technologies and products, insights can be gained:

  • Generate Design Structure Matrices and cluster them to better understand the interdependencies between technologies, projects, products and responsible teams.
  • Perform What-If Analysis, to understand how exactly a change in technology affects each of your products and calculate the sensitivity of your results to know how well you have to succeed with your reasarch for it to be meaningful for your products.
  • Benchmark your future products against the ones proposed by your competitors.
What Could evidence-based results look like?
Flight Range of a specific Drone depending on its landing gear mass

This chart shows the Flight range of a specific drone, dependent on its landing gear mass. Very quickly now can be assessed:

  • Is this potential improvement in flight range worth the money and time that we plan to spend on 3D printing?
  • Or in reverse: How much better do we have to get in our landing gear design so that is has a meaningful impact?
  • During the development: When the original target cannot be met, you can immediately see what it means for your future products.

From your Design Structure Matrix (DSM) you can identify the impacts of Technologies to products and immediately visualize secondary effects: e.g. here, which constellations are affected by which drone or which products could benefit from a new technology.

Market analysis based on technical evidence

With all the information stored and linked, it is an easy task to extract benchmarkings along many axis within seconds. Comparisons of in-house products and competition can be analyzed at any given time, and with a change in R&D result projections be updated and the strategy revised.

TakeAways
  • A Technology Roadmapping excersise can be made far more meaningful by taking evidence-based decisions.
  • Linking all technical interdependencies is the key to generating meaningful insights.
  • These insights don’t only help to make initial R&D funding descisions, but also to monitor progress and react to unforseen breakthroughs or delays.

If you are looking for a tool which can help you perform evidence-based Technology Roadmapping, have a look at our software Valispace. The Airbus Chief Technology office is using it for their roadmapping activities.

 

Why Valispace?

“Why?” might be the most intriguing question about any startup.

Independently of the technical or economic viability of the idea or product, the answer to the motives of creating something new reveals so much more than a landing page will ever tell you. It sets the foundation as well as the vision, which leads the startup like a guiding star every step along the way.

So why do we build Valispace?

Almost every object a human interacts with in our modern world, has been built by a team of engineers. The screen you are reading this blog from, as well as the microchips inside your device; the wifi that connects you to a network of servers, which make this text available; as well as the power plant which provides the power for them. But also the chair you are sitting on, the lamp that illuminates your room and the window you look through have been carefully designed by engineers. Every screw in your refrigerator has been simulated, selected and its position debated and optimized. Its door has been tested in a refrigerator-door-open-close-simulator machine for thousands of cycles.

The reason why we engineers are able to design objects today that would have seemed like magic to people living only 100 years ago (think of rockets or smartphones), is because we are standing on the shoulders of giants: applying the knowledge of millions of scientists who lay the groundwork as well as millions of fellow engineers who developed tools and methods which we shamelessly apply and improve.

But the limits of today’s engineering are not the imagination of engineers, but the tools for collaboration: the more complex a product becomes, the more engineers need to work together on it. Designing a complex satellite for example involves usually several hundred engineers, spread over dozens of companies and its complete development and testing can take more than a decade.

Valispace aims to radically streamline the engineering process of hardware projects. This will enable small teams to design highly complex systems fast and cheap and big teams to build things which seem like magic to us today.

Software engineering already went through this revolution: while in the 70s you would need hundreds or thousands of engineers to create the simplest programs, small and lean teams nowadays are able to create amazing and scalable software[1] at a rapid pace. In our opinion, mainly two things drove this revolution:

Collaboration tools such as git or svn, which made teamwork truly feasible and are the backbone for practically every small and big[2] software project today.

Open source, which allowed the creation of reusable building blocks and tools, of which millions of software engineers benefit daily in their open or closed projects.

Today hardware engineering works exactly in the opposite way. What seems unimaginable in the software engineering world (such as emailing source code, instead of managing it in one central place), is common practice for collaboration in complex hardware engineering. Outside of the limited world of CAD, there is no common representation of engineering data, to allow collaboration or reuse. In the past 50 years, academic efforts such as SysML or other ModelBased theories have not proven to be suitable for industrial real world problems, which is why there are no useful, widespread tool implementations of these ideas in companies who actually design complex hardware.

The sad reality is that today’s collaborative engineering of the most complex hardware in the world, such as satellites, power plants, robots, etc. is managed with Word and Excel.

Expensive inconsistencies, low reuse, hardly manageable technical complexity and frustrated document updating engineers are only some of the results.

At Valispace we believe that there is a hardware revolution to come: harnessing the power of low costs for electronics, the availability of methods such as 3D printing as well as ever-growing connectivity, we expect the hardware world to make a huge leap forward in the coming years. And Valispace is aiming at becoming the backbone of this revolution.


Read more:

[1] Think of WhatsApp, who – before their acquisition – built and maintained the app for 900 million monthly users with a team of only 50 engineers.

[2] Even google stores all of its code in one single repository.