This technical information has been contributed by
Machine Downtime May Have Met its Match
Glassbeam Founder and CEO Puneet Pandit calls Glassbeam’s machine data analytics platform a “major breakthrough” for the manufacturing industry. “Our solution will collect, organize, and convert large complex log files in real time to generate actionable business intelligence for manufacturers,” he said in a recent release announcing Glassbeam’s selection by Romi, a Brazilian maker of machine tools and plastic processing machines.
Photo courtesy of Glassbeam Inc.
Most manufacturers already know that their machines generate tremendous amounts of data that can be useful for improving their performance. But far fewer know how to effectively analyze that data and translate the resulting insights into greater efficiency or productivity. That’s the gap that Glassbeam, Inc., a Santa Clara, Calif.-based machine data analytics company, is seeking to remedy with its cloud-based analytics platform.
“We’re very focused on analyzing machine data from any connected machine or connected asset, broadly in the context of IoT today,” said Puneet Pandit, founder and CEO of Glassbeam, in a recent phone interview. “We have largely the industrial business IoT market, as opposed to the consumer IoT market, so we deal with complex machines and complex data.”
When machines get connected, Pandit said, they generate all kinds of machine log data, which has immense value for manufacturers who want to understand how the machines are working, when parts may be about to fail, or how customers are using their products. It all adds up to what Pandit calls “a very rich use case on machine log data.”
“We have a software platform in the cloud that can ingest this data as streams or batch files and convert that into actionable information through our proprietary language and compiler that can design for machine log data,” he explained. “Once the data is structured, we then pull out different kinds of analytics, reporting, and charts and graphs to our customers, through our web based portal.”
For manufacturers, Glassbeam’s machine data analytics platform can help them increase overall equipment efficiency (OEE) and boost the productivity of their machines, the company says. It is also said to calculate and monitor key metrics, such as mean time between failure (MTBF), and minimize downtime by proactively monitoring and maintaining machines. These capabilities are seen as increasingly important at a time when sophisticated manufacturing equipment, ranging from CNC machines to industrial robotics, is beginning to stream more and more data to company managers and executives.
Glassbeam CEO Puneet Pandit and his team are committed to creating what the company calls “a new category of Big Data business app around machine data analytics.” Their aim is to help manufacturers mine vast amounts of rich, unstructured machine data that they say will unlock “tremendous business and operational insights.” In a recent interview with Design-2-Part Magazine, Pandit spoke at length about Glassbeam’s ability to provide manufacturers an easy way to analyze valuable machine data. Following are edited excerpts of our conversation.
D2P: It’s been said that Glassbeam “brings structure and meaning to the data from any connected device.” How is that actually accomplished?
Puneet Pandit: There are two parts of our IP. One is the language that we’ve invented, called SPL, which stands for semiotic parsing language. Think of SPL as a very rich contextual grammar to understand machine log data. That’s the language part. So let’s say we come across a new product family from our customer. For example, Siemens comes to us and says ‘I’ve got these 10,000 CT scan machines and I’m collecting data on a daily basis, an ASP sample log data. We will create an SPL connector, write language, a quick program in SPL, which will really understand how the data is coming, what kind of attributes are coming, and how to connect those parsed attributes into a scalable data model. So all of that is defined in the SPL language.
And then the second part of our IP is the compiler, which has a loader and parser, which is really our platform. It takes the SPL that we built, combines that with the incoming machine log data, and then runs it through the program and puts that into a back-end data structure, which, in this case, is Cassandra, which is a highly scalable no-sequel database. So, in a sense, what we’ve done with SPL is input the unstructured from a structured log data. That’s where our core IP gets into the structure of Cassandra and a few other data structures, and then we basically take it forward.
D2P: What would you say is the biggest benefit that manufacturers stand to gain from machine data analytics?
Puneet Pandit: There are multiple layers of benefit. The first core benefit is just storing all these incoming data streams and logs into one single repository in the cloud, which is basic log management. You’ll be surprised to find that many times in large companies, the logs come in, data is collected from these machines, but it is just sitting in isolated islands. It’s not centralized, and people lose track of it. So there’s a big benefit around log management, and that’s the first key benefit.
The second benefit is to now be able to act on that incoming data with certain rules that we embed into our engine with the help of our clients. And as that data is coming in, we apply certain rules. What that does is it basically drops certain errors right off the bat. So as opposed to some technical support engineer looking at the log data, when the customer says ‘I’ve got a problem,’ and then they spend hours or, sometimes, even days, looking at yesterday’s logs and historical logs on that machine, our engine, through applying rules, can very quickly decipher what is wrong with the machine, what kind of bugs may be in this machine. Maybe they’re not running the right software revision code, maybe they’re out of configuration, or maybe they’re using it in a different way that is not recommended. All of that is very quickly analyzed through our engine. So in a sense, we are making our customers, manufacturers, more proactive in solving problems.
I would say there are two more benefits. One is the predictive analytics, which is catching errors before they happen, catching issues before they happen. And that is done through machine learning, which could be supervised or unsupervised. You need a lot of data to derive those statistical outputs, but that is one area of predicting what might be going wrong with this machine—not today, but a week down the line, or a month down the line. And you can alert the field, alert the customer.
And the last benefit is the descriptive analysis, which [enables you to tell] your customers through a web based portal or reports, that ‘Look, we found these five issues, or this configuration in your system, and here are the five potential resolutions.’ Either you change the way you’re using the machine, or you download a new piece of software, or you call in our field engineer to solve the problem. So you’re prescribing a solution, and not just stopping it from happening, but also taking action to solve it right away, with the help of the customers.
D2P: What types of manufacturers do you target? Are they OEMs or product manufacturers more than, say, contract manufacturers?
Puneet Pandit: That is right. Largely our customer base and prospect list is centered around OEMs and product manufacturers. We also are going after operators, as opposed to makers. In IoT, people tend to use the broad terms ‘makers’ and ‘operators,’ and ‘users.’ Makers are the manufacturers, operators are the service providers in the middle, and the users are the end users who buy the equipment and use it. We are more in the ‘makers’ category. We also are going after the operators.
D2P: I understand that you recently announced a successful collaboration with ThingWorx.
Puneet Pandit: ThingWorx is a great partnership. It’s an awesome partnership, actually. We spent more time yesterday with one of their partnership executives in our office. That partnership opened up three or four months ago. We are working with IBM today. IBM has been a customer for a number of years; they’re also a big IoT player now. Cisco is another one that we are talking to. And all of these are natural extensions of our solution offering. There are probably five to ten big software companies out there who are eagerly looking to partner with technology solutions like us.
Through its custom analytics offering, Glassbeam creates custom applications—built to a customer’s specifications—to provide insights on parsed machine data.
Photo courtesy of Glassbeam Inc.
D2P: Romi, the manufacturer, recently selected Glassbeam’s IoT analytics. What did they like most about Glassbeam, and what kind of specific benefits would they expect to achieve with it?
Puneet Pandit: Manufacturing, as a market, is a huge market. Just think of the CNC machines that go on a factory shop floor and are used to manufacture end products. Romi, as an example, has 150,000 CNC machines. These machines are engineered to generate data which is used by the monitoring systems, by the operators, to set certain parameters on the machine to operate the machine, and so forth.
Let’s assume that there’s a factory shop floor which is producing, let’s say, a piece of an automobile, which is a very complex machine, really, at the end of the day. There may be 10 or 15 different kinds of machines on the factory shop floor, making certain parts, assembling certain parts of that automobile. If one of the machines goes down, the whole assembly line breaks down, and the downtime is very, very expensive for that whole production value chain.
So one of the things that Romi and many other CNC manufacturers are trying to do, is they are trying to collect data from each of these disparate, different types of CNC machines, which are doing one collective function at the end of the day. They’re dumping all of the data into one single console, and correlating data and trying to understand how to optimize the whole operation from the perspective of, if something is going down, then how do you solve that proactively? How do you optimize a certain machine when the other machine may be a bottleneck? There are a lot of those benefits you can accrue by looking at the output from these machines, from log data, and that’s a grand vision that companies like Romi have.
We just did a pilot with Romi. These machines are not connected yet. That means they’re not sending data back home to Romi on a regular basis, but the data is there in these machines. So Romi gave us some sample data and we did a quick pilot. We showed them value with Glassbeam software and how to analyze that data and be more proactive on problem resolution and root cause analysis. Now, the next step is to build more extensive dashboards on that data, and then the third step would be to connect these machines. Once that happens, there’s a whole slew of data that will flow out of these machines to provide more proactive and predictive analysis for Romi.
D2P: Can you tell us about your collaboration with ThingWorx and why you’re excited about these possibilities created by the collaboration?
Puneet Pandit: Absolutely. ThingWorx is very interesting because they are now acquired and are part of the PTC family. And PTC has roughly 25,000 plus customers from the last 30 years—some of the biggest product manufacturers across the world, across verticals like manufacturing, industrial, energy, automotive, high tech. These are many, many, many companies—large OEMs, mid-size OEMs—who are customers of PTC.
Now, ThingWorx’s acquisition is a very strategic acquisition for PTC because what PTC is trying to do is build the connectivity design from ground up into their customers’ product portfolio. So PTC has been historically in the business of helping their customers, through PLM software or CAD CAM or design automation, build better products. So how do you instrument the machine to convey the right data at the right time, so that you can take action on that data? This is where ThingWorx comes into play.
Where we come into play is kind of extending that whole value proposition from an analytics perspective. ThingWorx will provide the connectivity to its large installed base for x number of years, and we can become the extension of that whole value chain and provide analytics on the data, once it’s collected through ThingWorx. That’s the huge opportunity for us—it’s like a huge channel. Instead of hiring 50, 100, or 200 sales people the next few years, and going off with these thousand, or two thousand companies, we can provide our solution through PTC/ThingWorx as a channel.
D2P: What sectors do the manufacturers who are using Glassbeam’s machine data analytics platform represent?
Puneet Pandit: I would say there are three sectors we are [involved in]. We have more than three, but three are predominant. One is storage, second is wireless networking, and third is medical. We also have recently gone beyond these three into electrical vehicle charging stations, which is more in the context of smart grid, smart energy.
We are also executing a very interesting kind of a project with a partner of ours for ATM machines. These ATM machines generate all kinds of data, and are pretty high tech. They have a printer, they have a computer, they have a keyboard, and they have other kinds of gadgetry. And downtime in ATM machines is very expensive, again. You can lose revenues because customers get turned off if the machine is unable to dispense cash. So that’s another kind of machine vertical that we’re going after, but the predominant ones, I would say, are storage and wireless.
D2P: Today, collaboration is increasingly seen as being necessary to the development of new and innovative products. Do you see this need for collaboration as becoming more important to manufacturers recently, or has it always been necessary?
Puneet Pandit: There was a recent report released by McKinsey, analyzing the huge IoT market—they’re quoting a $311.5 trillion economy being created by 2025 just around IoT. One of the key things they mentioned in that report is that collaboration among different players, whether they are makers or users or operators, is absolutely critical to realize that economic potential in the next 10 years. Because at the end of the day, these machines can be analyzed only in the context of the environment they’re running in. Analyzing a machine on its own, just as a silo, will not create the benefit that you want. In a data center, for example, you want to analyze data from a storage machine connected to a networking switch, connected to the data center fabric of HVAC, air conditioning, humidity, temperature, power—everything together, in the grand picture, to provide cost benefits to the data center operator. So, without that collaboration, it’s going to be very difficult to realize the immense value of the IoT.
|Home | About Us | Back To Technical Library | Contact Us|
Copyright © 1996-2010 JobShop.com. All Rights Reserved.
General or Technical Questions? E-mail support@JobShop.com