Category: AI Edge Engineer

Wouldn’t it be great if every beer bottle that left your brewery met with your exacting quality standards, if overflowing and excessive foaming were yesterday’s problem, and not a persistent, wasteful expense?

What if you could be consistently proactive in your quality control process, and not reactive to cases of returned beer and disappointed customers?

Independent breweries like Sugar Creek Brewing faced just such a problem, and according to this Forbes article, the problem cost the brewery $30,000 a month.

Independent, hand-crafted breweries are discovering what manufacturers have understood for decades. You can measure product quality with IoT. But many breweries, like Sugar Creek Brewing, are using AI to predict quality issues before they happen allowing brewers to proactively manage the quality of their beer.

The Smart Factory for the Independent Brewery

The smart factory for breweries is here. Connecting your brewery with IoT devices capable of streaming real-time data to the cloud, analyzing that data with AI, and creating a responsive feedback loop that your employees can use to consistently improve the brewing process is no longer a tool meant only for the larger brewers with immense IT teams and budgets.

To help pinpoint the specific causes of overflow and excessive foaming during the brewing and bottling processes, you can connect your equipment with embedded flow meters and sensors that allow you to collect a high volume of accurate data on pressure, temperature, pH balance, carbonation, fill time and more.

By collecting all this data and applying the correct analytics, you can discover insights about your operations, improve efficiency, stop equipment bottlenecks, and help your employees accomplish their daily tasks.

Working around high-pressure tanks presents big risks to employees working on the production line. With the right pressure sensors on every kettle, still, brewing pump, and with object detection in place, the factory floor can become a much safer place to work, as production managers can track performance issues that could lead to serious injuries.

The fermentation process is the heart of the brewery factory. What if you could get constant feedback on this process without opening that tank?

Breweries See a Return on their Investment with AI and IoT

Not only does the Smart Factory for Breweries improve efficiency, beer quality, and employee safety, but it also helps improve your brewery’s financials.

• Save on returns from poor quality products
• Reduce waste of materials for bad batches
• Fully use capital equipment
• Protect the health and safety of employees and customers
• Reduce energy use
• Monitor and track batches effectively for compliance and regulations
• Save on parts and repairs using AI predictive maintenance
• Control inventory with RFID tracking

The cloud has allowed the cost of analytics systems to drop significantly, and many IoT platforms can right-size their solutions to fit with your consumption, giving you the option to ease into your IoT and AI journey affordably.

Connecting KPIs as a Measure of Success

As a brewer, you’ll want goals and specific metrics tied to your Smart Factory for Breweries. Some of these include the following depending on your solution approach:

• Equipment downtime for repairs
• Predictive cleaning schedule vs set schedule
• Returns from retailers because of quality
• Bbl brewed per hour
• Bbl packaged per hour
• Bbl lost per hour
• Bbl to energy costs

Your Business Processes will Need to Adapt

There are daily processes undertaken in the brewery that add to waste and risk contamination. The typical small brewery will draw liquid from the tanks daily to check for quality, consistency, and the state of the fermentation process. This is the process of monitoring specific gravity. It’s not a great process when managed manually, because only checking once or twice a day isn’t likely to catch a problem quickly enough, and with each test the product is exposed to possible contamination.

To reduce the risk of contamination and increase regular monitoring, it makes sense to add a sensor to the tank to collect data. The data collected from the tank can then inform operations.

But how is this continuous monitoring done on large tanks that are usually made of copper or industrial steel? This is where it helps to understand the problem and find the right IoT device for the job. In this case, low energy Bluetooth devices floated in the liquid can communicate with gateways positioned nearby. This allows IoT to completely replace a manual process, as well as do a better job of monitoring for regular consistency.

This is why it’s important to work directly with a partner who can help match the right product with the situation. A great partner will learn your business processes and won’t be afraid to point out where your business process will need to adapt to fit with the innovation the technology brings.

A Use Case: Automate the Manual Method of Measuring Specific Gravity

Brewers are well aware of how this process works, but for this use case I’ll summarize it:

Specific gravity will indicate the stage of fermentation. Brewers use this measurement to determine the state of fermentation—is it on track, done, or perhaps stuck. Yeast can stop consuming sugar too early, and when this happens the brewer needs to take steps to save the batch. This is one of the reasons continuous monitoring is required for the modern brewery. The sooner you’re aware of a stalled batch, the sooner you can undo the damage.

This is typically measured using the manual method described above. The actual measurement is looking at the liquid’s density in comparison to water. As beer ferments and coverts sugars to alcohol and gas, the specific gravity falls and the liquid gets less dense. Eventually, there are few sugars left and the fermentation process slowly stops.

It’s clear to see why putting a machine in charge of monitoring this process consistently helps improve this task.

But what does this look like, technically? Are there tools and systems in place today to make this easy to install and monitor without risking batches of beer or installing expensive equipment?

Tilt Pro, Bluetooth, and IoT Central

If we focus on one particular use case, like measuring specific gravity, there are a number of ways to accomplish this, but the following architecture is one that I recommend. This architecture is built for a small brewery with just a few fermentation tanks. It uses off-the-shelf devices and requires minimum programming to get the solution up-and-running. But it also has room to grow into a robust platform for managing more than just fermentation monitoring.

FERMENTATION TANKS

The assumption is that you’re using industry standard fermentation tanks. These are, without a doubt, a challenge for electronics equipment. Most are made of steel or copper, which means that they basically act like a Faraday Shield capable of blocking or dampening signals.

Some IoT devices will not work in this situation. So either you need a wired solution or something clever like the Tilt Hydrometer. When I first discovered this device I thought that perhaps it was a toy for homebrewers. Surely, nothing as low in price as this device could have professional applications, right? Well, after reviewing the device and seeing it’s capabilities on paper, I’m willing to include it in this architecture.

Especially, backed up by a custom device to work with the iBeacon signal generated from the Tilt. For this architecture, I’m recommending the Tilt Pro. The standard Tilt device might be adequate, but I think it’s worth the extra money for the added battery life and sensor capabilities.

The Tilt device goes inside the tank, but it’s enclosed in a protective shell. Cleaning and care instructions are available on their website. Here’s the marketing breakdown on the device:

At 3x the volume and weight of our original Tilt hydrometer, the Tilt Pro’s larger size allows us to more than double the battery life with preinstalled Energizer® AA lithium batteries that will keep it powered for 3 to 5 years depending on use. So much time you may forget your Tilt actually runs on a battery! The extra size and weight comes with an extra bonus: improved specific gravity stability at high krausen so you will see a steadier read-out on your iPhone, Android, or Raspberry Pi. In addition, we have boosted range with a high-gain antenna that in testing increased range by 20% vs. our original Tilt. Now you can see your Tilt through thicker heads of foam and heavier stainless steel walls. To top it off we’ve included the extra decimal of precision available from the sensors and to the read-out so there are now decimal degrees Fahrenheit as well as higher resolution Celsius. Specific gravity now reads out to the 10,000th place (for example 1.0000). You can now see fermentation activity dropping less than a brewer’s point per day. In all we’ve made several adjustments we believe the craft beer professional (and serious home brewer) is sure to appreciate.

The Tilt’s cloud backend is basically writing data to Google Sheets. We definitely don’t want to write our data to Google Sheets. No offense to Google Sheets. It’s a great app for many purposes, but we want something a bit more robust. I also do not want a mobile app. I want a device dedicated to capturing the Bluetooth signal, converting the raw message from that signal to MQTT, and delivering that to a server on the Brewery’s internal network.

DATA FLOW FROM TANKS TO EDGE GATEWAY

The basic flow of data from tank to IoT gateway looks like the following:

The Tilt is a one-way leaf device and it’s tightly coupled to a nearby Raspberry Pi Zero W. By nearby, I mean as close to the tank as possible, other than inside of it.

The Pi gives us a lot of development options. We could, if we wanted to, connect the Pi directly to Azure IoT Central. This would make it a gateway device. It would still need a WiFi router or hub to connect to, but it could manage two way communication with IoT Central.

For this architecture, though, I want to couple a Pi with each Tilt. Considering that the cost of each device is minimal and the power you get when you add a more robust gateway, like the Ectron Edge Computer ECT-ECI, this allows us to remain technically small and simple, but with the option of applying more sophisticated Edge Patterns.

For instance, there might come a future date when I want to I want to deploy machine learning modules to the Pi. These modules could be specific to each tank’s beer, allowing brewers the ability to monitor for different types of beer.

THE ARCHITECTURE

The following is our more robust design.

I want to continue with the idea that this is just a proof-of-concept set with one use case. However, I’ve expanded out to manage three different tanks. Each tank as been coupled with its own Tilt device, Raspberry Pi Zero W, and all three of these connect to the Ectron Edge Computer.

Where we start to flesh out the solution, is when we collect the data in the cloud. Here we’re using Azure’s IoT Central, but we could just as easily use any other cloud IoT platform. I like IoT Central because it’s quick to get an app up-and-running, it makes device management simple, and includes all the analytics you need to prove out an IoT POC.

It’s also scalable, so building out to more devices and more use cases is easy. IoT Central is meant for smaller to mid-sized IoT projects, but it can still support multiple organizations in production scenarios.

When you start to work with more than one million devices, though, IoT Hub, AWS IoT SiteWise, or Google’s IoT Core are better choices. Though I designed this with Azure in mind, the Gateway is flexible enough to work with any cloud backend IoT service.

A couple of future use cases we could explore once this platform is built:

• Use Machine Learning to predict fermentation stages for specific beers
• Use time series analysis to detect anomalies in the brewing process
• Collect data specific to certain recipes, build guidelines and intelligent monitoring around these recipes, and alert Brew Masters to tanks that drift outside the confines of the established guidelines

Who do I need to put this solution into place?

The team to build this out could be relatively small. One person with the right skills could build this entire solution, but what skills will they need?

1. The Tilt devices are relatively complete. As an off-the-shelf solution they provide a lot of flexibility. To work with this device, you’ll need to understand how to collect Bluetooth Beacon data. Lucky for us, Tilt includes a link to work with this data from their website’s FAQ. https://kvurd.com/blog/tilt-hydrometer-ibeacon-data-format/
2. Creating the coupled Pi Zero will pose a slightly bigger challenge. This isn’t an out-of-the-box solution, but it isn’t undiscovered country either. Basically, we want to run Containerd and the IoT Edge Modules on the device. This will allow us to deploy to the Pi as though it were an Edge Device. We could also use the IoT Central SDKs to create a basic device and send data. We have a few choices here but ideally, you want to capture the BLE message, format it to something that best fits the use case, and then wrap that in an MQTT message to send to IoT Central.
3. The Edge Gateway is probably one of the easier pieces to implement. This article covers the process, so I won’t dive into it here: https://docs.microsoft.com/en-us/azure/iot-edge/how-to-connect-downstream-iot-edge-device?view=iotedge-2020-11&tabs=azure-portal
4. Setting up the Edge Gateway and devices in IoT Central, is also trivial, but that might be a different person completing the set up than the device developer.

Basically, if you have someone with experience developing IoT Edge modules, lite Python programming skills, and IoT Central administration basics, you could have a solution up-and-running quickly.

Is this Secure?

If you keep the signals within the brewery on their own dedicated network and authenticate the devices through IoT Central, it should be secure. The real trick here is to make sure you provide a separate network on your brewery floor from the rest of the company’s network. For instance, many breweries are also open to the public for dining and tours. You don’t want the public to even be able to see the brewery floor network, much less access it from their devices.

In this scenario, the endpoints for IoT Central are the only exposed URLs. You can protect these with device authentication. Logging into IoT Central as an administrator would be protected by Azure Active Directory using MFA.

What about my data? Is it backed up?

IoT Central only stores about 30 days worth of data, but there are means of integrating storage solutions with your IoT Central App. This isn’t reflected in this architecture, but at a minimum I would recommend adding an Azure Storage Account to retain messages. You’ll likely want this data for future data science and trend analysis.

What if I have more than one brewery?

This solution isn’t designed for extreme robustness or resilience. I definitely put this solution in the category of proof-of-concept or minimal viable product. If you set this up, and you like your initial tests, and you feel like this is something that could expand to multiple physical locations, IoT Central can definitely handle this up to 1 million devices. If your device requirements look like they will spread beyond that number, it’s time to consider Azure’s IoT Hub solution with a more complete Azure Analytics backend. Synapse or Databricks with Azure Time Series Insights make great additions to this type of architecture.

How do I manage this environment

As long as this is kept small, it’s probably ideal for a single developer or administrator. IoT Central has a few built in roles that the administrator or App owner can assign to users so they have access to dashboards. At a minimal, I would consider these roles in IoT Central:

• Application Administrator can manage and control every part of the application – give this role to someone in your IT operations department
• App builder has many of the same abilities as the Admin, with the exception that they can’t make administration changes or connect to data exports
• The Operator role is ideal for someone who needs to monitor device health and status. This is for anyone working the brewery floor

Those are basic roles. If you begin to build your solution out to multiple locations, these can be defined as Organizations. If I own a brewery in three different cities in my state, I might have my Fort Worth brewery, my Dallas brewery, and my Austin brewery. I would separate those by Organization and assign the above roles to users in those areas.

What can I do with this platform, once it’s up-and-running?

IoT Central comes with many tools needed to run operational IoT.

• Robust Dashboards that can be customized for many different views
• Device life-cycle management and Gateway management
• Rules triggered based on telemetry with built in messaging to email, webhook, Azure Monitor Action Groups, Microsoft Power Automate, and Microsoft Azure Logic Apps
• Time-series Analytics allows you to create and save queries
• Run jobs on device groups that can send commands to devices or change device properties
• Build and store device templates
• Export to multiple data sources
• Complete administration of the app, the users, and pricing

In addition to the IoT Central benefits, you can build upon the IoT Edge devices by integrating device specific logic or machine learning models.

If I have an IT team, will they need a lot of training?

Perhaps, if they aren’t familiar with IoT. However, Azure does make some of this fairly easy. It helps to work with a partner who understands how to make cloud solutions operational. Knowing and understanding the burdens and responsibilities of a lights-on operations team will help your partner understand what things the team needs to monitor for and integrate into your team’s existing playbooks and operations procedures.

I’m interested to see how IoT can help my brewery

If you aren’t already using IoT in your brewery, it’s important to understand that there is a learning curve. As I’ve outlined here, it takes some time to understand how the technology can help you and how you might need to adapt to the technology. If you have the right partner, this can be a smooth transition. There are definitely some key stages I think a partner should walk you through before you make a large financial commitment.

1. Be careful with any solution that appears to be a one-size fits all. Your brewery is unique. Your product is unique. You didn’t become an independent, craft brewer because you wanted to mimic a certain company with the initials of AB. So AB’s amazing, multi-million dollar IT solution probably isn’t right for you.
2. This isn’t a quick, one-and-done project. IoT is a journey best taken in measured steps. The right partner is going to understand that process and they will help guide you along the path.
3. The right partner takes a wholistic view of you, your brewery, and cares about your bigger picture goals. Selling gadgets is great, if all you do is sell gadgets, but a solutions partner wants to make sure you have the right gadget for your needs.

Solution Visioning

Expect a couple of workshops from your solution partner. Believe it or not, these aren’t just a way to make more money off of you. These are a necessary part of the process. A partner needs to learn about you, your business, your aspirations and your challenges. These envisioning workshops are a way to achieve that.

Architectural Understanding

A good partner will want to understand your network topology, but also your enterprise architecture. Not just the technical systems, but your policies, procedures, your team structure, and even your company culture.

POCs and Pilots

Smaller, measured steps into a bigger picture solution is a great way to learn, without breaking the bank. The days of huge, multi-year projects only to deliver a technical nightmare that doesn’t meet your needs is gone. Be ready to work with a partner willing to take small steps toward bigger goals.

• Limit your use case to one single point of value
• Limit your scope to one particular area of the brewery
• Time-box your efforts. Don’t make the mistake of trying to enforce an artificial deadline, but don’t get mired in complexity.
• A POC won’t be production ready, but it will be testable with real-world scenarios
• Take this opportunity to learn and try to avoid assumptions
• It might require multiple POCs to arrive at a worthwhile pilot
• It will require your time, the time of your staff, and your feedback—you can’t hand off your standard operating procedures and expect a bespoke solution to your particular needs

How do we make this production ready?

When you’re prepared to take this solution to production, there are a few factors to consider.

1. How do we reach zero-trust security with this platform? Are our users authenticating using MFA. Does each user have the least privileges possible? Are my devices hardened against tampering? Are my Edge devices loaded with security modules?
2. Do I have system oversight? Is there the appropriate level of logging and log monitoring in place? Are there alert groups established in Azure. Are these alert groups receiving alerts from IoT Central, devices, and any other backend analytics systems?
3. Is data backed up securely? Is access to this data limited to integrated systems? Is access to the data monitored?

There are other considerations like building out staging layers for developers and testers. A more complete security monitoring tool like Microsoft Defender for Cloud should be a consideration. Hiring penetration testers to test your security is also a good idea.

Teaching your team how the system works

Hopefully your team has been involved from the beginning. A good IoT integration partner will talk to the brewery floor crew during development and system installation, get their feedback for alerts and dashboards, and familiarize everyone with the systems as it’s built.

Good formal training for the brewer floor will be something a partner can provide. Delivering this training in documentation, video, and one-on-one sessions is the typical method for handing off a production system. As your system grows and matures, you’ll eventually want staff trainers who have made your technical solutions part of your floor’s standard operating procedures.

What if I need continued help after the solution is built?

Let your partner supplement your time with ongoing support and monitoring. Many IoT developers and integrators either have a managed service team or they outsource that job to a trusted partner. If you don’t have the IT staff to manage and monitor your solution, your IoT partner should be able to provide one for you at an affordable rate.

Who do I include in partner meetings?

Independent brewers tend to keep the company lean. Titles like CEO, CTO, and CFO might not exist. That’s not a problem. The people who need to be at the table are decision makers, financiers, brewery operations experts, and trusted advisers.

Are you available to help us?

Yes, as a certified Azure IoT specialist and AI Engineer with twenty years of IT experience, I’m in the ideal place to help with this exact type of solution. I’m a strong advocate of cloud transformation. I’ve helped large and small organizations transform their operations using Azure and Big Data analytics platforms. I’d love the opportunity to work with a brewery interested in exploring how IoT could improve their operations and save them thousands of dollars a month.

What are the next steps?

Are you happy with your current brewery floor operations? Now that you know there are better ways to brew beer, can you continue to operate the same way moving forward? Inertia is difficult to break, but one thing that helps is a clear decision. Is a Smart Factory for Breweries something you’re ready to take on? Or can you continue on your present path?

When it comes to these type of decisions, it’s usually best to start small. This is one of the reasons I’m an advocate of the POC. With a small equipment investment, an Azure subscription that only charges based on consumption, and a hands-on consultant who can take you through every step of the process, it’s easier now than ever before to test your ideas without breaking the bank.

The cloud, IoT, and AI are tools that brewers use to improve operations, innovate, and save money. If you’re interested in seeing if this is the right path for your brewery, I’m here to help.