Tracking medication use on the farm

Medication use, specifically antibiotics on the farm are a spotlight issue for animal agriculture.  Consumers, restaurants, retailers and politicians are joining the discussion.  Traceability is a common buzzword that is heard on medications at the farm.  This is a particularly difficult issue to install an iot solution around because of the intrinsic need for people to execute at the ground level.  Since we are not medicating animals daily, it is not a part of the daily routine, which makes it more difficult to train people to execute these important processes.  We think the data platform and mobile app we have developed can help.  Here is what we are doing in FarmStreams:

The mobile app allows two different type of updates:

1.  A QR code or NFC type system where medications are bar-coded or QR coded before going to the farm.  Each doser or medicator in the farm site is also bar coded or NFC coded.  A barn worker would scan in the medicine, verify its correct.  Enter the amount.  Then scan the doser at the water line before setting it.  Repeat same process for stopping the medicine only just scan the doser code and STOP.

Some snapshots of what that looks like is below:

   

2.  If the QR/barcode/NFC type system is too complex we also have the option to just have the barn worker use the app to select from a drop-down menu what and where they are adding meds to the water.  Enter the amount.  When finished stop the medication in the app.

The mobile app can also be set up to do other process verification steps such as:

1.  Have the vet dispatch instructions for a specific group to be given a specific medication and a specific amount by a certain time.

2.  Send customized and frequently updated instructions/messages for the selected medication.  Withdrawal days, mixing instructions, other pertinent data that may not be on the bag itself.

3.  An email can be sent to the vet notifying them the medication event began / ended.

4.  If the medication event has not occurred within a specific timeframe then an alert notifying the vet and supervisor is also sent.

The mobile app cannot ensure that the medication that was entered into it actually got into the water and to the animals.  These processes break down quickly when you don't have participation from the workforce.  We have discussed labor quality previously in the blog.

On the sensor side most of the dosers in the market today do not have cloud connectivity or even local data storage.  This often makes people throughout the organization blind to what is going on inside the farm.  If there are issues like plugged diaphragms they must be troubleshot and recorded by the worker on site.  Data quality is often compromised and not uniformly gathered.

There are newer doser products that are sold at Farm Tech Store that connect to the cloud using the MultiSense sensor gateway.  These measure water consumption, medications injected (how much and when) to the cloud.  When combined with the data entered in the mobile app, you have a report of exactly what went in the water, how much and at what times.  Critical streams needed for a good traceability program.

One issue that is not resolved yet is having that medicine / medicator handshake on what product is going into the water.  You still need a human to enter that data, a potential weak link in the data chain.  In the future some type of cartridge or other medication reading technology may be placed within the medicator hardware; but for now that is futuristic thinking.  It can be done; but need someone to pay for it.

Another issue that is possible; but not practical today is to have the vet receive an email telling them that the medication is in the medicator ready to be inserted into the water.  The medicator can be started with an approval from the vet's mobile app.  100% possible, impractical to develop until someone says they want that level of control over their medications on the farm.

When we get over to the data side in Grovestreams this is where we can put some logic and alerts around checks / balances to ensure the processes are followed properly.

For example,

If a vet has dispatched a medication event, set an alert that lets you know if it has not been done within 4 hours in the barn.  Double check not only the app but also the medicator data stream.  The medication should be entered into the app and flowing to animals within X hours or vet/supervisor are being alerted.

Once the medication has started check to see if it has stopped by X hours.  If it is still medicating after 48 hours for example, send an alert or if medicator is set up to be turned on/off remotely, turn off remotely either with software or with an app.

Cross reference other data in the store.  If a specific medication shouldn't be given to birds greater than 24 pounds.  Check that, cross reference it.  If its greater, send an alert.

Produce reports that document how much and over what times were medications given by barn.  Further analytics from the feed sensors, scales and water lines would dive into whether they were effective and possibly deliver payback scenarios in real-time.

There is technology that can help organize the process around medications on the farm.  However there is not an easy button.  It still requires 100% buy-in throughout the organization in the value of the data and a commitment to the process.

What is Big Data?

Big Data was created or if not created at least taken mainstream by the powerhouse known as Google.  So, who better to turn to ask what the definition of Big Data is:

big da·ta

noun

COMPUTING

1. extremely large data sets that may be analyzed computationally to reveal patterns, trends, and associations, especially relating to human behavior and interactions.

   "much IT investment is going towards managing and maintaining big data"

I would say this is a fairly accurate representation of what it means to me with the caveat replacing "especially relating to human behavior and interactions" will soon be replaced with more industrial terms, like machines, robots, software or business processes.  

Big Data processes like FarmStreams can constantly be verifying conditions in barns and alerting when conditions are not ideal for maximum production.  In addition, additional "smart automation" can be installed where several dependents are monitored and when conditions are met then equipment is remotely controlled to perform different tasks.  

Some good examples of a simple smart command would be to automate a bin slide to switch bins when one runs empty, open the full bin and shut the empty bin and order new feed for the empty bin.  That is a simple automation that could be accomplished pretty easily with Big Data.

A more complex automation that could be done is to calculate weight of feed consumed per bird and once it exceeds a certain percentage of body weight for the day to switch feed rations or stop feeding for the day.  Additional complexity could be added to cross reference water consumption before making a decision.  I am not sure if this is a realistic scenario but it is illustrating that many different streams can come together to automate a variety of decisions inside a barn.

Furthermore, within animal agriculture this level of data collection from the farm has never been available.  Most data that is available remains on the site.  Once the smart professionals within the animal ag industry get their hands on this amount of production data I am certain the value they find and impact will be tremendous.

So, how big is our data?  The data platform, Grovestreams follows the Google Big Data architecture with its own wrinkles.  This is called Hadoop.

So, what can we do?

• Farmstreams can store 94,000,000 samples for each stream - this is 3 years of one-second resolution.
• 140 different statistics for each sample available for each rollup period from one second to many years, with customizable time definitions in between.
• Each sample can have unlimited derivations (formulas) occurring with them. 
• The derived samples can also have unlimited derivations (formulas) occurring with them.
• Near real-time processing and flow-through.
• Data in / out seamlessly from cloud data platform back to the enterprise level system.  Users can add us as another user interface; but don't have to.

Most livestock farms will not need single second resolution on their data; but it is nice to know the data platform was built to handle that heavy of a workload.  The ability to run derivations / formulas in real time converts several data feeds into valuable information that ensures the livestock barn is optimal.

That is what Big Data means to us.  We are focused on data quality from the sensors.  Connecting the sensors in rugged, tough environments.  The data piece on the back-end.  Optimizing the production environment for the best performance possible.

More feed line problems (setting delay timers)

Once we start gathering more and more feed data and keeping in the store we can go back and look at some interesting things.  We have already posted about alerts to solve feed problems, such as feed outages, bin rotations, and feed ordering, and what we can do to help solve them. This morning I decided to go back and look at frequency of feed events and discovered some alarming problems which we went over this morning with the site owner and are fixing.

In my analysis, I found that on some feed lines, the Feedmeter was reporting feed events very frequently, sometimes every minute.  See the underlying data in the store below:

In 18 minutes, we reported 16 feed events.  When I roll up the data to a day and also look at the sum, average, sample count I see the following data.

Several issues for us to work through here.

1.  At that small of a run time the Feedmeter is set up to fail.  A key variable in the Feedmeter's algorithm is the run time.  If a setting is off causing the start to end run time to be off by 0.25 seconds, it is rounding error on a 60 second run; but catastrophic on a 1.25 second run.  The Feedmeter's ability to estimate intake is not optimal.

2.  Feed motor life expectancy is slightly reduced for every time the motor runs.  A motor that runs 500 times a day versus 50 times a day probably will go bad much faster.  I did a quick inquiry with some equipment sales guys in town and a typical motor life is 5-6 years and costs $225 to replace.  Moisture (see future blog post about humidity sensors), feed quality and frequency of running will cause them to go bad faster.  

3.  Feed motors failing cause feed outages.  Without a full suite of monitoring it is easy not to recognize this problem until it is later than you would like.  By the time it is fixed, the feed outage has likely cost you money and hurt your animals.

So how does it happen?  Its a tough job and barn workers forget about it.  Perhaps a line didn't start in the past and the barn worker blames the delay timer and shuts it off.  There is a lot of things to train and its possible the barn worker was never trained in how to use the delay timers.

We could set alerts around delay timers if we wanted to it would be very easy.  Simply take the current value's timestamp in the store minus the previous value's timestamp in the store and see if it exceeds a minimum delay timer value.  If it doesn't send an alert.  I always hate having pointless, repetitive low-value alerts so I would be more prone to having a once a week check.  Grovestreams does a very good job of allowing you to add unlimited intelligence to the alerting engine.  Perhaps in this case put in the software to look at the average time between feed runs for the past week or count the total feed runs and alert if it exceeds a certain value.  

In addition, monitor your whole site for feed.  If a barn worker is worried about the line starting up on a delay timer we have an alert for that out of the box.  

In conclusion, putting a cost on improper settings on a feed line is nearly impossible and not that large in the grand scheme of things.  The bigger issue is the cost of poor on-site management to livestock operations.  If barn workers are not setting delay timers properly it is an indication of poor management which likely have bigger ramifications throughout that site's production.

Monitoring employee's time - Trust but Verify

Ask most livestock producers what the biggest difference between a high performing site and a low performing site and they would tell you the on-farm labor.  Sites with good managers perform well, sites with poor management perform poorly.  FarmStreams entire business exists because of this fact.  Drilling deeper a good indicator of management quality would be the amount of time an employee is on the site.  A 45 minute walk-through and leave is going to have different results than a site that has an employee on site for 10 hours each day.  Its just common sense.

One of the benefits of using FarmStreams is that we can watch over all of the systems for a farm worker while they are gone in a much more intelligent way than the existing alarm systems in the marketplace.  Its almost like extending the management time on a site with software and advanced automation instead of a person.  Its a tough concept to embrace; but it can really be true particularly in systems where the labor force is stretched thin or even the supervisor level is stretched thin allowing workers to take shortcuts because they know that they have a good chance of getting away with it.

So what is our solution?

First, monitor the entire site to ensure all systems are working at all times.  We talk about this at length in the blog and will continue talking about ways we do this.

In addition, install sensors such as door sensors or motion detectors in common areas.  Collect the information and transmit it to the cloud via the MultiSense sensor gateway.  The calculations would be very easy.  Last timestamp in epoch millis minus the first timestamp in epoch milis converted to a decimal hour should be a pretty accurate time on site.  The full suite of alerts for an upper manager would be available.  Less than a certain threshold triggers an alert.  Fully customizable.  Also the full reporting/dashboard graphs would also be available.

I have had personal experience with some barn workers and its often hard to believe them as much as I want to.  When we have feed problems on the site they are never there.  They left an hour ago.  They weren't planning to come until 3pm today, whatever the situation might be 9am or 6pm the story is the same.  Like any business you have good employees and you have bad employees.  I would not say that me calling them and them not being near the site means they are bad.  It could be coincidence.  However if the monitoring sensor kept track of human activity on the site each day you now have a way to go back and know what is going on with that employees time.

As one industry exec recently told me.  Trust, but Verify

Machine Learning

Usually I do not blog until the evening times here in Minnesota; but today I stumbled across a very good blog by a very good blogger I follow David White which I had to push out there.  He blogs at www.industrial-iot.com.  The blog I read is here.  He talks about machine learning in industrial applications and I especially like this quote:

"In the long run, a weaker algorithm with lots of training data will outperform a stronger algorithm with less training data.  That’s because machine learning algorithms naturally adapt to produce better results based on the data they are fed, and the feedback they receive."

This has really been true during our trials and tribulations of developing a product.  In the earlier days often we were trying to put a single feed sensor on a line and drive value from that.  We spent a ton of time on our algorithm of that single sensor and got it to be pretty good.  Even with that really good algorithm, the data would quickly get skewed, generate false alerts and leave customers unhappy.  This was painful to go through.  In reality, we needed to have several more sensors feeding in to a robust data platform (like GroveStreams) which then we use all the data streams to interact with each other to provide the valuable information.  As a result of these streams interacting we can then teach the sensors how to be more accurate.  If we use several different sources of data I have now taught devices to be much more accurate over time as David White suggests. This is a huge breakthrough for what we are working on accomplishing.

Often times users are on a budget and only wish to try out certain sensors to save money.  We need to really resist this urge for this very purpose.  All of the streams working together provide a more complete picture of performance and are needed to effectively reach the goals.  If we do go halfways at a site, users need to realize the picture painted may only be halfways completed.

Feed alerting solves costly problems.

In previous blog posts we talked about how remote monitoring can have provide the tools needed for precise feed ordering and feed budget execution.  We have also talked about feed outages in previous blog posts.  In this blog entry I will show you the ways that we handle alerts as it relates to feed.  We have several already set up standard out of the box.  In the coming weeks I will be adding some new alerts to the system.

Alert #1.  Line Empty.  The communication hub on farm is monitoring the Feedmeters on each line.  If it detects that a line is running for longer than a pre-set time (out of the box is 15 minutes) a message is sent to the cloud.  Every X minutes thereafter another message is sent and the Grovestreams alerting engine processes it and sends alerts when your alert conditions are met.  This is a very effective tool for knowing when a feed line is having issues on the farm.

In addition to the alerting, a calculated value counting the feed outages occurring during a turn will be a newly discovered KPI for livestock farms.  Furthermore, calculating the cost of feed outages with this data will begin to have more data behind it, potentially answering that question.

Alert #2.  Line Full.  This is a rare alert condition; but can happen.  This is when a tube or something in the barn comes loose and feed spills on the floor.  The Feedmeter can detect the difference between an empty and full line very well.

Alert #3.  Time between feed events.  Out of the box the setting is 5 hours but it can be customized. Grovestreams handles this issue very intelligently.  Using an interval stream, we look to see how many pounds of feed have been fed in a period of time, an hour but can be more granular if needed.  If the pounds is zero then add a 1 to the previous value.  Once the stream becomes equal to or greater than 5, send an alert.  In the alert settings, I set this to check in on the alert condition every hour and if the line still hasn't reported a feed event you are reminded of the issue.  This is an important function because on Alert #1 often times feed lines will time out after a period of time.  Alert #3 with the reminders makes sure that the issue is not forgotten.  

Furthermore on Alert #3 often times animals sleep at night.  This is particularly prevalent in turkey and poultry.  So, rather than getting alerts in the middle of the night that no feeding has occurred you simply set a time filter to ignore night time missed feed events.

The image below shows the previous 21 days of spikes in periods where a feed line has not run.  As you can see that overnights sometimes the pigs don't eat on this line until morning.  If the value exceeded 5 you would get an email.

Alerts 1-3 are problem alerts.  If these occur someone needs to fix it immediately.  

We also have other alerts that are possible and would be quite easy to model.  For example,

Concept #1 - Alerts around feed intake / animal as compared to a budget.  If the intake is different from the budget by X% an alert is sent, perhaps once a day or once a week.  It could even alert differently based upon how big the difference is from the budget.  

Concept #2 - Alerts around bin rotations, feed budget management.  Alert when the feed bins are not rotated properly.  We have identified these issues in previous blog posts.  Calculate how old feed is on site, perhaps you have some KPI's on that and we want to minimize age.

Concept #3 - Alerts around movements in feed intake.  This happens with every group of animals.  At some point they really take off and it can surprise you.  Their feed intake can jump up significantly and you run the risk of a bin running empty.  Run an alert to tell you when the previous day's intake differs from the previous 5 days by X% or more.  Investigate why that happened.

Concept #4 - Alerts around bin deliveries.  This can and will become more important as feed ingredient traceability becomes more important particularly when medications are in the feed.  Alert concepts here would include: If bin fill event occurs > X hours from feed production (API push from mill to Farmstreams).  Bin fill event occurs without a feed production event.  Anything that relates to exceptions in that handshake between the feed mill and the correct bin filling transaction is an alerting event.

In conclusion, the alerting engine has a lot of very useful tools for providing customized information to a user and weed out false alerts.  The first 3 alerts I identified that come standard with FarmStreams today are critical for livestock producers to have.  There are clear payback scenarios here and good management would require knowledge that animals have regular feed.  The others are provided to show you just how powerful our platform can be for advanced analytics.  Likely not immediate applications; but as we start improving the every-day management of barns by using tools like FarmStreams offers I am confident we will start diving into those details in the coming years.

Entering Temperature in Mobile App

In my last blog entry, we identified temperature sensors as being a low-cost sensor providing valuable information to barn workers, managers and analysts in a livestock operation.  If installing sensors and connecting them to the cloud is not in the cards but the data would be valuable to collect perhaps the mobile app is a worthwhile option.

Select the temperature icon from the customize-able list of options.

Select the Group reporting.  This is a drop-down based upon groups your user-id has access to update.  Timestamp is set at the current timestamp, cannot be edited in the standard option.  Then an decimal entry field for low (required), average (not required) and high (required).  Most barn controllers will have options for a user to look at the previous day's high and low reading.

Once the data is collected in the cloud, charts showing the data over time are standard.  Other analytics are also possible as we discussed in the last blog entry.

The mobile app option is an improvement over just leaving the data in the barn on the controller.  Some of the improvements are:

1.  A requirement for the barn workers to look at the temperature every day.

2.  Getting the data into the cloud gives you flexibility and centralization to your process improvements efforts.  Also an ability to correlate data to other interesting data like feed, water, and mortalities to name a few.

3.  If barn workers are not doing the work each day, you can structure alerts around missing data.  If a barn worker does not enter temperature for 30 hours an alert goes to their direct supervisor.  If 72 hours passes it escalates to a manager.  Rapid feedback of compliance issues.

The mobile app still has some potential flaws, including:

1.  Barn workers may falsify data entries if it makes them look bad.  Difficult to prosecute.

2.  Actionable alerts are delayed versus real time sensor data.

In conclusion, logging temperature data in a mobile app is a low-cost step for well-run producers to more easily collect on-farm data into the cloud for analytics or push into their own enterprise systems.  For poorer managed sites, it is a step in the right direction for behavior and process improvement.  FarmStreams mobile app runs on both Android and IOS.  Contact us to set up your site today.

Sensor Spotlight: Temperature

Air temperature is a very important measurement in hog and poultry barns around the world, impacting animal health, feed consumption and growth.  Every single barn controller system I am aware of has temperature readings which control various heating/cooling/ventilation equipment to maintain the optimal temperature within the barn.  Most of these controller systems either do not store the information or keep it locally which is not helpful for analytics.  Many producers also likely pay for an alerting service which temperature is likely a variable that is alerted upon if it is too high or too low.

Logging the temperature data in the cloud can be very valuable.  By connecting an existing controller to a cloud service like Grovestreams (which will be possible with the MultiSense Plus soon to be available) or by installing independent temperature sensors you now have that data logged forever.  You can simply produce charts which show the temperature over time, the average/high/low for a day over time or a variety of display options.  This is generally easy stuff for most data platforms to handle.  GroveStreams does it well; but so do others.

What we can do in FarmStreams using the GroveStreams platform is far more in-depth.  

One area is in the alerting engine.  Most alert systems are set up to be very simple.  Is it higher than a high threshold, call.  If its lower than a low threshold call.  A basic alert protocol in our alerting engine could be to set different thresholds based on the age of the animal.  At day 7, 92 degrees is the high threshold, at day 60, 84 degrees is the high threshold.  Other more involved alerts could include feed intake, if temperature is below 50 degrees in the barn and the animals have consumed over 6 pounds of feed in the last 24 hours.  That also could be a sliding scale: ex 3 pounds when they are 7 days old, 6 pounds when they are 67 days old.  It could be an alert based upon a percentage over or below budget also tied to temperature.  I am not saying these are practical alerts to execute; just showcasing how complex an alert condition could be set up as in our platform.

In addition to the more in-depth alerts we also can do some real interesting analytics with all of the data points.  Since we have accurate hourly intake information we can easily compare very granular actual versus budget for different temperature levels.  We can use time-filters to look at consumption at different times of the day actual vs budget across an entire system for different temperature levels in the barn.  I am not suggesting this is a priority task to do today or even next year.  As we have documented earlier, simple execution of feed orders and bin switching are low-hanging fruit to tackle now.  But for hardcore analytics like myself, it is easy to see how when a major livestock production system decides to adopt a Big Data IoT initiative the separation from the pack will be swift.  The way in which this industry leader will separate is through the process verification tools, like we have been talking about on the blog - which impacts things on the monitoring and total quality management side; but also from the robust and powerful analytics engine like GroveStreams on the data benchmarking side.  In my humble opinion this will quickly separate them as a low-cost, high-quality producer.

In conclusion, capturing and storing temperature data is very easy.  It is a very important part of hog and poultry operations in terms of animal health, feed intake and growth.  FarmStreams can 

• capture the information down to the second if so desired, 
• store it, 
• send it to your enterprise level system as a raw value or calculated value at customized intervals, 
• chart it, 
• perform simple or complex alerts, 
• engage in some hardcore analytics/benchmarking (in real time).  

In a data hungry industry this is far superior to sitting on a controller in the barn.  In my next blog post I will show how we created another weaker (but cheaper) solution for uploading the data via the FarmStreams mobile app.

Entering Water in Mobile App

Not everyone will have the budget to install water meters but still have a need to pull water information off the farm into the cloud and/or into their enterprise level software system.  In my previous blog I explained the way on farm water meters work and how a user might get a payback (beyond just the peace of mind that your animals are always getting water or that water is not spilling all over your barn).  Many of the same ideas can apply with entering water readings into the mobile app; but with far less real-time alerting or off-hours coverage.

The app has a water screen located here:

Once you select Water you open a screen that looks like this:

Select the barn you are inputting water data for.  

The times are not editable.  What you see is the current timestamp and the previous entry's timestamp.  

Enter the reading as you see it on the meter right in front of you.  The previous reading is shown on the screen for comparison purposes.  The water usage is calculated for you.

Select "Submit Water" and the gallons used for the previous period is submitted.  

Right now the total gallons with the current timestamp is what goes to the store.  This can cause bumps in the data when the data is not entered regularly at the same time.  A enhancement forthcoming soon in GroveStreams will allow us to submit an average gallons per minute rate (or per hour or whatever is easiest) and drop data in the store that is smoothed out for each unit of time between the previous timestamp and the current timestamp.  Although most enter water data daily, barn workers could enter the information at the beginning and end of the shift for more accurate time-series data.

Cons of the mobile app reporting water:

1.  Major leaks or other water-related issues that happen when no one is there will not be known until the barn worker arrives again the next day.

2.  If barn worker forgets to enter the data in a day or several days in a row the data can get bad quickly.  On that note, an administrator can set the maximum hours without a report as an escalated alert event.  For example, if 36 hours passes without the barn worker entering any water data their supervisor would get an email.

3.  Without a sensor you are relying on the barn worker to honestly report the information.  If a major leak occurred or animals go without water they may try and cover it up by falsifying the data entered.

Conclusion:

Most all livestock producers try and collect water records on a daily basis.  This is typically done with paper-based record keeping and maybe phoning or emailing the records in on a weekly basis.  Water meters connected to the cloud is the best way to monitor; but for budget-minded producers a mobile app can also be an effective way to collect water information remotely, analyze it and create alerts off the data.  If the goal is to get water data from a mobile app into your enterprise level system that can be accomplished also.  FarmStreams provides a simple water collection page in the app today with enhancements forthcoming that will make it even better.

Sensor Spotlight: Water Meters

Most of our discussion so far has been on the feed side of livestock agriculture: ordering properly, feed outages, feed budgets and overall execution of the feed side.  There is a lot of money left on the table in feed execution for a typical livestock producer, that is my belief.  Water consumption and air quality are right up there.  Water is something that can be monitored fairly easily and is growing in importance in a world that is more concerned about its water resources than ever before.  Monitoring can provide actionable alerts with payback as well as provide a lot of insight into animal health and should be a key performance indicator.

Basic Water Monitoring.

There are many different types of water meters in the market today.  Every barn I have been in has a water meter with a gallon count.  Most barn workers are required to write down the water reading each day, subtract from the previous day and attempt to keep track of daily water usage on a piece of paper at the barn.  Many water flow meters come with a digital, or pulse output, which can then be connected to a Multisense gateway and sent into the cloud.

Alerts.

Once the water reading is sent to the cloud it passes through the customize-able alert engine to determine if an alert condition has been met with the individual piece of data.  Furthermore, Grovestreams allows alert conditions to be set as data is aggregated, calculated or whatever you might wish to do with the data over time.  For example, a standard alert condition that I have set up for each water monitoring install is at 1201am, I look at the previous day total water / animal vs the previous 5 days average.  If that total is 90% or less or 110% or more it triggers an email alert.

I have not set it up; but a user could also check this total at different points of the day, perhaps at 10am it could be compared to the previous 5 days midnight to 10am total / animal.  Again at 4pm and finally at midnight.

A user may also wish to trigger an alert condition when so much water is flowing it can only be attributed to a leak.  If water consumption spikes > 500% in a given hour compared to the average hour.  Or if the barn manager wants to know if 50 gallons or more is recorded in an hour all of those alerts could be set up around water usage.

A 5 day water / animal average chart is shown below.  One day on Aug 4 there was a drop in water consumption by 9%.  I am not sure the cause.  Otherwise, most days have shown a 3-5% increase from the previous day.

Many users wish to see the data in a column chart versus the previous X days.  That can look something like this.  Or we can put it in a dashboard ready for viewing at any given time.

Payback opportunities.

With water there are quite a few payback opportunities that arise from having solid monitoring and analytics programs around it.  A few worth pointing out.

1.  Manure level and quality.  Water leaks or poor nipple settings that result in large quantities of water in the manure pit or on the poultry side in the litter can have costly consequences.

a.  With swine, having to empty a pit in the spring time is not ideal.  Even if a spring pumping is avoided, extra gallons to pump can add up.  Pumping manure pits is estimated to cost 1-2 cents per gallon.   50,000 additional gallons of water in the pit can cost $500-1000 annually.  Plus crop producers want to pay for nutrients, not water.  Diluted manure can be harmful to maximizing the value of your manure.

b.  With poultry, wet litter causes increased ammonia levels and a worse environment for the bird.  This can be costly to the production costs.

2.   Drops in water consumption can help a producer detect health challenges in a group.  Any tools that help detect health issues sooner can result in big savings for that group of animals.

3.  A serious malfunction like a well pump failure or leak that results in animals not getting water for a period of time can be very costly just like a feed outage can be.  The sooner that type of problem is identified and water is restored to the animals the more productive the animals will be.

Water meters themselves are not expensive.  Connecting them to the cloud/internet can be.  Installing water meters in addition to other sensors, such as bin sensors and feed line sensors is incrementally low-cost.

In conclusion, water meters are a proven well known technology.  Connecting them to the cloud and putting heavy analytics to the data is a logical next step for animal agriculture.  Particularly when water use in agriculture is a growing concern in many parts of the world.  Alerts that allow barn workers to fix leaked or broken water lines quickly can save money.  Trends that suggest health challenges can alert veterinarians to sick animals before it is too late.

Executing the feed budget is tough!

An issue on proper bin switching caught my attention earlier today.  What I am talking about is on farms with tandem bins installed the standard practice is to run Bin 1 empty with Bin 2 slide shut, open Bin 2 slide and close bin 1 slide, deliver feed to Bin 1 with slide shut, run Bin 2 empty, open Bin 1 slide and shut bin 2 slide, order feed for bin 2.  Repeat.  In reality this can be very difficult to do.  Bins can run empty in the middle of the night so barn workers will switch slides before they go home for the night.  They forget to close the empty slide the next morning and pretty soon the bin situation is chaos.

One of the farms I am monitoring I noticed a couple days ago that on one of the tandem bin groups we had the feed emptying out of the same bin two times in a row.  When you look at the chart from the bin sensors you can see it clearly.

A couple things are contributing to the mayhem on this site:

• Many of the feed deliveries have come in feed on feed causing problems.  If you look at Bin 21 in the chart above, you can tell that since the turn began the feed has yet to show below the low sensor (which is in the bin's cone).  We need to get bins 100% empty before putting new deliveries on top of them...especially early in the turn.  The feed on feed deliveries is also contributing to feed flow issues.  
• The feed flow issues are causing ratholing and side-sticking of the feed in the bins making it much more difficult to visually estimate inventory causing poor ordering.
• The primary barn manager has been out nursing a broken wrist (non-work related) so we have had fill-in workers on site.  This makes it harder to keep track of where the bins are at on deliveries and which one should be pulled from.

I believe feed outages are probably the biggest contributor to poor ADG and feed efficiency in animals.  I would be very interested in figuring out how a well executed feed budget versus a poor or average executed feed budget compares performance wise.  Fresh feed and the correct feed at the right time should make a big difference in how well a group of animals performs.  On this particular farm, we have not done a very good job so far on executing the budget but we are determined to turn it around over the next 90 days.  I hope its not too little too late.

Short term solution:

• Order feed remotely (already done)
• Completely empty feed bins before taking new feed deliveries.
• Generate email to barn manager daily (have software create email) on which bin slides should be open and which should be shut.  The algorithm will be rather simple.  Whichever bin had a feed delivery longer ago but is not in an empty state is the slide that should be open.
• Try and determine the performance difference between poor and excellent tandem bin and feed budget execution.

Longer term solution?

Can software and a bin slide open/close device solve?  Probably.  Just take the person on the farm with 100 other things to worry about and manage right out of it.  FarmStreams currently does not send any commands back into the barn; but we could.  Grovestreams is already capable of sending commands back out through its API.  The communications hub can receive commands and act upon them in the barn.  A google search shows that ChoreTime has developed this device already which hooks into their controller.  I have never been on a pig or poultry farm that had it installed so I can't say how it work but I will be looking into it.  Perhaps this or something similar could be something worth looking at including in the FarmStreams suite of solutions.

Sensor Spotlight: Feedmeter

Many years ago, Feedlogic began development of a feed line sensor with the purpose of estimating feed flowing through a standard auger line.  The way that it works is that a box with a sensor, small computer and a wireless RF communication chip is strapped to any size auger line between the feed line coming into the building and the first drop.  This is a very durable device with Feedmeters that have been in barns for over 3 years still working as they did when they were installed.  An image of the Feedmeter is shown below.

The Feedmeter can

1.  Fairly accurately estimate the duration of feed events, especially events that are longer than 10 seconds.
2.  Estimate feed weight delivered by event.
3.  Alert a user to a feed related problem.

How does FarmStreams use and improve the information from this sensor in our system?

1.  Upon arrival we pass each data point through our robust alerting engine.  Alert conditions are customizable by user; but the key ones are going to be around feed outages and then some of the analytics comparing actual feed usage to budget.

2.  Based upon other streams - mostly software streams (budget information) and BinLogic bin level sensor streams we can calibrate the weight data in the cloud, improving accuracy.

3.  The estimated feed intake predicts when feed orders will need to be placed.  Software can place the orders or a person reviewing the data can place the orders.

4.  Create actual intake curves and compare them to a customizable budget threshold curve.

5.  Estimate average weight of the animal and compare to a budgeted weight.  A future sensor spotlight will focus on scales that can be integrated with FarmStreams.

In conclusion, the Feedmeter can provide at its very basic level a very good alerting tool for feed problems.  Most experts believe that feed outages are the biggest contributor to poor ADG and feed conversion.  The reasons for the feed outages vary but the first step is knowing about them, recording them and fixing them as they happen.  Not all sites have these problems; but sites that do should strongly consider installing Feedmeters to get a handle on the situation.  In addition to the immediate alerting functionality, the Feedmeter provides the analytic guys a whole wealth of information in real time that can be used to improve processes within the system.  I would strongly recommend installing both the Feedmeter and BinLogic simultaneously for maximum accuracy in the data.  As Peter Drucker often said - "If you can't measure it, you can't improve it."

Ordering feed can get complex...fast!

One of the hog sites we are monitoring with the FarmStreams sensors and mobile app program is a 3 barn, 6 group site.  Each of the three barns is split in the middle to make 6 groups.  Each of the 6 groups has a set of tandem bins delivering feed to two feed lines in each group.  Total of 12 feed bins, total of 12 feed lines.

The barn worker at the site has been ordering the feed aggressively as we documented in a previous post.  The typical routine has been to order a half load (12T) for 1 of the bins on each side of a building.  Example, Bldg 1, East Side (12T) + Bldg 1, West Side (12T).  Bldg 2 orders always get grouped, and bldg 3 orders always get grouped.

We are only 6 weeks (42 days) into the turn and the feed inventory is already flipped upside down.  It can happen very fast.  Bldg 3 east side has 1 bin empty, and Bldg 3 west side is essentially full with both bins.  The software is accurate as was attested by the barn manager in a physical inventory taken earlier today.

How does this happen?

1.  Placements of animals is uneven in the groups.
2.  A group hit harder with mortalities or health challenges can also throw intake projections.
3.  Average weight of the animals starting is not the same between the groups.
4.  Poor feed ordering.  Feed deliveries when their is not room in the intended bin results in the truck dropping feed in other bins.
5.  Pigs, chickens and turkeys can eat differently.  Perhaps the environment is different or the feed quality differs between groups.
6.  Feed delivery trucks can make mistakes and feed that everyone thinks ends up in a bin can end up in a different bin all together or back at the feed mill.

So, why is this a problem or who cares?

1.  When the bins get out of whack it is more difficult to keep track of.  No one ever wants to run out of feed so there is a tendency to over-order feed.
2.  It is also very easy for the bin getting depleted much faster to run out of feed too quickly leaving the site with a set of bins out of feed for a period of time.
3.  Delivering feed on feed can cause flow issues.
4.  Delivering feed on feed can cause the nutrition plan to not be executed as well as planned.
5.  For companies keen on analytics, having this information available in a robust database to

What is the solution?

Use the BinLogic sensors and feed line sensors to keep track of inventory.  Integrate with feed mill software to ensure accurate and timely deliveries.

Feed inventory in mobile app

We discussed using various sensors to estimate on farm feed inventory in the last blog entry.  Several different sensors are on the market, we prefer the BinLogic sensor for its simplicity and cost.  Whatever sensor is used, getting the data into Grovestreams for analytics/alerting is crucial.  However, in some cases some farms may still want the person at the farm to climb bins each week or several times a week and take a feed inventory and report it more efficiently to the feed mill or management.  Even if bin sensors are installed it may be valuable to also get physical inventory entries on the site pushed into the data store.

Here is how it works.  The Farmsteams mobile app allows for a user to submit feed inventory based on a visual assessment of the feed bins on the farm.  The app has been programmed to allow setting up to 10 feed bins for each group housing unit.  Any more would be a custom programming request.  A user on the farm must submit feed inventory for all feed bins in the group during a single submission.  This ensures the most accurate data in the data store.  A screenshot of the app's interfact is shown:

The FarmStreams mobile app (which can be downloaded here:) 

       Android http://tinyurl.com/mqgskr3

       Apple http://tinyurl.com/pclejlh

Pros/Cons.

Pros:

1.  Writing down and calling/faxing in later has flaws.  Timing may be off, people lose their scribblings, the person answering the phone might not be available, etc.  Entering in the mobile app as it happens reduces lost communications and bad data.

2.  Grovestreams alerting engine allows us to set alerts around timing of entries.  If the rule is once a week to take an inventory, management can get an alert to their email on the 8th day without an entry.  Easier to manage than reviewing a spreadsheet or scribblings in an office.

3.  Phone calls and faxes are difficult to track for employee performance.  The database is set up to keep track of all the times it was entered but also can be set up to track all the times it was supposed to be entered but wasn't.  This can be included in the report at the end of the turn.

4.  Software streams can produce graphical feed curves, compare to a budget range, estimate feed orders, etc.  An example graphical feed curve chart that looks like below can be produced and delivered via email daily/weekly/etc. or accessed in real time by logging into the platform and viewing your dashboards.

Cons:

1.  People don't always climb the bins, perhaps they throw a rock or just guess.  Climbing the bins in January isn't fun.

2.  Feed, particularly mash feed is very difficult to estimate visually.  It doesn't always flow evenly and peering down the top of a bin can be visually tricky.  The bin sensors like BinLogic and ultrasonic have similar issues; but at least their error factors should be relatively consistent.  Meaning that a 1/2 ton discrepancy from actual is consistent every reading which minimizes the noise in the reading data and can be dealt with in software.

3.  May be difficult to enforce at some sites (likely your worst performers).  If they are not climbing bins now giving them an app and asking them to do it isn't going to change that.  Perhaps with the Farmstreams mobile app and the powerful back-end data system, management could implement incentive programs with measurable performance standards.

Strategy: Installing bin sensors is the best; but if that is not possible entering the inventory in a mobile app is better than the way it is normally done of calling or faxing data in.  If a user is not ready to go all the way with fully integrated sensor systems the mobile app can be a great way to get started collecting important information from the farm site and using cloud-based Big Data technology to find value in your livestock operation.

Sensor spotlight: BinLogic feed bin level sensor

BinLogic is a bin level sensor that is valuable for managing feed inventory in the FarmStreams feed inventory system.  There are many other types of bin level sensors such as ultrasonic, load cells among others.  Our focus is really about getting data streams out of the farm site into the cloud analytics engine and/or an enterprise level system.  So, we could integrate with other bin level sensors and are willing to do that if a farm feels very strongly or has already invested heavily in another type of sensor.  With that said, our preference right now is the BinLogic sensor because it is low cost and effective to help tell a user remotely when a bin needs a feed order.

The BinLogic sensor is very simple device.  It provides a boolean (1 or 0) output.  1 is when feed is covering it and 0 is when feed is not covering it.  My preference is to put two sensors in each feed bin.  One at the top and one at the cone.  A user could put several in each bin for improved accuracy or a single sensor at a single level.  A picture of two sensors from the top of the bin is shown here.

We encountered some initial issues with ratholing when the center sensor was down the center of the bin causing false empty readings.  We moved the lower sensor off from the sensor and that has largely eliminated false empty readings.  A 2nd lower sensor on the opposite side would eliminate false empty readings.  Software also is very important in interpreting the data.  A single top sensor combined with other information like population, age and especially a feed line sensor to estimate feed flow is also a very tidy effective solution to order feed accurately.

Primary benefits to BinLogic Sensor:

1. Send alerts to specific users when thresholds are met, triggering feed orders.
2. Provide an off-site party ability to easily see inventory levels on the farm.  People such as feed mill personnel, managers, or the barn worker.
3. Provide managers / nutritionists verification their nutrition plan is being executed.  Tandem bins are being switched, in order and on time.
4. When combined with other sensors and software streams effectively keep track of inventory and feed consumption.
5. Calibrate other sensors such as the feed line sensor in the cloud based on actual feed disappearance over time, improving the sensor's future accuracy.
6. Feed inventory values at specific timestamps to create very accurate feed curve charts.  Track vs budget.
7. Timestamp bin fill events to ensure feed is delivered to the correct bin at the right time.
8. Ensure there is room for feed in the bins when the truck leaves the mill.
9. Reduce need to climb bins improving safety and employee morale.

The BinLogic sensor is a valuable tool to manage feed ordering and ALSO provide very valuable feed intake data for your analytics.  All stored and accessible indefinitely in real time.  Provide alerts around the analytics or deliver meaningful, customizable reports to management on a schedule or on the fly to correspond with site visits.  Benchmark your performance internally or externally if you choose.  The sky is the limit...and it all starts with a simple bin sensor.  

Recording mortalities

The original purpose of why we created the FarmStreams mobile app was because we needed an easy way to upload daily mortalities into Grovestreams.  This was important because animal inventory was needed to run certain ratios, such as water/pig or feed/chicken.  Recording mortalities and keeping track of inventory is a task that is done today on paper or a local computer at nearly every farm.  Typically the barn worker calls or faxes in the results once a week to an office somewhere that enters them into a spreadsheet or database.

As the mobile app evolved we began adding more features that people wanted.  But its origin truly was only to give us a way to easily enter mortalities into Grovestreams for record keeping and as required for the analytics engines.

Here is how the mortalities work:

Step 1, open the app.

Step 2, click the "Mortalities" icon

Step 3, select the group the mortality occurred in, the date and enter the found / euth.

The app then uploads the data via API transfer into Grovestreams where we can do all sorts of things with it.  Some examples:

• Keep a running count of mortalities and inventory.  
• Push the data into an enterprise level system through the API Transfer
• Chart mortality %

• Chart mortality % versus budget

• Although I have not accomplished this yet it would be fairly easy to benchmark and chart each site's mortality against the entire system's mortality based upon the day/week in the turn.

• Accurately update various ratios such as: water gallons / pig with current inventory each day.

• Generate alerts from the mortality data.  For example, 
• alert when mortality goes above the budget and stays there for more than 7 consecutive days.
• alert when no mortalities are entered for 7 or more days.
• alert when more than 1% of the population is reported in a 3 day period.

I am outlining a lot of complicated reports and alerts that can be generated just to show readers the robustness of the system we have developed.  In reality, most producers just want a way to submit mortalities from their cell phone or tablet and easily view the running total online or get the data into their own enterprise level system easier or faster.  We are working on getting the running total back into the mobile app; but don't have that piece developed yet.  Stay tuned.

Mobile app for livestock producers

In my first blog post I discussed sensors and software and their usefulness in managing a confined livestock farm.  As great as sensors are that just sit and collect useful data on the site storing in the cloud with near real-time analytics (and that is very great) managers still need to collect data from the people working on the site each and every day.  Most systems require certain information to be recorded each day on the farm site.  Common pieces of data would include:

• mortalities for each group by type
• water readings from the meter 
• temperature from the controller (high and low) 
• feed deliveries
• medications
• feed estimates in the bin (this is perhaps a weekly or bi-weekly task)

Recording the information at the site is fraught with problems.  The information stays local.  Other people in the chain like vets, nutritionists, feed mill operators and animal owners experience delays in receiving the information.  What if the barn worker does chores from 6am-11am each day and the vet needs information at noon.  They have to wait until the barn worker can return to the site before they can get the information.  The exact timing of the information can be lost.  Employees can also sanitize the information before sending it out so that it doesn't make them look so bad.

Collecting data still comes down to how motivated the worker on the farm is, no different than when it is required to take the daily paper logs.  Collecting the data to the cloud will allow other people in the system to access information quickly and in a uniform manner.  Poor data collectors will quickly be identified with the alerting functions already available in Grovestreams.  The poor performers can be given measures to improve quickly.  

We created an app called FarmStreams to allow this information to be collected into the cloud for further processing or direct transfer into a larger enterprise level system.  The app can be found here for downloading: 

       Android http://tinyurl.com/mqgskr3

       Apple http://tinyurl.com/pclejlh 

Future blog posts will go into detail on each of the pieces of data that we collect, how we do it, why we do it and what it is used for.  Certain data is very cut and dry like mortalities.  Other data like medications is more abstract because they can be administered in so many different ways and in so many different units of measure.  We will work through how this is programmed in the app as we engage more producers with different needs.  Other data like temperature, water, and feed estimating are best collected with a sensor; but cost and connecting to cloud can be a constraint.

Dealing with Feed Interruptions

I woke up this morning and saw an email alert from a hog site that we had a feed line hung up and was running continuously in an empty state since about 2am the night before.  This is an indication of a feed interruption.  Grovestreams alerting engine has several different ways it processes alerts, which is nice because over-alerting is a big issue in remote monitoring applications.  Grovestreams has thought of this and has functions such as dwell, latency, time-filters and other factors to eliminate unwanted alerts and only gives you the ones you need.

The way our system is set up is that the gateway sends a message to Grovestreams every 15 minutes that this running empty feed line state is being met.  Upon arrival in the database, I get an email notifying me of the situation, reminding me every 15 minutes until the situation is eradicated.

This past weekend we had several issues like this.  2 on Saturday and 1 on Sunday.  It seems to happen most frequently when the tandem bins are being switched.  Indications of poor feed flow.  The barn manager believes the feed is ground to fine and doesn't flow properly.  Yours truly climbed each of the 12 feed bins on site and saw for myself feed clumping to the side of the bins and not flowing.  I even took a short youtube video to document what I was seeing in one of the bins.  In the video you can also see my high and low sensors.

One item worth noting on this issue at this site that gives me some pause before suggesting a coarser grind or a bin-beater is the solution to our feed flow issues:

1.  The bins are not being completely emptied between deliveries.  These pigs are still relatively small, 90 lbs.  Going through perhaps 2-3 tons per day per group.  The barn manager who is scared to run out of feed is ordering feed very aggressively and this has resulted in delivering feed on top of feed often.  See the chart below which represents a high and a low sensor in the bin.  100 means the bin is completely full of feed.  50 means its somewhere in the middle and 0 means feed is below the cone.  If you look at my video you can see about how it is laid out.

This shows me that since the turn started in July, one of the feed bins in the tandem set has not emptied below the cone before receiving the next order of feed.  This is potentially disrupting the nutrition plan and possibly contributing to feed flow issues.  Old feed clumps easier and doesn't flow.  Additionally, logistics issues arise when the feed in the truck doesn't fit in the bin it belongs to.  All breakdowns in the feed process which add cost.

Solutions:  

1.  Monitor feed flow issues, document what the cause is potentially include a feed flow documentation module to the mobile app.  

2.  Fix feed ordering.  Ordering feed is not an easy task for a barn manager.  There are many factors to consider and timing it correctly to fit the feed mill's schedule is not a simple task.  Going forward, having the software manage the feed ordering will improve that process a great deal.  By fully emptying the bins we can know for certain that older feed is not clogging the bin from flowing properly.  Grovestreams uses data from all types of streams - sensors, app, and software which gives us a remarkably accurate estimate of feed on site and predictions on when it may go empty.

What is a farm stream?

FarmStreams collects, stores, organizes, analyzes and displays data from livestock farms.  In our world a stream is a piece of data collected from a sensor, person or piece of software.  The data is stored in the cloud using the latest Big Data structures.  Grovestreams is a very powerful platform and analytics engine that we use to mashup the data in the cloud.  

Examples of remote sensors:

• bin level sensors to estimate feed bin level
• feed line sensors to estimate feed flowing through a flex auger
• water meters to estimate water consumption
• air temperature sensor
• humidity sensor
• CO2 sensor
• weight scales to get bird or hog weights
• door open/shut to estimate how many hours a day laborers are at the site

Example of data that can be collected from people:

• FarmStreams mobile app collects data visible by people on the site, like: 
• mortalities
• medications
• health conditions
• water (if sensors are not available)
• temperature (if sensors are not available)  
• feed bin estimates (if sensors are not available)
• customizable inputs for each operation

The app can be found here for downloading: 

       Android http://tinyurl.com/mqgskr3

       Apple http://tinyurl.com/pclejlh

Lastly, we can create streams from software, such as:

• day, week, etc. of turn
• Animal inventory and subsequently feed/animal, water/animal, cost/animal
• Estimated budget intake based on feed curve
• Weight estimates, market prediction dates
• Feed inventory and feed ordering predictions

The data platform can analyze each and every one of these streams for alert conditions for immediate action.  Totally customize-able to each producer's needs with time filters (only alert certain times of day).  Different people in an organization can receive different alerts.  Examples could include:

• feed outage, alerts site manager that the animals are not getting feed
• water / animal decreases or increases compared to previous 2, 3, 4, or 5 days.
• feed bin is empty
• temperature is too hot / cold
• many many more 

View the streams in real time, summarized by second, minute, hour, day, week, month, or turn.  Charts and graphs quickly show the data needed.  Customize-able charts and graphs provides users maximum value with their data streams.  Users with an intermediate or higher understanding of Excel of Sequel query can develop powerful real time tools in our platform.

Follow our blog!  In the coming posts we will dive into different sensors and observations from the world of animal agriculture.  We will have technical and industry experts contribute.  The internet of things has impacted many industries.  In the world of agriculture, crop agriculture more so than animal ag.  the livestock side of ag has big opportunities for improved processes by combining cloud connected sensors, data from people and software.