“Artificial intelligence and machine learning methodologies have already proven game-changers in many industries. These methodologies could be applied to the grain industry as well and transform the way that managers interact with the silo, receive information on grain condition, and make their decisions. They could improve the accuracy and efficiency of the existing grain management tools but also create new innovative ones.”
Dr. Efstathios Kaloudis
Effective storage management necessitates maintaining high crop quality and preventing loss. To achieve these goals, one must thoroughly evaluate factors such as grain temperature, moisture content, carbon dioxide concentrations, weather conditions as well as their interaction and impact on grain in a storage structure.
In most cases, silo and storage managers rely on their experience and intuition to manage grain, which occasionally leads to either costly human errors or overly conservative planning at an unnecessarily high cost.
New technologies, both in hardware and software, could transform the passive storage structure to an intelligence one, assisting the storage manager to detect and resolve efficiently challenging issues.
Artificial intelligence and machine learning methodologies have already proven game-changers in many industries. These methodologies could be applied to the grain industry as well and transform the way that managers interact with the silo, receive information on grain condition, and make their decisions. They could improve the accuracy and efficiency of the existing grain management tools but also create new innovative ones. The transition to an "intelligent" silo should include the following:
Example of wireless sensor installation inside a silo feeding the predictive algorithm with real-time data
• a grain level monitoring system that measures the grain mass and tracks grain movements automatically.
• a grain status monitoring system able to sense and detect accurately temperature, relative humidity, CO2 and O2 gases and provide early spoilage detection, grain quality forecast, and propose the best aeration strategies.
• a fumigation system that proposes the right dosage, monitor the fumigant concentration automatically and predicts if the treatment would be successful and when.
In the following sections, a description of these readily available technologies is presented.
Until recently, the typical method of monitoring a mass of stored grain utilized thermocouples attached to high-strength steel cables. These cables are subject to vertical frictional loading during filling, storing, and emptying operations. Since these cables are supported by the bin roof, the loads imposed on them have caused failures as reported by an Iowa State University study . Besides the mechanical stability risks, cable monitoring systems require advanced expertise for their installation and have high installation costs. Copper cables often get corroded by phosphine and stop working.
Recent advancements in grain storage technology enabled the development of wireless sensors (Figure 1), that measure accurately all the parameters that affect grain quality (e.g. temperature, relative humidity, O2, and CO2). They also transmit safely their data to the cloud in real-time, offering worldwide accessibility. Their advantages also include the ability to move with the grain, in case of grain transit, offering enhanced traceability.
Carbon Dioxide (CO2) monitoring
Monitoring gases in a silo offer significant advantages to early spoilage detection, since gases, like CO2, are associated with grain respiration, mold, and insect development. Furthermore, sensors can detect variations in gas concentrations 100 times faster than variations in temperature and relative humidity.
Nonetheless, the evaluation of the CO2 readings is not straightforward, and the assist of Machine Learning algorithms is needed to extract useful information. Figure 2 presents the CO2 concentrations of two silo bins as recorded by Centaur Analytics, Inc. sensors. Even though Bin B has higher CO2 values than Bin A (particularly for the first 16 days), the algorithm doesn't issue an alert since the grain condition is good. On the contrary, the silo manager of Bin A would receive an alert as early as the 17th day since a sudden rise of the CO2 is detected.
Figure 2: CO2 concentrations of two silo bins as recorded by Centaur Analytics, Inc. sensors. The silo manager of Bin A would receive an alert as early as from the 17th day since a sudden rise of the CO2 is detected.
Grain level monitoring
A fill level sensor is a contactless radar system able to measure distance accurately (+- 5 mm). Specifically, it measures the round-trip time of flight of microwaves to the grain surface. It can be easily installed on the top of a silo and it provides continuous monitoring. To translate the measured distance to the total grain mass stored in the silo one should consider the following variables:
• Grain bulk density (varies with moisture content)
• Pack factor (varies with grain height)
• Grain surface shape (varies with the operation: loading, emptying)
A user interface of a web platform that presents the grain conditions at every location in a silo, including forecast values (left). On the right the safe storage time (in days) is displayed based on several quality metrics, such as the dry matter loss, the appearance of visible molds, and the germination capacity.
An AI algorithm can correlate bulk density with the moisture content measurements, pack factor equations that compensate height, and determine the grains surface shape without user input. Thus, volume and grain mass could be calculated with improved accuracy. And without dust interruption!
Predicting Grain Condition
During grain storage, many factors influence the status of the grain which are difficult to know beforehand. For this reason, a new computational tool has been developed to perform simulations on stored grain. The software is based on the Computational Fluid Dynamics approach and Machine Learning methods and can evaluate sensor data, heat transfer effects that originate from the heat produced from grain respiration, the heating of the silo due to solar radiation as well as the cooling due to wind effects. Additionally, it implements equations predicting the variations of grain moisture content and the two gases involved in grain respiration (combustion of a carbohydrate), O2 and CO2. Furthermore, models evaluating grain deterioration are integrated into the software . These models offer predictions of safe storage time based on losses in grain dry matter, the possibility of mold appearance and the reduction of germination capacity (Figure 3).
To maintain grain quality, proper aeration is essential for most of the grain storage systems. During a typical aeration process, a fan forces cool air into the bottom of the silo. As the air moves upwards, it cools the grain and eventually exits from the aeration ducts at the top.
An intelligent system considers a variety of factors to propose a tailored aeration plan (Figure 5) and make a prediction of the outcome of aeration including inlet air temperature, silo geometry, aeration fan characteristics, current and predicted climate conditions, and grain condition constantly measured by wireless sensors.
User interface of a web platform predicting the outcome of a phosphine fumigation.
Utilizing the capabilities of advanced numerical modeling in fumigation applications and constant feedback from wireless sensors the phosphine concentration could then be determined for every location inside the storage volume and at any given time based on all the factors that occasionally affect the toxicity of the fumigant and prevent treatments to be successful . Centaur Analytics, Inc. have developed a unique system.
• The degassing rate of metal phosphides like Mg3P2, AlP which depends on temperature and relative humidity values.
• Weather conditions to evaluate accurately the storage interaction with its surroundings in terms of heat transfer, gas losses, and movement
• Absorption of Phosphine gas by grain at differing rates depending on the commodity which can reduce the concentrations of fumigation doses to sublethal levels before grain has been disinfected.
• Insect species and resistance
• Even pot-fumigation-aeration and people safety can be measured in real time.
A user interface of web platform that integrates all the relevant information concerning grain stored in a silo, e.g. current temperature and moisture content, a proposed aeration plan, user notifications and alerts.
As the cognitive software correlates phosphine exposure with insect mortality, a methodology for planning precision fumigations can be established (Figure 4). Users receive notifications for the fumigation duration and when success could be enabled.
As much as important is to create the information described above, it's also necessary to present it to the grain managers in a meaningful and user-friendly way. A web platform is an ideal way since data could be updated in real-time, 24/7. Furthermore, users could access the information from their pc or smart devices at any place.
Figure 5 shows an example of a web platform user-interface presenting an overview of the storage facility. It includes a storage facility representation showing the type of stored commodity, it's mass (which may be a user input or calculated automatically by an installed fill level sensor) and the installed locations of sensors. The user interface also provides information on the current temperature and moisture content of the commodity, a proposed aeration plan, a geographical map of the storage facility location, the current weather conditions at the same location and notifications about the facility’s operation.
Adopting as many as possible of the new technologies described in this article could transform a typical silo into an intelligent one. The benefits for the users could be unprecedented:
• prevention of mold development and mycotoxin contamination
• improved grain quality - higher grain market value
• lower operating costs and reduced aeration fan running times
• supply chain transparency – defensibility against quality claims from retailers
• cost reduction of the overall pest control application, avoiding excessive chemicals (e.g. overdosing, need to repeat failed fumigations) as well as excess labor
 Schwab et. al, "Vertical Loading of Temperature Cables" (1991). Agricultural and Biosystems Engineering Publications. 110. https://lib.dr.iastate.edu/abe_eng_pubs/110
 Kaloudis, E.; Bantas, S.; Athanassiou, G.; Agrafioti, P.; Sotiroudas, V. Modeling distribution of phosphine in cylindrical grain silos with CFD methods for precision fumigation. Proceedings of the 12th International Working Conference on Stored Product Protection (IWCSPP) in Berlin, Germany, October 7-11, 2018. https://ojs.openagrar.de/index.php/JKA/article/view/10771