Fiber, News, Processing, Technology

For optimum stalk processing, let AI run the system

Share this:
  •  
  •  
  •  
  •  
  •  

By Robert Ziner, CIHC, Canada

As hemp stalk processing gains in popularity, it faces a fundamental economic challenge at a fundamental level: Decortication.

Until now, decortication has been handled as a one-step, linear throughput process where stalk goes in one end, and the three outputs – bast, hurd and dust – come out the other. Each output is separated then appropriately packaged either in bags or in bulk loads, ready to be sold to commodity buyers who generally operate relatively close-by – in order to control transportation costs. The existing process relies on continuous human co-ordination and interaction, and generates very low margins.

When productivity cannot be tightly controlled throughout the production environment it inevitably gets impacted by “unexpected” production problems, primarily caused by jammed pipes and infeed conveyors throughout the production line. Most of the problems are caused by fiber jamming in the decortication unit itself. Anecdotally, some large volume decortication plants report losing up to 20% of their daily production due to unforeseen problems, including long delays in getting parts and having them replaced.

Predictive maintenance

Wouldn’t it be great if there was a way to predict when maintenance and service will be needed? A way to ensure that all preventative service was completed – and preferably shortly before a part would fail – causing the whole line to shut down.

The reality is that this opportunity has existed throughout the mega-processing and manufacturing world for over 25 years. It is proven; and proven to be invaluable.

Automated Predictive Maintenance has essentially transformed the art of production maintenance by leaving it to an AI system to monitor, control and synchronize data, communication technologies and digitally-driven production equipment.

Value of AI

AI’s functions help reduce downtime and boost productivity. They include:

  • Predicting how long it will take before a part breaks or “fails.”
  • Predicting when to change parts.
  • Relying on the system to remind you what needs to be done – step by step.
  • Monitoring cost, duration and all downtime associated with each part or operation failure in order to identify patterns and trends.
  • Scheduling the most appropriate time to perform each service – based on sales commitments and inventory on-hand availability.
  • Utilizing sensors to monitor the condition of operating equipment and analyze the data on an ongoing basis enabling organizations to plan and service equipment when needed, instead of relying on scheduled service times.

Potential in hemp

Most large-scale manufacturing facilities are Investing in AI-based predictive maintenance solutions. It is a sure-fire way to improve operating efficiency and has an immediate impact on productivity – and the bottom line. In fact, machines today can evaluate their own conditions, order their own replacement parts and schedule a field technician when needed.

McKinsey reported in a November 2019 article that AI-enhanced predictive maintenance of industrial equipment has demonstrated a 10% reduction in annual maintenance costs, as well as a 20% reduction in downtime and a 25% reduction in quality inspection costs.

Automated Predictive Maintenance could obviously be a big plus for hemp stalk processing. However, there is a limitation: it will only be possible in an AI-driven production environment.

Inevitably, advanced manufacturing technologies driven by AI and machine learning will appear in the hemp stalk processing industry – just as they have in every advanced manufacturing industry built in the last 20 years.


Robert Ziner is Founder & CEO, Canadian Industrial Hemp Corp. (CIHC), Toronto, which is developing an advanced hemp stalk processing and optimization system. Ziner has more than 30 years in the building materials distribution and secondary wood processing industries.

Get Hemp Industry Updates

* indicates required