A digital instructor can partially solve the instructor shortage in forestry. But to be able to develop one, accurate and reliable automatic detection of work steps and methods is required.
Digital Instructor-In forest engineering research, the concept of driver effect is often used. The problem is that there is no consensus on the difference between driver power and working methods. These concepts are often confused, but driver power should only include psychological, cognitive, and motor abilities. These abilities are linked to the individual’s innate talents and therefore cannot be quickly improved through education or practice. Working methods, on the other hand, include both roughly described work steps and technical driving issues concerning the operator’s way of manoeuvring the machine at a detailed level with control, as well as systematic procedures and strategies for performing a given task. The driver’s working methods are easier to change than their innate abilities. Therefore, driver power and working method are not interchangeable terms.
During field studies, the most productive forest machine operators have been found to be 40-100 percent more productive than their less productive colleagues. Productivity improves most rapidly at the beginning of one’s career and in the younger years. According to Nordic labour science literature, forest machine operators reach their highest productivity in their 40s, when they have about 20 per cent higher productivity than operators in their 20s. Drivers tend to maintain their achieved productivity relatively well over time. According to Central European research, a new harvester operator usually doubles his productivity in the first year, but after that the rate of improvement slows down significantly. Even experienced drivers can show a learning curve when faced with changes in the content of their work.
It is not entirely clear whether the performance differences between drivers are mainly due to driver power, different working methods, or both equally important.
-
That’s a remarkable amount of work hours for a single machine, the Norcar 600 owned by Erkki Rinne is taken well care of, it even has the original Diesel engine.
-
Kieran Anders is a forestry contractor working in the lake district. His work involves hand cutting and extracting timber using a skidder and tractor-trailer forwarder.
-
It is not possible to eliminate chain shot, but there are simple steps that can be taken to reduce the risk.
-
Arwel takes great pride in the fact that the mill has no waste whatsoever, “the peelings are used for children’s playgrounds, gardens and for farm animals in barns in the winter and the sawdust has multiple uses in gardens and farms as well.
-
Timber hauliers need to encourage young blood in, and also look after the hauliers we have, we need make the sector a safe and positive place to work.
FIND US ON
Related Posts
However, there are some studies that have tried to distinguish working method from driver effect. In these studies, it has been estimated that the performance of felling work can differ by up to 20 per cent between different working methods. Corresponding studies for forwarders are lacking, but standardized experiments show that correct working methods are a crucial factor for performance in forwarding. This is not surprising considering that forwarding requires strategic thinking above all.
Digital Instructor
According to existing research and proven experience, a method instructor can significantly rationalize forest machine operators’ working methods and thereby increase productivity. However, method instructors are a scarce resource and can only visit the drivers infrequently and for a short time. A digital instructor can be at least a partial solution to the instructor shortage.
Just like a traditional method instructor, a digital instructor should give machine operators real-time feedback on the work and guide in how the work can be made more efficient. Unlike a traditional method instructor, the support is computer-based. While a digital instructor is unlikely to replace human instructors in the foreseeable future, a digital instructor can be a great addition.
Various digital instructors have been studied in the transport sector for several decades. Studies show that a digital instructor can not only rationalize driving patterns and working methods, but also reduce the risk of drivers reverting to old inefficient work habits, presumably thanks to constant reminders and feedback.
However, implementing a digital instructor in forestry requires detailed and reliable automatic detection of work steps and methods. Therefore, the first step in the development of a digital instructor is to develop a system for automatic data collection and work step detection.
Development of systems for automatic data collection and work step detection
Regular CAN bus data, such as the operator’s control signals, in itself provides useful information about the forest machine’s movements. However, work element detection requires extensive data processing and/or additional data sources, especially for forwarders.
Proficiency for standardized work step detection for harvesters is available
The harvester’s computer system timestamps and registers the machine’s geographical position at each felling cut. This information is automatically saved in an HPR file. Although HPR data is valuable in itself, harvesting work cannot be divided into work steps based on a single timestamp per strain. However, regular HPR data can be supplemented with extended timestamps and information from mom files. Then, in addition to felling, events such as log cutting and tilt-up can also be timestamped. The harvester’s work steps can be automatically detected with reasonably high precision based on these data points. Extended timestamps have been tested in field studies, but are not yet formally part of StanForD.
Slow development of work torque detection for forwarders
Although the forwarder’s CAN bus provides data on driving and crane usage, this data does not provide any timestamped data points that can directly determine whether the forwarder is loading or unloading. It is even impossible to divide a forwarder dataset into separate loads. According to StanForD, it is possible to manually report the load’s start and end times as well as estimated volumes using FPR files, but this means that the data collection is not automatic.
However, there are some old solutions to automatically detect the forwarder’s work steps. Identifying loading and unloading cycles then enables identification of other work steps. You can try to do this based solely on control signals: close/open the grapple and turn the crane towards or away from the cargo area. Another option is to equip the forwarder crane with a scale and the crane column with an angle sensor. The basic principle is the same in both cases: first it detects whether the grapple is holding timber, and then whether the grapple is moving towards or away from the cargo hold.
Geofencing (geofencing), combined with speed enforcement, is another old solution. If the speed is below a certain limit value, either the loading or unloading operation is recorded depending on the position of the machine, inside or outside the yard (its corner coordinates). If the speed exceeds the limit value, either the journey with a load or without a load is recorded, depending on whether the previous operation was loading or unloading. However, research has shown that the forwarder’s work steps cannot be reliably detected based on geofencing and speed alone, but additional data sources are needed. Therefore, in recent years, researchers have begun to shift their focus from older to more modern solutions.
Artificial intelligence (AI) and machine vision open up new opportunities
AI, combined with fully sensorized cranes, offers new possibilities for detecting work tasks, especially for forwarders. During loading, logs are picked up at about the same height (ground level), and there is often extensive driving between crane cycles. When unloading, it is reversed. Machine learning could use these systematic differences to first detect loading and unloading, and then other work steps.
There is a lot of research on machine vision and machine learning, and scientific studies on forest technology applications are being published at an increasing rate. Integration of machine vision and machine learning has the potential to identify assortments and improve the quality of automatically collected forwarder datasets. For harvesters, full assortment information is already available at log level through the bucking system.
The ability of machine vision to gradually replace sensors has received too little attention. Machine vision should be relatively easy to detect whether the grapple is holding timber and its direction of movement. Together with a remote-measuring sensor that suffers, AI could also measure the crane’s reach when the forwarder grapple grabs a bundle of timber or when the harvester head grabs a log, thus identifying work zones. Mobile applications based on this technology are already being used for forest inventory, and forest machine manufacturer Ponsse is developing a solution for harvesters with lidar technology to locate trees around the machine.
Despite the potential of modern technology, it is unlikely that all working time can be correctly divided into different work stages, or that all ranges are correctly identified. However, it is not necessary in this context either. The important thing is to know when you don’t know. Automatic data collection generates large amounts of data, allowing selectivity without significantly impacting representativeness.
Feedback must be given in a relevant way
Working conditions should be taken into account when giving feedback on work. Working conditions affect performance and have an impact on which working methods and strategies are appropriate. The harvester’s bucking system generates large amounts of data related to working conditions, but the machine manufacturers’ existing performance monitoring systems do not take this into account. A simple factor like stem volume would significantly improve the usefulness of feedback. Other strain-wise easily accessible performance-affecting variables were: number of logs, assortment, compulsion (yes/no) and so on. For forwarders, however, it is more difficult. It is well known among forwarder operators and in the scientific literature that the performance of the forwarder is mostly due to the spatial distribution of the range volume. However, it has not yet been possible to model this basic effect in any practical way. The necessary data would be available after harvesting through HPR files, and in the future, laser scanning with high point density could offer useful data even before harvesting. Despite the limited knowledge of the underlying factors of productivity, a digital instructor remains relevant, as forwarding instructions are usually general, for example on efficient crane tip roads and work zones.
The situation today and the way forward
For harvesters, a technical skill to automate work step detection, and thus also to construct a digital instructor, has already been around for several years. And further development would not require extensive investments in new technical solutions either. However, the development of automatic work torque detection for forwarders has not made a real breakthrough. It’s time to abandon old solutions and consider new approaches, such as AI and machine vision, that can be a game-changer – not only for forwarders, but also for harvesters.
A further dimension in this discussion is the transition of resources to the development of the next generation of remote-controlled and autonomous forest machines. If these machine types gradually replace the existing ones, the question arises as to how long a digital instructor will be relevant. For harvesters, this does not have to be a problem, as a digital harvester instructor could be realized relatively quickly if there is demand. However, the slow development of forwarders raises the question of how long a digital forwarder instructor will be once in place.
However, it is important to distinguish between a digital instructor and the development of automatic data collection and work step detection. The digital instructor may become obsolete if forest machines change in the coming decades, but the technology developed for automatic data collection and work step detection can maintain its relevance for future generations of machines. The ability to detect work steps, volumes and assortment in real time will be a valuable asset, regardless of what future forest machines look like.
This text is a summary of an article published in Silva Fennica. The article itself and the referenced literature can be found here:
Manner J. (2024). Automatic work-element detection: the missing piece in developing intelligent coaching systems for cut-to-length logging machinery. Silva Fennica vol. 58 no. 1 article id 24004.
Forest Machine Magazine is written and edited by a forest professional with over 40 years hands on experience. We are dedicated to keeping you informed with all the latest news, views and reviews from our industry.
#homeoflogging #writtenbyloggersforloggers #loggingallovertheworld
