PETRA creates products which enable mining companies to turn their data into predictions and optimisations using engineered data science. PETRA is a pioneer in the application of machine learning and artificial intelligence in mining and has combined this with extensive mining industry knowledge and experience to bring a range of software solutions to the industry.
PETRA’s solutions are built to be used by miners in an operational context allowing the benefits of machine learning and artificial intelligence to be integrated into day to day workflows and existing processes and systems at work in the mining environment. This creates significantly greater, sustainable value faster. All of PETRA’s software solutions are supported via our helpdesk system and because we are committed to delivering robust, quality products our customers can be confident that any solution we deliver will continue to be updated, supported and delivering value for many years in the future.
Creation and delivery of PETRA’s solutions is managed by mining professionals with data science expertise (PhD qualified engineers and scientists). We are proud of the years’ of site experience and industry knowledge that enables us to engineer original data science solutions. Because we are mining professionals, we won’t take up valuable personnel time trying to understand the data or what it means in a practical mining context.
PETRA maintains platform agnostic interoperability points to allow you to deploy our solutions in your environment and has a strategic partnership with Maptek to enable integrated use of PETRA solutions within common mine planning and geological modeling systems. PETRA is widely recognised as a thought leader in the space of application of machine learning in mining and has delivered solutions to many mining operations, large and small, creating significant value for our customers.
PETRA’s MAXTA digital twin value chain optimisation software uses machine learning models built from historical performance data about the specific combination of mining operations, orebody geology and plant performance. MAXTA uses large quantities of these data representing at least one year of operation at the mine. This allows the machine learning algorithms within our MAXTA software to properly learn all the variation and complex, often non-intuitive, combinations and relationships and the downstream effects that will result specifically for each operation.
1) Geology model – a block model describing the geological resources at the mine. PETRA’s MAXTA solution will consider all parameters contained within the block model (typically may exceed 100 parameters).
2) Load and Haul data – typical fleet management data showing tonnes hauled, source (blast location) and destination and loading and haul cycle times
3) Process performance data – could be crusher or mill throughput data, blast design parameters, product yield or other process results in the plant or any other relevant parameter that is needed to be modelled, such as crusher energy use.
MAXTA enables the combination of large quantities of this historic data to be combined to create a data model representing the performance of the processes being evaluated with reference to all the variation in the orebody. The historic data is virtually ‘remined’ allowing the process performance to be characterised based upon hundreds of millions of tonnes of ore being treated through the actual process and the achieved historical results.
This is done using advanced proprietary techniques in ore tracking through load and haul, stockpile and material handling processes. These techniques have been developed over many years of research and development by PETRA and are core to enabling the creation of a true digital twin of the mines performance.
MAXTA employs machine learning techniques to create predictive algorithms describing the performance of the unmined rock in the orebody and creates valuable orebody knowledge which can be included in the geological block model and downstream process predictions to optimise performance based on vast quantities of historic performance data and non-intuitive relationships between a large number of geological, metallurgical and process parameters.
MAXTA allows engineers and plant operators to access practical, relevant process predictions via the MAXTA applications portal, and to generate geological block models with additional detail for use in planning and design functions via common workflows already used in the mine and plant and integrated with tools such as Maptek’s Vulcan or OSISoft Pi software. Packaging the machine learning and digital twin capabilities within a software product that can be used in day to day use by mining engineers and plant operators means that the customer does not need expertise in data science and machine learning and provides a sustainable, supported, commercial solution to value chain optimisation that can generate value from day one, significantly faster and with less expense than in house point solutions are able to achieve.
Originality of PETRA’s Innovations
PETRA is the only company that enables dynamic mine value chain optimisation through our MAXTA digital mine to mill and pit to port enterprise software. MAXTA is unique in that respect, and also in the fact that it is a packaged commercial software solution that can be easily deployed into a mining operation to deliver benefits of machine learning to day to day operations. We are the only company in the world that is optimising for ore variability across the whole value chain. This is a new concept that has not been achievable in the past due to the difficulty of combining all of the orebody knowledge with material tracking and process historian data in large quantities representing a significant sample of the orebody material. More traditional process prediction techniques are based on a very small number of sample data points and do not properly address the full range of variability that exists in all orebodies, and are therefore not accurate for all material domains present in the geology. The variability associated with an orebody is a challenge unique to mining and MAXTA has been developed specifically to help miners understand and manage that.
The concepts embodied in MAXTA of capturing large quantities of historical data and using this to predict the behaviour of ore variability in various processes extended from the PhD studies of PETRA founder and CEO Dr Penny Stewart. Penny has an extensive background and career in engineering and rock mechanics both within industry and research roles. Penny started PETRA to commercialise this work, seeing and opportunity for the industry to gain dramatic benefits by optimising processes to suits the specific rock being mined. MAXTA was developed over the past 5 years exclusively by PETRA and was first commercialised in 2017. It has evolved during that time to be a fully supported, scalable enterprise solution which is being used in some very large operations across large segments of the mine value chain now. The MAXTA solution is relevant to a number of different processes in mining. Other features of MAXTA that differentiate the solution include:
- Automated and near-real-time data correction, validation and calibration for dynamic live digital twin prediction and optimisation in the processing plant.
- Sensor health is built into our algorithms and this enables algorithms to be automatically switched in or out depending on the sensor health.
- Platform agnostic integrated with existing workflows and mining platforms for seamless workflows and decision support e.g. Maptek, OSISoft/AVEVA.
- Glass box solution with visibility for trust and transparency
MAXTAConnect API enables automated feedback of actual performance into mine planning every time the block model is updated – over a month models have accuracy of +/-1-2% for mill throughput and dig rates (based on actual customer reconciliations and validation of model performance).
For the first time mining companies can see how every blast performed downstream – MAXTADrill&Blast data visualisation allows each blast to be evaluated as a function of crusher throughput, mill throughput or similar downstream processes.
Impact of PETRA’s Innovations
MAXTA is now, and has the potential to, change the way mine optimisation and planning decisions are made, with significant positive impact on the bottom line performance of a mining operation. This is because MAXTA can predict, with great accuracy, how all the various characteristics of the ore being mined will combine and behave for a given downstream process, and can make this information known and usable at the planning and decision making stages.
For example, the throughput of a crusher can be predicted to less than 2% accuracy as a function of the specific rock type and the blast design details used. This allows planners to understand the production rates, as well as to selectively design blast patterns and charges to optimise for a given result at the crusher. Previously throughput predictions were either fixed numbers or simplistic estimates based on some empirical models and very few samples. MAXTA can be applied at various stages of the value chain, allowing for a dramatic increase in the precision of how a mining operation is run, as well as reducing variability and uncertainty in the mine planning process. Applied consistently, this can have huge positive impact on the production results and operating costs at a mine.
MAXTA can be applied at a large number of operational areas in the mine value chain. MAXTA is delivered as a number of modules allowing modelling and prediction of various processes and details. MAXTAGeomet describes the geometallurgical behaviour of ore in downstream processes such as beneficiation plants and adds this detail to the block model. MAXTADrill&Blast advises engineers of downstream process performance for blast design parameters selected for a given rock domain type. MAXTALoad&Haul predicts dig rates for certain rock based upon blasting and rock types. MAXTAProcess predicts material process performance for equipment such as crushers and mills and plants. These applications have been developed as a result of demand and opportunity in the market, and all have been successfully delivered and deployed commercially. MAXTA can be applied to hard rock mines and coal mines and to open cut or underground operations, and is scalable across very large, complex operations. Our largest customer now has over 20 instances running across a number of process stages at their operations in WA and is considering scope to expand this further.
There are a wide range of potential applications for MAXTA within other process areas. The fundamental concept of using machine learning to create insights about the behaviour of a certain process with a certain material is embodied within MAXTA, and there are a wide range of potential applications that could be considered, including energy use, water use, fleet schedule and utilisation optimisation and more.
Practicality of Application
MAXTA is delivered via the MAXTA Portal. This is a web application providing end users with an intuitive, simple interface to ensure that the correct processes, data and models are applied and run together depending upon the application area being considered. This allows scalable use across multiple orebodies, multiple processes, multiple sites and unlimited users. The MAXTA Portal also provides secure logins for users and secure data upload and download to ensure accuracy and security for a customers data. MAXTA runs all of the data processing and machine learning processes on Amazon Web services and provides user specific interfaces to functional solutions.
MAXTA consumes common data formats found in mine operations and needs no data cleaning or manipulation to apply the solution. Many data science projects require significant data cleaning to make the data ‘useable’. PETRA has solved this with MAXTA by applying years of experience with mining data of all types and building the ability into the various algorithms within MAXTA to work with the raw data and so the solution requires no interference with the data at any stage. This adds significantly to the sustainability and ongoing value of the solution as new data can be applied with consistent treatment to older data and continuity in the results can be assured. It also doesn’t require a team of data scientists to be employed just to maintain the solution.
The results are presented either in a UI specific to the user and application being considered, or are integrated directly into the formats where users will need to apply those using the existing tools they have available. This is all automatic and takes very little time to complete – typically minutes.
As an example, drill and blast users are presented with a handbook style list of recommendations for drill and blast design parameters together with MAXTA predictions for downstream performance. This could be crusher or mill throughput, and the results can be used immediately in the blast design process. PETRA has integrated MAXTA with Maptek’s Vulcan mine planning software so that engineers can access these predictions in the same environment as they are doing the blast design work.
The risk of deployment of MAXTA to an ongoing operation is very low, as existing workflows are not disrupted, but only benefit from additional, more accurate information and predictions. Uncertainty is reduced.
Orebody knowledge and process setpoint predictions generated by MAXTA are written in standard industry formats such as Vulcan BMF files and OSISoft Pi data so that the intended users of this information can immediately access it using the tools they are already using and generate value using the results immediately. PETRA maintains partnerships with Maptek and OSISoft to ensure the benefits to users of these systems are maintained and maximised.
MAXTA is the most cost effective method of achieving mine to mill optimisation we are aware of, both in terms of time and money and effort. A MAXTA solution is usually delivered in around 14 weeks from commencement of the delivery and costs a fraction of comparable studies and plant surveys and more complex empirical modelling techniques intended to simulate process performance.
Measurability of Benefits
There are distinct, measurable financial benefits inherent in the use of MAXTA. Besides the cost of implementation and licencing being significantly lower than alternate mine to mill optimisation techniques, the solution can have a dramatic impact on an operation’s financial and production performance.
MAXTA can be used to predict metal recovery in mines where variable recovery performance has been unpredictable. This variance and inability to predict the recovery performance has the effect of causing some ore blocks to be classified as ore when they may in fact be waste, and vice versa. The result of this is that mine planning practitioners will schedule what are effectively waste blocks to be processed without knowing that the actual metal recovery will be very low. Over the typical life of a mine, this may result in hundreds of millions of dollars in value being lost. MAXTA can help avoid this.
MAXTA is able to accurately predict the recovery performance in this example and enable reclassification of specific problematic ore blocks such that only ore is fed into the plant, and some blocks which may have been considered waste are also reclassified as ore if the recovery performance is predicted to improve. Over the life of a mine this can cause significant financial impact, not only due to the metal recovery directly affecting production results, but also on the orebody knowledge and mine optimisation with consideration for the actual economic value of each block.
Selection of blasting products and explosive products can be managed with reference to downstream performance using MAXTADrill&Blast, influencing explosives cost over the long term and MAXTALoad&Haul can predict dig rates accurately, enabling prevention of over trucking or under trucking at diggers.
Crusher and mill feed, blockages and throughput can be managed optimally using MAXTA to ensure these processes, typically the bottlenecks in a mining operation are optimised and planning can be done to maximise the use of these processes. SAG mill downtime can cost hundreds of thousands of dollars an hour. MAXTA enables feed rates to be optimised to avoid this downtime.
MAXTAProcess can predict and optimise plant setpoints to maximise metal recovery or coal yield through floatation circuits, CIP plants and more, saving on reagent cost and maximising metal recoveries or coal yield. Better ability to plan and meet product specification and lower waste using MAXTA means greater economic exploitation of the resource in general, adding significantly to a companies ability to meet NPV predictions in the long term.
The wider benefits for the industry in general are global in their relevance, and extend to the potential to predict and optimise the entire value chain using MAXTA. The ability for MAXTA to create feedback loops and relationships between the geological detail in the orebody and the performance of downstream systems and processes enables significantly more accurate management of mining operations to target quality through better certainty about performance, less variation from plan and more informed decision making.
MAXTA could be applied to other industries or disciplines, however PETRA’s domain expertise is in mining, and we have not attempted this to date. The concept of process control by use of a machine learning algorithm combining highly variable and large numbers of inputs could be applied to many other industries. The most logical would be the civil engineering / construction or oil and gas industries, where the fundamental variances in geology are still relevant.
MAXTA can be applied in mining operations anywhere in the world. PETRA has delivered solutions in Australia, Asia, North and South America. The solutions are all web applications and can be implemented and supported globally. The majority of mining processes are common in various combinations across the world and MAXTA’s ability to consume a wide range of existing mining data types means it is applicable to most mining operations immediately.
Industry processes across a number of areas in the mining value chain can be revolutionised by the use of MAXTA. Mine Planning and scheduling decisions can now be made with accurate reference to the process performance downstream of the mine. Whereas previously planning stopped at stockpile figures and production volumes, planning can now be done taking into account throughput and recovery – allowing a holistic view of the entire mine operation to be planned for. Processes such as drill and blast and load and haul can be optimised for the specific rock being worked on and loaded, and plant operations can be managed with setpoints recommended by MAXTA to target optimised process performance in near real time.
All of these advances are enabled by MAXTA’s use of the entire orebody knowledge about the mine. The investment a mining company makes in creating a geological block model is vast – many tens of millions of dollars, and long term, high value decisions are made on the basis of that information. MAXTA enables two key benefits which are not available otherwise. Firstly, the use of that orebody knowledge in conjunction with the performance data about other parts of the mine to learn and predict about future performance, and secondly, to enhance that orebody knowledge by making block by block predictions about the orebody and it’s behaviour in those processes for the benefit of better decisions at the planning stages.